Transcript
Pennsylvania
Economic Association
Annual Conference
May 30-June 1, 2013
The University of Scranton
Scranton, PA
Proceedings of the Pennsylvania Economic Association
PENNSYLVANIA
ECONOMIC
ASSOCIATION
ANNUAL CONFERENCE
May 30-June 1, 2013
The University of Scranton
Scranton, Pennsylvania
Visit the Pennsylvania Economic Association Home Page at
http://aux.edinboro.edu/pea/index.html Acknowledgements: Thanks to the Federal Reserve Bank of New York Cleveland and Dr.
Michael Mensah, Dean, Kania School of Management, The University of Scranton, and Fr.
Kevin P. Quinn, S.J., President, The University of Scranton for their support of this
conference.
Proceedings of the Pennsylvania Economic Association
Pennsylvania Economic Association: 2012-2013
Executive Board President: Tracy Miller, Grove City College
President-Designate: William Bellinger, Dickinson College
Vice President, Program and Proceedings: Sandra Trejos, Clarion University of Pennsylvania
Vice President, Publicity: Deborah Gougeon, The University of Scranton
Vice President, Membership: Vacant
Secretary: Stephanie Brewer, Indiana University of Pennsylvania
Treasurer: Steven Andelin, Penn State - Schuylkill
Co-Editors, Pennsylvania Economic Review: Thomas Tolin & Orhan Kara, West Chester University
Webmaster: Michael Hannan, Edinboro University of Pennsylvania
Immediate Past President: Orhan Kara, West Chester University of Pennsylvania
Board of Directors Ron Baker, Millersville University
Charles Bennett, Gannon University
Kosin Isariyawongse, Edinboro University
John McCollough, Penn State - Lehigh Valley
Brian O'Roark, Robert Morris University
Mark Schweitzer, Federal Reserve Bank of Cleveland
Yaya Sissoko, Indiana University of Pennsylvania
Luke Tilley, Federal Reserve Bank of Philadelphia
Ralph Ancil, Geneva College
Soma Ghosh, Albright College
Jolien Helsel, Youngstown State University
Ex-Officio Directors Abdul Pathan, Pennsylvania College of Technology
Andrew Economopoulos, Ursinus College
Andrew Hill, Federal Reserve Bank of Philadelphia
Bijou Yang-Lester, Drexel University
Brian O’Roark, Robert Morris University
Brian Sloboda, US Postal Service
Daniel Y. Lee, Shippensburg University
David Culp, Slippery Rock University
David Yerger, Indiana University of Pennsylvania
Donald Dale, Muhlenberg College
Donna Kish-Goodling, Muhlenberg College
Elizabeth Hill, Penn State-Mont Alto
Gayle Morris, Edinboro University
Gerald Baumgardner, Pennsylvania College of
Technology
Heather O'Neill, Ursinus College
Ioannis N. Kallianiotis, University of Scranton
Jacquelynne McLellan, Frostburg State University
James Dunn, Edinboro University
James J. Jozefowicz, Indiana University of
Pennsylvania
John A. Sinisi, Penn State University-Schuylkill
Johnnie B. Linn III, Concord University
Joseph Eisenhauer, University of Detroit Mercy
Kenneth Smith, Millersville University of
Pennsylvania
Lawrence Moore, Potomac State College of West
Virginia University
Lynn Smith, Clarion University
Margarita M. Rose, King's College
Mark Eschenfelder, Robert Morris University
Mehdi Hojjat, Neumann College
Natalie D. Reaves, Rowan University
Osman Suliman, Millersville University
Patrick Litzinger, Robert Morris College
Paul Woodburne, Clarion University
Robert D'Intino, Rowan University
Robert Liebler, King's College
Rocky Jui-Chi Huang, Pennsylvania State University -
Berks Campus
Roger White, Whittier College
Stanley G. Long, University of Pittsburgh/Johnstown
Tahereh Hojjat, DeSales University
Thomas O. Armstrong, Pennsylvania Department of
Community & Economic Development
William F. Railing, Gettysburg College
William Sanders, Clarion University
Editor’s Introduction and Acknowledgments
Sandra Trejos, Ph. D.
Editor of Proceedings
2013 Annual Conference of the Pennsylvania Economic Association
This volume contains papers presented at the 2013 Annual Conference of the Pennsylvania Economic
Association (PEA) held at the University of Scranton from May 31 to June 2, 2013. Only the papers
and discussions submitted according to manuscript guidelines are included in these proceedings. Not
every paper listed in this program was submitted for inclusion though. I thank Lori Klepfer, secretary
for the Department of Economics, and Yang Yang, my graduate assistant, in the College of Business
Administration at Clarion University of Pennsylvania for their excellent assistance in preparing these
proceedings.
University faculty, research professionals, graduate students, and undergraduate students gathered in
Scranton, Pennsylvania, to present their scholarly work from different corners of our country and the
world. Sessions allowed not only the presentation of fine work, but also the constructive criticism and
discussion present in any productive and meaningful professional meeting. The different concurrent
sessions were properly complemented by our Luncheon Speaker Mr. Peter Danchak, President of
PNC Bank, and the Federal Reserve Bank of New York lecture featuring Mr. Michael F. Silva, Senior
Vice President in the Financial Institutions Supervision Group. The annual conference ended with a
general meeting in which different aspects of our association were discussed with all members in
attendance.
The PEA would like to take this opportunity to thank Deborah Gougeon for hosting the 2013
meetings in such an organized and smooth manner. Her planning, energy, and organization skills
were crucial for another year of success for our conference. Thanks are also extended to the Federal
Reserve Bank of New York and Dr. Michael Mensah, Dean of the Kania School of Management, The
University of Scranton, and Fr. Kevin P. Quinn, S.J., President of The University of Scranton for their
support of this conference. Thanks, finally, to the PEA board and directors for their continued
commitment to and support of our organization, and all participants for their enthusiasm and
dedication to knowledge shown in their work and lively discussions. You make of this conference a
valuable event and we would like to invite you to continue being part of such a great group of
scholars. See you next year!
Table of Contents
Conference Agenda……………………………………………………………………………………………… page 1
The Effects of Power, Prestige, and Performance on Salary in the National Hockey League
Ashley M. Alt and Daniel R. Oberkofler, Indiana University of Pennsylvania …………………… …… page 16
Pennsylvania Tax Simplification: Nuisance Tax Credit, Obsolete Taxation and Administration Provision
Repeals, Including Proper Placement within the Tax Reform Code
Thomas O. Armstrong, Commonwealth of Pennsylvania, Department of Revenue……………… …… page 28
Business and Real-Estate Price Cycles across the Us: Evidence from a Markov-Switching Regression Exercise
Aram Balagyozyan, University of Scranton, Christos Giannikos, Baruch College, and Kyoko Mona,
Manhattanville College………………………………………………………………………………………… page 34
A Model of Relative Consumption
Chong Hyun C. Byun, Wabash College……………………………………………………………………… page 41
Peer Mentor Development as Secondary Leaders at the University Level
Dana D’Angelo and Susan Epstein, Drexel University…………………………………………………… page 48
Climate Neutrality in the Higher Education Sector: Making the Commitment
By Soma Ghosh, Discussion by Solomon T. Tesfu, Mount St. Mary’s University……………………… page 52
The U.S. Dollar as an International Currency Reserve and Its Continuous Depreciation
Ioannis N. Kallianiotis, University of Scranton……………………………………………………………… page 54
Trade Flows between S. Korea and the U.S.A.
Orhan Kara, West Chester University………………………………………………………………………… page 67
A Simple Model of Baseball Desegregation
Timothy F. Kearney & David Gargone, Misericordia University………………………………………… page 78
Impact of Migrant Remittances on Education Outcomes
Tanu Kohli, University of Illinois……………………………………………………………………………… page 87
Manufacturing Productivity in Pennsylvania
James A. Kurre, Penn State Erie, The Behrend College…………………………………………………… page 97
Force-Using Firms in the Competitive Equilibrium When the Only Publicly Provided Good is Envy
Gratification
Johnnie B. Linn III, Concord University…………………………………………………………………… page 109
The Effects of Increases in Cigarette Prices on Smoking Behaviors: Estimates Using MSA as a Natural
Experiment
Zhen Ma, Misericordia University…………………………………………………………………………… page 114
Mortgage Cash Flow Analysis and Pricing Using CAS (Collateral Analysis System)
Stephen M. Mansour and Riaz Hussain, University of Scranton………………………………………… page 123
Discussant Comments: The Geographic Concentration of Economic Activity
Stephen M. Mansour, University of Scranton…………………………………………………………………page 129
Credence Goods and State Mandated Vehicle Safety Inspections: How Non-Profit Inspection Services Can
Correct For Market Failure
John McCollough, Lamar University………………………………………………………………………… page 130
Some Possible Reasons for the Irrational Choice of Gift Cards
David Nugent, Robert Morris University……………………………………………………………………… page 137
Service Quality in the U.S. Airline Industry: Factors Affecting Customer Satisfaction
Logyn Pezak and Rose Sebastianelli, University of Scranton ……………………………………………… page 142
What Affects New Zealand Wine Prices? Estimation of the Effects of Sensorial, Reputational, Objective, and
Quality Factors in the Hedonic Price Model
Angela M. Rowland, Indiana University of Pennsylvania ………………………………………………… page 150
Author Index………………………………………………………………………………………………………page 162
Proceedings of the Pennsylvania Economic Association 1
Pennsylvania Economic Association
2013 CONFERENCE AGENDA
THURSDAY, May 30
4:00 pm - 9:00 pm Registration (Brennan Hall, 1st Floor Lobby)
5:00 pm - 8:00 pm Board of Director's Dinner/Meeting (Brennan Hall 509)
6:00 pm - 10:00 pm Reception (Brennan Hall 5th Floor Lobby)
FRIDAY, May 31
8:00 am - 12:00 pm, 2:00pm - 4:00pm Registration (Brennan Hall, 1st Floor Lobby)
7:30 am - 10:30am Continental Breakfast (Brennan Hall, 2nd Floor Lobby)
9:00 am - 10:15 am Concurrent Sessions (Brennan Hall)
10:15 am - 10:30 am Break (Brennan Hall, 2nd Floor Lobby)
10:30 am - 11:45 pm Concurrent Sessions (Brennan Hall)
12:00 pm - 1:45 pm Luncheon and Speaker - Mr. Peter Danchak, President of PNC
Bank, NE PA Region (Brennan Hall 509)
2:15 pm - 3:30 pm Concurrent Sessions (Brennan Hall)
3:30 pm - 3:45 pm Break (Brennan Hall, 2nd Floor Lobby)
3:45 pm - 4:45 pm Fed Lecture Featuring Mr. Michael F. Silva, Senior Vice President in
the Financial Institutions Supervision Group, Federal Reserve Bank of New York
(Pearn Auditorium, Brennan Hall 228)
5:00 pm - 8:00 pm Fed Sponsored Reception (Brennan Hall, 5th Floor Lobby)
SATURDAY, June 1
7:30 am - 10:30 am Registration (Brennan Hall, 1st Floor Lobby)
7:30 am - 9:00 am Continental Breakfast (Brennan Hall, 2nd Floor Lobby)
9:00 am - 10:15 am Concurrent Sessions (Brennan Hall)
10:30 am - 11:00 am General Membership Meeting (Pearn Auditorium, Brennan Hall
228)
11:15 am - Closing
Proceedings of the Pennsylvania Economic Association 2
FRIDAY, May 31, 2013
Conference Registration 8:00 a.m. – noon, 2:00-4:00 p.m., Brennan Hall Lobby, first
floor
7:30 a.m. – 10:30 a.m. Continental Breakfast (Brennan Hall, second floor)
Sessions F1: Friday, May 31, 2013, 9:00 a.m. – 10:15 a.m.
Session F1A: Housing Market
Location: Brennan Hall 102 Chair: Shuang Feng – Edinboro University
Mortgage Cash Flow Analysis and Pricing Using CAS
Stephen Mansour Rutgers University
The Effects of Housing Tenure on Labor Market Outcomes: Evidence from the USA
Md. Alauddin Majumder Middle Tennessee State University
Business and Real-Estate Price Cycles across the US: Evidence from a Markov-Switching Exercise Aram
Balagyozyan University of Scranton
Discussants:
Michael Hannan – Edinboro University
Patrick Litzinger – Robert Morris University
Steven Breslawski – College at Brockport, State University of New York
Session F1B: Economic Development I
Location: Brennan Hall 103 Chair: Yaya Sissoko – Indiana University of PA
Private Returns to Investment in Education in Cameroon; A Quintile Regression Analysis
Yaya Sissoko Indiana University of Pennsylvania
Wilfred Awung University of Buea, Cameroon
Antecedents and Consequences of the Aging Developed World: Implications for Business Systems
Abhijit Roy University of Scranton
Health Gap between Countries: Does Globalization Matter?
Khaled Elmawazini Gulf University for Science and Technology
Discussants:
Minh Tam Schlosky – West Virginia University
Tanu Kohli – Rutgers University
David Latzko – Penn State York
Proceedings of the Pennsylvania Economic Association 3
Session F1C: Women and Gender Economics
Location: Brennan Hall 105 Chair: Alberto Posso - RMIT University
The Latin American Mystique: Discrimination against Women in Ecuador's Labor Market Alberto Posso
RMIT University
An analysis of Gender Wage Gap over time in China’s urban labor market
Zhonghui Liu State University New York at Binghamton
Measuring Links between Labor Monopsony and the Gender Pay Gap in Brazil
Brandon Vick Indiana University of Pennsylvania
Discussants:
Charles Telly – SUNY Fredonia
Debarshi Indra – SUNY Buffalo
Zhonghui Liu – State University New York at Binghamton
Session F1D: Pricing and Market Reaction
Location: Brennan Hall 202 Chair: Travis Yates – Penn State Erie
The Effects of Increases in Cigarette Prices on Smoking Behaviors: Estimates Using MSA as a Natural Experiment
Zhen Ma Misericordia University
Psychological impacts on markets' response to earnings
Qian Hao and Anthony Liuzzo Wilkes University
A Pricing Course with Client-Based Experiential Learning
Robert Schindler Rutgers University - Camden
Alternatives for Highway Financing: Impact on Equity and Efficiency
Tracy Miller Grove City College
Discussants:
Ruttana Ruttanajarounsub – Edinboro University
Kosin Isariyawongse – Edinboro University
Travis Yates – Penn State Erie
Zhen Ma – Misericordia University
Proceedings of the Pennsylvania Economic Association 4
Session F1E: Student Session I
Location: Brennan Hall 203 Chair: Abdul Pathan – Pennsylvania College of Technology
Effects of Power, Prestige and Performance on Salary in the National Hockey League
Daniel Oberkofler and Ashley Alt Indiana University of Pennsylvania
Honor Killings: The Economic Parasite of Pakistan
Lauren Tassone Clarion University of PA
Challenges and Opportunities of Mass Customization: A Chinese Market Perspective
Huilan Zhang University of Toledo
Environmental Degradation and the Brazilian Economy
Erin Krotoszynski Clarion University of Pennsylvania
What Affects New Zealand Wine Prices? Estimation of the Effects of Sensorial, Reputational, and Quality Factors in the
Hedonic Price Model
Angela Rowland Indiana University of Pennsylvania
Discussants:
Deborah Gougeon – University of Scranton
Johnnie Linn, Concord University Margarita Rose – King’s College
Abdul Pathan – Pennsylvania College of Technology
John Kallianiotis – University of Scranton
Sessions F2: Friday, May 31, 2013 10:30 a.m. – 11:45 a.m.
Session F2A: Economic Development II
Location: Brennan Hall 102 Chair: Tracy Miller – Grove City College
Trade Flows between Korea and the U.S.
Orhan Kara West Chester University
Is There Moral Hazard in the HIPC Initiative Debt Relief Process?
Minh Tam Schlosky West Virginia University
Impact of Migrant Remittances on Fertility and Education Outcomes
Tanu Kohli Rutgers University
Bank Lending Channel in MENA Countries: Evidence from Dynamic Panel Model
Aram Belhadj University of Orléans
Proceedings of the Pennsylvania Economic Association 5
Discussants:
Jeremy Schwartz – Loyala University of Maryland
Sabri Yilmaz – Lycoming College
Orhan Kara – West Chester University
Tracy Miller – Grove City College
Session F2B: Taxation
Location: Brennan Hall 103 Chair: David Buehler – Penn State Harrisburg
Economic Impact of the Earned Income Tax Credit (EITC) for Erie County, Pennsylvania
Kosin Isariyawongse Edinboro University
Title Cost Function Estimations: Regularity Conditions and Cost Minimization
Kwami Adanu Gimpa Business School, Ghana
Pennsylvania Tax Simplification: Nuisance Tax Credit, Obsolete Taxation and Administration Provision Repeals
Thomas Armstrong PA Department of Revenue
Discussants:
James Kurre – Penn State Erie, The Behrend College
Aram Balagyozyan – University of Scranton
Elsy Kizhakethalackal – Bowling Green State University
Session F2C: Financial Economics
Location: Brennan Hall 105 Chair: Abhijit Roy -- University of Scranton
Informed Trading at Capitol Hill: Evidence from Congressional Trading over the 2004-2010 Period
Serkan Karadas West Virginia University
A Note on an Equation of Modigliani and Miller
Steven Andelin Pennsylvania State University
Help Your Students Realize Their Retirement Dreams by Quantifying the Cost of Procrastination
Jonathan Kramer and John Walker Kutztown University
Discussants:
Lee Siegel – PA Dept of Labor & Industry
Steven Breslawski – College at Brockport, State University of New York
Abhijit Roy – University of Scranton
Proceedings of the Pennsylvania Economic Association 6
Session F2D: Regional Economic Development
Location: Brennan Hall 202 Chair: Stephen Mansour - University of Scranton
Determinants of Economic Growth in ECOWAS Countries
Brian Sloboda US Department of Labor
Yaya Sissoko Indiana University of PA
The Geographic Concentration of Economic Activity across the Eastern United States, 1820-2010
David Latzko Penn State York
Credence Goods and State Mandated Vehicle Safety Inspections; How Non-Profit Inspection services can correct for market
failure
John McCollough Lamar University
Discussants:
John McCollough – Lamar University
Stephen Mansour – University of Scranton
Yaya Sissoko – Indiana University of PA
Session F2E: Economics of Education
Location: Brennan Hall 203 Chair: Sandra McPherson - Millersville University
Unders and Overs: Using a Dice Game to Illustrate Basic Probability
Sandra McPherson Millersville University
Students Perceptions of Hybrid Course Learning: A Case Study in Principles of
Microeconomics
Jui-Chi Huang Penn State Berks
Climate Neutrality in the Higher Education Sector: Making the Commitment
Soma Ghosh Albright College
Discussants:
Tufan Tiglioglu – Alvernia University
Huilan Zhang – University of Toledo
Solomon Tesfu – Mount St. Mary's University
Proceedings of the Pennsylvania Economic Association 7
Lunch Speaker
Friday, May 31
12:00pm Brennan Hall 509
Mr. Peter Danchak
President of PNC Bank, NE PA Region
Mr. Danchak joined PNC Bank in 1984 and has held various positions of responsibility in corporate
banking. He was named regional president of the Northeast PA market of PNC Bank in January 2001.
Mr. Danchak currently serves as co-chair of the Pennsylvania Early Learning Investment Commission and
is a member of the Executive Leadership Council of Pre-K Counts in Pennsylvania. He sits on the board
of trustees at Marywood University and Scranton Prep. He is a member of the board of directors of the
Pennsylvania Bankers Association, Blue Cross of Northeastern Pennsylvania, the ARC of Northeastern
Pennsylvania Foundation, the Luzerne Foundation, the Northeast Regional Cancer Institute, the Regional
Chamber Partnership and Scranton Lackawanna Industrial Building Company. He also serves on the
President’s Advisory Council for Keystone College and the Kania School of Management at the
University of Scranton.
Mr. Danchak previously served as a member of the board of directors of the F.M. Kirby Center for the
Performing Arts, King’s College, Keystone College, Johnson College, Junior Achievement of
Northeastern Pennsylvania, the Greater Pittston Chamber of Commerce, the Greater Scranton Chamber of
Commerce and the Greater Wilkes-Barre Chamber of Business and Industry.
Mr. Danchak received a Bachelor of Science degree in accounting from the University of Scranton.
Proceedings of the Pennsylvania Economic Association 8
Sessions F3: Friday, May 31, 2013 2:15 p.m. – 3:30 p.m.
Session F3A: Education Performance
Location: Brennan Hall 102 Chair: Zhonghui Liu - State University New York at Binghamton
Preliminary Results of Factors Contributing to Grade Changes
Robert Balough Clarion University of Pennsylvania
Rod Raehsler Clarion University of Pennsylvania
Teenage Socializing Behavior and Schooling Outcomes for American Youth
Solomon Tesfu Mount St. Mary's University
Analysis of Differences in Learning Outcomes: Online versus In-class Course
Sunita Mondal and Soma Ghosh University of Pittsburgh at Greensburg
Retention of Microeconomics Knowledge by Content and Cognitive Constructs
Steven Breslawski College at Brockport, State University of New York
Discussants:
Zhonghui Liu – State University New York at Binghamton
Brandon Vick – Indiana University of Pennsylvania
Steven Breslawski – College at Brockport, State University of New York
Serkan Karadas – West Virginia University
Session F3B: Health and Public Economics
Location: Brennan Hall 103 Chair: Khaled Elmawazini - Gulf University for Science and Technology
Force-Using Firms in the Competitive Equilibrium when the Only Publicly Provided Good is Envy Gratification
Johnnie Linn Concord University
Impact of Disaggregated Health-aid on Child Vaccinations: A Quantile Regression Analysis
Elsy Kizhakethalackal Bowling Green State University
Length of life, Individual Benefit Streams, and the Social Discount Factor for Long Lived Public Policies
William Bellinger Dickinson College
Discussants:
Khaled Elmawazini - Gulf University for Science and Technology William Bellinger -- Dickinson College
John Walker -- Kutztown University
Proceedings of the Pennsylvania Economic Association 9
Session F3C: Regional Economics
Location: Brennan Hall 105 Chair: Robert Schindler – Rutgers University
Identifying and Quantifying a Local Economy’s Exports
Travis Yates Penn State Behrend
Employment and Population Growth in Florida’s Counties
Sooriyakumar Krishnapillai American University of Nigeria
Manufacturing Productivity in Pennsylvania
James Kurre Penn State Erie, The Behrend College
Discussants:
Aram Belhadj -- University of Orléans
Dana D’Angelo -- Drexel University LeBow College of Business
Xuebing Yang -- Penn State Altoona
Session F3D: Economic Utility
Location: Brennan Hall 202 Chair: Margarita Rose – King’s College
A Model of Relative Consumption
Chong Hyun Byun Wabash College
Some Possible Reasons For The Irrational Choice of Gift Cards
David Nugent Robert Morris University
Service Quality in the U.S. Airline Industry: Factors Affecting Customer Satisfaction
Logyn E. Pezak and Rose Sebastianelli University of Scranton
Discussants:
Margarita Rose – King’s College
Abhradeep Maiti – Middle Tennessee State University
Sabri Yilmaz – Lycoming College
Session F3E: Labor and Demographic Economics
Location: Brennan Hall, 203 Chair: Zhen Ma -- Misericordia University
A Simple Model for Baseball Desegration
Timothy Kearney Misericordia University
David Gargone Misericordia University
Trials and Tribulations Of Analyzing Ferreted Pennsylvania Data: A Case Study
Lee Siegel Center for Workforce Information & Analysis
Recidivism of Juvenile Offenders
Proceedings of the Pennsylvania Economic Association 10
David Kalist Shippensburg University
Discussants:
Zhen Ma -- Misericordia University Steven Andelin -- Pennsylvania State University
Brian Sloboda -- US Department of Labor
Session F3F: Pedagogical Panel
Location: Brennan Hall 205 Chair: Abdul Pathan
Making the Principles of Economics Class Interesting to your Students
A group of faculty members will share methods they use to engage students in the classroom
Federal Reserve Lecture
Friday, May 31
3:45pm Pearn Auditorium, Brennan Hall 228
Mr. Michael F. Silva
Senior Vice President in the Financial Institutions Supervision Group, Federal Reserve
Bank of New York
Mr. Silva joined the bank in August 1992 as a law clerk in the Legal Group. He held positions with increasing
responsibilities in the legal group and in September 1995 was appointed an officer of the bank with the title of
Counsel. In December 1998, Mr. Silva was promoted to assistant vice president and became the lead counsel for
the Bank’s international account relationships and currency distribution. He was promoted to vice president in
December 1999 with those same responsibilities. In June 2006, Mr. Silva was promoted to senior vice president and
Proceedings of the Pennsylvania Economic Association 11
moved from the Legal Group to the Executive Group, where he served as chief of staff for president Geithner and
subsequently president Dudley. Mr. Silva’s tenure as Chief of Staff included all of the 2008 - 2009 financial crisis.
Mr. Silva concluded his assignment as chief of staff in September 2010 and moved to the Financial Institutions
Supervision Group, where he currently serves as the Senior Supervisory Officer for The Goldman Sachs Group. As
a collateral duty, he served as assistant corporate secretary of the bank from December 1995 to December 1999.
Mr. Silva holds a B.S. degree from the United States Naval Academy and a J.D. from Columbia Law School. He is
also a graduate of the Harvard Business School’s Advanced Management Program.
In 2004, Mr. Silva received the Department of Defense Joint Civilian Service Medal and the Secretary of the
Treasury’s Honor Award, both for his service in Iraq as a coalition advisor to the Central Bank of Iraq. Mr. Silva is
also the author of “A Central Banker in Iraq”, Journal of International Business & Law, Spring 2004.
Prior to attending law school and joining the bank, Mr. Silva served as an officer in the United States Navy from
May 1983 to September 1989. During that period, he was assigned to Fighter Squadron 142 as an F-14A Tomcat
Radar Intercept Officer and also to the On-Site Inspection Agency, a joint military and civilian agency in
Washington DC responsible for conducting the first ever arms control verification inspections in the former Soviet
Union. Mr. Silva is a 1986 graduate of the Navy Fighter Weapons School (a.k.a “Topgun”).
5 –8 P.M. Brennan Hall Lobby, fifth floor Reception hosted by the Federal Reserve Bank
Proceedings of the Pennsylvania Economic Association 12
SATURDAY, June 1, 2013 ~ 7:30 – 10:30 A.M. Conference Registration-Brennan Hall Lobby, first floor
Continental Breakfast-Brennan Hall Lobby, second floor ____________________________________________
Sessions S1: Saturday, June 1, 2013 9:00 a.m. – 10:15 a.m.
Session S1A: Minimum Wage and Labor Market
Location: Brennan Hall 102 Chair: Michael Hannan, Edinboro University of PA
Optimal Unemployment Insurance: When Search Takes Effort and Money
Jeremy Schwartz Loyala University of Maryland
Labor Demand Elasticity in the United States
Abhradeep Maiti Middle Tennessee State University
A Wage You Can Live On Minimum Wage Debate: Public Perceptions vs. Academic Findings
Jui-Chi Huang Penn State Berks
Discussants:
Robert Balough, Clarion University of PA
Michael Hannan, Edinboro University of PA
Sooriyakumar Krishnapillai, American University of Nigeria
Session S1B: International Economics
Location: Brennan Hall 103 Chair: Jonathan Kramer, Kutztown University
Patterns of Product Level Trade
Xuebing Yang Penn State Altoona
Trade Creation, Diversion, and Displacement: A Shift-Share Analysis Of EU Enlargement
David Buehler Penn State Harrisburg
The U.S. Dollar as an International Currency Reserve and its Continuous Depreciation
John Kallianiotis University of Scranton
Discussants:
Tufan Tiglioglu, Alvernia University
David Nugent, Robert Morris University
Jonathan Kramer, Kutztown University
Proceedings of the Pennsylvania Economic Association 13
Session S1C: General Economics
Location: Brennan Hall 105 Chair: William Bellinger, Dickinson College
The Mythological Hero and Its Development from Socrates to Dante, to Mandeville, to Adam
Smith, to Hayek and the Modern Manager
Charles Telly SUNY Fredonia
Developing University Peer Mentors as Second Chair Leaders
Dana D'Angelo Drexel University LeBow College of Business
Additive Manufacturing and Costs of Production
Patrick Litzinger Robert Morris University
Discussants:
Tracy Miller – Grove City College
David Latzko – Penn State York
David Kalist – Shippensburg University
10:30 a.m.-11:00 General Membership Meeting
Brennan Hall 228
This Annual Business Meeting of the General Membership of the Pennsylvania Economic Association is open to
the entire membership of the PEA, including all registrants at the conference. Door prizes will be awarded.
SATURDAY, June 1, 2013 CLOSING
Proceedings of the Pennsylvania Economic Association 14
Program Author & Participant Index
First Name Last Name Email Session
Kwami Adanu kadanu@gimpa.edu.gh F2B
Ashley Alt bbpq@iup.edu F1E
Steven Andelin sla7@psu.edu F2C, F3E
Thomas Armstrong thoarmstro@pa.gov F2B
Aram Balagyozyan aram.balagyozyan@scranton.edu F1A, F2B
Robert Balough balough@clarion.edu F3A, S1A
Aram Belhadj aram.belhadj@etu.univ-orleans.fr F2A, F3C
William Bellinger bellinge@dickinson.edu F3B, S1C
Steven Breslawski sbreslaw@brockport.edu F3A, F1A, F2C
David Buehler dlb74@psu.edu F2B, S1B
Chong Hyun Byun byunc@wabash.edu F3D
Dana D'Angelo dangeldc@drexel.edu S1C, F3C
Khaled Elmawazini k.elmawazini@alumni.uottawa.ca F1B, F3B
Shuang Feng sfeng@edinboro.edu F1A,
David Gargone dgargone@misericordia.edu F3E
Soma Ghosh sghosh@alb.edu F2E, F3A
Deborah Gougeon gougeond1@scranton.edu F1E,
Michael Hannan hannan@edinboro.edu F1A, S1A
Qian Hao qian.hao@wilkes.edu F1D
Jui-Chi Huang jxh74@psu.edu F2E, S1A
Debrashi Indra dindra@buffalo.edu F1C
Kosin Isariyawongse kisariyawongse@edinboro.edu F2B, F1D
David Kalist dekali@ship.edu F3E, S1C
John Kallianiotis ioannis.kallianiotis@scranton.edu S1B, F1E
Orhan Kara dearorhankara@gmail.com F2A
Serkan Karadas Serkan.Karadas@mail.wvu.edu F2C, F3A
Timothy Kearney tkearney@misericordia.edu F3E
Elsy Kizhakethalackal ekizhak@bgsu.edu F3B, F2B
Tanu Kohli tkohli@rutgers.edu F2A, F1B
Jonathan Kramer jkramer@kutztown.edu F2C, S1B
Sooriyakumar Krishnapillai sooriyakumar.krishna@aun.edu.ng F3C, S1A
Erin Krotoszynski E.Krotoszynski@eagle.clarion.edu F1E
James Kurre k12@psu.edu F3C, F2B
David Latzko dlatzko@psu.edu F2D, S1C, F1B
Johnnie Linn linnj@concord.edu F3B, F1E
Patrick Litzinger litzinger@rmu.edu S1C, F1A
Proceedings of the Pennsylvania Economic Association 15
Zhonghui Liu zliu8@binghamton.edu F1C, F3A
Zhen Ma zma@misericordia.edu F1D, F3E
Abhradeep Maiti am4m@mtmail.mtsu.edu S1A, F3D
Md. Alauddin Majumder mm4v@mtmail.mtsu.edu F1A
Stephen Mansour stephen.mansour@scranton.edu F1A, F2D
John McCollough jmccollough@lamar.edu F2D
Sandra McPherson smcpherson@millersville.edu F2E
Tracy Miller tcmiller@gcc.edu F1D, F2A, S1C
Sunita Mondal sum46@pitt.edu F3A
David Nugent davidanugent@hotmail.com F3D, S1B
Daniel Oberkofler danoberkofler22@gmail.com F1E
Abdul Pathan apathan@pct.edu F1E, F3F
Logyn E. Pezak pezakl2@gmail.com F3D
Alberto Posso alberto.posso@rmit.edu.au F1C
Margarita Rose MargaritaRose@kings.edu F1E,F3D
Angela Rowland a.m.rowland91@hotmail.com F1E
Abhijit Roy roya2@scranton.edu F1B, F2C
Ruttana Ruttanajarounsub ruttana@edinboro.edu F1D
Robert Schindler rschindl@camden.rutgers.edu F1D, F3C
Minh Tam Schlosky tammyschlosky@gmail.com F1B, F2A
Jeremy Schwartz schwarjf@msn.com F2A, S1A
Rose Sebastianelli sebastianer1@scranton.edu F3D
Lee Siegel LSIEGEL@pa.gov F3E, F2C
Yaya Sissoko ysissoko@iup.edu F1B, F2D
Brian Sloboda bsloboda@email.phoenix.edu F2D, F3E
Lauren Tassone L.Tassone@eagle.clarion.edu F1E
Charles Telly tabak@fredonia.edu S1C, F1C,
Solomon Tesfu tesfu@msmary.edu F3A, F2E
Tufan Tiglioglu tufan.tiglioglu@alvernia.edu F2E, S1B
Bradon Vick vick@fordham.edu F1C, F3A
John Walker walker@kutztown.edu F2C, F3B
Xuebing Yang xyang@psu.edu S1B, F3C
Travis Yates tmy5026@gmail.com F3C, F1D
Sabri Yilmaz yilmaz@lycoming.edu F2A, F3D
Huilan Zhang hzhang11@rockets.utoledo.edu F1E, F2E
Proceedings of the Pennsylvania Economic Association 16
THE EFFECTS OF POWER, PRESTIGE, AND PERFORMANCE
ON SALARY IN THE NATIONAL HOCKEY LEAGUE
Ashley M. Alt
Indiana University of Pennsylvania
2315 Franklin Street
Johnstown, PA 15905
Daniel R. Oberkofler
Indiana University of Pennsylvania
587 Saxony Lane
Yardley, PA 19067
ABSTRACT
The impact of individual player statistics on National Hockey
League (NHL) players’ salaries is analyzed empirically using
career statistics. A double-log regression equation is
estimated using ordinary least squares (OLS). Dependent
variables in this study are the 2011-2012 salaries.
Independent variables are the natural log of goals, assists and
shots on goal, points per game, plus/minus, hits, penalty
minutes, height, and weight, save percentage, wins, losses,
the natural log of games played, natural log of shutouts.
Dummy variables are also assigned. Shots on goal, points
per game, goals, and star, assists, hits, save percentage and
the natural log of games played are significant determinants
for respective groups.
JEL Codes: L83, M2, J30
Keywords: NHL Salary, hockey statistics, salary
determinants
INTRODUCTION
Team Profit Maximization
Basic economic theory asserts that the goal of every firm is to
maximize profit. Firms maximize profit; in part, through the
use of the wages they pay employees. For efficiency, the
harder an employee works, the more compensation they will
receive in return for their work. All rational, profit-
maximizing firms operate under this principle, and the 30
National Hockey League (NHL) franchises are no different.
Under the profit maximization principle, an NHL team will
strive to increase total revenues and decrease total costs.
Increasing total revenue includes creating an environment in
which fans will want to attend games, purchase merchandise,
and support their respective teams. Fostering a winning
culture within a franchise leads to higher revenues.
A fundamental additive in maximizing profit consists of
delving into how the General Manager (GM) of each team
decides on the salary each player receives. However, this is
not always an easy task for a GM because of the wide range
of skills amongst players. Player performance during the
season is used as a benchmark for determining a starting
point for salary negotiations. Since successful teams earn
higher revenues, and every team relies on its players to
perform at a certain level in order to be successful, the GM
must spend money in order to create that success. Yet,
spending more money does not necessarily create success.
Therefore, the goal of the GM is to create a competitive
franchise without overpaying the players. In other words,
maximizing total revenue and minimizing total cost. It is
important to note that while all firms operate under the profit
maximization principle, the same goes for all sports leagues.
Major League Baseball (MLB), the National Football League
(NFL), and the National Basketball Association (NBA),
among others, all function according to the same criterion.
That is, they all want to keep salary as low as possible while
still championing a successful team.
NHL Labor Market
To begin, we can draw an abundance of similarities between
the NHL and the conventional job market. For instance, just
as workers are represented by unions in large industries,
players in the NHL are represented by the NHL Players'
Association (NHLPA). The NHLPA serves as the go-
between for players on matters similar to unions; working
conditions, contractual rights, and the NHLPA also serves as
the collective bargaining agent for players. Both traditional
labor unions and the NHLPA help protect employees from
unfair working conditions.
Moreover, there is a set of restrictions built into the wage
system within professional sports leagues. All athletes must
earn at least the league approved minimum salary, and
though these differ across professional sports leagues, they
are very similar to minimum wage laws. By the same token,
rookie athletes generally earn less than league veterans
because they lack the experience and additional skills one
gains through multiple years of play. At the same time, new
employees at any firm often receive lower wages to start and
must climb their way up the wage ladder by gaining more
experience and expanding their skill set. It also makes
Proceedings of the Pennsylvania Economic Association 17
economic sense that those employees who contribute more to
the firm will be rewarded in kind, thus we can confidently
expect that the more a player contributes to a team, the more
they will be rewarded financially.
Furthermore, professional sports leagues are similar to the
traditional job market in that professional sports have the
concept of free agency. Free agency gives a player the right
to sign with another team if he feels dissatisfied with his
current team. An obvious parallel to this is the time it takes
for one person to find the right job. Workers are not forced
to keep the same job for the rest of their lives, and many
people change jobs or even careers several times in their
lifetimes. Workers may do this in search of higher wages,
better working conditions, or a better location. The same
goes for professional athletes, many of whom switch teams
several times as well. Clearly, because of the similarities one
can draw between the traditional job market and professional
sports leagues, it is apparent why sports franchises should be
studied using economic theory and econometrics.
At the same time, professional sports franchises can be easier
to study than conventional firms because of the massive
amount of statistics every player accumulates each year.
Another bonus to the accumulation of these statistics is that it
becomes easier to find which ones have the most significant
impact on salary determination. On the other hand,
employees in many firms do not amass large numbers of
individual statistics, making it difficult to show how their
input affects firm output. By analyzing the determination of
salary for NHL players, we can better understand how an
individual's impact on firm output is rewarded with salary.
This study examines the factors that are significant in
predicting NHL salary. In other words, how do performance,
prestige, and power factor into determining National Hockey
League players' salaries? This paper examines previous
literature on NHL salary determinants and identifies the
independent variables used in this regression. Based on the
results, we can draw conclusions about salary and which
predictors are significant. Additionally, we used many of the
same independent variables as Vincent and Eastman (2009);
career goals, assists, points per game, plus/minus rating, draft
round, penalties in minutes, height and weight. We also
included hits as part of our defensive model, something that
had previously not been done. Thus, our study contributes
something new to the pre-existing literature. Furthermore,
for goalies, shutouts were not included in any reviewed
literature. Moreover, the way the star variable is constructed
is completely original. Finally, our study examines data for
the most recent season available.
Outline of Paper
The second section of the paper reviews previous
studies of salary determinants in the NHL. Explanatory
variables and data are discussed in the third section. The
econometric model for this study is explained in section four.
Section five examines the regression equations and results.
The last section of this paper concludes by discussing the
findings and their implications.
LITERATURE REVIEW
Numerous empirical studies have been conducted on the
determinants of salary for professional athletes. There are;
however, only a smattering focused on the National Hockey
League. Zimbalist (2010) stressed the importance of
collecting the correct data in order to correctly define
compensation for players. In order to conduct research
efficiently, it is imperative to understand the National
Hockey League’s definition of salary, revenue, and how
those definitions are created. This study accentuates the fact
that different leagues define salary, and salary as a function
of total revenue, differently. Moreover, in Marchand,
Smeeding and Torrey (2006), the effects of overall team
performance and teamwork were analyzed. They used 2,866
player-season observations of 1,001 players over four NHL
seasons: 2000-01 through 2003-04. Specifically, they
discussed whether or not there is a "star effect" versus a
"journeyman effect," or, whether having a few star players
who earn above median salary leads to better performance.
In fact, their results showed that the presence of a player
deemed a "star" increased the spread of assists while the
goals scored spread remained relatively the same. In other
words, other lower paid players were jumping into the play
and contributing on the ice more.
Vincent and Eastman (2009) used a quantile regression with
data from the pre 2004-2005 lockout period to see how
changes in their explanatory variables would be reflected
within different quantiles; the 10th
, 25th
, 50th
, 75th, and 90
th.
In the end, they found that variables have different effects
across quantiles. For example, their results implied that one
more game played is much more valuable to a player in the
75th
or 90th
quantiles of games played than it is for a player in
the 10th
or 25th
quantiles. Furthermore, the study discussed
the difficulty in truly analyzing player performance because
of the combined offensive and defensive aspects of the game.
Lambrinos and Ashman (2007) explored the effects of
arbitration on salary for the 2001-02 season. They found that
often arbitrated salaries do not differ from negotiated salaries
for both forwards and defensemen. However, they did
acknowledge a possible sample size problem since only 17
forwards and 20 defensemen went through arbitration in
2001-02.
Chan, Cho and Novati (2012) discussed the contribution of
different player types to team performance. Their analysis
was based on Vincent and Eastman (2009), but they
expanded by putting additional emphasis on defensive
Proceedings of the Pennsylvania Economic Association 18
categories for forwards and defensemen. These categories
included: first line, second line, defensive, and physical
players for forwards, and offensive, defensive, average, and
physical for defensemen. They also included elite, average,
and bottom categories for goalies. They concluded that
goalies contribute the most to team performance, followed by
forwards, and then defensemen. Furthermore, they found
that player-types previously considered less attractive
become investments with a comparable return to more
offensive-oriented forwards. Another study done by Vincent
and Eastman (2012) analyzed player mobility. Their analysis
found that players who frequently change teams incur a
negative cumulative effect on their salaries.
Berri and Brook (2010) analyzed the most important position
in professional sports; the goaltender. The focus of their
study was mainly on whether general managers efficiently
ascertain a goaltender's worth in salary through the Vezina
Trophy (best goaltender), previous salary data, and a list of
independent variables, of which they found save percentage
to be most significant. Their results suggested that the
difference between a goalie considered "one of the greatest
ever" and "just average" is negligible.
DATA
The dependent variable in this study is the natural log of the
individual player salary, for the 2011-2012 NHL season.
Vincent and Eastman (2009) also used the natural log of
player salary in their analysis. Data is obtained from USA
Today, similar to Marchand et al. (2006). We separated the
dependent variable amongst the forwards, defensemen and
goalie groups. By using a large sample size of 375 NHL
forwards, 200 NHL defensemen and 48 goalies, we can
determine which statistics are the most influential in
determining players’ salaries. Since there is such a large
sample of NHL players, we do not have to be concerned
about potential degrees of freedom issues. Like Vincent and
Eastman (2009), forwards and defensemen are split because
each utilizes a separate set of skills in order to accomplish the
objectives of their position. Goalies also use a separate
regression, for obvious reasons. Therefore, a different set of
independent variables will be used for each in order to
generate a model that explains salary differences.
Salaries observed are for players who played at least 20
regular season games during the given season for the purpose
of consistency and the elimination of potential outliers.
Since the majority of teams are located in the United States
of America, with the exception of seven Canadian teams,
U.S. dollars are used to calculate salary. The player salaries
for the seven teams who use Canadian dollars are converted
into U.S. dollars.
Additionally, we used many of the same independent
variables as Vincent and Eastman (2009); career goals,
assists, points per game, plus/minus rating, draft round,
penalties in minutes, height and weight. The player
performance data was collected from the individual team
websites hosted by NHL.com. All of the data collected for
the study includes career as well as season totals. For the
purpose of consistency, statistics for the players’ entire
careers are used instead of just the 2011-2012 season. Career
statistics account for all years, including years of exceptional
performance as well as years of poor performance. Outliers
are a potential issue due to the variation in human
performance variables and the length of players’ NHL
careers. Players have varying degrees of experience, with
some having only a few years, and others like Jaromir Jagr,
have been playing for over two decades. Salaries, as well as,
human performance variables have a tendency to be right-
skewed. The natural logs of assists, goals, shots on goal,
shutouts and games played are taken in order to reduce this
potential outlier problem.
Expected Signs
Each season NHL players amass several statistics. Table 1 in
the appendix displays the variables used in this study and
their expected signs. Easily, the most important of these
statistics is goals scored (G_ln). We used the natural log in
order to address outlier issues. A goal is awarded to the last
player to touch the puck before it crosses the goal line. Goals
are the deciding factor in winning or losing. Typically,
forwards will have more goals than defensemen. A player
who scores more goals should theoretically earn a higher
salary. We expect that goals will have a positive coefficient
for both forwards and defensemen. There is an abundance of
studies that confirm our expected positive sign. For example,
Vincent and Eastman (2009) conclude points, which are
defined as the summation of goals and assists, have a
significant positive impact on salary for forwards and
defensemen.
Another useful statistic is assists (A_ln). Again, assists are
typically right-skewed, thus the natural log is taken. An
assist is attributed to up to two players of the scoring team.
They must have shot, deflected or passed the puck to the goal
scorer, or touched it before another player earned the goal.
For every goal, there typically is at least one assister, so we
can expect assists to have a positive coefficient for both
forwards and defensemen like goals scored. Vincent and
Eastman (2009), and Marchand et al. (2006) incorporate
assists into their models.
Shots on goal (SG_ln) can be a very influential aspect of
hockey. To correct for right-skew, the natural log is taken. It
is not possible for a player to earn a goal unless they are
actively shooting on the net. Shots supplement goals and
assists, and help to further explain variations in salary.
Typically, a player who contributes more shots on goal helps
create offensive chances, thereby increasing the likelihood of
Proceedings of the Pennsylvania Economic Association 19
his team achieving a win. Therefore, we expect the
coefficient for shots on goal to be positively related to salary.
It is also likely that a forward will have many more shots on
goal than a defensemen, for this reason shots on goal is
included only for forwards. Moreover, no existing literature
that was examined for this study included shots on goal,
which is another way this study contributes to the current
literature.
Another key statistic in determining salary for NHL
forwards, in particular, is plus/minus (PLUSMINUS).
Vincent and Eastman (2009) and Lambrinos and Ashman
(2007) also use plus/minus as an explanatory variable. A
player earns a plus one for being on the ice when a teammate,
or the player himself, scores a goal. Conversely, a player will
earn a minus one for being on the ice when a player of the
opposing team scores a goal. Naturally, a player with a
higher, positive plus/minus total will earn a higher salary.
Thus, it is logical to assume a positive sign for this
coefficient.
Points per game (PPG) statistics give insight into how well a
player is performing on a game-by-game basis. Points per
game are calculated by summing the number of career goals
and career assists, and then taking that sum and dividing it by
the number of career games that were played. We expect
points per game to have a positive sign because there is a
positive relationship between points per game and
goals/assists. Additionally, we expect that points per game
have a much more important impact on salary for a forward
than it does for a defenseman. Using this reasoning, points
per game is included in only the forwards model. Vincent and
Eastman (2012) also used points per game as an explanatory
variable.
Crucial statistics for defensemen, in particular, are penalty
minutes (PIMS). A penalty occurs when one team is awarded
an extra attacker due to a rule infraction from a player on the
opposing team. Defensemen often take more penalties
because the nature of the position requires more aggressive
play. Coaches generally approve because a defenseman’s
primary objective is to keep the puck from getting to the net.
Moreover, few other events rile up the crowd like a big hit or
a fight. On the other hand, a penalty always hurts a team
because it is put on the defensive, moreover missing one
man. Therefore, taking penalties can either carry a positive or
negative sign. Vincent and Eastman (2012) used penalties in
minutes as a covariate; however, they predicted a positive
relationship with salary.
Height (HEIGHT) and weight (WEIGHT) are used in several
other studies, such as Lambrinos and Ashman (2007) and
Vincent and Eastman (2009), as significant determinants of
salary for defensemen. This makes sense because bigger
players have the ability to hit harder, thus keeping the net
safer from the opposing team. For the purpose of determining
salary, we believe that the expected sign of both height and
weight will be positive.
When determining a player’s salary, for defensemen
specifically, hits are a vital statistic. Hits (HITS) are typically
recorded when a player body checks a player of the opposite
team causing him to lose control of the puck. As a
defenseman, hits are essential because it is necessary to play
rough in order to get the puck away from the net and out of
the defensive zone. Therefore, hits should have a positive
sign. Furthermore, we did not identify an existing study that
included hits in their model.
Dummy variables are created for the round a player was
drafted (DRAFT), and whether a player was either invited to
or played in the 2010-2011 All-Star Game, or had won the
Stanley Cup in the 2009-10 and/or 2010-11 seasons (STAR).
For the round drafted, a value of one was given to the players
that were drafted in either the first or second round of the
draft, and all others are represented by zero. For star power,
players are given a one if they either were invited to play, or
played in the All-Star Game, or won the Stanley Cup in the
aforementioned seasons, and those who did not are given a
zero. A first or second round draft pick should carry a
positive expected sign because the best players are selected
first, and therefore should be projected to have a higher
salary. Similarly, only the best players are invited to play in
the All-Star Game so we should expect those players to earn
a higher salary.
On the other hand, a prominent statistic for goalies is save
percentage (SAVE_%). Save percentage is the number of
shots a goalie saves divided by the number of shots faced. A
higher save percentage generally garners confidence from
both the front office and the players on the ice in their
goaltender because they trust the goalie to make the save.
Therefore, it is logical to expect that goalies with a higher
save percentage earn higher salary. Berri and Brook (2010)
used save percentage as an explanatory variable.
Another useful statistic for goalies is wins (WINS). A goalie
earns a win in a game for playing the entire game and
winning, or when he becomes the goalie of record by being
on the ice for the game-winning goal, regardless of the team
that scores. Naturally, wins are expected to carry a positive
sign. Conversely, losses (LOSSES) are also used. Goalies
earn a loss when the game-winning goal is scored on them.
We expect losses to have a negative sign. While Berri and
Brook (2010) included wins in their analysis, losses are not
used in any study we examined.
Goalies are also evaluated on their experience, thus the
natural log of games played is taken (LN_GP). The natural
log is taken because there are a plethora of goalies with very
few games played and only a few with a high number of
games played. The natural log helps control for this skew in
Proceedings of the Pennsylvania Economic Association 20
the data. Generally, with more experience there is a higher
salary, so we expect the natural log of games played to carry
a positive sign. Berri and Brook (2010) used minutes as a
measure of experience, instead of games played.
Also, for goalies, the natural log of shutouts is included
(LN_SHUTOUTS). A shutout is recorded when the
opposing team does not score. Again, this data is right-
skewed, so the natural log is taken to correct the skew.
Shutout victories are typically accredited wholly to the goalie
for their exceptional play. Again, we expect shutouts to carry
a positive sign. Shutouts were not used as an explanatory
variable in any existing study.
Finally, a goalie star dummy variable is created (STARG)
which indicates the 2009-2010 and 2010-2011 All-Star
goalies, the Stanley Cup winners, as well as the Vezina (best
overall goaltender) and Jennings (fewest goals against)
trophy winners from those years. We expect the goalie star
variable to carry a positive sign.
Descriptive Statistics
The mean salary for all NHL forwards is $2,500,508. Table
2.1 displays the descriptive statistics for forwards. The
maximum salary earner for forwards is Brad Richards,
bringing in a whopping $12,000,000, while 15 players earned
the league minimum $525,000. The mean number of career
goals is 103; similarly, the average number of career assists is
143. Jaromir Jagr has both the most career goals and assists
with 665 and 988, respectively. Averaging 1.40 points per
game, Sydney Crosby leads the pack in points per game. The
mean number of points per game is 0.48. Again, Jaromir
Jagr owns the highest plus/minus statistic at +280. The
league average is +8. The league leader in shots on goal is
Jaromir Jagr with 4,766. The mean number of shots on goal
for the league is 914.
Table 2.2 displays the descriptive statistics for defensemen.
Christian Ehrhoff leads NHL defensemen earning
$10,000,000. The mean salary for NHL defensemen is
$2,551,294. Nicklas Lidstrom leads defensemen in goals
with 264 and assists with 878. The average number of goals
for defensemen is 33 and 109 for assists. Zdeno Chara is a
monster on the ice at a towering 81 inches, the shortest player
comes in at 68, and the mean is 74 inches. The mean weight
for defensemen is 210 pounds. The player with the most hits
is Luke Schenn with 270. The league average is 88.
Table 2.3 displays the descriptive statistics for goalies. The
league leader is Ilya Bryzgalov earning $10,000,000. On the
other hand, the mean salary for goalies is $2,820,563.
Tuukka Rask commands the best save percentage with 0.926,
meanwhile the league mean is 0.912. The mean number of
wins is 157.75 and the average number of losses is 188.6.
ECONOMETRIC MODEL
This study utilizes ordinary least squares (OLS) regression
analysis to measure the importance of performance, power,
and prestige on a player’s salary. The dependent variable for
all equations is the natural log of salary for the 2011-2012
NHL season. For all equations, the X matrix consists of
performance variables and the Z matrix consists of power
variables, and STAR/STARG is the relevant prestige
variable. The empirical specification for defensemen is:
LnSali = 1*Xi + Zi3*STARi + i (1)
where X is a matrix of control variables, in relation to the
defensive position, consisting of G_ln, A_ln, HITS, PIM, and
DRAFT. Z is a matrix of control variables including
HEIGHT and WEIGHT.
The empirical specification for the offensive
position is:
LnSali = 1*Xi + Zi3*STARi + i (2)
where X is a matrix of control variables, consisting of G_ln,
A_ln, and SG_ln. Z is a matrix of control variables including
PPG, PLUSMINUS, and DRAFT.
The empirical specification for the goalie position
is:
LnSali = 1*Xi + i3*STARGi + i
where X is a matrix of control variables, consisting of
SAVE_%, LN_SHUTOUT, and WEIGHT. Z is a matrix of
control variables including WINS, LOSSES, and LN_GP.
Econometric Issues
The analysis is cross-sectional and therefore
heteroskedasticity may be present. Thus, the White test was
run on all models. The defensive, offensive, and goalie
equations were found to be heteroskedastic, and the models
were corrected for heteroskedasticity using White
heteroskedasticity-consistent standard errors. Final models
discussed in the following section reflect adjustments to the
specifications required to address multicollinearity issues.
Outliers are a potential issue due to the variation in human
performance variables and the length of players’ NHL
careers. Players have varying degrees of experience, with
some having only a few years, and others like Jaromir Jagr,
have been playing for over two decades. Salaries, as well as,
human performance variables have a tendency to be right-
skewed. The natural logs of assists, goals, shots on goal,
shutouts and games played are taken in order to reduce this
potential outlier problem.
Proceedings of the Pennsylvania Economic Association 21
RESULTS
Table 3.1 in the appendix displays results for the defensive
models. For all equations, assists are found to be positive
and significant at the 1% level. This corroborates past
studies which have also noted the significance of assists for
defensemen, including Marchand et al. (2006), among others.
Furthermore, hits are found to be positive and significant at
the 1% level for all models. As stated earlier, no literature
reviewed for this study included hits, hence a significant
factor in determining salary for defensemen has been
discovered by this study. The star variable also proved to be
positive and significant at the 1% level. Though we
constructed the star variable differently, previous studies like
Marchand et al. (2006) and Vincent and Eastman (2009) also
concluded that the star variable is positive and significant.
Goals is shown to be positive as expected but not significant,
perhaps because coaches do not necessarily expect
defensemen to score.
On the other hand, draft round selected and height show
negative and insignificant signs. One reason for the
unexpected sign on height could be the correlation it shares
with weight. Penalties in minutes is found to have an
insignificant negative coefficient, owing to the fact that when
a team is penalized they lose a player, thus making it more
difficult to defend the net. Likewise, since the 2005
Collective Bargaining Agreement, and especially since the
injuring of super-star Sydney Crosby, the league has been
stepping up its player safety efforts. Our adjusted R-squared
values, ranging from 0.557 to 0.572, are higher than Vincent
and Eastman (2009), whose values range from 0.299 to
0.533.
The regressions run for forwards can be seen in Table 3.2 in
the appendix. For all models, points per game is shown to be
positive and significant at the 1% level. Similarly, Vincent
and Eastman (2009) found points per game to be positive and
significant at the 5% level. Shots on goal is found to be
positive and significant. Goals are found to be significant at
the 5% level in Model 2, and significant at the 1% level in
Models 3 and 4. Moreover, the star variable is found to be
positive and significant at the 5% level. Previous literature,
such as Marchand et al. (2006), and Vincent and Eastman
(2009), all found their versions of the star variable to be
positive and significant. Although the draft dummy variable
and plus/minus variable did not turn out to have any
significance, they are left in the models in accordance with
Vincent and Eastman (2009). Moreover, the signs are
consistent both with literature and expectations. Vincent and
Eastman (2009) obtained adjusted R-squared values ranging
from 0.324 to 0.758. The range of our adjusted R-squared
values are more concentrated, ranging from only 0.681 to
0.692.
The equations for goalies are displayed in Table 3.3. For all
but one model, the natural log of games played is significant
at the 1% level and the circumstance in which it is not, it is
significant at the 5% level. Our findings corroborate Berri
and Brook (2010) who also found their experience proxy,
minutes, to be significant at the 1% level. Furthermore, save
percentage is found to be significant at the 5% level for all
models. Again, this corroborates Berri and Brook (2010),
who found save percentage to be significant at the 1% level.
Weight and the natural log of shutouts were found to be
insignificant. Wins and losses were not found to be
statistically significant; however, they are left in the models
in accordance with previous literature. The star variable
failed to garner significance in any model. Berri and Brook
(2010) obtained an adjusted R-squared range of 0.27 to 0.48.
In comparison, our adjusted R-squared ranges from 0.362 to
0.376.
CONCLUSION
Selected based on past literature, power, prestige and
performance variables are utilized to estimate player salary
for 200 defensemen, 375 forwards, and 48 goalies in the
NHL during the 2011-2012 hockey season. Several important
results are evident and show the importance of separating
defensive, offensive, and goalie equations. To begin, results
show the importance of lagging the star variable back to the
2009-10 and 2010-11 seasons to indicate that success from
past seasons has an effect on future salary. From our results
we are able to conclude that players, who are considered to
be of star quality, do garner a higher salary. As mentioned
earlier, other studies, such as Vincent and Eastman (2009),
also found a positive and significant relationship between star
and salary. Ultimately, our hypothesis was that there is a
positive and significant relationship between star and salary,
and the results support our presumptions clearly.
When comparing descriptive statistics it is important to note
that defensemen do earn a slightly higher salary, on average,
than forwards. We attribute this difference in salary to the
fact that defensemen play the harder position. Defensive
players must make sure that their skills are top notch since
overall defense is very difficult, and a player is required to
keep extra sharp, mentally and physically. We conclude that
defensemen should put added emphasis on hits and assists
while on the ice because these factors contribute the most to a
higher salary. Therefore, an aspiring defenseman should be
sure to practice both his offensive game, specifically passing
and setting up shots, and defensive skills like hitting and
using their body weight to keep opposing players away from
the net.
Goalies earn the highest salaries of all positions. We believe
this is because, as previously mentioned, goalies are often
described as playing the most important position in
professional sports. Goalies must constantly adapt to the
Proceedings of the Pennsylvania Economic Association 22
ever-changing nature of the game, adding an additional layer
of difficulty to the position. Furthermore, a majority of
hockey analysts believe that an outstanding goaltender is
paramount to a Stanley Cup Championship. We conclude
that goalies should focus their attention on stopping the puck
because this has implications on the save percentage variable,
which is found to be significant. Moreover, goalies should
continuously work to perfect their skills, in order to gain the
trust of the coaching staff because a higher number of games
played leads to a higher salary.
The findings of this study indicate that NHL franchise
owners seek to maximize profit by signing defensemen who
contribute on the offensive side of the ice, as well as, the
defensive side. Likewise, owners maximize profit by signing
forwards who put a large number of shots on goal, and who
accrue a large number of points per game. Finally, owners
maximize profits through the signing of goalies with a high
save percentage, and a large number of games played. The
results from this study can help predict how much a GM is
willing to spend on a specific player since the ultimate goal is
to create a winning culture without overpaying players.
These kinds of econometric techniques can be applied in
order to help GMs arrive at decisions.
Proceedings of the Pennsylvania Economic Association 23
Table 1: Variables and Expected Signs
Dependent Variable Definition
Salary_ln Ln of salary from 2011-2012 (Defensive, Offensive & Goalies)
Variables Definition Expected Sign
G_ln Ln of Career Goals (All Positions) +
A_ln Ln of Career Assists (All Positions) +
PIMS Penalties in minutes (Defensive) ?
HITS Career hits (Defensive) +
HEIGHT Height of player in inches (Defensive) +
WEIGHT Weight of player in pounds (Defensive & Goalies) +
PPG Career points per game (Offensive) +
SG_LN Ln of Career Shots on goal (Offensive) +
PLUSMINUS Career plus/minus (Offensive) +
DRAFT Dummy variables were used for round in which player was drafted. Value of 1
given to players drafted in first two rounds. All other rounds were given value of
0. (Defensive & Offensive)
+
STAR Dummy variable to indicate whether a player was asked to play in the All-Star
game for the 2010-11 seasons, as well as if they won the Stanley cup. If invited
and/or won, player was given value of 1. If not invited and/or did not win,
player was given a value of 0. (Offensive & Defensive)
+
SAVE_% Career save percentage (Goalies) +
WINS Career wins (Goalies) +
LOSSES Career losses (Goalies) -
LN_GP Ln of career games played (Goalies) +
STARG Dummy variable to indicate whether a player was asked to play in the All-Star
game in 2010-11, if that goalie won the Stanley cup, and if that goalie won the
Vezina and/or Jennings awards. If invited, won, and/or got an award, the player
was given a value of 1. If not invited, did not win, and/or the goalie did not win
an award, the player was given a value of 0. (Goalies)
+
Proceedings of the Pennsylvania Economic Association 24
Table 2.1: Descriptive Statistics – Offensive Players
Variables Mean Standard
Deviation
Maximum Minimum
Offensive Salary 2,500,508 213,840 12,000,000 525,000
Goals 103.3680 104.3502 665.00 1.00
Assists 142.7893 150.7759 988.00 0.00
Shots on Goal 914.4400 824.9344 4766.00 23.00
Points Per Game 0.476410 0.246206 1.403226 0.035714
Plus / Minus 7.970667 46.14261 280.00 -112.00
Observations
375
375
375
375
Table 2.2: Descriptive Statistics – Defensive Players
Variable Mean Standard
Deviation
Maximum Minimu
m
Defense Salary 2,551,294 1,888,990 10,000,000 525,000
Goals 32.62500 37.05319 264.00 0.00
Assists 108.885 109.2558 878.00 3.00
Hits 88.26500 55.14244 270.00 6.00
Penalties in
Minutes
292.2800 314.2796 1809.00 2.00
Height 73.8100 2.113346 81.00 68.00
Weight 210.2450 15.94794 270.00 180.00
Observations
200
200
200
200
Table 2.3: Descriptive Statistics - Goalies
Variables Mean Standard
Deviation
Maximum Minimum
Goalie Salary 2,820,563 2,189,261 10,000,000 570,000
Save Percentage .912 .0057 .926 .900
Wins 157.75 119 664 17
Losses 118.60 86.77 373 16
Games Played 331.90 235.34 1204 44
Weight 198.15 14.745 220 166
Shutouts 23.13 20.2 120 0
Observations
48
48
48
48
Proceedings of the Pennsylvania Economic Association 25
Table 3.1: Defensive Models
Independent
Variables
Model 1 Model 2 Model 3 Model 4
Constant 12.5473
(7.448485)
11.0072
(16.55968)
11.17213
(17.37732) 11.40563
(21.41883)
G_ln 0.1415
(1.284556)
0.1479
(1.347288)
A_ln 0.4173***
(3.404970)
0.4161***
(3.376466)
0.5191***
(10.88916)
0.5010***
(13.85968)
HITS 0.0021***
(3.265746)
0.0022***
(3.619744)
0.00195***
(3.139579)
0.00197***
(3.156443)
PIM -0.0002
(-1.083743)
-0.0002
(-1.107060)
-0.0001
(-0.652943)
HEIGHT -0.0284
(-1.007627)
WEIGHT 0.00795**
(2.077327)
0.0052*
(1.930346)
0.0044
(1.639127)
0.0035
(1.475723)
DRAFT -0.0052
(-0.063005)
STAR 0.3911***
(2.895259)
0.3764***
(2.823645)
0.4209***
(3.320626)
0.4294***
(3.410982)
Adjusted R2
Observations
0.570636
200
0.572124
200
0.557485
200
0.558237
200
Significance Level: ***1% **5% *10%
Parentheses contain t-statistics which are based on White heteroskedasticity-consistent standard errors.
Table 3.2: Offensive Models
Independent
Variables
Model 1 Model 2 Model 3 Model 4 Model 5
Constant 11.2995
(34.45103)
12.4239
(126.1133)
12.4725
(142.0312)
11.4943
(64.83570)
12.4609
(141.3301)
G_ln -0.0735
(-0.8537)
0.1795**
(2.5975)
0.285187***
(7.897713)
0.2844***
(7.867554)
A_ln 0.003
(0.0389)
0.1168
(1.601749)
SG_ln 0.3960***
(3.6114)
0.325294***
(8.899114)
PPG 1.6464***
(7.752388)
1.3784***
(6.539435)
1.4430***
(7.051820)
1.5842***
(8.838350)
1.4814***
(7.728525)
PLUSMINUS 0.0004
(0.561448)
0.0004
(0.533011)
0.0004
(0.5086)
0.0004
(0.555996)
DRAFT 0.0574
(1.086464)
0.0696
(1.260)
0.0697
(1.272083)
0.058739
(1.110243)
0.0677
(1.240681)
STAR 0.1703**
(2.220124)
0.1642**
(2.1005)
0.1619**
(2.051448)
0.16943**
(2.213890)
0.1651**
(2.136421)
Adjusted R2
Observations
0.690912
375
0.681463
375
0.680389
375
0.692272
375
0.680941
375
Significance Level: ***1% **5% *10%.
Parentheses contain t-statistics which are based on White heteroskedasticity-consistent standard errors.
Proceedings of the Pennsylvania Economic Association 26
Table 3.3: Goalie Models
Independent
Variables
Model 1 Model 2 Model 3 Model 4
Constant -27.05846
(-1.912256)
-27.40919
(-1.905393)
-24.75199
(-1.749214) -25.46925
(-1.575011)
SAVE_% 40.75025**
(2.524084)
40.385571**
(2.516581)
38.00340**
(2.365291)
39.26572**
(2.200933)
WINS 0.002028
(1.042287)
0.001983
(0.983937)
0.001601
(0.672149)
0.001970
(1.000411)
LOSSES -0.005684
(-1.500329)
-0.005627
(-1.446286)
-0.005656
(-1.486731)
-0.005550
(-1.438604)
LN_GP 0.861590***
(3.558112)
0.860657***
(3.487055)
0.905895***
(3.920409)
0.788406**
(2.119856)
WEIGHT 0.001313
(0.254666)
STARG
0.165882
(0.468582)
LN_SHUTOUT 0.059181
(0.289365)
Adjusted R2
Observations
0.376399
48
0.362184
48
0.364360
48
0.362201
48
Significance Level: ***1% **5% *10%
Parentheses contain t-statistics which are based on White heteroskedasticity-consistent standard errors.
Proceedings of the Pennsylvania Economic Association 27
REFERENCES
Berri, D.J., & Brook, S.L. (2010). On the evaluation of the
“most important” position in professional sports. Journal of
Sports Economics, 11(2), 157-171.
Chan, T. C.Y., Cho, J.A., & Novati, D.C. (2012).
Quantifying the contribution of NHL player types to team
performance. Interfaces, 42(2), 131-145.
Deutscher, C. (2009). The payoff to leadership in teams.
Journal of Sports Economics, 10(4), 429-38.
Jones, J.C.H., & Walsh, W.D. (1988). Salary determination
in the National Hockey League: The effects of skills,
franchise characteristics, and discrimination. Industrial and
Labor Relations Review, 41(4), 592-604.
Lambrinos, J. & Asham, T. (2007). Salary determination in
the National Hockey League. Is arbitration efficient? Journal
of Sports Economics, 8(2), 192-201.
Marchand, J.T., Smeeding, T.M., & Torrey, B.B. (2006).
Salary distribution and performance evidence from the
National Hockey League. Retrieved from
http://scholar.googleusercontent.com/scholar?q=cache:NPIU
uLCZKJcJ:scholar.google.com/+salary+distribution+and+per
formance+evidence+from+the+national+hockey+league&hl=
en&as_sdt=0,39.
Vincent, C. & Eastman, B. (2009). Determinants of pay in
NHL: A quantile regression approach. Journal of Sports
Economics,10(3), 256-277.
Vincent, C. & Eastman, B. (2012). Does player mobility lead
to higher earnings? Evidence from the NHL. The American
Economist, 57(1), 50-64.
Zimbalist, A. (2010). Reflections on salary shares and salary
caps. Journal of Sports Economics, 11(1), 17-28.
Proceedings of the Pennsylvania Economic Association 28
PENNSYLVANIA TAX SIMPLIFICATION:
NUISANCE TAX CREDIT, OBSOLETE TAXATION AND ADMINISTRATION PROVISION REPEALS,
INCLUDING PROPER PLACEMENT WITHIN THE TAX REFORM CODE
Thomas O. Armstrong*
Commonwealth of Pennsylvania,
Department of Revenue, Harrisburg, PA 17128
ABSTRACT The optimal taxation literature argues that successful tax
reform reduces economic losses to society where resources
can be transferred to higher valued uses. Taxing jurisdictions
can reduce inefficiencies by enacting tax simplification
processes resulting in administrators and taxpayers spending
fewer resources to administer taxes. Tax simplification gains
can be achieved by repealing nuisance tax credits, obsolete
taxation provisions, and obsolete administrative provisions as
well as placing relevant law within the tax code. The Corbett
Administration and the General Assembly in recognizing the
benefits of tax simplification have enacted for Fiscal Year
2013-14 tax reform measures including tax simplification
repeals.
I. INTRODUCTION
1
One purpose of taxation is to raise revenue to finance
government expenditures. Once a set of taxes are in place,
economic inefficiency results (Weimer and Vining, 1992).
From an optimal taxation perspective, minimizing
inefficiencies or deadweight losses of taxes relative to
benefits received will result in greater availability of
resources to enhance economic growth and development
(Slemrod, 1990; Scully, 1992).2
The optimal taxation literature places the administrative costs
of taxation into its research agenda when considering optimal
taxation possibilities (Slemrod, 1990). Tax simplification is
aligned with the classical objective of proper taxation policy
due, in part, to its efficiency objectives. With fewer
resources expended to administer taxes, greater resources are
readily available for more efficient market activities.
Successful tax simplification is when taxes are easy to
understand with relatively low compliance and administrative
costs, which promote a high degree of voluntary compliance
among taxpayers. Relatively low administration and
compliance costs with a simplified tax system occurs when
the system has relatively few taxes; simple tax laws; a limited
number of rates for each tax; limited exemptions, credits,
rebates or deductions; a broad base; and limited use of the tax
system to achieve not too many social goals (Handbook for
Tax Simplification, 2009). A higher degree of voluntary
compliance results in more returns processed, lower
administrative costs, and a higher degree of economies of
scale.
Gale and Holtzblatt (April 2002) consider complexity of a tax
system as the sum of compliance costs, incurred directly by
individuals and businesses, and administrative costs, incurred
by government. Compliance costs can include the time and
expenditures taxpayers spend preparing and filing tax forms;
researching the law to determine the application and if the tax
code is still in effect; maintaining records; taxes prepared by
third party preparers; respond to audits; and resources to
avoid or evade taxes.
Administrative costs include the resources allocated to the
tax agency, and resources required of other agencies to help
administer tax programs. It should be noted while
administrative costs are incurred by the government; it is
ultimately borne by taxpaying individuals along with the
compliance costs. Effective tax simplification results in less
compliance and administrative resources for collection of
certain revenue.3
The Corbett Administration and the General Assembly have
proposed and enacted a number of tax simplification
measures for Pennsylvania Fiscal Year (FY) 2013-14: Act 52
of 2013 (HB 465, PN 2211, Omnibus Tax Reform Code) and
Act 71 of 2013 (SB 591, PN 1328; Omnibus Fiscal Code).
The expectation is resources currently used to comply with a
complex state tax code will now be used in more efficient
economic activities as the result of a more simplified tax
code.4
II. TAX SIMPLIFICATION
Tax complexity increases the overall cost of taxation as well
as increases the likelihood that taxpayers make inadvertent
mistakes in calculating their tax liabilities (Kopczuk, 2006).
Other “…commonly recognized and negative effects of
complexity are (1) decreased levels of voluntary compliance
stemming from taxpayer confusion; (2) increased costs of
compliance for taxpayers; (3) reduced perceptions of fairness
in the …tax system; and (4) increased difficulties in the
administration of tax laws,” (Joint Committee on Taxation,
p.101, April 2001).5 Tax simplification refers to reducing
impediments or complexity of tax processes.
Proceedings of the Pennsylvania Economic Association 29
The goal for tax simplification is to increase the ease of
compliance, reflected in the costs and time saved for
taxpayers and tax administrators (Handbook for Tax
Simplification, 2009). The positive result is more resources
are available for other economic activities, thereby reducing
economic inefficiency (Final Report of the Pennsylvania Tax
Commission, March 1981).
The Joint Committee of Taxation Report (April 2001)
identified various sources of tax complexity relevant to
Pennsylvania’s FY 2013-14 tax changes that will reduce
complexity, thereby enhancing simplification. Three
provisions that have added to complexity within the
Pennsylvania Tax Reform Code of 19716 and the Fiscal
Code7 are 1) transactional complexity, 2) obsolete provisions
complexity, and drafting comlexity.8
“Transaction complexity” refers to the extent that tax laws
complicate the planning and execution of transactions by
taxpayers. State nuisance tax credits that increase the
complication of transactions and administrative costs relative
to few or no beneficiary recipients would fall into this
category of complexity.
“Obsolete provisions complexity” is the numerous
superseded or invalidated tax provisions within
Pennsylvania’s Tax Reform Code and Fiscal Code. While
most taxpayers are unaffected by the obsolete provisions,
these provisions require time and resources to determine
whether a particular provision has continuing applicability.
The tax and fiscal code bills repeal many obsolete taxation
and administration provisions contributing to better resource
usage as the result of tax simplification.
“Drafting complexity” occurs when a tax law is misplaced
within Pennsylvania’s statutes, where the time to locate and
associate the law can take an unnecessarily excessive time
period. The omnibus tax reform and fiscal code bills, Act 52
and Act 71 of 2013 remove tax code language from the fiscal
code and places the language within the Pennsylvania’s Tax
Reform code of 1971.
Tax simplification initiatives from a tax reform/budget focus
were last enacted in 2001 (Armstrong and Brehouse, 2001).
Governor Corbett proposed and the General Assembly passed
into law tax simplification initiatives of repealing state
nuisance tax credits and obsolete taxation and administration
provisions as part of Governor’s budget and tax reform
initiatives announced February 5, 2013.9,10
The tax
simplification measures are projected to have a minimal
fiscal impact of $500,000 out of a General Fund enacted
budget of $28.375 billion. The efficiency gains to the
Commonwealth are expected to be greater than the nominal
revenue loss.
III. STATE NUISANCE TAX CREDIT REPEALS
State nuisance tax credits are an accumulation of many prior
legislative session enactments, where the original intent of
certain tax credits may no longer be valid, no longer
consistent with current policies, or have very few recipients
where the benefits are less than the economic costs of the tax
credits. Upon reviewing the Commonwealth of
Pennsylvania’s taxation statutes, the determination was made
that the following four taxation provisions caused
unnecessary taxpayer and administrative burdens relative to
the little or no economic benefits by these state nuisance tax
credits where the effective repeal date for all the tax credits
listed below is July 1, 2013:11
Coal Waste Removal and Ultraclean Fuels Tax
Credit. This Credit was enacted in 1999 in the Tax
Reform Code (Article XVIIII-A) and available for
certain capital expenditures for companies that
produce synthetic fuels from coal, culm, or silt.
This credit is capped at $18 million per year. The
Credit can be used against sales and use tax,
corporate net income tax, and capital stock/foreign
franchise tax. From the Legislative Budget and
Finance Committee report (June 2010) and
Department of Revenue, no eligible developer has
applied for, or claimed, this tax credit. While
nominal, the Department of Revenue incurs the cost
of maintaining the administration of this credit.
Call Center Tax Credit (CCTC) Program. The
CCTC program was set up to entice call centers to
relocate into the state. $30M was set aside for this
program. A tax credit is available to call centers for
the sales and use tax paid on incoming and outgoing
interstate telecommunications. Beginning in on or
after January 1, 2004, the credit is equal to the gross
receipts tax paid by a telephone company on the
receipts derived from the incoming and outgoing
interstate telecommunications. This program has
been underutilized since the inception and the
number of applicants has decreased over the last
several years. While nominal, the Department of
Revenue incurs the cost of administering this
underutilized credit. Between 2005 and 2009, $6.4
million in tax credits (out of $30 million available
capped annually) were approved and awarded to 51
call centers applying for the credit. For 2010, 12
call centers applied and were approved for refunds
totaling $379,910 and for 2011, 9 call centers
applied and were approved for refunds totaling
$547,903. Currently, only 9 call centers applied for
2012 totaling $655,315 of the available $30M.
Proceedings of the Pennsylvania Economic Association 30
IV. OBSOLETE TAXATION AND ADMINISTRATIVE
PROVISIONS INCLUDING PROPER PLACEMENT
WITHIN THE TAX REFORM CODE
A review of the Commonwealth of Pennsylvania statutes has
indicated that a number of obsolete provisions exist that
should be repealed. It is expected that the repeal of these
obsolete provisions as well as proper placement of statute
within the Tax Reform Code will reduce the time and
resources expended by taxpayers to become familiar with the
Commonwealth’s applicable taxation laws. The obsolete
provisions can be classified into two categories: 1) Obsolete
Taxation Provisions, and 2) Obsolete Administrative
Provisions.
Obsolete Taxation Provisions Repeals—These taxes have
been superseded by other legislation or have been declared
unconstitutional by a court, yet the legislation remains in
Pennsylvania statutes. The effective repeal date for all the
obsolete taxation provisions is the date of enactment. There
is no fiscal impact for repealing the obsolete taxation
provisions.
Treasurer Report of Municipal Loans (1929 0
P.L.343 A176 §709). This law in the fiscal code
requires the treasurer to submit to the Department of
Revenue an annual report of municipal loans and
pay appropriate taxes. This was enacted April 4,
1929. This law is no longer actively imposed or
applied and is repealed. This repeal is part of the
Department of Revenue's recommendation to the
Local Government Commission concerning state
mandates on local governments.
Registers to File Monthly Inheritance Tax
Statements (1929 0 P.L.343 A176 §724). This law
in the fiscal code requires the Register of Wills to
file monthly statements of inheritance taxes paid.
This was enacted April 4, 1929. This law is no
longer actively imposed or applied and is repealed.
This repeal is part of the Department of Revenue's
recommendation to the Local Government
Commission concerning state mandates on local
governments.
County Treasurers Receive Use Tax (1971 0 P.L.6
A2 §226). This provision in the Tax Reform Code
designated county treasurer as receiver of use taxes
from any person other than a licensee and sets
procedure for transmittal to the Department of
Revenue. This was enacted March 4, 1971. This
law is no longer actively imposed or applied and is
repealed.
Inheritance Tax Poverty Exemption (section 2112).
This provision in the Tax Reform Code provides an
exemption from Inheritance Tax for transfers of
property for certain classes of persons, spouses.
This law is no longer actively imposed since June
30, 1995, and is repealed, due to being superseded
by an exemption for transfers to spouses.
Obsolete Administrative Provisions Repeals—These
provisions are non-functioning tax related administration
laws that remain in Pennsylvania’s statutes and should be
repealed. Repealing these provisions will reduce time and
resources expended by taxpayers to determine whether the
non-functioning tax related administrative laws are no longer
applicable. The proposed effective repeal date for all the
obsolete administrative provisions is the date of enactment.
There is no fiscal impact by repealing the obsolete taxation
provisions.
Motor License and Vehicle Operators’ License Fees
(72 P.S. §1206). This is an administrative provision
directing the Department of Revenue to collect fees
for the registration and titling of vehicles, and to
perform licensing and titling functions. The
Department of Revenue’s powers related to the
registration and titling of motor vehicles were
transferred to the Department of Transportation,
effective July 1, 1970 (Act of July, 1970 (P.L. 356,
No. 120)). Despite this transfer of responsibility,
this provision was never repealed and thus remains
in the Fiscal Code. Because the Department of
Revenue is no longer responsible for these
functions, this provision does not serve a function.
Collection of Amounts Payable to State Institutions
(72 P.S. §1209). These sections require the
Department of Revenue to place its agents in every
state institution for the purpose of collecting moneys
due the institution for any expenses accrued on
account of patients, pupils or inmates. This
procedure is not operative. The second paragraph of
section 210 of the Fiscal Code contains a similar
provision (72 P.S. §210).
Proper Placement Within the Tax Reform Code—The proper
placement of tax legislation will reduce the time and
resources expended by taxpayers to become familiar with the
Commonwealth of Pennsylvania’s applicable taxation laws.
There is no fiscal impact of transferring exactly existing
taxation language into the Tax Reform Code.
Neighborhood Improvement Zones, NIZ (Act 26 of
2011 and Act 50 of 2009, omnibus fiscal code
legislation). The Acts authorized a city of the third
class to designate a NIZ for the purpose of
improvement and development of a Zone and to
construct a facility or facility complex within the
Zone. Certain state and local taxes attributed within
Proceedings of the Pennsylvania Economic Association 31
the NIZ is allocated to the NIZ authority for
development.
Keystone Special Development Zones, KSDZ (Act
26 of 2011, omnibus fiscal code legislation).
Creates a new program for the designation of a
KSDZ for parcels of real property certified as
Special Industrial Areas by the Department of
Environmental Protection pursuant to the Land
Recycling and Environmental Remediation
Standards Act, and which as of July 1, 2011 had no
permanent vertical structures affixed to it. The Act
provides a tax credit for employers within a KSDZ
for new full time jobs created in the zone.
V. CONCLUSION
Tax complexity increases the cost in the form of time or
money imposed on taxpayers and tax administrators to
comply with the tax law. Tax complexity can enhance tax
evasion and avoidance that reduce economic efficiency. The
preferred method in reducing tax complexity is to correct the
tax code. The Corbett Administration and the General
Assembly has introduced and enacted tax simplification
measures into law that will improve the administration of
taxes for taxpayers and the Commonwealth. As a result,
efficiency gains can be expected.
ENDNOTES
* The author would like to thank Sally Fishel and discussant
for their assistance and comments. The conclusions do not
necessarily reflect the positions of the Pennsylvania
Department of Revenue. All possible errors are the author’s.
1. Information for the next two sections are generally from
Armstrong and Brehouse (2001), the Report (April 2001),
and Gale and Holtzbatt (April 2002). The Report and Gale
and Holtzbatt suggest that there is no general consensus on
the appropriate method of measuring the effects of
complexity; while recognizing that there is general agreement
that complexity has adverse economic effects.
2. It is recognized that there are other taxation policy
objectives. Some of the classic objectives are adequacy,
neutrality, equity, accountability, and ease of administration
(Armstrong, 2002).
3. Kopczuk (2006) argues that simplification can reduce tax
evasion and avoidance more efficiently than traditional
enforcement measures of costly detection and penalty
compliance administration.T ax avoidance is a reduction by
legal means to reduce the amount of tax that is payable whilst
making a full disclosure of the material information to the tax
authorities. Examples of tax avoidance involve using tax
deductions or changing one's business structure through. By
contrast, tax evasion is a reduction by illegal means to reduce
or evade the payment of taxes. Tax evasion usually entails
taxpayers deliberately misrepresenting or concealing the true
state of their affairs to the tax authorities to reduce their tax
liability, and includes dishonest tax reporting such as under-
declaring income, profits or gains; or overstating deductions.
4. Kopczuk (2006) argues that by reducing tax complexity,
taxpayers will be more responsive to changes in taxation.
5. It should be noted that it is not the overall level of
complexity within a tax system but the costs and benefits of
taxes including the degree of complexity relative to the
advancing of policy goals. The Handbook for Tax
Simplification, (2009) provides a list and rationale for
reasons that tax complexity arises.
6. The Pennsylvania Tax Reform Code of 1971 (Act of Mar.
4, 1971, P.L. 6, No. 2 Cl. 72), codifies and enumerates
certain subjects of taxation and imposing taxes thereon;
providing procedures for the payment, collection,
administration and enforcement thereof; providing for tax
credits in certain cases; conferring powers and imposing
duties upon the Department of Revenue, certain employers,
fiduciaries, individuals, persons, corporations and other
entities; prescribing crimes, offenses and penalties.
7. The Pennsylvania Fiscal Code (Act of Apr. 9, 1929, P.L.
343, No. 176 Cl. 72) includes relating to the finances of the
State government; providing for the settlement, assessment,
collection, and lien of taxes, bonus, and all other accounts
due the Commonwealth, the collection and recovery of fees
and other money or property due or belonging to the
Commonwealth, or any agency thereof, including escheated
property and the proceeds of its sale, the custody and
disbursement or other disposition of funds and securities
belonging to or in the possession of the Commonwealth, and
the settlement of claims against the Commonwealth, the
resettlement of accounts and appeals to the courts, refunds of
moneys erroneously paid to the Commonwealth, auditing the
accounts of the Commonwealth and all agencies thereof, of
all public officers collecting moneys payable to the
Commonwealth, or any agency thereof, and all receipts of
appropriations from the Commonwealth, authorizing the
Commonwealth to issue tax anticipation notes to defray
current expenses, implementing the provisions of section 7(a)
of Article VIII of the Constitution of Pennsylvania
authorizing and restricting the incurring of certain debt and
imposing penalties; affecting every department, board,
commission, and officer of the State government, every
political subdivision of the State, and certain officers of such
subdivisions, every person, association, and corporation
required to pay, assess, or collect taxes, or to make returns or
reports under the laws imposing taxes for State purposes, or
to pay license fees or other moneys to the Commonwealth, or
any agency thereof, every State depository and every debtor
Proceedings of the Pennsylvania Economic Association 32
or creditor of the Commonwealth (title amended June 21,
1984, P.L.407, No.83).
8. The Report (April 2001) identifies other sources of
complexity including a general level of complexity:
computational complexity. This refers to the complex
calculations to determine tax liability. The greater the
complex calculations, the greater resources required such as
hiring tax professionals or purchase software to assist in
preparation of liability.
9. Act 72 of 2013 repeals Corporate Loans Tax (CLT)
effective January 1, 2014. The treasurer of the corporation
assesses and withholds the CLT from interest paid at a rate of
4 mills on the nominal value of scripts, bonds, certificates,
and other evidence of indebtedness. Banks pay it as well as
other corporations, and LLCs are included. Technically, the
issuer of a bond is withholding this 4 mill tax for resident
individuals who earn the interest on the bond, but individuals
never have to pay it directly. Non-PA corporations only pay
it if they have their treasurer located in PA. They only have
to withhold the tax on interest paid to PA residents. The CLT
adds a lot of complexity to tax administration for a small
amount of revenue. A full fiscal year impact ranges between
about $11 to $14 million impacting about 7,000 taxpayers.
By repealing the CLT, tax simplification for taxpayers and
tax administrators is increased.
In addition, Act 52 of 2013 enacts Pass Through Business
Compliance legislation, effective January 1, 2014, where one
of the provisions is to authorize the assessment of tax at the
entity level for pass-through entities such as partnerships,
LLCs and S corporations. The change will simplify tax
administration and tax compliance. It is more cost effective
to issue one assessment to a large partnership than to
separately bill hundreds of partners for small amounts. One
assessment will create one appeal as opposed to hundreds of
appeals on the same issue. The provision will only apply if
the partnership as a whole understates income by at least $1
million, and has at least 11 individual partners. It does not
make any partner liable for another partner’s tax.
10. The omnibus tax reform code legislation, Act 52 of 2013,
includes the extension of the Capital Stock and Foreign
Franchise Tax (CSFT) phase-out. The CSFT is currently at
0.89 mills. The CSFT will be phased-out at 0.67 mills on
January 1, 2014; 0.45 mills on January 1, 2015; and repealed
on January 1, 2016. The CSFT is calculated as an average of
net income and net worth of a firm. From a tax
simplification perspective, compliance and administration
costs will be eliminated for 101,000 taxpayers and the state
government administrators after repeal (Armstrong and
Brehouse, 2000).
11. Governor Corbett proposed in his budget address on
February 5, 2013, for repealing the following additional state
nuisance tax credits enacted as free standing acts separate
from the tax reform and fiscal codes:
Alternative Energy Production Tax Credit (73 P.S. §
1649.701 et. seq.). Act 1 of 2008, Special Session Number 1,
created the Alternative Energy Production Tax Credit.
Taxpayers that develop or construct energy production
projects located within the Commonwealth, which have a
useful life of at least four years, may apply to the Department
of Environmental Protection for a tax credit beginning
September 2009. The total amount of the tax credit that can
be awarded is from $2 million to $10 million per fiscal year.
Act 2009-48 reduced the annual credits available to $0 for
Fiscal year (FY) 2009-10 and FY 2010-11. Afterwards, the
credit reverts back to the previous total amounts. Beginning
for FY 2011 and onwards, no taxpayers have received this
credit. This credit is generally not used. The costs to
administer this credit are borne by the Departments of
Environmental Protection and Revenue. As present, HB
1171 to repeal this Credit was voted out of the House Finance
Committee.
Organ and Bone Marrow Donor Tax Credit (35 P.S. § 6120.1
et, seq.). The freestanding Organ and Bone Marrow Donor
Tax Credit, Act 65 of 2006, provides for a tax credit for
expenses incurred when a business firm grants to any of its
employees a paid leave of absence for the purpose of
donating an organ or bone marrow. The Tax Credit expired
after tax year 2010. The economic rationale for repealing
this tax credit is the savings in time and resources expended
by taxpayers to determine the tax credit is no longer
applicable. Currently, no bill exits to repeal this credit.
REFERENCES
Armstrong, Thomas. 2002. “State Taxation Reform Proposals
for Pennsylvania,” Pennsylvania Economic Association
Proceedings, 1-10.
Ibid and Jason R. Brehouse. 2001. “State Tax Simplification:
State Nuisance and Obsolete Provision Tax Repeals,
Including Proper Placement within the Tax Reform Code,”
Pennsylvania Economic Association Proceedings, 171-175.
Ibid. 2000. “Capital Stock and Franchise Tax Phase-Out
Initiative for Pennsylvania,” Pennsylvania Economic
Association Proceedings, 18-28.
Final Report of the Pennsylvania Tax Commission. March
1981. Pennsylvania Tax Commission.
Gale, William G. and Janet Holtzblatt. April 2002. The Role
of Administrative Issues in Tax Reform: Simplicity,
Compliance, and Administration. In United States Tax
Reform in the Twenty-First Century, George R. Zodrow and
Peter Mieszkowski (eds.) Cambridge University Press.
Proceedings of the Pennsylvania Economic Association 33
Handbook for Tax Simplification. 2009. Washington, DC:
International Finance Corporation.
Kopczuk, Wojciech. 2006. In Max Sawicky (ed.), Bridging
the Tax Gap. Addressing the Crises in Tax Administration,
Washington, DC: Economic Policy Institute, 111-143.
Joint Committee on Taxation. April 2001. Study of the
Overall State of the Federal Tax System and
Recommendations for Simplification, Pursuant to Section
8022(3)(B) on the Internal Revenue Code of 1986.
Washington: U.S. Government Printing Office.
Pennsylvania’s Tax Credit Programs. June 2010. Legislative
Budget and Finance Committee. Harrisburg, PA.
Scully, Gerald, W. 1992. Constitutional Environments and
Economic Growth. Princeton, N.J.: Princeton University
Press.
Slemrod, Joel. Winter 1990. “Optimal Taxation and Optimal
Tax Systems.” Journal of Economic Perspectives. 4:1, 157-
178.
Weimer, David L. and Aidan R. Vining. 1992. Policy
Analysis: Concepts and Practice. Englewood Cliffs, N.J.:
Prentice-Hall, Inc.
Proceedings of the Pennsylvania Economic Association 34
BUSINESS AND REAL-ESTATE PRICE CYCLES ACROSS THE US:
EVIDENCE FROM A MARKOV-SWITCHING REGRESSION EXERCISE
Aram Balagyozyan
The University of Scranton,
Scranton, PA 18510
Christos Giannikos
Baruch College,
One Bernard Baruch Way, New York, NY 10010
Kyoko Mona
Manhattanville College,
2900 Purchase St., Purchase, NY 10577
ABSTRACT
This study examines whether house price cycles lead or lag
business cycles in state-level US data from 1979 to 2012.
Using a Markov-switching model, we test various lead/lag
scenarios in every US state as well as the aggregated US. For
the majority of states, we reject the hypothesis that house
prices did not lead the economy. Between 2002 and 2011,
house prices led the economy nationally as well as in twenty-
two US states. However, the co-evolution of the two series is
driven by factors that are specific to geography and time.
INTRODUCTION
Business and real-estate downturns can be devastating to the
welfare of households and investors. The US recession of
2007 was a bitter reminder of the strong ties between real
estate markets and the broader economy. The idea that these
two sectors are closely related is not new; scholars have been
actively investigating the nature of this relationship since the
eighteenth century.
It has generally been found that housing cycles lead
economic cycles at the city, state, and national levels (Case et
al. (2000), Iacoviello(2005), Leamer (2007), Ghent and
Owyang (2010), Strauss (2013)). Broadly speaking, the
literature offers four explanations for this observation. First,
the housing market may be a proxy for another, more
important variable that leads economic cycles. For example,
Strauss (2013) suggests that improvements in consumers’
expectations of future income may forecast improvements in
both housing and the general economy. If housing reacts to
these expectations faster than the economy, then it would
predict the economy. The second explanation casually
connects housing to the broader economy through residential
construction. When the housing market picks up, so does
employment in the construction industry. Rising incomes in
the construction industry are transferred with a multiplier to
the GDP and employment. The third explanation is also
casual, but operates through the credit channel. When the
housing market weakens, so does the strength of the lenders’
balance sheets. Foreclosures and mortgage defaults
undermine the financial strength of the banking sector,
causing the banks to tighten their lending standards, which in
turn weakens the consumer demand for goods and services
that are financed by credit. We suspect that the systemic
impact of this channel became stronger with the
securitization of the credit market. Finally, there is the wealth
effect. As real estate appreciates, homeowners become
wealthier. According to the life-cycle theory of consumption,
they spend more and thus increase equilibrium output and
employment.
While the wealth and credit effects of housing on aggregate
expenditure are the two most widely quoted channels, there is
no clear consensus about their validity and magnitude. Case
et al.(2005) studied a panel of US states and found that the
housing wealth effect has an important effect upon
consumption. They also found that the wealth effect of
housing exceeds that of the stock market. Muellbauer (2007)
studied a cross-section of countries and found that the wealth
and credit effects are strong everywhere but stronger in the
US and UK. Moreover, the wealth effect in the US became
much stronger after liberalization of the housing market.
However, despite such evidence there are good reasons to
argue against this interpretation of data. If the wealth and
credit effects are truly valid and significant, then they must
work through changes in real-estate prices. But real-estate
prices are often considered rigid in the downward direction,
and for this reason, many authors reject their ability to lead
the broader economy. Leamer (2007), for example, shows
that while national housing indicators such as residential
investment and volumes are powerful predictors of business
cycles, house prices are not. Ghent and Owyang (2010) and
Strauss (2013) reach similar conclusions for US cities and
states respectively. Hence, the results of these authors
indirectly imply that the wealth and credit effects are weak
and possibly invalid
In this paper we use a Markov-switching vector
autoregressive model to investigate the coevolution of
business cycles and house price cycles in the cross-section of
all US states as well as nationally. Our findings suggest that
at the national level, house prices sometimes lead and
Proceedings of the Pennsylvania Economic Association 35
sometimes lag business cycles. Moreover, in line with the
conclusions of Leamer (2007), Ghent and Owyang (2010),
and Strauss (2013) we do not find any consistent relationship
between state business and house price cycles. Yet, for the
overwhelming majority of US states as well as the aggregated
US, we fail to reject the hypothesis that house prices do not
lead the economy. We also find that the deterioration of each
state’s business conditions in late 2007 and early 2008 was
preceded by a decline in house prices in nearly half of US
states. Although the national recession of 2007 started
sometime after a downturn in real-estate prices, the housing
market recovery lagged the economic recovery by at least
two years. These results imply that even though we fail to
observe any systematic patterns in which house prices lead
state or national economies, neither can we prove that they do
not. Hence, the credit and wealth effects of housing on
consumption cannot be written off as weak or even non-
existent; they remain legitimate subjects for further
investigation.
The remainder of the paper proceeds as follows: the
following section describes the Markov-switching model and
hypothesis testing methodology. We describe our data in the
Data section. The Results section explains our findings, and
we offer some concluding remarks in the Conclusion section.
MODEL AND METHODOLOGY
In order to investigate the joint dynamics and turning points
of house prices and the economy, we rely on the
methodology employed by Hamilton and Lin (1996) and
Smith et al. (2000). The US business cycle turning points are
traditionally dated by the National Bureau of Economic
Research. On the other hand, a vast body of literature
investigates regime changes in time series using the Markov-
switching autoregressive model of Hamilton (1989). Our
model belongs to the latter category. Specifically, we assume
that the economy and housing prices in a given region evolve
according to the following vector Markov-switching model
with no autoregressive dynamics:
[
] [
] [
] (1)
The variables and represent the growth rates of the
economy and housing prices respectively. The innovation
terms and are assumed to be jointly normally
distributed, with zero means and time-invariant correlation
coefficient . Their variance-covariance matrix therefore has
the following form:
[
] (2)
The coefficients and are the mean growth rates of the
economy and house prices respectively. These coefficients
are assumed to be time varying and subject to discrete
switches between two regimes: low ( ) and high ( ). The
superscripts refer to the regime . It follows that
the growth rates of the economy and house prices in the
region jointly assume four regimes:
(3)
Hence, the vector of means,
, may also assume four
regimes. The individual and joint regimes are unobservable,
but we assume that they follow a Markov process.
Since forecasting is not our primary objective, we do not
include any autoregressive or moving average components in
the model. We are only interested in the mean coefficients
and . Since the asymptotic estimates of these coefficients
remain unbiased and are identical to those estimated under
more sophisticated linear specifications, we use the most
parsimonious model.
Shifts between different regimes are governed by a
transition probability matrix, which we estimate along with
the other parameters of the model (1)-(3). The matrix is:
[
] (4)
where , is the transition
probability that the regime was in the previous period and in the subsequent period. We assume that the matrix is time-
invariant. Each column must add to unity. Hence, when
estimating the model only twelve probabilities are left as free
parameters. A single element in each column is equal to one
minus the sum of the remaining elements in the column.
Using the procedures described in [8], [10], and [11] we can
obtain maximum likelihood estimates of the following
regression parameters:
(5)
as well as the transition probabilities in (4). Thus, the
unrestricted model (1) has a total of 19 free parameters.
When estimating the model, we enforce the constraints
and .
In principle, the transition probability matrix (4) should
describe all aspects of the joint dynamics of the housing
market and economy. A useful way of thinking of these
Proceedings of the Pennsylvania Economic Association 36
relationships is through the classification of shocks that may
affect both the economy and the housing market. The
transition probabilities reflect five types of shocks. First,
there may be no shocks. In this case inertia dominates both
sectors, so the regimes do not change. If regime-changing
shocks are rare in the sample, we expect to estimate
significant values of the probabilities on the principal
diagonal of the transition probability matrix, , , ,
and . Second, sterile shocks, that affect one sector but not
the other. If these type of shocks is dominant we expect to
estimate significant values of , , , and . Third,
systemic shocks affecting both sectors simultaneously,
although not necessarily in the same direction. If this type of
shocks is dominant we expect to estimate significant values
of probabilities along the minor diagonal of the transition
probability matrix, , , , and . Fourth, shocks
affecting the economy first, and housing prices after some
delay. If this type of shocks are dominant, then real-estate
prices would tend to follow the economy. In this case, we
would expect to estimate significant values of and .
Finally, shocks affecting housing first, and the general
economy with some delay. If this type of shocks are
dominant, then the economy would tend to follow real-estate
prices and we would expect to estimate significant
probabilities and .
The above interpretations imply that if housing prices in a
region tend to lead the economy then we should be able to
reject hypothesis : . Rejection of this
hypothesis implies that we are unable to reject the wealth and
credit effects of housing. Similarly, the rejection of
hypothesis , implies that the economy
in a given region may lead housing prices. Although the
rejection of hypothesis does not lead to any conclusions
about the wealth and credit effects, it would shed light on
how consistently house prices in a region lead the economy.
If for a given region both hypotheses ( and ) are rejected,
then we can conclude that both leading relationships exist in
the data, although not at the same time. On the other hand, if
we fail to reject either hypothesis, then neither sector leads
the other. For example, this may happen when housing and
business cycles coincide. Based the outcomes of tests and
, we can place each state into one of the four categories
. These four categories are summarized in Table 1.
Note that the number of states for which we reject (or not)
any of the two hypothesis can be obtained by combining the
states fallen in two categories along the rows or columns of
Table 1. If, for example, we are interested in obtaining the
states for which only hypothesis was rejected, we can
combine states which fall in categories 1 or 2.
We use the likelihood ratio test to determine the significance
of transition probabilities and reject (or not) hypotheses
and for each individual US state as well as the US as a
whole. Since each test imposes two restrictions on the
transition probability matrix, the likelihood ratio statistics has
a distribution with two degrees of freedom. We test these
hypotheses at significance level.
DATA
In order to estimate equation (1), we rely on two monthly
indicators from January 1979 to September 2012. First, is the
monthly coincident index of economic activity, which
proxies for the level of economic activity in a each US state
and the US as a whole. This seasonally adjusted index is
compiled by the Federal Reserve Bank of Philadelphia and
combines four state-level economic indicators: non-
agricultural payroll employment, unemployment rate,
average hours worked in manufacturing industries, and real
wage and salary disbursement.
Our second data set is the Freddy Mac housing price index,
which reflects the average level of real-estate prices in each
state. The Freddie Mac index includes valuation and location
data, and is based on the combined portfolio of loans that
were purchased by either Freddie Mac or Fannie Mae since
January 1979. The portfolio of loans covers every state,
although it reflects Freddie Mac and Fannie Mae’s collective
market coverage and thus is not random across states.
Furthermore, the loans are limited to one-family detached
and town-home properties financed by first-lien conventional
and conforming loans. We seasonally adjust the housing
price index using the Eviews (7th version) seasonal
adjustment utility which utilizes the U.S. Census Bureau’s
X12 seasonal adjustment program.
RESULTS
For each US state as well as the aggregated US, we estimate
the Markov-switching vector autoregressive model (1).
Parameter estimates for the unrestricted model are presented
in Table 2. Following this exercise we test hypotheses and
, and based on the outcome of the tests, place each region in
one of the four categories as described in the Model and
Methodology section. The results of this exercise are
presented in the first column of Table 2. When the entire
sample between 1979 and 2012 is used, we reject both
hypotheses and
for 40 out of 50 states. This implies
that for 80% of all US states, there is statistical evidence that
housing price cycles in the region may both lead and lag
business cycles at different times. Hence, consistent with
Leamer (2007), Ghent and Owyang (2010), and Strauss
(2013), we cannot establish that house prices are reliable
leading indicators of state business cycles. On the other hand,
since there are no states for which both hypotheses are
sustained, we can conclude that economic and real-estate
price cycles also do not have a consistent tendency of
evolving independently. North Dakota and Hawaii are the
only states where house prices tend to lead business cycles.
Proceedings of the Pennsylvania Economic Association 37
ND is largely an agricultural state, and in this case the test
decision seems to be dominated by a single housing recovery
that pre-dated an economic recovery in the early 80s. In eight
other largely agricultural and oil-producing states, real-estate
prices follow business cycles but not the other way around.
The aggregated US falls in Category 1. Thus, national real-
estate price cycles and business cycles each have some
probability of leading the other as suggested by the cross-
correlation analysis. Interestingly, for the US, the -value of
hypothesis is 0.008. Therefore at the 1% significance level
we would reject hypothesis only marginally. This implies
that the data are fairly close to placing the US as a whole
(along with Hawaii and North Dakota) in the 3 category,
meaning that real-estate prices lead the economy but not the
other way around.
Strauss (2013) suggests that during the US recession of 2007,
the worsening economy in most US states was preceded by a
decline in building permits. We are interested in whether the
pre-recession period was also accompanied by a decline in
house prices. If house prices are downward rigid, as for
example Gao et al. (2009) suggest, then the housing decline
prior to the recession of 2007 must have impacted the
economy through channels other than the wealth and credit
effects. This would obviously go against conventional
wisdom. We re-estimate the model and test hypotheses and
for the sub-sample starting in January 2002, right after the
end of the national recession in 2001, and ending in
September 2012. The results of the test for this sub-sample
are presented in the second column of Table 2. Although
Category 1, in which housing prices and business cycles may
each lead the other, remains the most populous (19 states),
the number of states in which house prices lead business
cycles (Category 3) increased from 2 to 10. The US as a
whole remained in Category 1, but it is again a borderline
case, close to the Category 3 region in which real-estate
prices lead the economy but not the other way around.
Category 2, in which the economy leads house prices but not
vice versa, now consists of 17 mostly agricultural and oil-
producing mid western and southern states
Being intrigued by the small number of states in Category 3
(house prices lead economies but not vice versa), we note
that our sample may be influenced by the recent signs of
housing recovery that started appearing in early 2012, more
than two years after the national recovery. Hence we
conjecture that some states ended up in Category 1 solely on
the basis of these recent data, but should be in Category 3 if
we focus the analysis on years prior to and during the
recession. That is, if the housing downturn in a region started
before an economic downturn and ended after an economic
recovery, we are likely to (properly) conclude that house
prices both lead and lag the economy, and label this region as
Category 1. However, if we what to answer the specific
question of whether house prices led the economy before the
recession of 2007, we should exclude the post-recession
housing recovery from our sample. To restrict our viewpoint
to the years prior to and during the recession of 2007, we
consider a second, shorter sub-sample running from January
2002 to September 2011. The test results based on this sub-
sample are presented in the third column of Table 2. Between
January 2002 and September 2011, house prices led state
economies but not the other way around in 21 out of 50 US
states, as well as nationally. During this period, state
economies led house prices in 15 mostly agricultural states.
This confirms our suspicion that in many US states, declining
house prices led the economic downturn. However, it is hard
to draw a single conclusion about the interplay between
house prices and the economy that holds for all US states and
the aggregated US. This result is a confirmation of one of the
main conclusions reached by Hamilton and Owyang (2011),
namely that there is no single pattern that characterizes the
behavior of business cycles across different US states.
CONCLUSIONS.
In this paper we investigated whether real-estate price cycles
can be written off as precursors of business cycles. If so, then
the validity of the wealth and credit channels as transmission
mechanisms from housing to the economy should also be
dismissed. Using a Markov-switching vector autoregressive
model, we test the order of precedence of state-level house
price and economic cycles in the US over the sample period
between 1979 and 2012. We fail to reject the hypothesis that
real-estate prices did not lead business cycles in 42 out of 50
US states as well as nationally. In a sub-sample preceding the
recession of 2007 and the recent housing market recovery
(2002 to 2011), house price cycles preceded regional
economies but not vice versa in 21 US states and the
aggregated US. This result implies that real-estate prices and
hence the wealth and credit effects of housing should not be
omitted in theoretical and empirical research
A word of caution is due. Although our results indicate that
house prices should not be dismissed as important precursors
of business cycles, they do not imply that house prices are
reliable predictors of state and national recessions and
expansions. If anything, the opposite is true. Consistent with
the findings of Ghent and Owyang (2010) and Strauss
(2013), our results indicate that house prices may lead, lag, or
coincide with regional economies, and that the dominant
patterns depend upon place and time. Thus, the interplay
between real-estate prices and economic activity requires
further theoretical and empirical research.
Proceedings of the Pennsylvania Economic Association 38
Table 1: Possible decisions based on the outcomes of hypotheses A and B
Test A:
Reject Do not reject
Test B
:
Reject
Category1
Business cycles
follow housing
price cycles
&Housing price
cycles follow
business cycles
Category 3
Business cycles
follow housing
price cycles &
Housing price
cycles do not
follow business
cycles
Do not
reject
Category 2
Business cycles do
not follow housing
price cycles &
Housing price
cycles follow
business cycles
Category 4
Business cycles do
not follow housing
price cycles
& Housing price
cycles do not
follow business
cycles
Proceedings of the Pennsylvania Economic Association 39
Table 2: Test decisions based on the outcomes of hypotheses A and B.
Jan.1979-
Sep.2012
Jan.2002-
Sep.2012
Jan.2002-
Sep.2011
Category 1
Business cycles follow
housing price cycles
&
Housing price cycles
follow business cycles
(A: Reject, B: Reject)
AK,AL,AR,
AZ,CA,CO,
CT,DE,FL,
GA,IA,ID,
IL,IN,KY,
LA,MA,MD,
ME,MI,MN,
MO,MS,NC,
NE,NH,NJ,
NM,NV,NY,
OH,OK,
OR,PA,RI,
SD,US,VT,
WA,WI,WV
(41)
CA,CO,CT,
IL,IN,MD,
ME,MN,NH,
NY,OH,OR,
RI,US,VA,
WA,WI,WV,
WY
(19)
CO,CT,IN,
KS,ME,NY,
OH,OR
(8)
Category 2
Business cycles do not
follow housing price
cycles
&
Housing price cycles
follow business cycles
(A: Do not Reject;
B: Reject)
KS,MT,SC,
TN,TX,UT,
VA,WY
(8)
AL,AR,AZ,
GA,IA,KS,
KY,MO,MT,
ND,NE,OK,
SC,SD,TN,
TX,UT
(17)
AL,AZ,GA,
IA,ID,KY,
MN,MO,ND,
NE,OK,SD,
TN,UT,WV
(15)
Category 3
Business cycles follow
housing price cycles
&
Housing price cycles
do not follow business
cycles
(A: Reject;
B: Do not Reject)
HI, ND (2)
AK,DE,HI,
LA,MA,MS,
NC,NJ,NV,
PA
(10)
AR,CA,DE,
HI,IL,LA,
MA,MD,MI,
MS,NH,NJ,
NM,NV,PA,
RI,US,VA,
VT,WA,WI,
WY(22)
Category 4
Business cycles do not
follow housing price
cycles
&
Housing price cycles
do not follow business
cycles
(A: Do not Reject;
B: Do not Reject)
(0)
FL,ID,MI,
NM,VT
(5)
AK,FL,MT,
NC,SC,TX
(6)
Proceedings of the Pennsylvania Economic Association 40
REFERENCES
Case, K., Glaeser, E., and Parker, J. 2000. Real estate and the
macroeconomy. Brookings Papers on Economic Activity,
2000(2):119-162.
Case, K. E., Quigley, J. M., and Shiller, R. J. 2005.
Comparing wealth e¤ects: The stock market versus the
housing market. Advances in Macroeconomics, 5:1-32.
Gao, A., Lin, Z., and Na, C. F. 2009. Housing market
dynamics: Evidence of mean reversion and downward
rigidity. Journal of Housing Economics, 18: 256-266.
Ghent, A. and Owyang, M. 2010. Is housing the business
cycle? evidence from US cities. Journal of Urban Economics,
67:3: 336-351.
Hamilton, J. D. 1989. A new approach to the economic
analysis of nonstationary time series and the business cycle.
Econometrica, 57(2): 357-384.
Hamilton, J. D. 1994. Time Series Analysis. Princeton
University Press, Princeton, NJ.
Hamilton, J. D. and Lin, G. 1996. Stock market volatility and
the business cycle. Journal of Applied Econometrics, 11:
573-593.
Hamilton, J. D. and Owyang, M. T. 2011. The propagation of
regional recessions. NBER Working Paper Series, w16657.
Iacoviello, M. 2005. House prices, borrowing constraints,
and monetary policy in the business cycle. The American
Economic Review, 95(3):739-764.
Kim, C.-J. and Nelson, C. R. 1999. State-Space Models with
Regime Switching: Classical and Gibbs-Sampling
Approaches with Applications.The MIT Press.
Krolzig, H.-M. 1997. Markov-Switching Vector
Autoregressions: Modelling, Statistical Inference, and
Application to Business Cycle Analysis. Springer.
Leamer, E. E. September 2007. Housing IS the business
cycle. NBERWorking Paper, No. 13428.
Muellbauer, J. 2007. Housing, credit and consumer
expenditure. Proceedings of the Jackson Hole Symposium on
Housing, Housing Finance, and Monetary Policy.
Smith, R., Sola, M., and Spagnolo, F. 2000. The prisoner’s
dilemma and regime-switching in the greek-turkish arms
race. Journal of Peace Research, 37(6):737-750.
Strauss, J. 2013. Does housing drive state-level job growth?
building permits and consumer expectations forecast a state’s
economic activity. Journal of Urban Economics, 73:77-93.
Proceedings of the Pennsylvania Economic Association 41
A MODEL OF RELATIVE CONSUMPTION
Chong Hyun C. Byun
Department of Economics
Wabash College
301 W. Wabash Ave.
Crawfordsville, IN 47933
ABSTRACT Habit formation and Keeping Up with the Joneses are two
specifications of relative consumption in which a household’s
utility depends, not only on its own current consumption, but
also on previous consumption levels, or on the consumption
levels of their neighbors. In such an economy, externalities
are generated by such actions. These externalities are not
accounted for when a household chooses its current
consumption level. Therefore, the competitive solution is not
the same as the social planner’s. Consequently, there is room
for improvement in the form of an optimal taxation policy.
I. INTRODUCTION
Relative consumption, in the form of either keeping up or
catching up with the Joneses, or habit formation, has now
been established as a relevant analytical tool in modifying
standard preference specifications to more adequately
represent consumer behavior. A household’s utility can
depend not only on its own current consumption level, but
also on its previous consumption levels, as well as on the
consumption levels of their neighbors. However, in such an
economy, externalities are generated by such comparisons.
Households do not realize that in choosing their
consumption level today, they will be influencing what their
neighbors do both today and tomorrow. And if they are
strongly influenced by habits, then their choice today will
also affect what their choices will be tomorrow. These
externalities are not accounted for when a household
chooses what to consume today. As a result, households
will consume too much, compared to the case where a
social planner chooses optimal consumption levels for all
households. Consequently, there is room for improvement
in the form of an optimal taxation policy that will induce
households to choose consumption levels that are identical
to the social planner’s outcome.
In this paper I will analyze a model that includes both habit
formation and catching up with the Joneses in the
household preference specification. Compared to the social
planner’s solution, where the planner takes the externalities
into account, the competitive solution is not the same as the
efficient one. Government intervention is required to ensure
the efficient solution, and this comes in the form of taxes on
capital and labor. Taxation on either capital or labor would
change the relative price of consumption either today or in
the future, and would encourage agents to shift
consumption accordingly.
Ljungqvist and Uhlig (2000) consider a model with
catching up with the Joneses preferences and find an
optimal tax policy for this economy. For an economy where
individual welfare depends on relative consumption and
income levels, Boskin and Sheshinski (1978) derive a
taxation system that will optimally redistribute wealth and
consumption. Alonso-Carerra et al. (2004) formulate a
model that includes both habit formation and both keeping
up and catching up with the Joneses preferences. Based on
these consumption spillovers, they derive an optimal tax
rate that will restore the efficient outcome to the economy.
My work builds upon that of both Ljungqvist and Uhlig and
Alonso-Carerra et al. In Ljungqvist and Uhlig’s model, the
utility function exhibits catching up with the Joneses
preferences. Based on this specification, they develop
optimal tax policies in an economy with productivity
shocks. Then they calculate welfare gains due to taxation by
introducing a stochastic productivity shock into three types
of economies (laissez-faire without taxes, the social planner
outcome with optimal taxes, and an economy where the tax
is kept constant at the steady state value). These welfare
gains are compared to those of two nonstochastic
economies: the laissez-faire and the social planner
outcomes.
However, Ljungqvist and Uhlig do not incorporate capital
accumulation in their model. In Alonso-Carrera et al,
preferences are modified to include habit formation and
both keeping up and catching up with the Joneses
preferences. Capital accumulation is included in the model,
and they analyze the optimal taxation rates on both
consumption and capital. But they do not include labor
supply, nor do they consider the welfare gains from the
optimal taxation schemes. My focus is to build on these two
prior works and extend them by including both capital
accumulation and labor supply in a model that includes both
habit formation and catching up with the Joneses in the
specification for household preferences. The next objective
is to compare the competitive equilibrium outcome with the
social planner’s solution for this model. Based on these
different outcomes, an optimal tax rate on labor and capital
can be determined.
Proceedings of the Pennsylvania Economic Association 42
The remainder of the paper is organized as follows. Section
II outlines the model and the household’s choice problem
when preferences are modified to include both habit
formation and catching up with the Joneses. Then the social
planner’s problem is outlined and I compare the results
from the competitive solution versus the social planner’s
solution. Since the preference specification generates
consumption externalities, these two outcomes will be
different. Consequently, there is room for improvement and
welfare gains if the government steps in and imposes an
optimal tax on consumption. This tax induces households
to choose a consumption level that is in accordance to the
socially optimal level. The derivation of the optimal tax is
described in Section III. In Section IV, the equilibrium
efficiency outcome is discussed. Finally, Section V
concludes.
II. THE MODEL
2.1 The Household’s Problem
Consider an economy with infinitely many identical
households, each with the same utility function, facing an
infinite lifetime. Here, a household’s preferences depend
not only on the current value of their consumption, but also
on the lagged value of their consumption, and also on the
lagged value of average consumption over all households in
the economy. The utility function is the standard constant
relative risk aversion format.
(1)
The variables are defined as follows: c t is consumption at
period t for the agent, ct-1 is lagged own consumption at t-1
(habit formation), and is the consumption in period t-1
averaged over all households, that is, catching up with the
Joneses. The parameters are defined as follows: > 0 is the
coefficient of relative risk aversion, measures
how important own lagged consumption is to an agent (i.e.,
the strength of habit formation), and measures
the importance to the household of lagged average
consumption. If 0 < < 1, this represents catching up with
the Joneses, that is, marginal utility of lagged average
consumption is negative. Note that if –1 < < 0 is true, this
represents lagging behind the Joneses, in that marginal
utility of lagged average consumption is positive.
The household faces the budget constraint:
(2)
where kt is capital, wt is wage rate, lt is labor, rt is rental
rate, kt is capital, is depreciation of capital, and
represent taxes on labor and capital respectively, and S t is
the lump sum transfer payment from the government.
The government faces a budget constraint of
(3)
where it receives income from taxes collected on both
capital and labor income. The production technology for the
economy is a standard Cobb-Douglas specification:
(4)
where Y is the output, K is the capital input, and L is the
labor input. Equation (4) can be rewritten as
(5)
where kt is the capital to labor ratio. In a competitive
equilibrium, the factor prices will be equal to their
respective marginal productivities, which are given by
(6)
where MPK stands for the marginal productivity of capital.
The rental rate on capital is equal to its marginal productivity.
The same is true for labor, and the wage rate is equal to the
marginal productivity of labor, given by
(7)
where MPL stands for the marginal productivity of labor. I
will define the marginal utility of current consumption as
(8)
and the marginal utility of lagged consumption as
(9)
Then for the household’s maximization problem, the first
order conditions with respect to consumption, capital, and
labor are
0t
θ1
t
σ1
1t1ttt
θ1
lA
σ1
)cαγc(cβu(t)
1tc
(0,1)γ
1,1)(α
t
k
ttt
l
tttt1ttS)τ(1kr)τ(1lwδ)k(1kc
k
t
l
t τand τ
tt
l
ttt
k
ttlwτkrτS
1η0 ,LKY η1
t
η
tt
η
tttk)f(ky
)(kf'ηkL
KηMPKr
t
η1
t
η1
t
t
t
tttt)k(kf')f(kMPLw
t
1tt1tt
1c
)c,c,c,u(c(t)u
1t
1tt1tt
2c
)c,c,c,(c(t)u
Proceedings of the Pennsylvania Economic Association 43
respectively, where t is the Lagrange multiplier for the
budget constraint. Equation (10) is the first order condition
with respect to current consumption, and represents the
marginal utility of consumption in period t. Equation (11) is
the standard consumption Euler equation for the household,
based on the first order condition with respect to capital.
Finally, Equation (12) shows that the marginal rate of
substitution between consumption and leisure are equal to
the after tax wage rate. An intuitive explanation of this last
equation is that it shows the rate at which a household is
willing to substitute consumption for leisure. The price of
leisure is the foregone income that could have been earned
by working, and ultimately represents the foregone amount
of consumption.
The competitive equilibrium for this economy is defined by
the positive paths of ct, kt, lt, and t satisfying (10), (11) and
(12) along with (2) and (3), and the following transversality
conditions.
(13)
(14)
In a symmetric equilibrium, is true. Define the
gross rate of growth of marginal utility as
(15)
Then from the first order conditions (10) and (11), I derive
the Euler equation
(16)
Defining the gross rate of growth of consumption as
(17)
then Equation (15) becomes
(18)
Combining the individual’s and government’s budget
constraints, I derive the resource constraint of the economy:
(19)
The first order difference equations (16), (17), (18), and
(19) with the transversality conditions and the initial
conditions on k0 and c-1 fully describe the equilibrium path
of the variables xt, t, ct, kt.
Assume now the government follows a stationary fiscal
policy so that taxes are constant throughout time, i.e.
. Then at the steady state, Equations
(16), (17), (18) and (19) respectively become:
(20)
(21)
(22)
To characterize the steady state, let . An
interior steady state exists if and only if the following
condition holds:
(23)
2.2 The Social Planner’s Problem
The social planner now considers the problem of
maximizing utility for the entire economy. In order to do so,
the planner must account for the consumption externality
generated by the household’s preference specification. The
planner internalizes the consumption externality such that
, i.e. that each household’s past consumption is
the same as the average consumption for the entire
economy. Here, catching up with the Joneses no longer
matters, and the utility function the social planner now
faces is the following:
(24)
where are the variables for consumption in
period t, consumption in period t-1, and labor in period t for
(12) 0)τ(1)k(kf')f(kλAl
(11) 0δ)(1)τ)(1(kf'βλλ
(10) 0γ)()cαγcβ(cλ)cαγc(c
l
ttttt
θ
t
k
1t1t1tt
σ
tt1tt
σ
1t1tt
0kλlim1tt
t
0(t)cuβlimt1
t
t
1t1tcc
(t)u
1)(tu
1
1
t
δ)](1)τ)(1(kβ[f'
1
βγ1
βγ1k
1t1t
1t
1t
t
tc
cx
α)(γ
x
11α)(γx
t
σ
1
t1t
tcδ)k(1)f(kk
tt1t
t τ τand ττ ll
t
kk
t
1x
)τβ(1
δ)β(11(k)f'
k
δkf(k)c
(0,1))τβ(1 k
δ)β(11
11 tt cc
θ1
lA
σ1
α)(γccβ(t)u
θ1
t
σ1
1tt
0t
t
tttlcc ˆ and ,ˆ ,ˆ
1
Proceedings of the Pennsylvania Economic Association 44
the social planner’s problem. The resource constraint the
social planner is
(25)
I define the following notation for the partial derivatives of
the utility function with respect to the two variables as
(26)
(27)
The first order conditions with respect to consumption,
capital, and labor are
(28)
(29)
(30)
respectively, where is the Lagrange multiplier for the
resource constraint. The solution to the social planner’s
problem is defined by the positive paths of ,
satisfying (28), (29), (30), the resource constraint, the initial
conditions , and the following transversality
conditions.
(31)
(32)
Define the gross rate of growth of marginal utility in the
social planner’s economy as follows.
(33)
Using first order conditions (28) and (29), and (33), we get
the Euler equation:
(34)
Define the gross rate of growth of consumption under the
social planner’s economy as follows.
(35)
Thus, from (33), we can derive the following.
(36)
The set of first order difference equations (34), (35), (36),
and (25), with the transversality conditions, and the initial
conditions completely describe the dynamics of the
variables At the steady state:
(37)
(38)
(39)
These three steady state conditions are exactly equivalent to
Equations (20), (21), and (22) from the competitive
equilibrium. The only difference between the competitive
equilibrium solution and the social planner’s solution is
between Equations (16) and (34). When k = 0 we get
When tax on capital is zero, the steady
state of the competitive solution coincides with that of the
efficient solution.
III. OPTIMAL TAXATION POLICIES
Inefficiencies may arise from the competitive equilibrium
outcome since agents are not taking into account the
spillover effects of their own consumption. The
externalities arising from an agent’s own lagged
consumption (the habit formation) will influence marginal
utility today. Agents are also influenced by past
consumption averaged over all the agents (the catching up
effect). When agents choose consumption today, they will
affect marginal utility of consumption tomorrow for all
agents in the economy.
Since the competitive equilibrium outcome can be
inefficient, there is a role for the government to step in with
a tax policy. An optimal tax will force consumers to shift
their consumption between periods such that the
competitive equilibrium outcome will be the same as the
social planner’s solution. We consider in this section what
the optimal taxation rates should be on labor and capital.
ttt1tckδ)(1)kf(k
t
1tt
1c
)c,c(u(t)u
1t
1tt
2c
)c,c(u(t)u
0λα)(-γα)(γccβα)(γcct
σ
t1t
σ
1tt
0)]k(f'δ)[(1λβλ1t1tt
0]k)k(f')k[f(λlAtttt
θ
t
tλ
ttttλ and ,l ,k ,c
10c and k
0kλlim1tt
t
0c(t)uβlimtt
t
t
(t)u
1)(tuˆ
1
1
t
δ)](1)k(β[f'
1ˆ
ˆα)β(γ1
ˆα)β(γ1
1t
t
t
1t
1ˆ
ˆˆ
t
t
tc
cx
α)(γ
x
11ˆα)(γx
t
σ
1
t1t
10c ,k
.k,c ,x ,ˆtttt
1ˆx
β
δ)β(11)kf(
kδ)kf(c
k.k c,c ,ˆ
Proceedings of the Pennsylvania Economic Association 45
To characterize the optimal tax on labor, I combined
Equations (12) and (30) to derive an expression for tax on
labor.
(40)
Using the first order condition for the competitive
equilibrium, Equation (10) and the first order condition for
the social planner’s problem, Equation (28), the tax on labor
can be expressed as the following equation.
(41)
Based on this equation, I can characterize the optimal tax on
labor, according to it.
Proposition 1: Optimal Tax on Labor
The optimal tax on labor is defined by
represents the rate of time preference
represents catching up with the Joneses
represents habit formation.
Consequently, the sign of βα determines the direction of the
tax, either positive or negative.
If α > 0, then βα > 0, which implies a positive tax on labor
income. Since the household has catching up with the
Joneses preferences, this positive tax is required to
discourage excess consumption. This is because the
marginal utility of one additional unit of consumption is
greater under the decentralized case compared to that in the
social planner’s economy. Households have to be
discouraged from consuming too much in the decentralized
economy, so this positive tax is necessary.
On the other hand, if α < 0 then βα < 0, which implies a
negative tax on labor. For α < 0, this implies that
households have a preference structure such that they
receive positive marginal utility with increases in average
lagged consumption. This is termed “lagging behind the
Joneses”, in that the household’s utility increases when their
neighbors’ consumption in the past increases. Thus the
households are not consuming enough; they are neither
keeping up nor catching up, but instead falling behind. In
this case, a negative tax (that is, a subsidy) is required on
labor income. The households have to be encouraged to
consume more. Note also that for α < 0, this means that the
marginal utility of one additional unit of consumption is
greater under the social planner’s case than in the
decentralized case. Once again, households have to be
encouraged to increase their consumption in the
decentralized economy, so a negative tax is necessary.
The size of the tax on labor is influenced by the parameter
γ, which represents the strength of habit formation. This
parameter γ is restricted to be between zero and one. For
bigger values of γ, habits are stronger, and as a
consequence, will be larger. On the other hand, if γ is
smaller, then habits matter less to the household, and the tax
on labor income will be smaller. Finally, it is worth noting
that if α = 0 or both α = 0 and γ = 0, then . No taxes
would be necessary in this case since catching up with the
Joneses preferences and habit formation do not exist.
To characterize the optimal tax on capital, we combine (11)
and (29) and evaluate them along the efficient path, to get
the optimal taxation condition for capital.
(42)
where the superscripts denote d=decentralized economy and
p=planned economy.
Proposition 2: Characterizing the optimal capital tax
rate
(a) For MRSd> MRS
p, we have
(b) For MRSd<MRS
p, we have
(c) For MRSd=MRS
p, we have
Part (a) means that consumers in the decentralized economy
want to shift their consumption to period t+1. To discourage
them from doing so, it is necessary to put a tax on capital
income. Part (b) means that consumers in the decentralized
economy are less willing to shift their consumption to the
future. In this case, the government must subsidize (i.e. use
a negative tax) on capital to encourage people to consume
less now and more in the future. Part (c) means that when
MRS for both the planned and the decentralized economies
are the same, so no tax is necessary.
IV. EQUILIBRIUM EFFICIENCY
Compare the equilibriums for an individual, when
. Comparing (6) and (34), we see that
(43)
t
tl
tλ
λ1τ
βγ1
βα1τ
βγ1
βατ
1β0
1,1)(α
(0,1)γ
0
)τ)(1k(f'δ)(1
)k(f'δ)(1
)c,c,c,c(MRS
)c,c,c,c(MRS
k
1t1t
1t
1tt1t2t
p
1tt1t2t
d
.0k
t
.0k
t
.0k
t
0ττ lk
1)(tuβ(t)u
2)(tuβ1)(tu
1)(tβu(t)u
2)(tβu1)(tu
21
21
21
21
Proceedings of the Pennsylvania Economic Association 46
So the competitive equilibrium is efficient if and only if
(44)
for all t and some along the competitive equilibrium path
of consumption.
Since the externalities enter the utility function
subtractively, are linearly dependent for all t.
In fact, it holds that along the competitive
equilibrium path of consumption. So by using (15), now
(41) becomes the efficiency condition
(45)
From this we have the following proposition.
Proposition 3: Let
(a) The competitive equilibrium is Pareto Optimal at the
steady state if and only if = 0.
(b) The competitive equilibrium is Pareto Optimal off the
steady state if and only if = = 0.
Since along the competitive equilibrium path
of consumption, the tax on labor at the steady state becomes
(46)
For = 0, we get . If = = 0, then utility will only
depend on current consumption, and there will be no
externalities, so the competitive equilibrium will be
efficient off the steady state as well.
V. CONCLUSION
A utility function exhibiting habit formation and catching
up with the Joneses will result in inefficiencies in the
competitive equilibrium outcome. Households do not
internalize the externalities their consumption choices may
cause. Thus, their consumption decisions today will
influence their own and all other households’ marginal
utilities today and tomorrow. Due to the distortions
introduced by this externality, the competitive equilibrium
will not be identical to the social planner’s solution.
Therefore, there is a role for government intervention in the
form of taxes on labor and capital. Optimal taxes will
induce agents to choose consumption levels that are
identical to the social planner’s solution.
Based on the work of Ljungqvist and Uhlig (2000) and
Alonso-Carerra et al. (2004), I extend their models to
include capital accumulation and labor supply in the
specification. For this new model, I derived the outcomes
under the competitive equilibrium and the social planner’s
problem. From these outcomes, I further derived optimal
taxes on capital and labor that will induce households to
choose consumption levels that are in accordance with the
social planner’s choice. It turns out that the optimal tax on
labor depends on the parameters that represent both the
catching up with the Joneses preferences and habit
formation. Depending on the direction (positive or negative)
and size of the “catching up” parameter, the tax on labor
will be either positive to discourage consumption, or
negative to encourage consumption. The habit formation
parameter plays a role in determining the magnitude of this
tax. If habits are stronger (i.e. the habit formation parameter
is larger), then taxes will be larger in either direction.
For my future research, I would like to consider the welfare
gains associated with these optimal taxation policies.
Ljungqvist and Uhlig determined the welfare gains from the
optimal taxation in their model, using the following three
stochastic economies: laissez-faire without taxes, the social
planner outcome with optimal taxes, and the social planner
outcome with taxes held constant at the steady state value.
These three are compared to two economies with no
productivity shock: the laissez faire and the social planner
to see the possible gains under these types of economies.
Based on the optimal taxes I derived for my model, I plan to
examine the welfare gains for the same economies as above.
This future work will help establish the importance of
relative consumption models for welfare gains compared to
the competitive equilibrium outcome.
REFERENCES
Abel, Andrew B., “Asset Prices Under Habit Formation and
Catching Up with the Joneses,” American Economic
Review, May 1990, vol. 80, no. 2, pp. 38-42.
Alessie, Rob and Lusardi, Annamaria, “Consumption,
Savings, and Habit Formation,” Economics Letters,
February 1997, 55, pp. 103-108.
Alonso-Carerra, Jaime and Caballé, Jordi, “Income
Taxation with Habit Formation and Consumption
Externalities,” Working Paper, July 2001.
Alonso-Carerra, Jaime, Caballé, Jordi, and Raurich, Xavier,
“Consumption Externalities, Habit Formation, and
Equilibrium Efficiency,” Scandinavian Journal of
Economics, Vol. 106, no. 2, 231-251, 2004.
Boskin, Michael J. and Sheshinski, Eytan, “Optimal
Redistributive Taxation when Individual Welfare Depends
1)(tβu(t)uψ1)(tuβ(t)u2121
(t)u and (t)u11
(t)u(t)u11
ψ1ψ)]γ(1β[α t
.0 k
t
l
t
)()(ˆ11
tutu
βγ1
α)β(γ11 τ l
0τ l
Proceedings of the Pennsylvania Economic Association 47
Upon Relative Income,” Quarterly Journal of Economics,
November 1978.
Campbell, John Y. and Deaton, Angus S., “Why is
Consumption so Smooth?” Review of Economic Studies,
July 1989, 56(3), pp. 357-73.
Carroll, Christopher D., “Solving Consumption Models
with Multiplicative Habits,” Economics Letters, 2000, vol.
68, pp. 67-77.
Carroll, Christopher D.; Overland, Jody and Weil, David
N., “Savings and Growth with Habit Formation,” American
Economic Review, June 2000, vol. 90, no. 3, pp. 341-355.
Constantinides, George M., “Habit Formation: A
Resolution of the Equity Premium Puzzle,” Journal of
Political Economy, June 1990, 98(3), pp. 519-543.
Duesenberry, James S., “Income, Saving, and the Theory of
Consumer Behavior,” Harvard University Press, 1952.
Dynan, Karen, “Habit Formation in Consumer Preferences:
Evidence from Panel Data,” American Economic Review,
June 2000, vol. 90, no. 3, pp. 391-406.
Faria, Joao R., “Growth and Labor Supply in the Presence
of Habit Formation in Consumption,” Economics Bulletin,
2002, vol. 4, no. 5, pp. 1-5
Ferson, Wayne, and Constantinides, George, “Habit
Persistence and Durability in Aggregate Consumption,”
Journal of Financial Economics, 1991, vol. 29, pp. 199-
240.
Fisher, Walter H. and Hof, Franz X., “Relative
Consumption, Economic Growth, and Taxation,” Journal of
Economics, March 2000, vol. 72, no. 3, pp. 241-262.
Frank, Robert, “The Demand for Unobservables and Other
Nonpositional Goods,” American Economic Review, March
1985, vol. 75, no. 1, pp. 101-116.
Fuhrer, Jeffrey C., “Habit Formation in Consumption and
its Implications for Monetary-Policy Models,” American
Economic Review, June 2000, vol. 90, no. 3, pp. 367-390.
Gali, Jordi, “Keeping Up with the Joneses: Consumption
Externalities, Portfolio Choice, and Asset Prices,” Journal
of Money, Credit, and Banking, February 1994, vol. 26, no.
1.
Grinblatt, Mark, Keloharju, Matti, and Ikaheimo, Seppo,
“Interpersonal Effects in Consumption: Evidence from the
Automobile Purchases of Neighbors,” Working Paper,
January 2004.
Gurdgiev, Constantin T., “Life-Cycle Model of Leisure
Demand with Habit Formation,” Working Paper, June 2001.
Hall, Robert E., “Stochastic Implications of the Life-
Cycle/Permanent Income Hypothesis: Theory and
Evidence,” Journal of Political Economy, December 1978,
86(6), pp. 971-87.
Harbaugh, Richmond, “Falling Behind the Joneses:
Relative Consumption and the Growth-Savings Paradox,”
Economics Letters, September 1996, 53, pp. 297-304.
Hayashi, Fumio, “The Permanent Income Hypothesis and
Consumption Durability: Analysis Based on Japanese
Panel Data,” Quarterly Journal of Economics, November
1985, 100(4), pp. 1083-113.
Lettau, Martin and Uhlig, Harald, “Can Habit Formation Be
Reconciled with Business Cycle Facts?”, Review of
Economic Dynamics, 2000, vol. 3, pp. 79-99.
Ljungqvist, Lars and Uhlig, Harald, “Tax Policy and
Aggregate Demand Management Under Catching Up with
the Joneses,” American Economic Review, June 2000, vol.
90, no. 3, pp. 356-366.
Mehra, Rajnish and Prescott, Edward C., “The Equity
Premium: A Puzzle,” Journal of Monetary Economics,
March 1985, vol. 15, pp. 145-161.
Modigliani, Franco, “Life Cycle, Individual Thrift, and the
Wealth of Nations,” American Economic Review, June
1986, 3(76), pp. 297-313.
Muellbauer, John, “Habits, Rationality, and Myopia in the
Life Cycle Consumption Function,” Annales d’Economie et
de Statistique, January-March 1988, (9), pp. 47-70.
Rhee, Wooheon, “Habit Formation and Precautionary
Savings: Evidence from Korean Panel Data Studies,”
Journal of Economic Development, v. 29, n. 2, 2004
Seckin, Aylin, “Consumption-Leisure Choice with Habit
Formation,” Economics Letters, 2001, 70, pp. 115-120.
Skinner, Jonathan, “A Superior Measure of Consumption
from the Panel Study of Income Dynamics,” Economics
Letters, 1987, 23(2), pp. 213-16.
Toche, Patrick, “Savings and Growth with Habit Formation:
A Comment,” Working Paper, May 2004.
Wendner, Ronald, “Capital Accumulation and Habit
Formation,” Economics Bulletin, vol. 4, no. 7, pp. 1-10,
2002.
Proceedings of the Pennsylvania Economic Association 48
PEER MENTOR DEVELOPMENT AS SECONDARY LEADERS AT THE UNIVERSITY LEVEL
Dana D’Angelo, Clinical Professor
Susan Epstein, Associate Clinical Professor
LeBow College of Business (Drexel University)
ABSTRACT
There has much discussion and use of the term "peer
mentor". Another way to consider what a peer mentor
represents is to recognize that peer mentors are, in fact,
"secondary leaders". The Peer Mentoring Program in the
LeBow College of Business at Drexel University was
developed over 15 years ago to provide leadership growth
opportunities for its undergraduate, upper-class students, as
well as mentoring benefits to new students, and support
to their partnered faculty. The program has seen tremendous
success over time and has been modeled by other programs
within the university, and at other
institutions. Undergraduate Teaching Assistants (UTAs) are
recruited based on highly selective criteria. Thus, on an
annual basis, up to 40 high achieving students participate in
the credit earning program. The peer mentoring program
combines both theory and practice, which supports the
university's experiential learning foundation. It is primarily
based upon academic research including that of Daniel
Goleman and Steven Covey. The focus on Emotional
Intelligence and Leadership Styles and Roles provides a basis
for the tasks, behaviors, and assignments that lead to personal
growth for the UTA and the mentoring for new students, As a
part of the evaluation process, a questionnaire is used to
compare perceptions among the peer mentor, the new
students, and faculty, to assess the effectiveness of the
application of specific leadership roles, skills, and
behaviors. The goal of an analysis of the results is to
recognize that peer mentoring is fundamentally secondary
leadership development.
PEER MENTORING IN ACADEMIA
Peer mentoring is a widely recognized approach for a
positive way to develop less experienced performers in
business and industry. Higher level academia has accepted
this mentoring relationship as one that can also create a
motivational environment and thus promote success among
new students. Traditionally, the benefits of the use of peer
mentors as peer leaders in the classroom has focused on the
outcomes achieved by the mentee, the more inexperienced or
new student. Peer mentoring provides these students the
opportunity to receive guidance, support and instruction from
someone other than the faculty member but who is also a part
of the classroom and course. The relationship pairs
collegiate upperclassmen with underclassmen, allowing the
underclassmen to have a role model from an experienced
peer who has met the challenges of the academic
environment. Peer mentors provide another foundation for
what new students need to be successful in the college and
university settings - general support and understanding,
positive role modeling and encouragement, and guidance and
instruction for a different station of life. Active peer mentors
are able to motivate and drive mentees to reach the next level
of success, while offering advice on how to get there.
Based on three mentoring projects, in the United Kingdom,
Korea and New Zealand, retention gains of up to 20% with a
return on investment of the order of magnitude of several
hundred per cent may be possible when peer mentoring is
used. (Boylea, Kwong, and Simpsond, 2010). Feedback
measures student responses to motivation, study skills, study
goals, and belongingness. The outcomes are as follows.
Motivation - Mentors helped motivate 92% of the
students to keep going with their studies
Study skills - Many students (90%) agreed that
mentors helped them with their study skills, and the
majority gave this a rating of 4–5 in terms of
importance to their learning
Study goals - The majority of students (86%)
considered that mentors helped them to achieve their
study goals and rated this help as 4–5 in importance
and 87% stated that the mentoring supporters had
helped them with strategies to manage workload
Belonging - Almost all (98%) of the students who
were surveyed stated strongly that belonging to a
learning group
Abraham Maslow identified Safety and Social needs in his
Hierarchy of Needs, aimed primarily at increasing
motivation. The sense of belongingness provided by the
relationship between peer mentors and their students clearly
contributes to creating affective commitment to the
university, thereby motivating students to academic success.
In addition, academic peer mentoring is often credited with
increased retention rates among mentees. The fostering of
mentoring relationships may assist organizations in
simultaneously promoting effective knowledge transfer and
the affective commitment that assists in the retention of key
knowledge workers. (Fleig-Palmer, 2009). In addition to the
mentee benefits, the positive outcomes to the mentor in an
academic environment are significant. The widely accepted
outcomes of increased communication skills and personal
Proceedings of the Pennsylvania Economic Association 49
leadership roles are clear. In addition, satisfaction of helping
a student reach her academic and professional goals,
recognition at work for participation in a job-related activity,
an expanded network of professional colleagues, recognition
for service to the community, increased self-esteem, self-
confidence and affirmation of professional competence are
all seen as mentee benefits. (Brainard and Ailes-Sengers,
1994).
THE PEER MENTORING PROGRAM IN LEBOW
COLLEGE OF BUSINESS AT DREXEL UNIVERSITY
Drexel University’s LeBow College of Business developed
its undergraduate peer mentor program over 15 years ago,
building on the university’s foundation of experiential
learning and leadership. The program provides the
mentoring students with a sequencing of growth over the
course of two mentoring relationship opportunities, focused
on creating and developing leaders. In LeBow’s peer
mentor program, second year/sophomore students are given
peer mentoring roles in an introduction course for all
incoming freshmen business students, and upper-classmen /
junior-senior student, having previously developed basic peer
mentoring skills in that course, are given the role of
Undergraduate Teaching Assistant (UTA) in the Foundations
of Business I and II classes, also taken by all incoming
freshmen business students. Applicants for the peer mentor
program must first be recommended by a faculty member
from LeBow College of Business. Grades, interviews and
peer recommendations are considered in the evaluation
process. Selected students are then “trained” in a classroom
setting, taking a course based primarily on the understanding
and development of various leadership styles. Beyond this
coursework, peer mentors are guided by their faculty partners
in developing self-confidence and assertiveness in the
classroom.
In his Leadership Pyramid, Steven Covey showed the
progression from Modeling to Mentoring to Teaching. He
described the role of modeling to be building trust and
credibility, the role of mentoring to be developing
relationships, and the role of teaching to be the pinnacle of
the leadership pyramid; it requires a foundation in modeling
and mentoring. (Covey, 1992). Emotional Intelligence has
also been identified as a key component of any leadership
development; and empathy is considered a platform for
human understanding, communication, and relationships as
well as being the foundation for effective leadership, both
personally and professionally. (Goleman, 2005). Based on
Goleman’s Five Components of Emotional Intelligence, peer
mentor learning outcomes within LeBow are designed to
include empathic listening, self-awareness, self-regulation,
motivation and social skills. (Goleman and McKee, 2008).
LeBow peer mentors are challenged with strengthening and
using leadership skills even beyond the basic expectations
normally associated with peer mentoring at the collegiate
level. The Big Five personality traits are explored, helping
the peer mentors identify their comfort zones, and the
students are challenged to demonstrate flexibility in these
areas and application with their mentee groups. At the
conclusion of the second peer mentoring relationship in the
six month UTA course, a 360 degree evaluation is used. The
questionnaire is given to and collected from the peer mentor,
the faculty partner, and the mentee student.
PEER MENTORS AS SECONDARY LEADERS
The sequencing of the mentoring programs at LeBow leads to
the development of secondary leaders, who are trained
beyond the relationship role to become leaders whose goal is
to promote independence among their mentees. A second
chair leader is a person in a subordinate role whose influence
with others adds value throughout the organization. (Bonem
and Patterson, 2005). A second chair leader fundamentally
works under a primary leader as a subordinate; in the case of
our peer mentors, that individual is a faculty member. But
together, these two leaders make up a team, one that
significantly impacts the overall program and the freshmen
students in it. The resulting value added by the second chair
makes the organization much better than it otherwise would
be. The relationships that exist between the primary and
secondary leader will vary, and as such outlooks of goals and
ideas may as well. Additionally, the perceptions of the led
group are also an important part of the dynamic and results
for both leaders.
As part of the evaluation process of the peer mentor
in LeBow’s Foundations of Business I and II course
sequence, a questionnaire was given to the students in the
peer mentor’s class, as well as to the peer mentor him or
herself, and to the supervising faculty member. The results
were primarily used to provide feedback for discussion with
the peer mentor regarding development and future goals, and
well as for overall performance evaluation purposes. Two
faculty members then observed that additional insight may be
provided by comparing results among the participants of the
questionnaire, in particular the primary leader (faculty), the
Proceedings of the Pennsylvania Economic Association 50
secondary leader (peer mentor) and the constituents
(freshman students). The questionnaire (Table 2) was
designed to elicit feedback based on the four roles of a leader
identified by Steven Covey: Modeling, Aligning, Pathfinding
and Empowering. The roles are an extension of principle-
centered leadership, where leaders are viewed as individuals
who do the right thing. The concept was also in direct
strategic placement with the objective of ethical awareness
and development of LeBow.
Four three-item statements were used to gain feedback on the
use of and effectiveness in these major roles of a leader, and
each was rated on a scale of 1 to 5 (1-Rarely, 2-Occasionally,
3-Sometimes, 4-Quite Often, 5-Almost Always). Table 1
shows the averages for the four roles for each of the
participants.
Table 1
Model Align Pathfind Empower
Student
Mentees
4.39 4.21 4.37 4.44
UTA/Peer
Mentor
4.14 3.92 4.25 4.53
Faculty 3.93 3.67 3.97 4.07
For three of the four roles, the mentees perceived stronger
behavioral roles in the peer mentors than either the peer
mentors did themselves or the faculty members did. The
student mentees also rated all four roles well above a 4.0,
indicating they experienced the behavior from the peer
mentor quite often to almost always. In all four roles, the
peer mentor perceived their behavior stronger than the
faculty member did. All three participant groups rated
aligning as the least practiced role and empowering as the
most practiced. In all roles and from all participants, the
averages were well above 3.0, suggested that the behaviors
were indeed perceived on a frequent and observable and
impactful basis.
The mentees clearly see the benefits of the inclusion of peer
mentors in their curriculum, and the peer mentors themselves
have positive outlooks on their roles and their achievements
as secondary leaders. Although the faculty perceptions are
not as strong as the other two groups, they are still positive
overall, and the information provided by the evaluation
questionnaire as a whole can assist faculty in identifying and
working on specific areas with the peer mentors for their
continued growth as secondary leaders in the classroom and
program.
The results of the questionnaire will have direct impact
on LeBow’s Peer Mentor Program going forward. Based on
the feedback, specific tasks and assignments within the peer
mentoring course can be developed to address roles and
behaviors that require more attention, including readings,
group discussions and hands-on activities. The responses to
the questionnaire also are able to give the faculty and peer
mentors collaborative information to use in setting goals for
themselves and in overall course designs for both the peer
mentors and the student mentees. Finally, the evaluation
process can be enhanced to draw more connections to these
goals, activities and results within the program.
Peer mentoring has proven itself to be a valuable aspect in
academia. Most of the benefits traditionally focus on the
mentees’ growth. In looking at the peer mentor as also a
secondary leader, additional benefits can be observed for and
by the mentees. More importantly, however, the
consideration of the peer mentor as the secondary leader as
well allows attention and focus to also be on the benefits to
the peer mentors themselves. Whatever the goals of the
sponsoring program or of the leadership development may
be, whether in roles skills, or behaviors, peer mentors are
indeed secondary leaders, and should be both acknowledged
and developed as such.
Table 2
The UTA
1. Set a personal example for expectations of
students in the class
2. Challenged students to innovate and think
creativity
3. Initiated ideas and activities that supported class
goals
4. Was enthusiastic and upbeat about the class and
LeBow
Proceedings of the Pennsylvania Economic Association 51
5. Conveyed a positive message about future
opportunities as a business student
6. Followed through on commitments regarding
the class, its students and instructors
7. Communicated open-mindedness and
encouraged success
8. Displayed effort toward ensuring students met
guidelines for the class
9. Created an interest and understanding about the
role of business within varying aspects in the
world
10. Connected learning in the class to outside
happenings
11. Showed respect and support for students and
faculty
12. Provided feedback and guidance to students
ACKNOWLEDGEMENT
The authors would like to thank Jordan Kenney,
Undergraduate Student, LeBow College of Business, for his
work in the collection and analysis of the questionnaire data.
REFERENCES
Allen, T. D., Russell, J. E., & Maetzke, S. B. (1997). Factors
related to protégés‘ satisfaction and willingness to mentor
others. Group & Organization Studies, 22(4), 488-507.
Black Issues. Mentoring: The Forgotten Retention Tool
(2002). Diverse Issues in Higher Education
Bonem, M. & Patterson, R. (2005). Leading from the Second
Chair. San Francisco: Jossey-Bass.
Boylea, F, Kwong, J, Rossc, C.,& Simpsond, O. (2010).
Student-Student Mentoring for Retention and Engagement.
Open Learning, 25(2) 115-130.
Brainard, S.G. &Ailes-Sengers, L.A. (1994). Mentoring
female engineering students: A model program at the
University of Washington. Journal of Women and Minorities
in Science and Engineering, 1, (2), 123-35.
Covey, S. R. (1992) Principle-Centered Leadership. New
York: FIRESIDE
Fleig-Palmer, M.M. (2009). The impact of mentoring on
retention through knowledge transfer, affective commitment,
and trust. ETD Collection, University of Nebraska –
Lincoln. Research Commons website:
http://digitalcommons.unl.edu/dissertations/AA13366037
Goleman, D. (2005). Emotional Intelligence: 10th
Anniversary Edition; Why It Can Matter More Than IQ .
New York: Bantam Dell
Goleman, D, Boyatziz, R., McKee, A (2008) What makes a
Leader, Emotionally Intelligent Leadership, Harvard
Business Review, HBR Article Collection
Huizing, R. L. (2010) Mentoring together: A literature
review of group mentoring. Paper presented at the
Northeastern Association of Business, Economics, and
Technology Proceedings, State College, PA. Retrieved from
http://www.nabet.us/Archives/2010/NABET%20Proceedings
%202010.pdf
Proceedings of the Pennsylvania Economic Association 52
CLIMATE NEUTRALITY IN THE HIGHER EDUCATION SECTOR: MAKING THE COMMITMENT
By Soma Ghosh
Discussion by Solomon T. Tesfu, Ph.D
Mount St. Mary’s University
SUMMARY
The paper analyzes the factors that influence the
decision to sign the college presidents’ climate
commitment (PCC) by colleges using descriptive
statistics and a probit model.
Analysis is based on data from a sample of 669 PCC
members for descriptive analysis and 303
institutions for probit model estimation.
The results show that PCC signatories are primarily
public, four-year institutions in urban campuses
located in the northeast region.
The results also indicate that public and land grant
institutions are more likely to join the PCC while
research-intensive institutions are less likely to join.
The research question addressed in the paper is
timely and well motivated.
Comments
It appears that the sample of 303 institutions used
for estimation of probit model for adoption consists
of voluntary responders to a survey. Since a
voluntary subsample may not be representative, the
conclusions about all the institutions based on
sample results in the paper may not be valid. For
example, it is stated that “…while public colleges
are more heavily represented in the network, the
results from the probit model do not indicate that
public institutions are more likely to join the PCC.”
(p.14). This is most likely because the sample used
for probit estimation is not representative of all the
relevant institutions. Therefore, It will be more
informative to include and compare the summary
statistics for the entire group and the sub-sample of
the PCC signatories used in probit analysis to see if
the subsample is representative of the entire group
(for example by public-private and urban-rural
divide of the colleges in the sample vs. all the
colleges).
The time of the decision to sign the PCC by some
colleges and the time period for which the data were
collected (2005-2006) for some of the covariates are
different. But it is possible for some of the
characteristics of the institutions to have been
different at the time they signed the PCC and the
correlation between the dependent variable and the
covariates observed at different points in time may
not be fully meaningful.
The author also mentions that potential endogeneity
might be avoided by backdating the data for the
college characteristics. It might be good to explain
how this is the case since it is not apparent. In fact I
believe, quite a few of the covariates like adoption
of Talloires Declaration (TD) and Green Power
Partnership (GPP) are endogenous (decision to join
these could be motivated by similar unobserved
factors which motivated the decision to join PCC
but it doesn't necessarily mean that membership in
TD and GPP are causing membership in PCC.). If
the author’s interest is in just correlations (and not
causation), then endogeneity may not be a concern.
It is stated that“…the results from the empirical
model imply that land grant and research/doctoral
schools are more likely to join the initiative.” (p.14).
But the coefficient for the dummy indicating
research-oriented institution is negative and highly
significant. It might be informative to explain why
this could be plausible.
It might also be useful to re-scale some of the
covariates to make the coefficients neater for
presentation (e.g. endowment/1000, full-time
equivalents/100, etc) and include the number of
observations for probit model in table III.
Proceedings of the Pennsylvania Economic Association 53
QUESTIONS
The author reports that “… the PCC signatories are
primarily white campuses (60%) with a low
percentage of international students (3%), and have
a balanced representation of women (57%)” (p.11).
Are the PCC signatories more ‘white’ than the non-
signatories? Is the proportion of international
students in these campuses ‘low’ compared to the
non-signatories? Is the proportion of women in these
colleges significantly different from the proportion
in the non-member colleges? I believe the paper will
be more informative if these questions are answered
(explained).
Proceedings of the Pennsylvania Economic Association 54
THE U.S. DOLLAR AS AN INTERNATIONAL CURRENCY RESERVE AND ITS CONTINUOUS DEPRECIATION
Ioannis N. Kallianiotis
Economics/Finance Department
The Arthur J. Kania School of Management
University of Scranton
Scranton, PA 18510-4602
ABSTRACT
The current account deficit (because consumption exceeds
production in the U.S., due to movement of U.S. MNCs
abroad), which causes a capital account surplus (capital
inflows in the U.S. to finance the deficit) and the scale of
financing needed to support the U.S. debts (fiscal deficit,
national debt, and the private sector’s and households’ debts)
together with the Federal Reserve’s policy of keeping U.S.
interest rates low to ward off deflation, to stimulate the
financial markets, and to revive growth, which is impossible
without fiscal policy, has revived concerns about a sudden
and sharp depreciation of the U.S. dollar since 2003.
Americans have lost an enormous amount of their purchasing
power and wealth. As a debtor, the U.S. is benefited from the
devaluation of its currency and it seems that the “market”
knows better than us and keeps the dollar “undervalued”, but
this cannot continue for a long time.
INTRODUCTION AND STATE OF THE
ECONOMY
The U.S. dollar has shown a large volatility even before
1971, when the gold exchange standard was abandoned by
President Nixon and continued since 1973, when the
exchange rate became flexible. The value of the dollar can be
seen by looking at the different exchange rate indexes. The
Trade Weighted Exchange Index (USXRI) has during this
period a mean value, 39919.97IUSXR and a standard
deviation, 39571.14USXRI . Its maximum value was 138
(1985:M03) and its minimum value was 69 (2011:M08). In
2002:M02, the index was at a value of 111 and in 2013:04
was 76 (Graph 1). After that date, the dollar continues to
depreciate with respect the other major currencies. The
question, which rises, here, is: What are the causes of this
depreciation? What is the cost and what are the benefits for
the U.S.? Graph 2 shows many factors, which have caused
the dollar’s depreciation. The current account deficit, the
capital account surplus, the huge national debt, the inflation
in the country, the price of oil, and the tremendous
uncertainty, due to the Middle East crises (Iraq and
Afghanistan, and there are new ones coming, soon) have
created for the U.S. and its currency a serious undermining
and creeping chronic instability. Also, the low (zero) interest
rate in the U.S. (enormous liquidity) is depreciating the dollar
and caused the bubble in the financial assets, in the precious
metals, and in the housing market. This was followed by the
financial crisis (August 2007), which affected the U.S.; and if
the Euro-zone had not experienced its unique debt crisis and
its disastrous common currency, the dollar could have been
worse.
The devaluation of the dollar reduces the current account
deficit ( MXCA ), but at the same time the capital
account ( KA ) surplus is falling because international
investors are selling off dollar-denominated investments, due
to their loss of value; then, the CA continues to be equal
with the KA ( KACA ). The value of the dollar is
determined by its demand and supply, as it happened with all
the currencies after 1973. The dollar has become weaker
because its supply exceeds the demand. Foreigners are selling
their products to the U.S. and the money (dollars) that they
receive is invested back to the U.S. by buying bonds and
stocks.1 What will happen if the international investors
(Chinese, Japanese, OPEC, etc.) are unwilling to keep and
start selling their American bonds and stocks? The capital
account surplus will diminish and consequently, the current
account deficit will fall. A devaluation of the dollar at this
point will reduce the current account deficit and equalize it to
the capital account.
Thus, there is a dollar crisis in the world, due to the
enormous level of the U.S. deficits and debt: [Federal
Debt=$16.776 trillion, Social Security Liability=$20.5
trillion, Medicare and Medicaid Contingent Liabilities=$98
trillion, State and Local Governments=$5.71 trillion,
Business Sector Debt=$11.63 trillion, Financial Sector
Debt=$13.6 trillion, Total Personal Debt=$13.22 trillion,
Financial Sector Bail-out=$2.5 trillion, Other Debts=$2.74
trillion: Total Debt (Public and Private)= $184.676 trillion].
The GDP (2013:Q1) was $13.750 trillion. Then, the total
debt is 1,343.1% of the GDP.2 The Federal Reserve Bank
tries to keep the interest rate low to affect positively the
financial markets, but this policy did not help so much the
real economy because we have reached a liquidity trap. This
Fed’s policy is only pro-market and not pro-social. Thus, the
social benefits are insignificant. Also, this policy of
enormous liquidity caused the bubbles in the financial market
and in the housing market and finally, it will induce inflation,
when the unemployment will reach the natural level. The
U.S. dollar has declined from its pick point USXRI=138
(1985:M03) until now USXRI=76 (2013:M04) by more than
-45%. With respect to the euro, the dollar has declined from
Proceedings of the Pennsylvania Economic Association 55
0.8530 $/€ (2001:M06) to 1.6001 $/€ (2008:M04), which is -
87.76%. Now (5/15/2013), it is 1.2921 $/€, a loss of -51.48%
since its pick value.
The depreciation of a currency means a decline in the
purchasing power of the currency, which leads the citizens of
the country to a decline in their living standards. The U.S.
consumers had lost 87% of their purchasing power compared
to the European ones. An investment in dollars has suffered
from this decline. The price of importables is going up with
the same proportion. All commodities that we import (i.e.,
oil, coffee, BMW, French wine, etc.) are setting record
prices, due to the depreciated dollar. The OPEC countries
watch the deprecation of the dollar, which is the currency that
they receive for their sales of oil (except Iran) and they raise
its price to keep their revenue constant. The trade deficit (
0MX ) causes the dollar to depreciate, so exports can be
stimulated and slightly improve the trade account (price
elasticities for exports and imports play a role, here, too). As
a currency declines in value and interest rate parity does not
hold, investors go to other countries with stronger currencies,
i.e., euro, yen, pound, etc.
A forecast from the IMF predicted that China’s economy will
surpass that of the U.S. in five years. The U.S. Fed, to keep
the interest rate close to zero ( %25.0FFi ), is doing the
“quantitative easing” (printing money).3 Then, the value of
the U.S. dollar will continue to be at a low level and inflation
will rise, especially, as long as the employment will be
improved. At the moment, the unemployment rate is still very
high ( %6.7u ). Europeans and other nations are very
critical of this easy money (printing money) policy of the
U.S. This is negatively affecting the competitiveness of the
foreign countries because their currencies are appreciated,
when the dollar is willingly depreciated (“beggar-thy-
neighbor” policy).
It seems that the IMF is interested in a new world reserve
currency. The U.S. dollar was playing this role since World
War II and many resources, like oil, were purchased with the
dollar (petro-dollars). The U.S. was increasing its debt and
countries with surplus (Germany, China, Japan, OPEC
nations, etc.) were financing the U.S. debts. Since 1980s with
all these deregulations in the U.S. financial market, the frauds
in investment banks, the corruption, the housing bubble, and
the financial crisis of 2007-2013, the U.S. dollar continues to
depreciate. The Special Drawing Rights (SRD) and the euro
are competing with the dollar for their share as world reserve
currencies. The G-20, the IMF, the World Bank, and the
Bank of International Settlement are supporting the role of
SDR. The U.S. has lost most of its manufacturing
infrastructure (the most of the U.S. MNCs have become
foreign firms, today) and it is importing much of the natural
resources, oil, high tech, food, clothes, etc. If SDR or euro
will become world’s reserves, the demand for dollar will fall
and the dollar will depreciate further. Thus, it will be very
costly for Americans to pay for their importables; their
purchasing power and their domestic wealth will be
restrained even more with the current financial crisis and the
“fiscal cliff”. There are about 50 million Americans, who
receive Federal Assistance (Medicare, social security,
disability, and unemployment). This is a sign showing the
trend of this “economic power”. As the dollar will be
depreciated, there will be higher cost of imports, more
unemployment, lower income, confinement of wealth, more
home foreclosures, and more bankruptcies.
The U.S. economy was depending on credit to invest,
consume, and grow; but, this “advantage” turned to a big
disadvantage, an unmanageable debt. This economic
expansion, fueled by debt-based capital markets, gave a
temporary advantage of the West (capitalism) toward the
East (communism). Now, with the latest financial crisis and
the debt crises, the West seems to be worse than the East
because it went a step forward. It altered capitalism to a new
system, the globalism and its first social cost is obvious now.
What it will follow is difficult to be predicted from now. The
credit-driven expansion will be restricted in the future and the
debt-driven contraction will take over. The two large U.S.
bubbles (the “dot.com” and the “housing” one) have caused
serious problems to the “laissez-faire, laissez-passer”
economic system. This system recently caused a global
contraction and serious social welfare problems (from
bankruptcies to national destructions to suicides).
Global demand is falling as credit contracts and employment
and income are falling, too. Then, from where the expected
growth will come? The U.S., as the world’s largest debtor,
might have difficulties paying what it owes in the future,
except by rolling its debt forward and borrowing more.
Today, the debt is everyone’s problem because it has reached
an un-payable level. A default by the U.S. will have global
consequences, especially in China, Japan, and the other large
creditors of the U.S.; and of course, to the global peace. The
amount of outstanding U.S. debt (public and private) has
reached a level that can never be paid off. Historically,
inflation destroys the value of money. Debts are paid back
with inflated dollars, a process, which benefits the debtor and
injures the creditor (U.S. usually inflates its way out of debt
by printing what it owes). Depreciation of the dollar is
another option to benefit borrowers at the expense of lenders.
Taxation is a different option that has similar results as the
two previous ones, but already the U.S. middle-class is
paying very high taxes. Corporations, in the U.S. and all over
the world, do not pay taxes. This is another unfair social
policy, too.
Actually, corporations and wealthy people are paying
relatively less taxes compared to the middle class, which is
unfair and unethical and their tax evasion is very high, too.
This illegal capital flight is a large proportion of deposits in
Proceedings of the Pennsylvania Economic Association 56
offshore centers and tax havens.4 Also, GE paid no taxes;
Goldman Sachs paid $14 million in 2010. The GAO reported
in 2008 that “two out of every three United States
corporations paid no federal income taxes from 1998 through
2005”. Companies have become all too astute at paying for
loopholes, which allow them to shift profits abroad or move
their gains (on paper) to foreign low-tax/no-tax nations
(money laundering). As the data show, the change in
corporate taxes — not merely rates, but what they actually
paid — over the past half century are astounding.5 This
practice increases the national debt and the dollar is
depreciated even more.
Also, inflation, through monetary expansion, could raise
aggregate U.S. debt and prices further. With the current
enormous money supply and zero interest rate, we reached a
liquidity trap and this policy has not so far and might fail to
promote growth in the near future. Europe is in trouble; due
to lack of liquidity (the ECB’s overnight deposit rate was 1%
and became 0.75%). Lending has declined drastically in the
U.S. and even more in Europe. The problem is lack of
demand. The question is now: from where will the expected
stimulus come? Troika believes that austerity generates
growth (sic). Of course, if this U.S. monetary expansion had
been used by the economy, we could have experienced
hyperinflation in the U.S. Thus, it is safe for the economy, at
the moment, because the high risk, the low income, and the
high unemployment have made banks reluctant to lend and
people unwilling to borrow. The U.S. has experienced high
rates of inflation in the past, even though that the number one
objective of the Fed is “price stability” ( 0e ). The dual
mandate, it is not “full employment and price stability”, but
“maximum employment and price stability”. The monetary
expansion, the last three years, has far exceeded any previous
ones. Then, what follows, when the unemployment will fall,
it will be a high inflation. This liquidity and uncertainty in
the market has increased the price of gold and silver (another
big bubble in precious metals and a depreciation of the
dollar).
Furthermore, depreciation of the dollar helps partially the
U.S., but not the creditors of the country because their return
is falling from their U.S. investments (translation exposure).
China has linked its currency to the U.S. dollar, so with any
depreciation of the dollar, the yuan is also declining in value
and China becomes the major beneficiary by increasing its
exports. Thus, the peg of the Chinese yuan to the U.S. dollar
prevents the U.S. from altering its trade deficit by currency
devaluation, but it helps the U.S. relative to the other
countries (i.e., Euro-zone), where their currencies are
appreciated.
In 1980s, the U.S. needed Japanese savings to finance
Reagan’s multi-trillion dollar debt-based military buildup
(the famous “star war”).6 If Japanese rates were raised,
Japanese savings would stay at home, but Japanese rates kept
low (with the “help” from the U.S.). These low interest rates
and high earnings of Japanese firms, due to their exports
ignited a speculative frenzy in their stock markets causing the
then largest stock market bubble in history. In 1990, the
bubble collapsed and Japan fell into a deflationary trap, from
which it has never fully emerged.
Today, the economic power has shifted to China from Japan.
The U.S. imports from China are enormous7 and it is
financing its debt with Chinese capital. Thus, the U.S.
dominance is being challenged by China. China has, now,
reduced its holdings of U.S. debt, which will affect the U.S.
deficits, interest rates, and the value of the dollar. A
depreciation of the U.S. dollar by 30% would impact the debt
held by foreigners; their losses would be significant.8 Of
course, after the above economic impacts, geopolitical
considerations are taking place. Expenditures are rising in the
U.S., production is falling, and the need to borrow is
increasing. Then, the total U.S. debt (public and private)
would prove difficult of even being repaid. The future of all
nations might be difficult with all these absurd decisions
(“irrational exuberance”) of the last 30 years. In 1973, when
gold was taken out as a monetary asset, balance sheets of
every central bank of the world suffered, as their dollar-
denominated assets sank in value, in terms of dollar.
Unfortunately, lately, spreads on sovereign debt are rising
and credit default swaps (CDS) reflect the higher premia
being charged to protect against default. Investors compare
risk ( D ) to reward [ )( DRE ] and try to maximize the
reward to variability (RV) ratio of their bonds’ investments
with respect to debt; when the reward is believed to
compensate for the risk (max RV), the bond is bought and the
bet is placed. Some countries are paying 34% interest rate,
due to high austerity measures that have increased the
probability of bankruptcy, as it happened, lately, with Greece
and other Euro-zone countries (PIIGS nations). We hope that
this will not take place (will be prevented) in the heavily
indebted U.S.A.
The most recent literature review on this area is as follows.
Uri and Boyd (1991) found that a devaluation of the U.S.
dollar have caused an increase in output of the agricultural
sector and all producing sectors, except the financial one.
Glain (2003) says that a cheap dollar makes U.S. exports
more competitive, but discourages investors from holding
dollar-denominated assets. Siegman (2008) says that the huge
U.S. deficits are unsustainable and will lead to disruptions in
international financial markets, to a global financial crisis,
and to worldwide recession. Kallianiotis and Bianchi (2009)
show speculation and monetary policy (risk and rate of
return) play a major role in the determination of the exchange
rate between the dollar and euro. Bonitsis (2011) shows that
the introduction of the euro changed the competitiveness in
Proceedings of the Pennsylvania Economic Association 57
Euro-zone nations and made Germany the most competitive
economy. Amstad and Martin (2011) say that after the
financial crisis the Fed increased its liquidity drastically, but
not the ECB, which has affected the value of the currencies.
Nechio (2011) states explicitly that the common target rate
(overnight deposit rate) of the ECB helps only Germany and
does not help at all the peripheral countries of the Euro-zone.
Sharma (2011) argues that the U.S. exploding debt and
deficits have raised concerns about the future of the dollar
and the competitive devaluation is a growing problem
globally. Krugman (2012) recommends fiscal policy for the
U.S. and not only zero interest rate monetary policy by the
Fed.
A THEORITICAL MODEL OF DOLLAR’S
VALUATION
The correlation and the causality tests between the exchange
rate and some macro-variables (Graph 2) reveal that the
exchange rate [e ($/€)] is correlated and affected by the
following variables:
),,,,
,,,,,,,,,(
, EDCDWDOPDiii
DJIACPINDISEYKACAfe
tGBTSFF
tttttttttt
ttt
(1)
where, e=exchange rate, CA=current account, KA=capital
account, Y=income, E=expenditures ( GICE ),
S=saving, I=investment, ND=national debt, CPI=consumer
price index, DJIA=Dow Jones Industrial Average, iFF=federal
funds rate, iS-T=short-term interest rate, iGB=government
bonds rate, OPD=oil price domestic, WD=Iraqi war dummy
(0 before March 2003 and 1 after March 2003), and
EDCD=European debt crisis dummy (0 before September
2009 and 1 after September 2009).
The total national income of the economy (Y) is used for
paying taxes (T), for consumption (C), and for saving (S):
tttt SCTY (2)
The national production (Y) can be used for consumption
(C), for investment (I), for government spending (G), for
exports (X), and a proportion from this aggregate demand
(AD) is imported (M):
tttttt MXGICY (3)
The trade account (TA) of the country can be presented with
the following equation:
tttt
ttt YY
P
PeTA *
32
*
10 )( (4)
where, TA=trade account, P=U.S. price level, P*=foreign
price level, Y=domestic income, and Y*=foreign income.
The capital account can be written as:
ttttttt sfiiCAKA )()( 2*
10 (5)
where, KA=capital account, i=domestic interest rate,
i*=foreign interest rate, f=ln of the forward rate, and s=ln of
the spot exchange rate.
Through substitution of eqs. (2), (3) ,(4), and (5), we receive:
ttt
ttttt
YY
KANDPPe
*
1
3
1
2
11
*
1
0
lnln
ln1
ln1
lnlnln
(6)
where, tKA can be substituted with the tCA , with (*tt ii ),
and with ( tt sf ), too.
Further, we can test the effects of the price of oil (toilP ),
national debt ( tND ), current account ( tCA ), war dummy (
WD ), and European debt crisis dummy ( EDCD ) on the
exchange rate.
t
ttoilt
EDCD
WDCANDPet
5
43210 lnlnlnln
(7)
Thus, eqs. (1), (6), and (7) can be used to determine the
factors that affect the exchange rate (the value of the dollar).
An increase of the direct quoted ($/€) exchange rate ( te )
means a depreciation of the U.S. dollar.
EMPIRICAL RESULTS
It is important to test the above theory by applying data from
the two economies, U.S.A. and Euro-zone. The data, taken
from economagic.com, imfstatistics.org, Eurostat, and
Bloomberg.com are monthly from 1992:01 to 2011:12. They
comprise, spot exchange rate (e), consumer price indexes in
U.S. and EMU (CPI and CPI*), federal funds rate (iFF), 3-
month T-bill rate (iRF), ECB overnight rate (iOND), 3-month
deposit rate (LIBOR) (i3mdl), nominal (Y and Y*) and real
GDP (Q), private consumption (C), private investment (I),
exports (X), imports (M), current account (CA), capital
Proceedings of the Pennsylvania Economic Association 58
account (KA), taxes (T), government expenditures (G),
national debt (ND), personal saving rate (psr), Dow Jones
Industrial Average (DJIA), oil prices (OPD), price of gold
(PGold) for the U.S. and the Euro-zone economies.
First, the correlation coefficients and a Granger causality test
between all these variables are presented in Tables 1 and 2.
There is a very high positive correlation between the
exchange rate (dollar’s devaluation) and the exports, imports,
current account deficit (capital account surplus), U.S. and
foreign income, investment, national debt, U.S. and foreign
prices, taxes, price of oil, consumption, and the war dummy;
also, a negative correlation (dollar’s appreciation) with
respect the domestic and foreign interest rates. The domestic
and foreign income, consumption, and foreign prices cause
the depreciation of the U.S. dollar. The U.S. national debt
causes the depreciation of the dollar, measuring with the
exchange rate index (USXRI), too.
Then, Table 3 shows the estimates of exchange rate by using
eq. (1). Domestic income, consumption, price of oil, the war
dummy, and personal saving rate have a significant positive
effect on the exchange rate (dollar is depreciated). The price
level and the government bond rate have a significant
negative effect on the exchange rate (dollar is appreciated).
Table 4 represents eq. (6). The U.S. national debt, the interest
rate differential, domestic income, and foreign income have a
significant positive effect on the exchange rate (dollar is
depreciated). The foreign price level has a significant
negative effect on the exchange rate (dollar is appreciated).
The foreign price level has a significant effect on the
exchange rate (dollar is depreciated). Table 5 presents the
price and income elasticities (domestic and foreign) of the
current account and the effects of interest rates and forward
discount of the dollar on the capital account.
Now, we do different tests for eq. (7) and the first results of
them are in Table 6, which presents the Augmented Dickey-
Fuller and Phillips-Perron unit root tests. These results show
that only the tnd contains no unit root [ )0(I ]. The other
series are nonstationary [ )1(I ]. Table 7 gives a cointegration
test of the series of eq. (7). It reveals that there is one (1)
cointegrating equation at the 5% level and one (1)
cointegrating equation at the 10% level. Then, the linear
combination of these non-stationary series is stationary; they
are cointegrated. There is a long-run equilibrium relationship
among the variables of eq. (7).
Table 8 displays the regression results of eq. (7). The price of
oil, the national debt, and the war in Iraq are depreciating the
U.S. dollar (spot rate increases). The European debt crisis
appreciates the dollar and depreciates the euro. Table 9
exhibits the correlogram and Q-statistics for testing high-
order serial correlation of the residuals of eq. (7). The
correlogram has spikes at lags up to six and the Q-statistics
are significant at all lags, indicating significant serial
correlation in the residuals. Lastly, Table 10 demonstrates the
serial correlation LM test (Breusch-Godfrey) and rejects the
hypothesis of no serial correlation up to order four. Thus, the
residuals are serially correlated and eq. (7) should be re-
specified before using it for hypothesis tests and forecasting.
In addition, Graph 1 presents the value of the dollar by using
an exchange rate index from BIS. Graph 2 shows the
causality and the two-way causation between the most
important variables in the U.S. economy (
CPIandOPDNDiiiiSKACA TLTSGBFF ,,,,,,,,, ). As
Graph 2 (2 lags) reveals that there is a significant causal
relationship among the economic variables (the arrows show
the direction of causality). The spot exchange rate is caused
by national debt, current account deficit, capital account
surplus, and consumer price index.9 Graph 3 shows the (€/$)
spot exchange rate. The dollar is losing value since 2002 with
the Iraqi invasion of the U.S. (it seems that the Muslim
countries are now investing in Europe).10
Graph 4 reveals the
depreciation of the dollar with respect the gold. In early
1970s, the dollar had a catastrophic downfall. Finally, Table
11 presents the global currency reserves. Since the
introduction of the euro in 1999, the U.S. dollar is losing as a
global currency reserve (from a 71% share, it fell to 61.9%).
CONCLUDING REMARKS
The objective of this analysis is to determine the factors that
have caused the depreciation of the U.S. dollar the last ten
years, and its effects (positive and negative) on trade, wealth,
and social welfare. The factors that have been determined,
here, are the U.S. and European income, the U.S.
consumption, the EMU prices, the U.S. national debt, the
current account, the personal saving rate, the price of oil, the
Middle East turmoil (war dummy), which has increased the
demand for euro, and speculation about the two economies
(U.S. and Euro-zone).11
The exchange rate dynamics is based
on shocks on the economy and on current account, due to oil
prices, debts, and risk, between the U.S. dollar and the euro.
Lately, the U.S. dollar has been losing value with respect the
euro and other major currencies of the world. The last two
years, the dollar is appreciated with respect the euro, due to
the Euro-zone debt crisis (European debt crisis dummy). We
want to see if this depreciation depends on economic shocks
and economic fundamentals or it is just speculation from
individuals and countries, which hold large amounts of
foreign assets denominated in different currencies or due to
the current global financial crisis, recessions, instability, and
the risk that the U.S. might freeze the foreign funds invested
in its assts.
The conclusion from this analysis can be that international
investors are investing in countries with higher return or
lower risk, and safety, depending on their utility function and
Proceedings of the Pennsylvania Economic Association 59
the modern systemic risk. This increase in demand for these
assets increases the demand for currency in that country and
its currency is appreciated; the oil prices, the high risk and
the enormous debts are affecting the currency, too. Before
2001, people were invested in the U.S. and Japan, so the U.S.
dollar and the Japanese yen were appreciated. After 2001,
they invested in the Euro-zone and the U.K. and the dollar
and yen lost their value. Of course, due to, high risk and wars
(in Iraq and Afghanistan) and the creeping ones (in Syria and
Iran), political conflicts, a unique financial crisis, and low
returns, many speculators have invested in euros and other
currencies, instead in dollars denominated assets. After the
last months of 2008, we see a change in this trend because of
the Euro-zone debt problems. The current account is affected
by risk and high debts, too. Historically, the American
governments have frozen the foreign assets inside the U.S.,
when a conflict arises.
Finally, the decline of the U.S. dollar occurs because the
country has failed to correct the macroeconomic distortions
in its economy (production, consumption, self-sufficiency,
international relationships, debts and deficits, low return,
high risk and unemployment, etc.) The response of the two
policy rates (*ONDFF iandi ) on the exchange rate is a
negative one for three months and then, it stays constant,
which means [ euroandSMi s ($)( )] a lot for
3 months (overshoots)12
and then, it is stabilized at a lower
level. Testing the effectiveness of monetary policy on the
exchange rate, we found it non-effective. [
euroSiandSi ONDFF ()($ *)].
13
Taking into consideration the effect of the freezing funds risk
premium (FFRP)14
on the exchange rate, we found that: [
)($ euroandSFFRP ],15
which is reasonable for
our state of the economy, due to the Middle-East crises, the
historic memory with Japan in the past, and the continuation
of the cold war. After all this analysis, we can say that the
dollar could appreciate with respect to euro, if we will have
any other domestic (like, increase of the federal funds rate or
ECB rate, due to fear of inflation) or external shocks on the
two economies (oil prices, Euro-zone member default, new
wars, etc.). Still the forecasting of the exchange rate remains
as a problem and the depreciation of the dollar will continue
if the U.S. socio-politico-economic system stays the same.
1 On May 15, 2013, the U.S. national debt (ND) was $16.776
trillion and foreigners were holding $5.758 trillion, which is 34.32% of the total ND. See, http://www.treasury.gov/resource-center/data-chart-center/tic/Documents/mfh.txt
2 Thus, for the beginning of 2012, the Social Distress Index
(SDI) for the U.S. was:
%91.361,1%03.350,1%58.3%3.8 duSDI
and for the end of March 2013, the SDI was:
%82.353,1%1.343,1%12.3%6.7 duSDI ,
which show that the country is improving a little, but it is still extremely distressful (risky). See, Kallianiotis (2011, p. 344) for this index. The U.S. needs 14 years to pay off its debt, if all the other spending would be zero. Then, it is impossible! 3 The monetary base (MB) from $846 billion in 2007 reached
$3,034.009 billion in 5/1/2013. See, http://research.stlouisfed.org/fred2/series/BASE/ . Also, the Fed is buying $85 billion of securities (government bonds and mortgage back securities) per month. This liquidity keeps the interest rate at zero level and hurts small savers, who face a nominal deposit rate closed to zero and a real one negative for four years. This policy is a redistribution of wealth from savers to banks and the financial market. See, http://www.usatoday.com/story/money/business/2013/05/01/fed-maintains-stimulus/2126381/ 4 See,
http://www.boston.com/business/globe/articles/2004/04/11/most_us_firms_paid_no_income_taxes_in_90s/. But, it was completely unethical for Troika to go against one offshore center, the Cyprus one in March 2013 and not against the big ones. 5 (1) Corporate Taxes as a Percentage of Federal Revenue
were; in 1955: 27.3% and in 2010: 8.9%. (2) Corporate Taxes as a Percentage of GDP; in 1955: 4.3% and in 2010: 1.3%. (3) Individual Income/Payrolls as a Percentage of Federal Revenue; in 1955: 58.0% and in 2010: 81.5%.
See,
http://www.ritholtz.com/blog/2011/04/corporate-tax-rates-then-and-now/. 6 The total military cost of wars since World War I until 2010
for the U.S. was, in constant 2011 dollars: $6,724 billion. The total Iraq-Afghanistan (2001-2010) war cost was $1,147 billion. See, Stephen Daggett, “Costs of Major U.S. Wars”, Congressional Research Service, June 29, 2010. 7 Trade deficit with China: in 2010, it was $273.1 billion; in
2011, it was $295.5 billion; and in 2012, it was $315.054 billion; in 1985 it was only $6 million. See, http://www.census.gov/foreign-trade/balance/c5700.html 8 China currently owns about $1,151.9 billion in U.S. dollar
denominated securities; and a depreciation of the dollar by 30%, it will cost China’s $345.57 billion in losses of its investment. 9 Graph 2, Tables 1, 2, 5, 6, 7, 9, 10, and 11 are omitted, due
to space limitations. All these and the F-statistics of this Granger causality tests and the correlation coefficients between the two variables are available from the author upon request.
Proceedings of the Pennsylvania Economic Association 60
10
See, Kallianiotis and Petsas (2008). 11
Economy of Euro-zone (2011)
(1) GDP = $12,460 billion
(2) Growth of GDP = -0.3%
(3) Inflation rate = 2.7%
(4) Unemployment rate = 10%
(5) Gross External Debt = (N/A)
(6) Public Debt = 86% of the GDP
(7) Budget Deficit = 4.1% of the GDP
(8) GDP as a % of the U.S.A. = 92.83%
Economy of U.S.A. (2011)
(1) GDP = $13,423 billion
(2) Growth of GDP = 1.7%
(3) Inflation rate = 2.98%
(4) Unemployment Rate = 8.5%
(5) Gross External Debt = $8,400 billion
(6) National Debt = $15,251 billion; 113.62% of the
GDP
(7) Budget Deficit = $1,300 billion; 9.68% of the
(8) GDP as a % of Euro-zone = 107.73%
(9) Total Debt (Public and Private)= $115.7
trillion+$40.8 trillion=$156.5 trillion; 1,165.91% of
the GDP
Then, U.S.A. is not doing better than the Euro-zone.
Why so much noise for the Euro-zone? There is no economic
explanation!..
12
See, Dornbusch (1976) and Kallianiotis and Bianchi (2009).
13
The regression is:
467.226,424.0,882.0
)023.0(
947.0
)025.0()016.0()009.0()051.0(
503.1001.0024.0218.0
2
2***
1**********
FSSRR
iis
t
tondFFt tt
14 The freezing funds risk premium must be:
tEUUS FFRPfdiittt $
* , as Kallianiotis and Bianchi (2009)
mention for the U.S. to attract capital from the Muslim countries. 15
These results are as following:
425.647,3,019.0,995.0
)089.0()090.0()018.0(
356.0342.1990.0
)001.0()002.0()002.0()506.0(
001.0007.0005.0357.0
2
2***
1***
1***
$*********
FSSRR
s
fdFFRPis
ttt
tONDt tt
Note: ts = ln of spot exchange rate, *
tONDi = ECB overnight
deposit rate, tFFRP = freezing funds risk premium, t
fd$ =
forward discount of the U.S. dollar, and t = the error term.
Proceedings of the Pennsylvania Economic Association 61
Graph 1
The Depreciation of the U.S. Dollar with Respect the Major Currencies
Note: USXRI = Trade Weighted Exchange Index: Major Currencies: Index March 1973 = 100.
Source: Economagic.com.
Table 3: Factors Affecting the U.S. Dollar (Exchange Rate Determination) [Eq. (1)]
----------------------------------------------------------------------------------------------------------------------------- -----------------------------
Variables teln teln teln teln teln teln
----------------------------------------------------------------------------------------------------------------------------- -----------------------------
C -14.166**
-0.710 -9.930***
-0.824* -9.659
*** -11.965
(6.752) (0.472) (1.201) (0.463) (1.203) (1.981)
tKAln -0.185***
- -0.014 -0.016 -0.033 -0.022
(0.060) (0.067) (0.021) (0.041) (0.042)
tYln 2.497**
- 3.227***
0.164 0.995**
0.867**
(1.197) (1.009) (0.334) (0.404) (0.369)
tCln -2.106***
- -1.707**
0.549* -0.159 0.312
(0.758) (0.861) (0.283) (0.390) (0.455)
tpsr 0.005 - -0.004 0.004* 0.001 0.003
(0.006) (0.008) (0.002) (0.002) (0.002)
tIln 0.277 - 0.021 -0.022 0.170 0.059
(0.216) (0.205) (0.066) (0.163) (0.183)
tNDln -0.025 0.040 0.024 -0.058 -0.034 -0.239
(0.149) (0.054) (0.144) (0.046) (0.142) (0.168)
tPln 1.443 - -0.964 -0.934***
0.209 0.557
(0.901) (0.968) (0.309) (0.743) (0.653)
tDJIAln 0.081 - - - - -
(0.079)
tFFi -0.008 - - - - -
60
70
80
90
100
110
120
130
140
50 55 60 65 70 75 80 85 90 95 00 05 10
USXRI
Proceedings of the Pennsylvania Economic Association 62
(0.031)
tRFi 0.003 - - - - -
(0.035)
tGBi -0.027 -0.024**
- - - -
(0.019) (0.010)
toilPln 0.138***
0.145***
(0.038) (0.020)
WD 0.227***
0.092***
(0.032) (0.022)
1ln te - - - 0.938***
- -
(0.027)
)1(AR - - - - - 0.902***
(0.030)
)1(MA - 1.187***
- - 1.337***
0.206**
(0.071) (0.077) (0.092)
)2(MA - 0.972***
- - 1.289***
-
(0.089) (0.113) 2R 0.866 0.969 0.773 0.977 0.971 0.978
SSR 0.598 0.140 1.002 0.101 0.126 0.098
F 70.480 658.558 70.465 763.622 435.230 692.873
WD 0.284 1.667 0.167 1.783 1.931 1.964
N 156 157 153 152 153 152
----------------------------------------------------------------------------------------------------------------------------- -----------------------------
Note: e = spot exchange rate, X = exports, M = imports, Y = U.S. nominal income, Y*= EMU nominal income, S = personal
saving rate, I = private investment, ND = national debt, CPI = consumer price indexes in U.S. CPI* = EMU consumer price
index, T = taxes, C = private consumption, CA = current account, KA = capital account, i = 3-month T-bill rate, i* = ECB
overnight rate, Poil = price of oil, WD = war dummy, and USXRI = U.S. exchange rate index.
Source: economagic.com, imfstatistics.org, Eurostat, and Bloomberg.com.
Proceedings of the Pennsylvania Economic Association 63
Table 4: Factors Affecting the Devaluation of the U.S. Dollar [Eq. (6)]
----------------------------------------------------------------------------------------------------------------------------- -----------------------------
Variables teln teln teln teln teln teln
----------------------------------------------------------------------------------------------------------------------------- -----------------------------
C -9.372***
-9.780***
-12.350***
-10.287***
-11.706***
-11.006***
(0.634) (0.778) (2.014) (0.647) (2.062) (2.135) *ln tP -1.444
** -1.284
* -0.067 -0.317 -0.263 -0.346
(0.611) (0.720) (0.740) (0.819) (0.755) (0.744)
tPln 0.848 0.423 1.034 0.834 1.109* 0.971
(0.560) (0.671) (0.664) (0.524) (0.663) (0.696)
tNDln 0.180***
0.121 -0.268* 0.095 -0.221 -0.185
(0.061) (0.076) (0.155) (0.062) (0.153) (0.161)
tKAln -0.002 -0.023 -0.029 *tt ii 0.008
** -0.009 -0.012
(0.024) (0.029) (0.038) (0.004) (0.008) (0.008)
tYln -0.364 0.076 0.827***
-0.723***
0.777**
0.905***
(0.266) (0.312) (0.311) (0.260) (0.302) (0.292) *ln tY 0.928
*** 0.818
*** 0.147
** 0.923
*** 0.130
* 0.054
(0.043) (0.053) (0.072) (0.042) (0.070) (0.068)
)1(AR - - 0.910***
- 0.916***
0.907***
(0.030) (0.030) (0.034)
)1(MA - 0.481***
- - - 0.194**
(0.076) (0.090) 2R 0.952 0.963 0.977 0.954 0.977 0.978
SSR 0.210 0.162 0.101 0.204 0.027 0.097
F 475.498 530.866 856.406 489.853 860.797 771.157
WD 1.042 1.718 1.741 1.089 1.696 1.966
N 150 150 149 150 149 149
----------------------------------------------------------------------------------------------------------------------------- -----------------------------
Note: See, Table 3.
Source: See, Table 3.
Proceedings of the Pennsylvania Economic Association 64
Graph 3: The Depreciation of the U.S. Dollar with Respect the Euro
Note: 1/EUS= the exchange rate between dollar and euro (€/$). Source: Economagic.com.
Graph 4: The Depreciation of the U.S. Dollar with Respect the Gold
Note: 1/GOLD=the value of U.S. dollar with respect the price of gold. (troy ounces of gold/$).
Source: Historical Gold Prices-1833 to Present. http://www.nma.org/pdf/gold/his_gold_prices.pdf
0.6
0.7
0.8
0.9
1.0
1.1
1.2
99 00 01 02 03 04 05 06 07 08 09 10 11 12 13
1/EUS
.000
.004
.008
.012
.016
.020
.024
.028
.032
40 45 50 55 60 65 70 75 80 85 90 95 00 05 10
1/GOLD
Proceedings of the Pennsylvania Economic Association 65
Table 8
U.S. Spot Exchange Rate Regression [Eq. (7)]
--------------------------------------------------------------------------
Variables teln teln
--------------------------------------------------------------------------
0 0.897 -2.662***
(0.566) (0.612)
oilPln 0.111***
0.130***
(0.020) (0.021)
tNDln -0.129**
0.252***
(0.066) (0.072)
tCAln 0.686***
-0.116
(0.132) (0.149)
WD 0.264***
0.058**
(0.026) (0.025)
EDCD - -0.026*
(0.026)
)1(MA - 1.257***
(0.076)
)2(MA - 1.048***
(0.100)
2R 0.849 0.969
SSR 0.614 0.126
WD 0.265 1.729
F 190.552 513.154
N 141 141
--------------------------------------------------------------------------
Note: teln = spot exchange rate, oilPln = price of oil, tNDln = ln national debt, tCAln = current account balance, WD =
war dummy, EDCD = European debt crisis dummy, MA = moving average process, 2R =R-squared, SSR = sum of squared
residuals, WD = Durbin-Watson statistic, F = F-statistic, N = number of observations, (*), (**), and (***) = significant at
the 10%, 5%, and 1% level, standard errors in parentheses. All variables are in natural log. Column 2 shows the correction for
the first-order serial correlation of the error term (D-W from 0.265 becomes 1.729).
Source: Economagic.com. Data from 1999:01 to 2010:12.
Proceedings of the Pennsylvania Economic Association 66
REFERENCES
Allen, Franklin and Douglas Gale. 2009. Understanding
Financial Crises. Clarendon Lectures in Finance, New York,
N.Y.: Oxford University Press.
Amstad, Marlene and Antoine Martin. 2011. Monetary
Policy Implementation: Common Goals but Different
Practices. Current Issues in Economics and Finance, Federal
Reserve Bank of New York, Volume 17, Number 7: 1-9.
Bonitsis, Theologos Homer. 2011. Eurozone
Competitiveness: An Analysis of the PIIGS. In 2011
Conference Proceedings Northeast Business & Economics
Association, Rajneesh Sharma (editor), Thirty-Eighth Annual
Meeting, November 3-5, 2011, Sheraton Society Hill Hotel,
Philadelphia, PA, U.S.A.
Dornbusch, Rudiger. 1976. Expectations and Exchange Rate
Dynamics. Journal of Political Economy. 84: 1161-1176.
Eiteman, David K., Arthur I. Stonehill, and Michael H.
Moffett. 2010. Multinational Business Finance, 12th
Edition,
Boston, U.S.A.: Prentice Hall.
Glain, Stephen J. 2003. Stocks Sour on Cheap Dollar World
markets Fall on Concerns over Devaluation Moves. Boston
Globe, September 23: D1.
Hale, David. 2005. The Risk of Dollar Devaluation. The
International Economy. 19 (2), Spring: 32-33.
Heath, Allister. 2004. Pressure Grows on G7 to Agree
Acceleration Dollar Devaluation. The Business, September
26: C2.
Kallianiotis, I. N. 2012a. The Single European Financial
Market and its Uncertain Future. Unpublished manuscript,
University of Scranton, March, pages 51.
Kallianiotis, I. N. 2012b. Privatization and its Anti-Social
Effects on National Economies. Hellas on the Web, February
2, 2012: 1-29. http://www.hellasontheweb.org/2010-04-05-
22-20-08/2010-04-06-19-07-44/12035-privatization-and-its-
anti-social-effects-on-national-economies.
Kallianiotis, I. N. 2012c. Greece’s Imposed Privatization
(Denationalization) and its Effects on Individuals’ Utility and
Social Welfare. Christian Vivliografia, February 20, 2012:
1-13.
http://christianvivliografia.wordpress.com/2012/02/20/greece
s-imposed-privatization-denationalization-and-its-effects-on-
individuals-utility-and-social-welfare/
Kallianiotis, I. N. 2011. Is the Imposed Global ‘Laissez-faire’
Socio-economic System Responsible for the Latest Financial
Crisis. Journal of Business and Economics. 2 (5), May: 325-
353.
Kallianiotis, I. N. 2010. Were the Current Bear Market and
the Recession Predictable? International Journal of Applied
Business and Economic Research. 8 (1): 37-64.
Kallianiotis, I. N. and Karen Bianchi. 2009. Speculation,
Uncertainty, Financial Assets Expected Return and Risk, and
Exchange Rate Determination. Unpublished manuscript,
University of Scranton, November 2009: pages 31.
Kallianiotis, I.N. and Dana M. Harris. 2010. What Went
Wrong with our International ‘Laissez-faire, Laissez-passer’
Economic System? International Research Journal of Finance
and Economics. 50, October: 96-122.
Kallianiotis, I.N. and Iordanis Petsas. 2008. Interdependence
between U.S. and EU Goods, Money, and Foreign markets
and Spillover Effects. Spoudai. 58 (1-2): 115-147.
Krugman, Paul. 2012. What Went Wrong? Folio, The
Magazine of the Graduate Center, The City University of
New York, Winter 2012: 20-23.
Morrissy, John. 2011. Currency Wars are Heating up: Dollar
Policy of Competitive Devaluation Slammed. The Windsor
Star. January 11, 2011: C8.
Nechio, Fernanda. 2011. Monetary Policy When One Size
Does Not Fit All. FRBSF Economic Letter, Federal Reserve
Bank of San Francisco. 6/14/2011: 1-5.
Pindyck, R.S. and Rubinfeld, D.L. 1981. Econometric
Models and Economic Forecasts. New York: McGraw-Hill.
Sharma, Shalendra D. 2011. U.S. Debt, Deficit, and the
Falling Greenback: Does it Mean Currency Wars and an End
to the Dollar’s Reign? SERI Quarterly. 4(3), July: 67-78.
Siegman, C.J. 2008. The Global Recession Risk: Dollar
Devaluation and the World Economy. Choice. 45(5),
January: 868.
Straubhaar, Thomas. 2004. The Euro is our Currency, and
the Dollar is your Problem. Intereconomics. 39(6),
November/December: 286-287.
Uri, Noel and Roy Boyd. 1991. The Impact of a Devaluation
of the U.S. Dollar on the Exports of the U.S. Agricultural
Commodities. Journal of World Trade. 25 (4), August: 5-27.
67
TRADE FLOWS BEETWEEN S. KOREA AND THE U.S.A
Orhan Kara
Department of Economics and Finance
West Chester University
West Chester, PA 19383
ABSTRACT This study examines the trade flows between Korea and the
U.S. after Korea adopted flexible exchange rate system. It is
found that the volatility of the exchange rate has a negative
impact on imports while it has a positive effect on exports.
Furthermore, the Korean exports are highly responsive to
changes in income; whereas, a percent change in income in
Korea leads to a one percent change in imports from the U.S.
Similarly, exports are more sensitive to changes in income
than that of imports. Finally, the relative prices have greater
impact on exports relative to imports.
INTRODUCTION
Korean foreign exchange market faced its first financial crisis
in 2008—2009 since the Asian financial crisis of 1997-1998.
Unlike the Asian finance crisis, until which Korea had an
exchange regime that was either freed or managed floating,
the new crisis took place since Korea adopted flexible
exchange rate system. In addition, the last crisis started in the
U.S., which is one of Korea’s top trading partners. What
followed the last financial crisis was the implementation of
fiscal and monetary policies in both countries. In the U.S.
the budget deficit soared, causing the national debt to reach
one hundred percent of the gross domestic product. The
federal funds rate was reduced to almost zero percent.
Similarly, Korea implemented policies to counter the
negative effect of the downturn. For example, the policy rate
fell to two percent from 5.25% and the Korean government
implemented new measures referred to as the Financial
Market Stabilization measures (Shin and Yoo, 2012).
Not only do stabilization policies affect exports and imports,
they also influence the exchange rates. On the one hand,
government policies cause a change on domestic price level,
leading fluctuations in the relative prices of imports and
exports. On the other hand, monetary policies, especially
interest rate changes, may result in volatility in exchange
rates in the short run. Depending upon the effect, trade
balance may improve or worsen. Therefore, it is important to
know how exports and imports respond to policies and how
sensitive the trade flows to changes in prices, exchange rates,
and volatility. Moreover, for a country such as Korea that has
been an example of export promoting growth strategy, it is
equally significant to find out how and what factors affect the
trade flows and if the trade flows respond any differently in
flexible exchange rate regime. This study aims at answering
those questions by investigating trade flows with export and
import demand functions.
To that and, the next section reviews literature. The section
after that presents the model and methodology. Then the
results are given in the following section. The last section
concludes the study.
LITERATURE REVIEW
An exemplary economic growth of Korea has served as a
model to developing nations, which involved heavy
government intervention with restrictions on trade and capital
flows as the early years of industrialization included
subsidies, credit, and a favorable exchange rate policy to
increase exports (Dornbusch and Park, 1987). When the
Bretton Woods exchange rate system was abolished, Korea
pegged its currency, won, to dollar and strictly regulated the
foreign exchange transactions from 1974 to 1980 (Lee,
2007). When the U.S. currency started appreciating in the
1980s, Korea switched to a controlled, floating effective rate
regime, which consisted of a basket of major trading
partners’ currencies and the Special Drawing Right (Hsing,
2009). After the Plaza Agreement, the Japanese currency, a
major trading partner of Korea, started appreciating, which in
turn increased the value of the won, leading to the Korean
government to allow the exchange rate to fluctuate within a
certain percent band in either direction in 1989 (Hsing,
2009). However, the won kept appreciating, resulting in
trade deficits, the Korean government announced a new
exchange rate policy called market average rate (Black,
1999). Black (1999) called this as a switch to an instrument-
based approach to stabilize the currency as Korea finally
achieved stabilization of the exchange rate. Between 1990
and 1995, the won was allowed to fluctuate by market forces
in a band of 0.4 percent to 2.25 percent (Hsing, 2009). As
the Asian financial crisis started, the Korean government
responded by widening the band to ten percent and
eventually allowing the won float freely on December 17,
1997 (Hsing, 1999; and Lee, 2007). Since then, Korea has a
68
flexible exchange rate system, same as the U.S., which is one
of Korea’s major trading partner.
Although adopting a flexible exchange rate regime has
certain advantages, such as independent monetary policy,
automatic adjustment to trade shocks, and an ability to avoid
speculative attracts as suggested by Frankel (2003), but it
also introduces more volatility in the exchange rate. Since
many countries switched to flexible exchange rate, a rich
literature in the volatility of exchange rates has been
emerged. However, no consensus has been reached as to
whether volatility has a positive or negative impact on trade
flows. A detailed survey of literature, conducted by Ozturk
and Kalyoncu (2009), showed that some studies found
negative effect while other studies discovered a positive
effect of exchange rate volatility on trade flows. For
example, Brada and Mendez (1988), McKenzie and Brooks
(1997), and McKenzie (1999) claimed that volatility
positively affected trade flows while Akhtar and Hilton
(1984), Chowdhury (1993), Vergil (2002), and Chit (2008)
found a negative effect of the volatility on trade. In their
study, Ozturk and Kalyoncu (2009) also examined the
Korean trade by employing Engle-Grainger based
cointegration method from 1980 to 2005, and concluded that
volatility had a significant negative effect. In addition, the
studies on the volatility of exchange rate on the Korean
economy also produced contradicting results (Doroodian,
1999; Doganlar, 2002; Poon, Choong, and Habibullah, 2005;
Park, 2007, Baak, Al-Mahmood, and Vixathep, 2007; Shin
and Yoo, 2012; and Bahmani-0skooee, Harvey, and Hegerty,
2012).
Focusing on the real exchange rates, Doroodian (1999)
investigated the effect of volatility in Korean, Indian, and
Malaysian international trade with quarterly data ranging
from 1973 to 1996. Doroodian (1999) asserted that the
volatility measures used by existing studies had certain
shortcomings and employed the GARCH procedure instead
of measures such as the standard deviation, deviation from
trend, difference between forward and spot rates, Gini mean
difference coefficient, coefficient of variation, and the scale
measure of variability. The empirical evidence illustrated that
Korean exports were negatively affected by the volatility of
the exchange rate. On the other hand, Doganlar (2002)
proxied the volatility with the moving sample standard
deviation of the growth of the real exchange rate and
analyzed the Turkish, Korean, Malaysian, Indonesian and
Pakistani quarterly data between 1980 and 1996. He found
that the volatility of the exchange rates decreased the real
exports (Doganlar, 2002). Similarly, Poon et al. (2005) used
the moving sample standard deviation of the growth of the
real exchange rate for the volatility measure and extended the
analysis to Japan, Singapore, Thailand, Indonesia, and Korea.
They also reached to the same conclusion; exchange rate
volatility was negatively correlated with exports. The inverse
relationship between the volatility in the exchange rates and
trade was also confirmed by Baak et al. (2007) who
investigated the bilateral trade between four Asian countries,
namely Hong King, Korea, Singapore, and Thailand, with the
U.S. and Japan from 1981 to 2004 by using the standard
deviation of the monthly real exchange rates as the measure
of volatility.
However, Park (2007) examined the relationship between
volatility in response to positive and negative shocks by
employing a quantile regression method to the exchange rate
between Korean won and the U.S. dollar and found a
stronger effect of positive shocks. Likewise, Bahmani-
Oskooee et al. (2012) studied ninety-six export industries and
twenty-nine import industries in the short and long run and
concluded that bilateral exports and imports in a majority of
the industries responded positively to volatility in the short-
run. In contrast, Shin and Yoo (2012) could not reach any
significant conclusion as to whether volatility had a negative
or positive effect.
In short, there is no consensus emerged with respect to the
volatility of the exchange and its effect of trade flows.
Studies often produced conflicting results as some found a
positive effect of the volatility on trade flows and others
indicated a negative effect while the rest could not reach any
conclusion. Therefore, this study further contributes by
empirically analyzing the bilateral trade flows between Korea
and the U.S. under the flexible exchange rate regime.
THE MODEL AND METHODOLOGY
In order to determine the effect of the volatility in exchange
rates on the trade flows between Korea and the U.S., we
formulate the functional form to be used in the empirical
analysis. Based on the studies investigating the determinants
of the trade flows, researchers agreed that the following
factors are the main determinants of trade flows: volatility,
relative prices, income, and exchange rates (Kara, 2012).
Therefore, following Bahmani-Oskooee and Kara (2008) and
Kara (2012), functional relation is formulated as follows:
Trade Flows = f(VOL,P, Y, EXC,) (1)
69
where TF is trade flows (imports and exports), P is relative
price index, Y is income, and EXC is exchange rate.
Exchange rate is defined as the number of units of Korean
currency per dollar, which means that an increase in EXC is a
depreciation of the Korean currency.
In order to apply the above functional form to the imports
and exports, we adopted the followings:
ttttt EPD
PMYVOLM ln)ln(lnlnln 4321
(2)
where M is the volume of imports from the USA, VOL is the
volatility, PM = the price of imports, PD = price of domestic
goods, Y = domestic income, E = nominal effective exchange
rate, and is an error term.
ttttt eEPXW
PXYTPVOLaX ln)ln(lnln 4321
(3)
where X is the volume of the Korea’s exports; VOL is the
volatility; (PX/PXW) is the relative price of a country’s
exports (PX) compared to the world export prices(PXW),
YTP is income of the trading partner, E is the nominal
effective exchange rate and e is an error term. ln before the
variables refers to natural logarithm of the variables.
The volatility variable is estimated from the following
formula (Baak et al., 2007):
2
1
2])(1
1ln[ i
tn
tmkikit RERRER
nVOL
(5)
where t is quarter, k is month, RER is reel exchange rate,
k=tm is the first month, and k=tn is the last month of the
quarter. Similar to Baak et al. (2007), the volatility is
computed the standard deviation includes current and the
previous quarter monthly exchange rates.
Equations (2) & (3) outline the long-run relationship between
trade flows and their determinants. In order to test the effect
of the variables, we incorporate the dynamic adjustment into
equations, which requires a specification in an error-
correction format. Following Pesaran et al. (1996, 2001, and
2009), we specify equations (2) & (3) in Autoregressive
Distributed Lag (ARDL) formats as follows:
tutMtEtPD
PM
tYtVOLitMn
ii
itEn
iiit
PD
PMn
iiitY
n
iitVOL
n
iitM
1ln51ln41)ln(31ln21ln1ln1
ln0
)ln(0
ln01
0
ln
(6)
tvtXtEtPXW
PX
tYTPtVOLitXm
ii
itEm
iiit
PXW
PXm
iiitYTP
m
iitVOL
m
iitX
1ln51ln41)ln(31ln21ln1ln1
ln'0
)ln('0
ln'01ln
0''ln
(7)
Again, the variables are defined as before. The error terms
are assumed iid( 0,2 ).
In this study, since we use ARDL approach developed by
Pesaran et al. (1996, 2001 & 2009) due to a relative ease in
estimations, we proceed to the two stages of the estimation.
First, by using an F-test the existence of cointegration is
determined by testing the significance of the lagged
variables. According to the asymptotic properties of the
distribution, an F-test table is provided by Pesaran et al
(2009). A decision is reached by comparing the test statistics
to the two critical values given in the table. When a test
statistic falls between the critical values, the result is
inclusive and other unit-root tests will be used. When a
statistic is larger than the upper bound of the critical values,
the existence of cointegration is accepted, and if test statistic
is below the lower bound, then the existence of cointegration
is rejected.
First, two null hypotheses (stating the non-existence of
cointegration) are constructed to determine the cointegration
among variables in each equation.
H0: 1=2=3=4=5=0 H1: Not H0
H0: 1=2=3=4=5=0 H1: Not H0
70
The above hypotheses are tested according to the test
statistics from the table.
After the completion of the first stage, the second stage
involves estimating the equations, applying error correction,
and making inferences for the estimated coefficients. We
estimate equations (6) and (7) based on the best model
chosen by the R-bar square criterion, Akaika information
criterion (AIC), Schwarz Bayesian criterion (SBC), and
Hannan-Quinn criterion (HQC). In addition, our selection
involves diagnostic statistics.
The quarterly data for this study covers the period from 1998
to 2012 (1998Q1-2012Q4). Part of data used in this study
was extracted from the IMF Financial Statistics, online and
the rest was obtained from IMF Direction of Trade Statistics.
The following data are used in this study:
o Korean imports from the U.S.
o Korean exports to the U.S.
o Unit value of the U.S. exports
o Unit value of exports of Korea
o Unit value of world exports
o Industrial production in Korea
o Industrial production in the U.S.
o Producer prices in Korea
o Nominal Effective Exchange Rates
o Real Effective Exchange Rates
RESULTS
Using quarterly data from 1998 to 2012 for the imports and
exports between Korea and the United States, the equations
(6) and (7) were estimated by applying an error-correction
model, specifically ARDL. As explained above, we obtained
F-statistics for testing the joint significance hypotheses in
equation (8). As the testing can be carried out based on; a) no
intercept, no trend variable, b) intercept and no trend
variable, and c) intercept and trend variables, we performed
six F test analysis. In most cases the F-statistic was
significant. For the case of intercept and no trend variable,
the F statistics were bigger than upper bound critical value,
which led to the existence of cointegration since the null
hypotheses were rejected (Table 1).
After establishing the existence of cointegration, we
proceeded to the next stage and estimated equations (6) and
(7) by first imposing four lags on each first differenced
variable. As a result, two groups of estimates are obtained
for: 1) imports and volatility, relative prices, income, and
exchange rate; and 2) exports and relative prices, income, and
exchange rate. For the import equation, Akaika information
criterion provided better results according to diagnostics
tests, and R-bar criterion indicated best result for the export
equation.
Table 2 illustrates the long run estimates as well as ARDL
estimates for the equation (6). Long run estimates confirm
our expected signs except for the relative price variable. An
increase in volatility reduces the imports, indicating a
negative effect. However, the volatility variable did not turn
out to be statistically significant and when measured in
absolute value, the effect is relatively small. Income variable
is highly statistically significant. Given that we use the log
of the variables, the coefficient can be interpreted as the
elasticity of income. Therefore, a one percent increase in the
Korean income is associated by a 1.057 percent increase in
the imports from the U.S., indicating almost a unit elasticity.
We expected that depreciation in currency makes imports
more expensive, which was supported by the estimated sign
as the exchange rate was defined as the number of Korean
currency per dollar. Ten percent depreciation in won
decrease Korean imports from the U.S. by 0.67 percent.
ARDL estimates indicate that adjustment process is very
short for the relative prices compared to the other variables
which can take as long as four quarters. Based on the
changing signs in the lags, the variables adjust over time to
long run values. For example, if there is an overshoot in one
period, there will be undershooting in the following period.
Diagnostic tests show the robustness of the estimates. For
instance, the value of R-bar squared was relatively high and
F-statistics was highly significant. Under the variables
column in the table, the number in parenthesis shows the
number of lags, which is the same in the other tables in the
article.
Table 3 provides the error correction estimates for the Korean
imports from the United States. In the variables column in
the table, the symbol Δ in front of the variables refers to the
differencing. For example, ΔImports means one period
difference in imports [imports-imports(-1)]. The number after
the variable refers to the further differencing in the variables:
Δimports1 indicates one period difference of ΔImports
[imports(-1)-imports(-2)], and so on. This notation is the
same in all.
Most of the variables are statistically significant in Table 3.
Ecm(-1) stands for the error correction coefficient, which
shows how fast the economy returns to the equilibrium (long
run values) once it is shocked. After a shock occurs (e.g., a
change in exchange rate), an adjustment process takes place
during the transition for the economy returning to its long-run
71
equilibrium values. This speed of this adjustment process is
indicated by the error correction term. We expect the sign of
the error correction coefficient is negative for the following
reason. When the estimated short- run values overshoot the
long-run equilibrium values, then the adjustment is
downward or vice versa (Greene, 2008). The magnitude of
the error correction coefficient is another important feature.
The larger the value of the error correction term in absolute
value, the faster the economy goes back to its long run
equilibrium values. Similarly, smaller value of the error
correction coefficient means a slower adjustment process for
the long-run equilibrium values.
With respect exports, Table 4 presents the long run and
ARDL estimates of the equation (7). All variables are
statistically significant at 95% significance level. Unlike the
import case which has a negative relationship with the
volatility, exports respond positively to the volatility in
exchange rates. Specifically, a one present increase in the
volatility increases Korean exports to the U.S. by 0.19
percent. Given the size of the coefficient for the income
variable, the U.S. imports from Korea is highly sensitive to
income. A one percent increase in the U.S. income leads to
over nine percent increase in the value of goods purchased
from Korea. Due to the high income elasticity, recessions in
the U. S. would cause a relatively big decrease in Korean
exports. Likewise, Korean exports to the U.S. are also
responsive to the changes in the Korean exports prices to the
world export prices. The estimate indicates a 1.81 percent
decrease in the exports if the relative prices go up by one
percent.
Moreover, exchange rate also has a large impact on Korea’s
exports. The estimated exchange rate coefficient has an
expected positive sign, meaning that depreciation in currency
makes the prices of Korean goods cheaper for American
consumers leading to an increase in Korean goods
consumption in the U.S. when won depreciates or dollar
appreciates. In terms of percentage effect, one percent
deprecation in Korean currency is associated with a 3.56
percent increase in the U.S. imports from Korea. Table 4
also gives the ARDL estimates illustrating how long the
response for changes in the variables given by the number of
lags. Given that the number of lags is as high as nine, the
effect of changes in the exchange rate on Korea’s exports to
the U. S. takes up to nine quarters.
The relative long adjustment process is also evident in Table
5 that provides the error correction estimates for the equation
(7). Many variables are, again, statistically significant
including the error correction coefficient (ecm(-1)).
Furthermore, diagnostic statistics show support for the
robustness of the analysis, except for the presence of serial
correlation. We have checked several diagnostic statistics for
the models in this study in addition to a visual examination of
the plot of residuals and fitted values.
CONCLUSIONS
This study analyzed the trade flows between Korea and the
U.S. after Korea adopted flexible exchange rate system. The
effects of exchange rate volatility, income, relative prices,
and exchange rate on trade flows were estimated with Import
and export demand functions. The main conclusions of this
study are as follows. First, exchange rate volatility has a
negative effect on imports, which is in line with the studies
by Akhtar and Hilton (1984), Chowdhury (1993), Doroodian
(1999), Vergil (2002), Baak et al. (2007), Chit (2008), Ozturk
and Kalyoncu (2009). With respect to exports, the volatility
has a significantly positive effect, lending support to
findings by Brada and Mendez (1988), McKenzie and Brooks
(1997), and McKenzie (1999), and Bahmani-Oskooee et al.
(2012). Second conclusion of this study is that Korean
exports to the U.S. are more sensitive to changes in the
exchange rate than that of imports. Third, imports have a
unit income elasticity while exports are very responsive to
income as the income elasticity was about nine in exports.
With respect to external balances, policy makers should
implement policies to counteract the decrease in exports
during downturns in the U. S. economy. Lastly, unlike
imports which are less sensitive to changes in relative prices,
exports are more sensitive to changes in relative prices.
72
TABLE 1. F-TEST RESULTS FOR COINTEGRATION
Equation F-Statistics
Import 7.724
Export 9.321
Critical values for F-Test at 95% Significance Level (intercept, No Trend):
Lower Bound=2.649 & Upper Bound= 3.805
TABLE 2. LONG-RUN COEFFICIENT ESTIMATES FOR IMPORTS
Variables Coefficient Standard Error T-Ratio P-value
Volatility -0.022 0.049 -0.447 0.658
Income 1.057 0.153 6.898 0.000
Relative Prices 0.373 1.699 0.219 0.828
Exchange Rate -0.666 0.389 -1.712 0.095
Constant 20.720 10.032 2.065 0.046
ARDL Estimates for Imports
Coefficient Standard Error T-Ratio P-value
Imports (-1) 0.073 0.117 0.623 0.537
Imports (-2) 0.064 0.108 0.590 0.558
Imports (-3) 0.511 0.117 4.363 0.000
Volatility -0.008 0.017 -0.451 0.654
Income 0.467 0.391 1.196 0.239
Income (-1) 1.375 0.587 2.344 0.024
Income (-2) -0.498 0.528 -0.943 0.352
Income (-3) -0.415 0.527 -0.787 0.436
Income (-4) -0.557 0.304 -1.830 0.075
Relative Prices -3.434 1.266 -2.712 0.010
Relative Prices (-1) 3.566 1.325 2.692 0.010
Exchange Rate -0.516 0.329 -1.567 0.125
Exchange Rate (-1) -0.350 0.388 -0.900 0.373
73
Exchange Rate (-2) 0.424 0.371 1.145 0.259
Exchange Rate (-3) -0.269 0.365 -0.737 0.465
Exchange Rate (-4) 0.476 0.234 2.039 0.048
CONSTANT 7.304 3.986 1.832 0.075
R-Bar-Squared
F-Stat.
Equation Log-likelihood
Akaike Info. Criterion
Schwarz Bayesian Criterion
DW-statistic
0.94041
55.2455
89.433
72.433
55.217
2.316
TABLE 3. ERROR CORRECTION ESTIMATES FOR IMPORTS
Variables Coefficient Standard Error T-Ratio P-value
ΔImports1 -0.575 0.127 -4.539 0.000
ΔImports2 -0.511 0.117 -4.363 0.000
ΔVolatility -0.008 0.017 -0.451 0.654
ΔIncome 0.467 0.391 1.196 0.238
ΔIncome1 1.470 0.390 3.766 0.001
ΔIncome2 0.972 0.389 2.500 0.016
ΔIncome3 0.557 0.304 1.830 0.074
ΔRelative prices -3.434 1.266 -2.712 0.010
ΔExchange Rate -0.516 0.329 -1.567 0.125
ΔExchange Rate 1 -0.631 0.284 -2.225 0.032
ΔExchange Rate 2 -0.207 0.271 -0.765 0.448
ΔExchange Rate 3 -0.476 0.234 -2.039 0.048
ecm(-1) -0.353 0.102 -3.457 0.001
R-Bar-Squared
F-Stat.
Equation Log-likelihood
Akaike Info. Criterion
Schwarz Bayesian Criterion
DW-statistic
0.706
11.397
89.433
72.433
55.217
2.316
74
TABLE 4. LONG-RUN COEFFICIENT ESTIMATES FOR EXPORTS
Variables Coefficient Standard Error T-Ratio P-value
Volatility 0.190 0.058 3.268 0.010
Income Trade
Partner 9.326 1.751 5.328 0.000
Relative Prices -1.810 0.231 -7.831 0.000
Exchange Rate 3.557 0.524 6.784 0.000
Constant -36.241 12.518 -2.895 0.018
ARDL Estimates for Exports
Coefficient Standard Error T-Ratio P-value
Exports (-1) 0.169 0.173 0.978 0.354
Exports (-2) -0.183 0.167 -1.090 0.304
Exports (-3) -0.547 0.181 -3.021 0.014
Exports (-4) 0.571 0.129 4.423 0.002
Exports (-5) -0.395 0.132 -2.997 0.015
Exports (-6) -0.238 0.187 -1.277 0.234
Volatility -0.067 0.021 -3.118 0.012
Volatility (-1) -0.039 0.022 -1.752 0.114
Volatility (-2) -0.011 0.019 -0.584 0.574
Volatility (-3) 0.021 0.026 0.811 0.438
Volatility (-4) 0.092 0.035 2.614 0.028
Volatility (-5) 0.106 0.041 2.587 0.029
Volatility (-6) 0.098 0.037 2.623 0.028
Volatility (-7) 0.052 0.032 1.656 0.132
Volatility (-8) 0.056 0.024 2.352 0.043
Income, Trading
Partner Income
6.896 2.409 2.862 0.019
Income (-1) 5.101 3.180 1.604 0.143
Income (-2) 2.999 2.746 1.092 0.303
Income (-3) 3.397 3.392 1.001 0.343
Income (-4) -9.033 2.934 -3.078 0.013
Income (-5) 9.397 1.958 4.798 0.001
Income (-6) -3.614 1.947 -1.856 0.096
Relative Prices -3.264 0.891 -3.665 0.005
Relative Prices (-1) -1.063 1.302 -0.816 0.435
Relative Prices (-2) 2.073 1.148 1.805 0.104
Relative Prices (-3) -2.502 0.927 -2.700 0.024
Relative Prices (-4) 1.014 0.835 1.214 0.256
Relative Prices (-5) -0.622 0.811 -0.767 0.463
75
Relative Prices (-6) 2.180 0.821 2.657 0.026
Relative Prices (-7) 1.176 1.025 1.148 0.281
Relative Prices (-8) -1.931 0.758 -2.549 0.031
Exchange Rate 2.096 0.620 3.378 0.008
Exchange Rate (-1) 1.133 0.664 1.708 0.122
Exchange Rate (-2) 0.854 0.676 1.263 0.238
Exchange Rate (-3) 1.027 0.774 1.326 0.218
Exchange Rate (-4) -0.120 0.736 -0.163 0.874
Exchange Rate (-5) 1.190 0.775 1.536 0.159
Exchange Rate (-6) -0.678 0.588 -1.153 0.278
Exchange Rate (-7) -0.446 0.575 -0.774 0.459
Exchange Rate (-8) -0.721 0.606 -1.189 0.265
Exchange Rate (-9) 1.442 0.340 4.245 0.002
CONSTANT -58.847 29.471 -1.997 0.077
R-Bar-Squared
F-Stat.
Equation Log-likelihood
Akaike Info. Criterion
Schwarz Bayesian Criterion
DW-statistic
0.962
31.582
137.743
95.743
55.175
3.187
TABLE 5. ERROR CORRECTION ESTIMATES FOR EXPORTS
Variables Coefficient Standard Error T-Ratio P-value
ΔExports1 0.793 0.285 2.786 0.015
ΔExports2 0.611 0.258 2.366 0.034
ΔExports3 0.063 0.209 0.302 0.767
ΔExports4 0.634 0.194 3.261 0.006
ΔExports5 0.238 0.187 1.277 0.224
ΔVolatility -0.067 0.021 -3.118 0.008
ΔVolatility1 -0.414 0.133 -3.107 0.008
ΔVolatility2 -0.425 0.136 -3.133 0.008
ΔVolatility3 -0.404 0.133 -3.032 0.010
ΔVolatility4 -0.313 0.112 -2.797 0.015
ΔVolatility5 -0.206 0.080 -2.592 0.022
ΔVolatility6 -0.108 0.048 -2.266 0.041
ΔVolatility7 -0.056 0.024 -2.352 0.035
ΔIncome 6.896 2.409 2.862 0.013
76
ΔIncome1 -3.146 3.632 -0.866 0.402
ΔIncome2 -0.147 3.434 -0.043 0.967
ΔIncome3 3.250 3.093 1.051 0.312
ΔIncome4 -5.783 2.386 -2.424 0.031
ΔIncome5 3.614 1.947 1.856 0.086
ΔRelative Prices -3.264 0.891 -3.665 0.003
ΔRelative Prices1 -1.389 0.961 -1.445 0.172
ΔRelative Prices2 0.685 0.852 0.803 0.436
ΔRelative Prices3 -1.818 0.925 -1.965 0.071
ΔRelative Prices4 -0.803 0.866 -0.927 0.371
ΔRelative Prices5 -1.426 0.759 -1.879 0.083
ΔRelative Prices6 0.755 0.747 1.010 0.331
ΔRelative Prices7 1.931 0.758 2.549 0.024
ΔExchange Rate 2.096 0.620 3.378 0.005
ΔExchange Rate1 -2.547 1.344 -1.895 0.081
ΔExchange Rate2 -1.693 1.244 -1.361 0.197
ΔExchange Rate3 -0.667 1.082 -0.616 0.548
ΔExchange Rate4 -0.786 0.964 -0.816 0.429
ΔExchange Rate5 0.403 0.718 0.562 0.584
ΔExchange Rate6 -0.275 0.568 -0.484 0.636
ΔExchange Rate7 -0.721 0.440 -1.637 0.126
ΔExchange Rate8 -1.442 0.340 -4.245 0.001
ecm(-1) -1.624 0.354 -4.593 0.001
R-Bar-Squared
F-Stat.
Equation Log-likelihood
Akaike Info. Criterion
Schwarz Bayesian Criterion
DW-statistic
0.802
6.570
137.743
95.743
55.175
3.187
REFERENCES
Akhtar, M. and Hilton, R. S. (1984). Effects of Exchange Rate Uncertainty on German and U.S. Trade. Federal Reserve Bank
of New York Quarterly Review, 9, pp. 7-16.
Baak, S. J., Al-Mahmood, M. A., and Vixathep, S. (2007). Exchange Rate Volatility and Exports from East Asian Countries to
Japan and the U.S. Applied Economics, 39, pp. 947-959.
Bahmani-Oskooee, M. and Kara, O. (2008). Relative Responsiveness of Trade Flows to a Change in Prices and Exchange Rate
In Developing Countries Journal of Economic Development, 33 (1), pp. 147-163
77
Bahmani-0skooee M., Harvey, H; and Hegerty, S. W. (2012). Exchange Rate Volatility and industry Trade Between The U.S.
and Korea. Journal of Economic Development, 37(1), pp. 1-27.
Black, S.W. (1999). Issues in Korean Exchange Rate Policy. Changes in Exchange Rates in Rapidly Developing Countries:
Theory, Practice, and Policy Issues. pp.269-283.
Brada, J. C. and Mendez, J. A. (1988). Exchange Rate Risk, Exchange Rate Regime, and the Volume of International Trade.
KYKLOS, 41, pp. 263-280.
Chit, M. M., (2008). Exchange Rate Volatility and Exports: Evidence from the Asian-China Free Trade Area. Journal of
Chinese Economic and Business Studies, 6,(3), pp. 261-277.
Chowdhury, A. R. (1993). Does Exchange rate Volatility Depress Trade Flows? Evidence from Error Correction Models. The
Review of Economics and Statistics, 76 , pp. 700-706.
Doganlar, M. (2002). Estimating Impact of Exchange Rate Volatility on Exports: Evidence from Asian Countries. Applied
Economies Letters, 9, pp. 859-863.
Doroodian, K. (1999). Does Exchange Rate Volatility Deter International Trade in Developing Countries? Journal of Asian
Economics, 10, pp. 465-474.
Dornbusch, R. & Park, Y. C. (1987). Korean Growth Policy. Brookings papers on Economic Activity. No. 2. pp. 389-444.
Frankel, J. A. (2003). Experience of Lessons From Exchange Rate Regimes in Emerging Markets. NBER Working Papers No
10032.
Hsing, Y. (2009). Modeling the Behavior of the KRW/USD Exchange rate and Policy Implications. Global Economic Review,
38 (2), pp. 205-214.
Greene, W. H., 2008. Econometric Analysis, 6th
Ed. New Jersey: Pearson/Prentice Hall.
Kara, O. (2013). Balancing the U.S. Trade Deficit with China: How Much Does Renminbi Need Appreciating against Dollar?
Journal of International Finance and Economics, 13 (2). pp. 71-82.
Lee. B. (2007). Economic Fundamentals and Exchange Rates Under Different Exchange Rate Regimes: Korean Experience.
Journal of Applied Economics, 10 (1)., pp. 137-159.
McKenzie, M. D., (1999). The Impact of Exchange Rate Volatility on International Trade Flows. Journal of Economic
Surveys, 13(1), pp. 71-106.
McKenzie, M. D. and Brooks, R. (1997). The Impact of Exchange Rate Volatility on German-US Trade Flows. Journal of
International Financial Markets, Institutions, and Money, 7, pp. 73-87.
Ozturk, I & Kalyoncu, H. (2009). Exchange Rate Volatility and Trade: An Empirical Investigation from Cross-country
Comparison. African Development Review, 21(3), pp. 499-513.
Pesaran, M.H., Y. Shin, & R.J. Smith (1996). Testing for the existence of a long-run relationship. In DAE Working paper, V.
9622, Department of Applied Economics, University of Cambridge.
Pesaran, M.H., Y. Shin, & R.J. Smith (2001). Bound Testing Approaches to the Analysis of Level Relationships, Journal of
Applied Econometrics, 6, pp. 289-326.
Pesaran, B., & Pesaran, M.H. (2009), Time Series Econometrics. Oxford: Oxford University Press.
Park, B. (2007). Trading Volume, Volatility, and GARCH Effects in the South Korean won/US Dollar Exchange Market:
Evidence from Conditional Quartile Estimation. The Japanese Economic Review, 58(3), pp. 382-399.
Shin, H. K. and Yoo, B. H. (2012). The Volatility of the Won-Dollar Exchange Rate During the 2008-9 Crisis. Journal of
Economic Development, 37(4), pp. 61-78.
78
Vergil, H. (2002). Exchange rate Volatility in Turkey and Its Effect on Trade Flows. Journal of Economic and Social
Research, 14(1), pp. 83-99.
A SIMPLE MODEL OF BASEBALL DESEGREGATION
Timothy F. Kearney & David Gargone
Department of Business
Misericordia University
Dallas, PA 18612
ABSTRACT Baseball began to desegregate in 1947 when Jackie Robinson
with the Brooklyn Dodgers broke the color line, a process
that would take until 1959. There is considerable research
about the positive impact from integration but no research
about the reasons that led to desegregation. We posit that
baseball’s weak financial position following the Depression
and WWII created pressures to desegregate. Using
attendance as a proxy for financial performance, we find the
more poorly performing team in the seven instances where
teams shared cities predicted which would integrate first.
INTRODUCTION
After decades of intransigence, baseball finally began the
process of desegregation in 1947 when the Jackie Robinson
with Brooklyn Dodgers of the National League and Larry
Doby with the Cleveland Indians of the American League
finally broke the color line. Official segregation had deprived
baseball of a significant pool of baseball talent, as 18 Hall of
Fame players spent time in the segregated Negro Leagues.
As would be expected, tapping that talent pool by major
league baseball (MLB) had significant impacts. In his article
The Cost of Discrimination: A study of Major League
Baseball, “A sizeable and statistically significant relationship
between winning and the number of black players is found
for the 1950s” (Hanssen 1998, p. 603). Nonetheless, MLB
retained segregated teams until 1959 when the Boston Red
Sox finally integrated.
In that article, Hanssen (1998) notes that “Why the color line
was broken remains an open question” (p. 605), and posits
changes in attitudes following WWII. We find that the Great
Depression, the subsequent World War II area and
demographic shifts West and South provided the incentives
for integration. In 1947, there were five cities with 11 teams
which shared MLB teams (Philadelphia A’s/Phillies; Boston
Braves/Red Sox; Chicago Cubs/White Sox; St. Louis
Browns/Cardinals; New York Yankees/Giants/Dodgers).
We consider each of these cities as economic pairs, and find
that the team in the city which was underperforming in terms
of attendance and had other financial difficulties integrated
first.
ECONOMICS OF DESEGREGATION
Gary Becker is well known for his path breaking studies of
labor market discrimination. In his work “The Economics of
Discrimination” (Becker, 1971), Becker identifies three
sources of discrimination: Employer discrimination, co-
worker discrimination and consumer discrimination. There
have been numerous studies of discrimination in sports,
generally showing returns to teams that ended discrimination.
Kahn (1991) surveys the literature on discrimination. He
finds that a number of studies show that teams which
integrated faster had better on-field results as a result. He
concludes this is what would be expected in a market that
eliminates employer discrimination. In terms of co-worker
discrimination, he notes that both members of teams
integrating as well as teams they would oppose balked at
playing on an integrated team/game). Oh and Buck (2012)
looked at wage differentials based on race and marginal
productivity and finds at least some evidence of all three
sources of discrimination in sports. Kuper and Szymanski
(2001, pp. 90-95) finds that there was discrimination, and
that integrating improved team records
Anecdotally, we know that both the Dodgers and Indians (the
first teams to integrate) strongly came down in favor of
integration. Dodger players presented Manager Leo
Durocher with a petition to jettison Robinson. Durocher told
them that “I’m the manager and I say he plays”. (Eig,
Opening Day, p. 44).
Kahn (1991) notes studies which demonstrate that consumer
discrimination was an important consideration. However,
given that integration of the field was followed by integration
of the grandstands, this effect is difficult to estimate.
Hanssen expressed it as “Because of the change in attitudes,
and the complementary rise of the black middle class, the
integration of baseball presented the prospect of a profit to be
made” (p. 605)
The elimination of employer discrimination presumes a free
market, where the inefficiencies of hiring lower ability
workers can be bid away. Baseball as an institution is
somewhat protected, as Major League Baseball (MLB)
enjoys anti-trust protections, which allows baseball to act as a
79
cartel, with their markets protected in general from
competition.
However, in the post WWII era, baseball teams were not
even distributed across the country. In fact, five cities had
multiple teams; Philadelphia, Boston, Chicago and St. Louis
had two teams and New York City had three teams. These
teams are geographically distributed across the eastern part of
the country, covering New England (Boston), the Upper Mid-
West (Chicago), the Mid-Atlantic (Philadelphia and New
York) and baseball’s West (St. Louis), though did not cross
the Mississippi. While anti-trust legislation protected the
leagues from competition, within cities there was strong
competition.
General Demographic Trends Favoring Integration
While the Dodgers and Jackie Robinson are rightfully well
known for integrating major league baseball, it is important
to recognize that American sports – and society – had been
desegregating throughout the 20th century. Baseball was a
late integrator. Of the major league sports in the USA, only
hockey integrated later, not until 1958, when Walter O’Ree
laced up for the Boston Bruins and became known as
“hockey’s Jackie Robinson”. We note that major sports were
integrated long before baseball:
• George Poage was the first American black to win a medal
in the Olympics, in track at the 1904 Olympic Games
• Boxing began to be fitfully integrated in the early 20th
century and by 1908 Jack Johnson won the main prize, the
heavyweight title.
• NCAA football began to be integrated in 1918
• Bobby Marshal integrated the NFL in 1920
Segregation was clearly not the simple result of ignorance by
baseball talent scouts of the abilities of African-American
players. Hall of Fame manager John McGraw tried to sign a
black player as early as 1901. “claiming that he was actually
Native American” (Chadwick, p. 28). Major league owners
were often very knowledgeable about their Negro League
counterparts. Negro League teams such as the New York
Cubans (NY York Giants Polo Grounds), Newark Eagles
(Newark Bears’ Rupert Field, home of the Yankee AAA
team), Homestead Grays (both Pittsburgh Pirates Forbes
Field and Washington Senators Griffith Park), and the East
West All Star Game (Chicago White Sox Comiskey Park)
among others gave white owners a chance to evaluate black
talent.
Importantly, lucrative barnstorming tours between white and
black players were common in the 20th century. In fact, in
the 1930s Future Hall of Fame Negro League pitcher Satchel
Paige was earning some $50k, reportedly second only to
Babe Ruth. (Chadwick, p. 124). Hall of Fame National
League St. Louis Cardinal pitcher Dizzy Dean barnstormed
with Paige from 1934 to 1945. Their relationship was an
important step forward, as “The color coded pairing of starts
gave a human face to the battles between white and black
teams which had been playing out in California for 25 years”
(Tye, p. 94) Importantly in a segregated society, “on the
barnstorming tour, ballparks that normally walled off blacks
let they where they wanted. It brought in white reporters
with white fans” (Tye, p94-95). Bob Feller barnstormed with
Paige, including a 1941 “big money matchup against Satchel
Paige and the (Negro league Kansas city) Monarchs at
Sportsman’s Park in St. Louis” and an important 1946 tour.
That tour, aided by the Flying Tigers Airlines, “provided
blacks with a ‘chance to prove ourselves against white
players’” (Gay, p 224).
In the early-1940s, the owner of the Philadelphia Phillies Bill
Veeck saw the Negro Leagues as a great, untapped area of
baseball talent. In 1943, Veeck made a bid to buy the “down
in the dumps” Philadelphia Phillies. “Veeck’s transformation
plan for the Phils included a secret weapon: Negro Leaguers”
(Eig, pp. 181-182). A few years later, Dodger GM Branch
Rickey recognized “the Negro Leagues contained a gold
mine of big league players “and hence he “saw inefficiency
and exploited it at the expense of his competitors”.
(Bradbury, p 130).
The Economic Downturn Leads to Integration
We contend that the economic pressures of the 1930s sowed
the seeds for integration. Discrimination is expensive;
economic research has shown that “discrimination is
expensive and its cost may reduce its incidence” (Hanssen,
p1998 p. 603). And while baseball as an entity enjoys anti-
trust protections, the structure of baseball in the 1930s
created competition within cities. This created competition
negated to some extent the benefit that monopoly granted.
Baseball had been static since 1908, when the Baltimore
Orioles moved to New York and became the Yankees. In all,
New York had three teams (Yankees, Giants and Dodgers);
two teams shared Philadelphia (A’s,Phillies), St. Louis
(Browns, Cardinals), Chicago (White Sox, Cubs) and Boston
(Braves, Red Sox). Single team cities were found only in
Cleveland, Cincinnati, Pittsburgh, Detroit and Washington.
The economic losses of baseball in the 1930s and WWII era
began to pressure ownership. MLB attendance fell from over
10mn in 1930 to a low of 6mn by 1933. It wasn’t until 1945
that the 10mn mark was reached again. This was matched
by demographic trends. The 1940 census recorded the first
drops in population for Boston, St. Louis and Philadelphia.
The 1950 census saw population peaks for Boston, St. Louis,
80
Philadelphia and Chicago, while the 1960 census saw the
first ever drop in population for New York City.
The 1930s saw the only decade on record where the average
MLB team sales price dropped in real 2002 dollars. In the
1930s, there were four sales, and the average price change
was a 33% reduction or about 4.1% per annum as compared
with a 0.3% annual drop for the Dow Jones Index. (Haupert,
http://eh.net/encyclopedia/article/haupert.mlb).
Our analysis is based on seven pairs in five cities: four cities
(Boston, Philadelphia, St. Louis and Chicago) with two
teams, and the two New York pairs (both the Dodgers and
Giants compared with the later integrating Yankees). We
also consider the Philadelphia situation in 1943when Veeck
intended to integrate (along with the actual 1953 integration
of the Athletics).
By comparing teams within a market, we are able to
recognize important variables which would exert similar
pressure on each team: local demographics including
population changes; state/local pressures on teams to
integrate; income growth and transportation issues.
Financial data for the pre-integration era are notoriously
difficult to find. There are no team salary statistics, nor are
there financial statements. We use attendance as a measure
of financial performance, in the absence of other potential
sources of financial information. However, we also consider
team histories to see if there is evidence of poor management
or financial performance to go along with attendance data.
We hypothesize that financial turmoil arising first during the
Depression and then during World War II sowed the seeds
for some teams to innovate and turn to integration to tap that
‘greatest font of talent. We use both relative attendances in
the 1930s until integration, and also team histories to see if
there is a developing pattern to the pace of integration. The
pattern of significant differences in attendance matching is
well established in the 1930s, and the underperforming team
in fact would ultimately integrate first. We note that the
1930s data predicts that the Phillies would integrate first, and
that the Phillies were stopped by baseball authorities from
doing so. (See Table 1).
World War II
Right after the 1941 Pearl Harbor attack, Commissioner
Landis petitioned the government for approval of the 1942
season. In what has become known as the “Green Light
Letter”, President Roosevelt replied in part “I honestly feel
that it would be best for the country to keep baseball going.
There will be fewer people unemployed and everybody will
work longer hours and harder than ever before. And that
means that they ought to have a chance for recreation and for
taking their minds off their work even more than before.”
Interestingly the President foresaw that “Even if the actual
quality to the teams is lowered by the greater use of older
players, this will not dampen the popularity of the sport”. A
chance to nudge baseball towards integration was bypassed
over the exigencies of the war effort.
But the draft took most white men into the service, leaving a
pool of players too young for military service (including one
15 year old pitching appearance), players who failed to
qualify for military service and players who were too old to
serve and past their baseball primes. These inconsistencies
intersected in with the sacrifices made by black troops during
the war. President Truman made the decision to integrate the
military, which made segregation impossible to keep as a
policy. The 1945 death of Commissioner Landis opened the
door to integration.
Model of Integration
With the door open, we still see an uneven pace of
integration. We posit that competition within markets led the
underperforming team to integrate first. Brooklyn Dodger
GM Branch Rickey famously said that the greatest font of
untapped talent was in the Negro Leagues, and that “The
Negroes will make us winners for years to come, and for that
I will bear being called a do-gooder…” We believe that this
impulse, the need to win and attract fans, led to integration.
We compare both financial histories and relative attendance
records.
Table 2 shows when each team played its first African
American player. We note that of the first seven teams to
integrate, five were in trouble in the 1930s and each of those
teams had changed cities within 10 years or fewer of
integration (Browns 1953; Braves 1953; Athletics 1956;
Dodgers, 1957; Giants 1957).
Team Histories and Relative Attendance
We consider cumulative attendance, from 1930 until
integration (as measured above). The charts below are
measured from 1930 (year 1) forward. In each case, we find
that the team that is underperforming at the box office, and
has a weak financial position, integrated first. We also
consider the team’s history to validate if it in fact had
financial troubles leading up to integration.
Philadelphia Phillies (Attempt to integrate, 1944)
The Phillies franchise foundered from the Depression into the
WWII years. By the early-1930s, deferred maintenance at
the Phillies home field the Baker Bowl was such that “Rather
than use lawn mowers, groundskeepers used goats instead”
(Gershman, p. 144). Into the 1930s, there was considerable
“agitation about the state of the Philadelphia facility in
general and the franchise in general” (Dewey and Acella, p.
81
462). National League owners “had enough votes to force
change in the ownership”, and the team went into new hands
by 1933. In 1938 the Baker Bowl closed, the team moved to
the Athletics home grounds Shibe Park and attendance fell
further. (Dewey and Acella, p. 463).
National League ownership again pressured the Phillies, and
in 1943 owner Nugent seemingly agreed to sell to Bill Veeck,
the impresario and then-owner of the, AAA Milwaukee
Brewers. Veeck made the mistake of informing
Commissioner Landis that he would stock the team with stars
from the Negro Leagues, and the sale was voted down. The
League agreed to a low bid from Mr. Cox, a local business
owner. (Eig, p. 182). Veeck followed through in July 1947
when, as owner of the Cleveland Indians, broke the American
Leagues’ color line with Hall of Fame player Larry Doby,
just months after Jackie Robinson’s introduction to the
National League.
Brooklyn Dodgers (Integrated 1947)
Brooklyn last showed a profit in 1930, and the phones at
times were shut off for non-payment (Dewey and Acella, p.
112). When Walter O’Malley become involved in the team
in 1938, “the Dodger situation...was so frightfully tangled
that no commentator was quite able to explain it. (Robert E.
Murphy, p. 44). Basically, the Dodgers had less money than
the New York Yankees, who performed strongly both on the
field and in the box office. “If the Dodgers wanted a winning
team, he had to tap new talent” (Kuper and Szymanski, p.
95).
St. Louis Browns (Integrated 1947)
St. Louis Brown attendance was the lowest in all of baseball
in the 1930s, and didn’t crack 100k in three separate years.
Grass maintenance was turned over to a goat. (Golenbock,
The Spirit of St. Louis, p. 322). In 1933, when owner Ball
died “No one wanted to buy the team” (Golenbock, p 270).
By 1941, owner Don Barnes ‘concluded he could not
successfully fight the Cardinals” (p 280). The team began
discussion to move to Los Angeles (Dewey and Acella,
p.557). Those plans were interrupted by WWII.
By 1947 when Sportsman’s Park was sold, the park was
‘badly run down, with a leaky roof, broken chairs and a
dilapidated clubhouses…despite a fix up, the Browns drew
only flies” (Golenbock, p. 320). In 1947 the Browns
integrated, just weeks after the Indians.
Boston Braves (Integrated 1948)
In the 1930s the Boston Braves began to sink under its debts
to what was “a desperate situation” (Dewey and Acella,
p.62). In 1933, the team was in financial distress which
ultimately led to a financial restructuring under league
auspices ((William Craig, p. 85).
New York Giants (Integrated 1948)
By the end of the 1930s, the New York Giants were ‘treading
water more seriously than any time…since the start of the
century…the team was beset by uncertainties about the
managerial abilities of the son” (Dewey & Acella, p. 383).
Chicago White Sox (Integrated 1951)
White Sox ownership struggles in the 1930s turned into
“years of wrangles that made the club’s efforts to stay above
water in the standings of almost secondary importance”
(Dewey and Acella, p. 168) This included the team’s
bankers determining that it was a ‘financial risk” in 1940.
This situation was in stark contrast with cross-town rival
Cubs, who were backed by Wrigley family money.
Philadelphia Athletics (Integrated 1953)
In this instance, we consider the post-1943 sale of the
Phillies. As noted, in 1943 the National League forced the
sale of the Phillies to new ownership, ultimately backed by
money from the DuPont Company. With DuPont money
behind the National League squad, the two Philadelphia
teams began to undergo a metamorphosis. Ownership
squabbles “erupted in the boardroom” (Dewey & Acella, p.
485) in 1949. The Mack and Shibe families constantly
fought. A 1950 deal which saw the team go into the hands
of the Macks ‘insured that the franchise would continue to be
run on a shoe string, never sure how the next payroll would
be met” (Jordan, p. 171).
CONCLUSIONS
We believe that the available evidence is that teams which
are poor financially and are doing poorly at the box office
were faced with an incentive to integrate. Once management
decided to integrate, co-worker integration is mooted. And
while consumer discrimination may remain a problem, the
data show that teams which integrated first performed better
on the field and were able to tap into a growing African
American consumer market. We believe our study helps
answer Hanssen’s question about “why do teams integrate”.
We believe it is competitive economic pressures.
82
Table 1: Attendance for Team Pairings from 1930 To 1939
Phillies Athletics Braves Red Sox
2,289,666 4,086,775 3,894,275 5,070,229
Browns Cardinals White Sox Cubs
1,393,827 4,054,118 4,109,937 8,791,668
Giants Yankees
7,516,744 9,089,9536,554,663
Boston
St. Louis Chicago
Brooklyn Dodgers
New York
Philadelphia
Table 2: Player Integration Dates by Team
Player Team Date Player Team Date
Jackie Robinson Dodgers 4/15/1947 Elston Howard Yankees 4/14/1955
Hank Thompson Browns 7/17/1947 Tom Alston Cardinals 4/13/1954
Monte Irvin Giants 7/8/1949 Elston Howard Yankees 4/14/1955
Sam Jethroe Braves 4/18/1950 Pumpsie Green Red Sox 7/21/1959
Minnie Minoso White Sox 5/1/1951 Ernie Banks Cubs 9/17/1953
Bob Trice Athletics 9/13/1953 John Kennedy Phillies 4/22/1957
Larry Doby Indians 7/5/1947 Carlos Paula Senators 9/6/1954
Curt Roberts Pirates 4/13/1954 Ozzie Virgil Tigers 6/6/1958
Nino Escalera Reds 4/17/1954
Figure 1: Philadelphia Phillies & Philadelphia Athletics Attendance (1930-1943)
Phillies: Dark, Athletics: Light
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Att
en
dan
ce
Years
83
Figure 2: New York Yankees & Brooklyn Dodgers Attendance (1930-1946)
Dodgers: Dark, Yankees: Light
Figure 3: Saint Louis Cardinals & Saint Louis Browns Attendance (1930-1946)
Browns: Dark, Cardinals: Light
0
2,000,000
4,000,000
6,000,000
8,000,000
10,000,000
12,000,000
14,000,000
16,000,000
18,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Att
en
dan
ce
Years
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
8,000,000
9,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Att
en
dan
ce
Years
84
Figure 4: Boston Red Sox & Boston Braves Attendance (1930-1947)
Braves: Dark, Red Sox: Light
Figure 5: New York Yankees & New York Giants Attendance (1930-1947)
Giants: Dark, Yankees: Light
0
2,000,000
4,000,000
6,000,000
8,000,000
10,000,000
12,000,000
14,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Att
en
dan
ce
Years
0
2,000,000
4,000,000
6,000,000
8,000,000
10,000,000
12,000,000
14,000,000
16,000,000
18,000,000
20,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Att
en
dan
ce
Years
85
Figure 6: Chicago White Sox & Chicago Cubs Attendance (1930-1950)
White Sox: Dark, Cubs: Light
Figure 7: Philadelphia Athletics & Philadelphia Phillies Attendance (1944-1952)
Athletics: Dark, Phillies: Light
0
2,000,000
4,000,000
6,000,000
8,000,000
10,000,000
12,000,000
14,000,000
16,000,000
18,000,000
20,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Att
en
dan
ce
Years
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
8,000,000
1 2 3 4 5 6 7 8 9
Att
en
dan
ce
Years
86
REFERENCES
Becker, Gary S. The Economics of Discrimination. 1971.
Chicago, IL: University of Chicago Press.
Bradbury, J. C. The Baseball Economist. 2007. New York,
NY: Dutton Press.
Chadwick, Bruce. When the Game Was Black and White.
1992. New York, NY: Abbeville Press.
Craig, William J. Boston Braves: A Time Gone By. 2012.
Charleston, SC: History Press.
Dewey, Donald and Nicholas Acocella. 2005. Total
Ballclubs. Wilmington, DE: Sport Classic Books.
Eig, Jonathan. Opening Day. 2008. New York, NY: Simon
& Shuster.
Gay, Timothy. Satch, Dizzy and Rapid Robert. 2011. New
York, NY: Simon & Shuster.
Gershman, Michael. Diamonds. 1993. Boston, MA:
Houghton Mifflin.
Golenbock, Peter. The Spirit of St. Louis. 2000. New York,
NY: Avon Books.
Hanssen, Andrew. The Costs of Discrimination: A Study of
Major League Baseball. The Southern Economic Journal,
vol. 64, no. 3, January 1998: 603-627.
Haupert, Michael J. The Economic History of Baseball.
http://eh.net/encyclopedia/article/haupert.mlb.
Jordan, David M. The Athletics of Philadelphia. 1999.
Jefferson, NC: McFarland & Co.
Kahn, Lawrence. Discrimination in Professional Sports: A
Survey of the Literature. Industrial and Labor Relations
Review, vol. 44, no. 3, April 1991. 395-418.
Kuper, Simon and Stefan Szymanski. Soccernomics. 2012.
New York, NY: Nation Books.
Murphy, Robert E. After Many a Summer. 2009. New
York, NY: Sterling Publishing.
Oh, Derek and Andrew Buck. Consumer Discrimination in
Baseball. Temple University Working Papers.
Overmyer, James. Queen of the Negro Leagues. 1993.
Lanham, MD: Scarecrow Press.
Tye, Larry. Satchel: The Life and Times of an American
Legend. New York, NY: Random House.
87
IMPACT OF MIGRANT REMITTANCES ON EDUCATION OUTCOMES
Tanu Kohli
LAS Global Studies, University of Illinois
03 S. Wright St, MC 301, Champaign, IL 61821
ABSTRACT
Remittances help recipient households to earn extra income
and increase their standards of living over time. These higher
standards are sometimes reflected in greater expenditures
appropriated to the education of each child in the household,
thereby promoting better human development outcomes for
the migrant-sending households. This paper evaluates the
impact of remittances on education expenditures made by
remittance receiving households, and compare these
outcomes with households that do not receive remittances.
The dataset used for this analysis is the 64th
Round of
National Sample Survey conducted by the Government of
India. It is seen that remittance incomes increase the share of
education related expenditures in the household and
education investments in each child.
Keywords: education expenses, schooling, remittances, India
I. INTRODUCTION
Economic migration allows households to expand their
income possibilities and achieve a higher standard of living.
Migration also acts as a tipping point for households to adopt
cultural practices that bring the source communities and
destination communities closer to each other. This paper is an
extension of the existing studies on migration, focusing on
the impact of remittance incomes on the investments in
human capital in the source community. If remittances are
devoted towards higher investments in human capital,
sustained contributions to education in the present will aid
the creation of a better, productive workforce in the future;
thus promoting the economic development of the country in
the long run. The current paper attempts to understand these
long-term human capital investments by studying the
schooling expenses made by remittance receiving households
as compared to non-remittance receiving households. The
dependent variable of interest is the schooling expenditure as
a share of total expenditure that is made by a given
household. The primary objective to see whether remittance
receiving household tend to invest more in child schooling as
compared to other households. The data utilized for this
study is the 64th
round of the National Sample Survey (NSS)
of the Government of India conducted in 2007-08. This data
stands as an outlier to the usual remittance-education studies
focusing on the Mexico-USA migration corridor. The results
from ordinary least square (OLS) analysis show that
remittance receipt has a positive impact on education
expenditures, thus leading to higher human capital outcomes
for these households. The results from instrumental variables
(IV) analysis are inconclusive and warrantee the use of better
instruments to deal with the problem of endogenous
variables. Rest of the paper is arranged as following: section
II presents the literature review; section III introduces the
hypothesis and the model, elaborating on the dataset and the
variables used for the analysis; section IV summarizes the
results of the OLS analysis; section V introduces the
instrumental variables (IVs) and presents the results and;
section VI concludes with data shortages and future work in
this direction.
II. LITERATURE REVIEW
In the past decade, there has been an increased focus on the
impact of migration and remittances on the schooling
outcomes of children in migrant sending households.
Empirical studies address these effects by observing different
parameters of education such as retention rates, academic
performance and gender-based differences in school
enrolments. Studies concentrating on the impact of migration
on schooling outcomes are seen to reflect an ambiguous
impact of parental migration on educational attainment such
as reduction in college aspirations but an increase in
educational aspirations and retention. On the other hand, the
literature concentrating on the impact of remittances on
schooling outcomes usually finds a positive impact of the
receipt of remittances on schooling outcomes. This
discrepancy arises in part because increased expenditure
towards schooling indicates the choices made by remittance
receiving households with respect to building future human
capital; but it is not indicative of the choices made by
households in the long run, with respect to the migration
aspirations of the children. The ambiguity is rendered from
the positive impact of remittances on the ability to make
educational investments (financial effect), combined with a
negative impact of parental absence on the academic progress
of the children (non-monetary effect). McKenzie and
Rapoport (2011) study the effect of Mexican migration to the
United States but find a negative effect of migration on
school attendance and high school completion rates. These
results are indicative of possibility that when a parent
migrates, the onus of taking care of the household falls on the
older child. McKenzie and Rapoport (2011) find support for
88
this argument observing higher workforce participation for
younger males, or a greater rate of migration of these and
greater participation in housework by female children.
Antman (2011) makes similar observations as with respect to
father’s migration from Mexico to the U.S. Her results show
a reduction in study hours and increase in work hours for
both boys and girls, with the effect being more severe on
younger boys and girls, ages 12 to 15 years. Meyerhoefer and
Chen (2011) focus on the impact of parental migration on the
schooling lags created for school children in rural China and
reveal a similar story. The migration of a parent from rural
area to urban area exhibits a greater likelihood of the female
child falling behind by more than half a year of schooling.
On the other hand, the literature focusing on the relationship
of remittances and schooling decisions reflect clear and
positive impact of former on the latter. Edwards and Ureta
(2003) study the impact of remittance incomes on schooling
in El Salvador and find a significantly large impact of
remittances on schooling retention rates. Remittance
receiving households exhibit a lower hazard of dropping out
of school. Acosta (2006) also reaches similar conclusions, as
the receipt of remittances increases the probability of
remittance receiving households to keep children enrolled in
school and reduce the labor force participation of children.
Amuedo-Dorantes and Pozo (2010) study the impact of
remittance receipt on school attendance in the Dominican
Republic by comparing the outcome for households that have
migrants with those that do not have migrants. Isolating the
effect of remittances on children in non-migrant sending
households, the authors predict better schooling outcomes for
these children, compared to the households where one of the
members undertook migration, thus leaving the child
susceptible to hardships. Returns to education and
expectations about future migration also affects school
attendance, with low returns to education in the destination
countries and higher expectation to migrate leading to lesser
school attendance. The positive financial impact of
remittances is thus overcome by the negative, non-monetary
impact of migration. Perhaps an outlier to this generally
positive impact of remittances on education expenses is the
study of Albanian households by Cattaneo (2012) finds that
remittance incomes do not have any impact on education
expenditures. The author attributes this non-preference of
education to spending conditions put forward by migrants to
send remittances and to low returns to education in the
Albanian labor market combined with the importance given
to other more urgent consumption expenditures by the
households.
Three broad conclusions that can be made from the studies
reviewed above. First, the impact of remittances on schooling
outcomes is still a less explored area even though the impact
of remittances on education is more or less predictable.
Second, most of the studies exploring the relationship
between remittances and schooling (or even migration and
schooling) are concentrated in exploring the Mexican
education outcomes, making studies for other countries
virtually negligible. Third, the studies focusing on
remittances and education outcomes focus on the receipt of
remittances and not the amount of remittances. Thus, within
the remittance receiving households, the magnitude to which
remittances effect schooling is not explored and can
definitely be utilized for comparisons of outcomes for
different remittance receiving households.
III. HYPOTHESES AND MODEL
In this paper, the primary dependent variable is the share of
total consumption expenditure devoted to schooling expenses
in the survey year; henceforth referred as share of schooling
expenses. It is measured as the ratio of the household’s
annual schooling expenses to the household’s annual
consumption expenditure. Share of schooling expenses
variable is indicative of the choice made by remittance
receiving households towards higher human capital
investments, as compared to non-remittance receiving
households. While most of the studies reviewed above focus
on schooling enrolment, this data set provides information of
grade of schooling completed only. At any given time thus,
the continued status of school enrolment is not known.
Hence, annual schooling expenses become the most plausible
tool to estimate a household’s preference for education.
Consequently, the following hypotheses is developed- do
remittance receiving households contribute a greater share
of annual consumption expenses towards schooling expenses
and thus, are more conducive to the creation of human
capital.The general form of the education expenditure model
can be summarized as below-
……………………… (1)
Here, the dependent variable is given by, log share of
schooling expenditure = [
] and,
the primary independent variable of interest , is
remittance receipt, a dummy variable which assumes value 1
if the household receives remittances and 0 if the household
does not receive remittances.
Other control variables include economic variables such as
employment status of the head of the household and
employment status of the adults in the household;
demographic variables capture the household and individual
characteristics; education variables measure the educational
attainment of adults in the household and; migration
variables study the strength of influence migrating members
of the household have on the household.
89
Economic variables include employment status of the head
of the household and employment status of adults in the
household. An employed household head can ensure
continued flow of income, enabling the household to spend
more money on tuition and school supplies and satisfying
schooling requirements of each child. If the head of the
household is employed and responds to the economic status
as self-employed or as working in a household enterprise or
as a regular salaried/ wage employee or reported to have
worked in casual wage labor, the dummy takes the value 1. If
however, the employment status of the household head
includes responses such as did not work but was available for
work, attended educational institution, attended domestic
duties, retirees and remittance recipients and disabled, they
were included as unemployed and their employment status is
coded as 0. Since many family structures in India can be
multi-generational families, the head of the household might
not always be employed. For example, the head of the
household can be a retired grandfather with working sons and
daughters. In order to account for this possibility, proportion
of employed adults in the household is also used as an
economic variable. This variable is calculated as the ratio of
number of employed adults to total adults in the household.
The higher the proportion of employed adults in the
households, greater is the assurance that schooling of the
children in the household will not be disrupted.
Demographic variables influence the consumption patterns
of a household via the caste the household belongs to, family
structure of the household, location of the household and;
female participation in household decisions. A household
residing in the rural area will spend lesser of schooling
because of two reasons. First, the concentration of schools is
generally lower in rural areas than in urban areas. Second,
rural areas tend to have more government-run schools that
are completely funded and do not require students to spend
anything extra. Meanwhile in the urban areas, private schools
exist along with government schools which tend to tip the
balance of schooling expenditure in favor of urban areas
further. If the household is in a rural area, the dummy
assumes the value 1, otherwise 0. The caste system in India
reflects the difference in economic opportunities among
households from the reserved castes and the general castes.
Since India has free and compulsory schooling for children
from ages 6 to 14 years, the caste of household will not affect
the enrolment of children in schools. Caste, via inherent
difference in economic opportunities can however, affect the
access to non-tuition education expenses on school supplies.
The household’s caste is thus included as a dummy variable
which equals 1 if the household belongs to any of the
reserved backward castes and 0 otherwise. The dummy
variable for multigenerational family (=1) or not (=0) is
expected to have a negative relationship with share of
schooling expenses, as schooling children can pool their
schooling resources and use them more efficiently. It is also
possible that the presence of a greater number of household
members diverts consumption to other kind of consumption
needs, thus reducing educational expenditures. To further
address decision making process in a multi-generation
family, where spending decisions can be influenced by more
than one parent or couple, a variable measuring the
proportion of adult females in the household is added. It is
calculated as the ratio of total adult women and total adults in
a household. If a larger proportion of adults are women, they
can influence spending decisions with greater bargaining
power. Additionally, three variables addressing the role of
children in the household are added to the analysis. Total
number of school going children is expected to increase the
share of schooling expenses; proportion of female children is
expected to be negatively related to the share of schooling
expenses due to the preference to educate a male child and;
ratio of total children to total members in the household is
expected to exhibit a positive relationship with share of
schooling expenses since the household comprises of more
children than adults, naturally tipping the expenditure in
favor of education expenses.
Education variables in the model are divided in two main
groups- maximum education level of the household and
proportion of educated adult females at the primary,
secondary and graduate levels. The household member who
has completed the highest level of education will influence a
household’s perspective towards education spending.
Maximum education is added in lieu of education level of the
parent in the household as a multigenerational family will
have more than one parent who can influence the spending
decisions of the household. The maximum educational
attainment is divided in three dummy categories- primary
education takes the value 1 if the maximum education
attained by any household member is the completion of
primary schooling (up to 8 years of schooling) otherwise 0;
secondary education takes the value 1 if the maximum
education attained by any household member is completion
of secondary school (9 to 12 years of schooling) otherwise 0
and; graduate education takes the value 1 if the maximum
education attained by any household member is the
completion of graduate or post-graduate education otherwise
0. The reference category for the maximum education
variable is given by no educational attainment or illiteracy of
adults. The second group of education variables account for
the proportion of women educated at each education level.
The education expenditure outcomes will be worse for a
household with a greater proportion of women who
completed primary education than with a household with
greater proportion of women who completed secondary
education or graduate education.
Lastly, the migration variables are used to measure the
strength of relationship between the migrant and the
remittance receiving household. The survey design allows
creating a migration history variable which measures the
average years the household has witnessed migration. This
90
variable is indicative of the changing preferences of a
household that has been exposed to more developed societies.
Since more developed communities also exhibit better human
capital, the effect of the migration history variable on human
capital expenditure in the source community should be
positive. The second migration variable is the proportion of
employed migrants in the household. If there are more
employed migrants, they will remit more money, which in
turn will have a positive impact on the spending abilities of
the household.
The descriptive statistics for all variables is given in Table 1.
The sample breakdown is provided according to remittance
receiving and non-remittance receiving households to get an
estimation of the characteristics of each kind of household.
IV. RESULTS FROM OLS ANALYSIS
Results from the OLS method are summarized in Table 2
with robust standard errors reported in the parentheses and
the results significant at the 1% level. Columns 1 and 2 of
Table 2 summarize the results using different combinations
of independent variables listed above. Variables excluded
from column 2 are employment status of the household head
which is replaced by the proportion of employed adults in the
household to capture the multi-generational household effect.
The multi-generational household dummy in column 1 is
replaced by proportion of adult women in the household in
column 2 to account for the multi-generational effect, as well
as measure the relative bargaining power of females versus
males in the household. Instead of using the total number of
school going children, used in column 1, the ratio of total
children in the household is used in column 2. This variable,
along with the proportion of female adults and proportion of
employed adults can account for the effects of a multi-
generational family. Other variables such as dummy for
remittance receiving household, social characteristics of the
household, proportion of female children in the household
and education and migration variables are included in both
the columns.
It is seen that remittance receipt has a consistently strong
impact on the share of schooling increasing the share of
consumption expenditure devoted to education expenditures
anywhere from 9.6% to 16.6%. Thus, the remittance
experience not only relives credit constraints but also
encourages households to invest in the human capital of the
children in the household in order to secure a better future.
Such tendency might come from the exposure of the migrant
to a better living environment, thus pushing the family left
behind to aspire for similar standards via long run human
capital investments. It is also possible that these households
already give importance to education and the extra income
helps them realize their education goals. Households with an
employed head (column 1) and a larger proportion of
working adults (column 2) seem to have a negative impact on
the two dependent variables by 17.9% and 35.7%
respectively. This result goes against conventional wisdom of
consistency of incomes and higher investments in education.
The proportion of employed migrants in a household also
exhibits a similar negative relationship with the dependent
variables. A greater number of employed migrants would
thus reduce share of education expenditure by approximately
6%. One plausible reason for such unexpected behavior of
the employment variable could be that households prefer that
the children get into the labor force as soon as possible,
instead of investing many years in obtaining education. Such
expectations could lead to lesser investment in schooling. It
is also plausible that if the migrant from the household did
not acquire higher education but is economically successful,
the household might not give importance to education as well
and groom the children to be economic agents instead. For
example, 17% of the households had no literate adult in the
household while 37% households had adults who completed
primary education. On the other hand, only 16.29% of the
household had adults with bachelor degree or higher. Thus,
the household can give more importance to entering the labor
force rather than obtain education.
Among the demographic variables, the coefficients behave in
the expected manner. Households in rural areas tend to spend
approximately 23.9% less on schooling expenses as
compared to a household residing in the urban area. This
difference in spending pattern can arise due to two reasons.
First, rural areas have more government sponsored schools
that do not require any additional investment on schooling or
schooling supplies from the parents. This shrinks the share of
schooling expense in rural areas compared to households in
urban areas where the children might go to private schools
and spend on their own books, extra tuition and other school
fees. Second, rural areas in general have lesser number of
schools, which along with a household’s requirement for
farm and non-farm labor can lead to lesser children enrolled
in schools and thus, lower education expenditure for rural
areas. Households from reserved backward castes devote
12% lesser consumption towards education expenditure than
a non-reserved caste household. As mentioned in section IV
above, this can be due to differences in economic
opportunities of the household because of being a lower caste
household instead of lack of access to education per se. A
multi-generation household also contributes 43% lesser to
education expenditure as compared to a nuclear household.
This however, does not imply that multi-generational
households assign lower importance to human capital. A
more likely explanation is that the household shares
education resources and thus has to spend lesser portion of
the consumption budget on school supplies. For example,
siblings can share school supplies, recycle the same books for
years before discarding them and the teach each other thus
eliminating the need for tutoring.
91
The variables related to children in the household behave
more or less as expected. A larger number of school-going
children in the household leads to a greater share of
consumption expenditure devoted to education expenses
(19.2%). In column 2, the total number of school going
children is replaced by the proportion of children in the
household and it is expected that a greater share of children
in the household will increase the share of consumption on
schooling expenditure. It is seen that with greater proportion
of children, the share of schooling expenditure increases by
almost 119%. Lastly, higher the number of female children in
the household lesser is the share of schooling expenditure
(average 11.8% lesser) in that household. This reflects the
preference for investing in the human capital of a male child
compared to a female child. The general opinion is that while
a male child will have to be the bread-winner of his family in
the future, a female child will be fine without work since she
can get married and secure her future.
Variables related to the education of adults in the household
also behave as expected of them. As the maximum education
obtained by any member in a household increases, the share
of schooling expenditure increased as well. Therefore, if
compared to household with illiterate adults, the share of
expenditure on education is 76.3% to 92.6% more for
households with maximum educational attainment at the
primary level; 130.3% to 149.4% for maximum educational
attainment at the secondary level and; 131.9% to 147% for
maximum educational attainment at the graduate level. As
the household’s education benchmark increases, they lay
higher premium on obtaining schooling for their children. It
is also seen that as more women acquire higher education in
the household, the education outcomes for the children in the
household also improve. Thus, while households where a
greater number of women completed secondary education
spend an average of 31.2% of their consumption expenditure
on education expenses while households with a greater
number of women with graduate education spend an average
of 40.4% of their consumption expenditure on education
expenses.
Migration history of a household has a small but positive
impact of 0.6% on share of education expenditures. Thus, the
exposure to a more developed society encourages households
to reach human capital outcomes similar to those societies.
However, the weak relationship shows that this variable
might not be a crucial determinant of education outcomes.
On the other hand, as the proportion of employed migrants
from a household increases, the schooling expenses incurred
by the household decreases by approximately 5.9%. If the
households see the economic benefits of migration, they
might substitute away from investing in schooling and push
children to become migrants, thus not requiring investments
in schooling. This negative impact of migration but positive
impact of remittances on education aspirations is similar to
the results of studies by Kandel and Kao (2001), Hanson and
Woodruff (2003) and Amuedo-Dorantes and Pozo (2010).
V. RESULTS FROM THE IV ANALYSIS
OLS estimates provide an extremely optimistic picture
regarding the effect of remittance receipt on the dependent
variables. This model however, suffers from potential
endogeneity issues. Remittance receipt will change the
consumption patterns and increase schooling investments, but
in some cases, where the migrant might be a close relative,
remittances might be received specifically to improve
schooling outcomes for the children (to pay for a tutor for a
poor performing child or to buy a computer). The IV analysis
is built on the results presented by OLS regression analysis in
column 1 of Table 2. The Durbin-Wu-Hausman test for
endogenous variables yields an F statistic greater than 10 and
p-value less than 0.05, thus requiring an IV analysis.
The first instrument is district-wise concentration of post
offices, obtained from the Indian Postal Services in the year
2007-08. Post offices allow easy transfer of monies and have
a deeper concentration of branches than commercial banks in
India. The second instrument is the state-wise and sector-
wise unemployment rate in the survey year 2007-08. While a
stronger network of post office will facilitate the transfer of
remittances, the unemployment rates, high unemployment at
the source will encourage the migrant to remit money to the
household. There is however, a strong possibility that high
unemployment will affect the household income, and thus
schooling expenses. To test that unemployment and the
dependent variable do not share a strong relationship, the
correlation between them is calculated. Unemployment and
share of schooling expenses exhibit a weak correlation, thus
allowing the use of these two instruments. The results from
these tests are summarized in Table 3 below. The
independent variables are the same as in Table 2, column 1
and include, remittance receipt (0/1 dummy), employment
status of the head of the household, rural or urban location of
the household, reserved caste status of the household, multi-
generational household and total children and proportion of
female children in the household. Education variable include
maximum education dummies with illiterate as the reference
category and proportion of educated females at each level of
schooling completed. Migration variables include average
years the household has witnessed migration and the
proportion of employed migrants in the household.
IV results show a positive impact of remittance receipt on
share of schooling expenses out of the total household
budget. That is, if the household receives remittances, it tends
to invest 141.5% more than non-remittance receiving
household towards share of education expenses. While this
positive relationship is encouraging, the value of the
coefficient is extremely high, which seems to raise some
concerns. Among other economic variables, an employed
92
household head is expected to devote 3.5% more towards
education expenses out of the total household budget, but this
value is insignificant. This result is expected, as an employed
head will be able to invest more in the educational attainment
of the children. The remaining variables do not behave
differently from the OLS results in Table 2. If the household
resides in a rural area, it will invest 27.5% lesser
consumption expenditure towards schooling and membership
in the reserved caste shows that households contribute 11.1%
lesser towards share of schooling expenses as compared to an
urban household and a non-reserved household respectively.
Multi-generational households spend a lesser portion of their
entire consumption expenditure on schooling expenses which
supports the previous assumption that such households might
be pooling resources and older children, might be helping
their younger siblings which reduces the need to spend more
on education expenses. As the number of school going
children increases, the share of schooling expenditure
increases by approximately 18%. A household where a
higher number of children are female, the share of schooling
expenses reduces by 12.8%.
As the years of completed schooling by a household
increases, the percentage share of consumption expenditure
on schooling expenses also increases. Thus, while households
where at least one adult completed primary school will spend
82.5% more on share of schooling expenses, than household
where none of the adults were educated; for households that
had at least one graduate the share of schooling expenses is
approximately 141% higher. Similarly, as the proportion of
women with completed primary, secondary and graduate
education increase, the share of schooling expenses of the
household increase. Therefore, a household with greater
proportion of female with graduate education would devote
35.9% of the consumption expenditure to schooling, as
compared to a household where the larger proportion of
women completed only secondary education. These
education variables present a picture similar to the
expectations that were set for them earlier in the paper. A
household with higher educational attainment will place
higher premium on schooling.
The two migration variables, migration history and
proportion of employed migrants do not change substantially
from the results reported in Table 2. As a household’s
average year of exposure to migration increases, it invests
approximately 1.1% share towards schooling. In order to test
the validity of this result, migration history of the household
is divided into three periods. If the household has had
exposure to migration in the last five years, the migration
history of the household is short, while medium term
exposure implies an average of five to 10 years since the
household sent a migrant. Breaking down this variable
provides a clearer picture of household’s schooling expenses.
It is seen that a household with short history of migration
would in fact reduce the share of schooling expenses as
compared to a household that has been exposed to migration
for a longer time. That is households with a long term
migration history spend 17.7% more on share of schooling
expenses as compared to a household with recent exposure to
migration. These results seem to indicate that as soon as a
migrant leaves the household, there is a disruption in the
household budget, which would affect the schooling
expenses as well (this disruption could occur if the household
had to divert resources from regular consumption towards
costs of migration). However, as the migrant settles at the
destination, and sends regular remittances, the share of
schooling expenses tend to increase. Antman (2011),
McKenzie and Rapoport (2011) and Meyerhoefer and Chen
(2011) find similar disruptions and reduction in schooling
attainment in migrant sending households. Additionally the
household, witnessing the benefits of migration (especially if
the migrant has high human capital), will tend to increase the
schooling investments of the current generation.
Post-estimation Tests- Over identification tests for the
instruments listed at the end of Table 3 show that the first
model for schooling expenses with share of schooling
expenses as the dependent variable is correctly identified by
using state-wise and sector-wise unemployment rates and
district-wise concentration of post offices as instruments. The
Sargan-Basmann scores are reported in column 1. For the
second model with schooling expenses per child however,
these instruments fail to correctly identify the model (column
2). This warranties the use of alternative instruments that can
better predict the impact of remittance receipt on schooling
outcomes.
VI. SUMMARY AND CONCLUSIONS
The primary objective of this paper was to observe if the
receipt of remittances by surveyed households leads to higher
investments in education in the household. A positive impact
of remittances on education expenditure would mean that not
only remittance incomes enable households to enjoy a higher
level of consumption, but also enable them to enjoy sustained
development by assisting the creation of higher human
capital of the children. The dependent variable chosen to
explore this impact of remittances was the share of schooling
expenditure out of total consumption expenditure. This
variable was chosen as a measure of educational attainment
in the household due to lack of data available on enrolment
rates. The results from the OLS analysis showed that
remittance receiving households devote more money towards
educational expenditure. The OLS model however, was seen
to suffer from endogeneity and to correct for this error, state-
wise sectoral unemployment rates and district-wise
concentration of post offices are introduced as instruments.
The model is recalculated and it is seen that remittance
receipt has a positive impact on the share of schooling
expenses. The model is correctly identified but the value of
both the coefficients is extremely high, which begs for
93
further investigation in terms of better instruments that can
provide more accurate results. Possible instruments could
include the district-level data on natural calamities such as
rainfall, as used by Munshi (2003) or droughts. Many studies
use destination community unemployment rates, but the lack
of data on migrant destination stops the use of this IV.
Table 1 - Descriptive statistics for schooling models
Variable Remittance receiving
households
Non- remittance receiving
households
Mean S.D. Mean S.D.
Annual schooling expenses 5011.67 10040.35 4345.40 10232.46
Share of schooling expenses
in total consumption
expenditure
0.0670 0.0769 0.0556 0.07115
Schooling expense per child 2989.19 6109.25 2650.34 7314.48
Remittance receipt 1 0 0 0
Amount of remittances 27102.13 49732.58 0 0
Employment status of the
head of the household
0.6601 0.4736 0.8582 0.3487
Proportion of employed
adults in the household
0.4722 0.3452 0.6017 0.2729
Rural household 0.6845 0.4647 0.6753 0.4682
Reserved caste 0.6550 0.4753 0.6745 0.4685
Multigenerational household 0.4295 0.4950 0.4750 0.4993
Sex of the head of the
household, male=1
0.6484 0.4774 0.8766 0.3287
Proportion of female adults in
the household
0.6282 0.2493 0.5053 0.1811
Number of school-aged
children (6 years to 17 years)
1.0490 1.3119 1.0450 1.3016
Proportion of female children
in the household
0.4680 03722 0.4582 0.3763
Ratio of total children 0.2913 0.2612 0.2492 0.2224
Maximum education
Primary schooling
Secondary schooling
Graduate education
0.4056
0.3145
0.1723
0.4910
0.4645
0.3776
0.4121
0.3126
0.1808
.4922
.4636
.3849
Education of female members
Primary schooling
Secondary schooling
Graduate education
0.2070
0.1632
0.0636
0.2868
0.3092
0.2084
0.1557
0.1370
0.0561
0.2075
0.2844
0.1948
Migration history 5.9871 5.8646 7.0317 7.0284
Proportion of employed
migrants
0.8398 0.2391 0.3982 0.4494
94
Table 2 - OLS estimates for share of schooling expenses
and schooling expense per child
(1) (2)
Economic Variables
Remittance receiving household 0.1667***
(0.0159)
0.0969***
(0.0166)
Employed household head -0.1791***
(0.1617)
--
Proportion of employed adults in the household -- -0.3579***
(0.0226)
Demographic Variables Household resides in rural area -0.2402***
(0.0153)
-0.2385***
(0.0156)
Household belongs to a reserved caste -0.1259***
(0.1427)
-0.1152***
(0.0146)
Household is multi-generational -0.4302
(0.0133)
--
Proportion of adult women in the household -- -0.1157***
(0.1926)
Total children of school age (6 to 17 years) 0.1928***
(0.0053)
--
Proportion of total children in the household -- 1.1980***
(0.0478)
Proportion of female children in the household -0.1221***
(0.0185)
-0.1157***
(0.0192)
Education Variables I- Maximum education dummies with illiterate as the reference category
Dummy for primary schooling as maximum education 0.7634***
(0.0748)
0.9267***
(0.0740)
Dummy for secondary schooling as maximum education 1.3035***
(0.0757)
1.4940***
(0.0752)
Dummy for graduate education as maximum education 1.3194***
(0.0789)
1.4701***
(0.0790)
Education Variables II- Proportion of educated adult females from each education group
Adult females with primary schooling 0.4102***
(0.0250)
0.2797***
(0.0268)
Adult females with secondary schooling 0.3680***
(0.0272)
0.2544***
(0.0285)
Adult females with graduate education 0.4680***
(0.0482)
0.3404***
(0.0510)
Migration Variables- Migration history of the household 0.0086***
(0.0009)
0.0044***
(0.0009)
Proportion of employed migrants from the household -0.0608***
(0.0182)
-0.0585***
(0.0187)
Number of observations 26436 26436
R-square 0.2171 0.1722
Standard errors are in the parenthesis; ***Significant at 1% level
95
Table 3 - IV estimates for share of schooling expenses and schooling expense per child using
unemployment and district-wise concentration of post offices as instruments
2SLS regressions
Economic variables
Remittance receiving household 1.4155***
(0.2026)
Employed household head 0.0358
(0.0390)
Demographic variables
Household resides in rural area -0.2754***
(0.0176)
Household belongs to a reserved caste -0.1117***
(0.0159)
Household is multi-generational -0.4570***
(0.0156)
Total children of school age (6 to 17 years) 0.1798***
(0.0060)
Proportion of female children in the household -0.1287***
(0.0196)
Education Variables I- Maximum education dummies with illiterate as the reference category Dummy for primary schooling as maximum education 0.8255***
(0.0696)
Dummy for secondary schooling as maximum education 1.3632***
(0.0705)
Dummy for graduate education as maximum education 1.4101***
(0.0751)
Education Variables II- Proportion of educated adult females from each education group
Adult females with primary schooling 0.2644***
(0.0372)
Adult females with secondary schooling 0.2586***
(0.0350)
Adult females with graduate education 0.3591***
(0.0555)
Migration variables
Migration history of the household 0.0114***
(0.0019)
Proportion of employed migrants from the household -0.8276
(0.1255)
Number of observations 26434
First stage correlation tests-
F- statistic 101.66
Prob > F 0.0000
Over-identification tests
Sargan score 2.4832
(p = 0.1151)
Basmann score 2.4818
(p = 0.1152)
Standard errors in the parentheses; *** Significant at 1% level; ** Significant at 5% level
96
REFERENCES
Acosta, Pablo. "Labor Supply, School Attendance, and
Remittances from International Migration: the Case of El
Salvador." World Bank Policy Research Working Paper 3903
(2006).
Amuedo-Dorantes, Catalina, and Susan Pozo. "Accounting
for Remittance and Migration Effects on Children’s
Schooling." World Development 38, no. 12 (2010): 1747-
1759.
Antman, Francisca M. "The Intergenerational Effects of
Paternal Migration on Schooling and Work: What Can We
Learn from Children's Time Allocations?."Journal of
Development Economics 96, no. 2 (2011): 200-208.
Becker, Gary S., Kevin M. Murphy, and Robert Tamura.
"Human Capital, Fertility, and Economic Growth." Journal of
Political Economy 98, no. 5 part 2 (1990): S12-S37.
Cattaneo, Cristina. "Migrants’ International Transfers and
Educational Expenditure: Empirical Evidence from Albania."
(2010).
Das, Saswati, and Diganta Mukherjee. "Role of Women in
Schooling and Child Labour Decision: the Case of Urban
Boys in India." Social Indicators Research 82, no. 3 (2007):
463-486.
Das, Saswati, and Diganta Mukherjee. "Measuring
Deprivation Due to Child Work and Child Labour: A Study
for Indian Children." Child Indicators Research 4, no. 3
(2011): 453-466.
Edwards, Alejandra Cox, and Manuelita Ureta. "International
Migration, Remittances, and Schooling: Evidence from El
Salvador." Journal of Development Economics 72, no. 2
(2003): 429-461.
Hanson, Gordon H., and Christopher Woodruff. "Emigration
and Educational Attainment in Mexico." Documento de
Trabajo del IR/PS. Disponible en http://irpshome. ucsd.
edu/faculty/gohanson/working_papers. htm (2003).
Kandel, William, and Grace Kao. "Impact of Temporary
Labor Migration on Mexican Children's Educational
Aspirations and Performance." International Migration
Review 35, no. (2001): 1205-1231.
Katz, Eliakim, and Oded Stark. Desired Fertility and
Migration in LDCs: Signing the Connection. Center for
Population Studies, Migration and Development Program,
Harvard University, 1985.
Katz, Eliakim, and Oded Stark. "Labor Migration and Risk
Aversion in Less Developed Countries." Journal of Labor
Economics (1986): 134-149.
McKenzie, David, and Hillel Rapoport. "Can Migration
Reduce Educational Attainment? Evidence from Mexico."
Journal of Population Economics 24, no. 4 (2011): 1331-
1358.
Meyerhoefer, Chad D., and C. J. Chen. "The Effect of
Parental Labor Migration on Children’s Educational Progress
in Rural China." Review of Economics of the Household 9,
no. 3 (2011): 379-396.
Munshi, Kaivan. "Networks in the modern economy:
Mexican migrants in the US labor market." The Quarterly
Journal of Economics 118, no. 2 (2003): 549-599.
Stark, Oded, and David E. Bloom. "The New Economics of
Labor Migration." The American Economic Review 75, no. 2
(1985): 173-178.
Stark, Oded, and J. Edward Taylor. "Relative Deprivation
and International Migration." Demography 26, no. 1 (1989):
1-14.
Stark, Oded, and Robert EB Lucas. "Migration, Remittances,
and the Family." Economic Development and Cultural
Change (1988): 465-481.
Yang, Dean. "International Migration, Remittances and
Household Investment: Evidence from Philippine Migrants’
Exchange Rate Shocks." The Economic Journal 118, no. 528
(2008):591-630.
Proceedings of the Pennsylvania Economic Association 97
MANUFACTURING PRODUCTIVITY IN PENNSYLVANIA
James A. Kurre
The Sam and Irene Black School of Business
Penn State Erie, The Behrend College
5101 Jordan Road
Erie, PA 16563-1400
ABSTRACT
Productivity is a crucial issue for any economy, a key
determinant of the standard of living of residents in an area.
And one thing about productivity is clear: it varies
dramatically across places and across industries. So how
does Pennsylvania and its metro areas fare in terms of
manufacturing productivity compared to the nation and other
areas? This paper documents how Pennsylvania’s
manufacturing sector stacks up—or doesn’t—compared to
other states and the nation for the manufacturing supersector
and for selected subindustries. And it further details the
productivity data for PA’s metro areas.
INTRODUCTION
“Whoever could make two ears of corn or two blades of grass
grow upon a spot of ground where only one grew before,
would deserve better of mankind, and do more essential
service to his country than the whole race of politicians put
together.”
-Jonathan Swift, 1667-1745
Three hundred years ago Jonathan Swift understood the
importance of productivity. Economists, too, value
productivity and have focused much research on the topic,
especially the productivity of labor. Spatially, most of the
work has been done at the national or international level, for
example Dall’erba et al (2005) and the NBER Productivity,
Innovation and Entrepreneurship program. The Organization
for Economic Co-Operation and Development (OECD)
publishes rankings of countries based on labor productivity,
and it is clear that productivity varies dramatically across
countries of the world. There has been some work at the sub-
national level, most of it at the state level in the United States
such as Carlino and Voith (1992), Primont and Domazlicky
(2005), Smoluk and Andrews (2005), and Iranzo and Peri
(2006).
This paper focuses specifically on one state—the state of
Pennsylvania. And it goes below the state level to explore
productivity in the state’s metro areas as well. If productivity
is crucial to progress—and to competition, it is important for
us to know how our state and its metro areas perform.
This paper is intended to be descriptive, to present some
aspects of the state of productivity in the state, without
attempts to explain why it is what it is.
PRODUCTIVITY CONCEPTS
Measuring Productivity
Most fundamentally, productivity is “some measure of output
per some measure of input.” We choose to measure
productivity as value added per hour worked by production
workers in manufacturing industries. In their study of
regional comparative advantage, Hill and Brennan (2000) use
a similar measure. We opt not to use a measure of the value
of goods sold, such as value of shipments, since that would
involve double-counting of inputs. As the Census Bureau
says: “Data for cost of materials and value of shipments
include varying amounts of duplication, especially at higher
levels of aggregation. This is because the products of one
establishment may be the materials of another. The value
added statistics avoid this duplication and are, for most
purposes, the best measure for comparing the relative
economic importance of industries and geographic areas.”
(U.S. Census, 2007B)
An economy that produces steel sheets, steel fabrication
(turning the steel sheets into fenders), and automobiles would
have a total value of shipments that double-counts the steel
fabrication and triple-counts the steel itself. Thus the value
of shipments is inflated compared to the true value
produced—the final product, the car. Using value added at
each step in the production process will avoid this problem.
And we focus on the productivity of a single input, labor,
rather than other factors of production. One reason why this
is appropriate is that labor costs account for the lion’s share
of costs for most businesses. And labor is the source of most
income for most Americans. Nationally, employee
compensation accounted for 63.4% of national income in
2007, compared with 12.2% for corporate profits, 8.8% for
proprietors’ income, 5.9% for net interest income, and 1.2%
for rental income. (U.S. BEA, 2013) And employment is a
key focus of government policy, both at the national and the
local level. This is certainly not to say that the other factors
of production aren’t important; we just choose to focus on
this specific factor.
Proceedings of the Pennsylvania Economic Association 98
For this paper we measure productivity using hours of work
by production workers as the denominator. It would have
been possible to calculate productivity as output per worker
instead of per hour of labor, but that would be a less accurate
measure since not all workers work full time. To the extent
that an area’s industries tend to use more part time workers or
to use overtime labor, their measures of labor input (hours of
work) may not correlate closely with employment. For
example, if area A’s firms only hire workers who work 20
hours a week, while area B’s firms only hire workers who
work 40 hours a week, area A would need twice as many
workers to produce the same output as B, although they use
the same amount of labor input (hours of labor). Clearly, use
of output per worker data to measure productivity could be
misleading. For that reason, we elect to use output per hour
as our measure of productivity. Choice of the “per hour”
variable constrains us to use production workers rather than
total employment in this analysis, since hours of work are
only available for production workers.
In fact, measures of productivity per worker and productivity
per hour are highly correlated across metro areas. Kurre
(2004) found a correlation of 0.991 across 327 metro areas in
the 1997 Economic Census, and similar results obtained for
both the “hours” and the “workers” measures in that
regression analysis, suggesting that either approach is
acceptable. 2002 Economic Census data yielded a
correlation coefficient of 0.994 for value added per
production worker and value added per production worker
hour across 273 metro areas (Kurre and Miseta, 2008). And
data from 2007 for the manufacturing supersector (Brunot
and Kurre, 2012) yielded a correlation of 0.995 between
productivity per production worker hour and per production
worker across the 332 Metropolitan Statistical Areas (MSAs)
for which there are data for both. Further examination of the
2007 Economic Census data show that there is a correlation
of 0.973 between value added per worker (including all
employees) and value added per production worker across
those 332 MSAs. And at the state level (including the
District of Columbia as a state), the correlation is 0.960
between productivity per worker and productivity per
production worker, 0.942 between productivity per worker
and productivity per production worker hour, and 0.994
between productivity per production worker and productivity
per production worker hour.
Although this suggests that “employment” or “production
workers” may be an acceptable proxy in the denominator of
the productivity measure, we prefer to use “hours worked”
since that is more appropriate on a theoretical level.
The current paper focuses on productivity for the state of
Pennsylvania as well as its metro areas, so it is necessary to
have a data source that provides geographical data below the
national level. Hammill (2002) explored possible data
sources for metro-level productivity, concluding that the
Geographic Area Series of the Economic Census is the
preferred source. It provides data for all metro areas for a
single point in time from a single data source, permitting the
kind of cross-sectional study that we wish to do. We note
that the Economic Census also provides data at the state and
national levels, which is crucial for purposes of comparison.
Further, the Census Bureau has a well-earned reputation for
the quality of its data, and this is an important factor for any
database.
Of course, the Economic Census is not a perfect database.
We would prefer to have data for as recent a period as
possible, but the Economic Censuses are taken only every
five years. As of this writing, the most recent data are for
2007 and the manufacturing portion of the Geographic Area
Series were released in early 2010, a lag of over two years
from the data point. 2007 was six long years ago, and the
world is surely different now than it was in 2007. The Great
Recession has come and (mostly) gone since then, changing a
lot in our economies. Data for the 2012 Economic Census
are being collected currently, but it will be several years
before those data become available. So 2007 data are what
we have currently. But what the Economic Census lacks in
timeliness, it makes up for in data coverage and detail.
Specifically, it is one of the few sources with input and
output data by industry at the metropolitan level, and that’s
important if we want to know about the Pennsylvania
economy and its metro areas.
Productivity Data
We chose the Economic Census as the source for the key data
for this study since it gives value added and production
worker hours, for detailed manufacturing industries, by metro
area. But as with all spatial and industrial data, there are
tradeoffs between coverage and detail.
Geographically, we selected Metropolitan Statistical Areas
(MSAs) as one unit of analysis since metro areas are defined
to be small economies—actually, labor markets. Not all
MSAs are independent economies; there is undoubtedly some
interdependence among MSAs that are adjacent to other
MSAs--which is the reason for designation of Combined
Statistical Areas (CSAs), after all. However, MSAs are
generally more logical geographical units for economic
analysis than counties (one level down the geographical
spectrum) or states (one level up), both of which use
historical political boundaries that often do not reflect current
economic forces—although they are certainly relevant for
policy purposes. The Economic Census presents data at both
the state and MSA level, making it a logical choice as a data
source.
The 2007 Economic Census uses MSA definitions as of
December 2006 from the Office of Management and Budget
Proceedings of the Pennsylvania Economic Association 99
(Executive Office of the President, 2006). All the PA MSAs
were included in the 2007 Census of Manufactures, as well as
data for the state as a whole. Not all the state’s MSAs had
data for all industries; some industries simply do not exist in
some MSAs. And when an MSA had a small number of
firms in one industry, the data were not disclosed. That led
to varying numbers of observations for different industries.
While the cost of living clearly varies across metro areas, we
do not attempt to adjust the nominal productivity data with a
spatial price index. The argument can be made that
manufacturing firms sell primarily to a national or
international market and thus these firms must compete on
the basis of nominal price—regardless of the local costs of
local inputs. Of course, this implies that low cost-of-living
places may have an advantage as a location for national-
market goods that have low transportation costs.
WHAT THE DATA SAY
The State Level
Productivity, measured as value added per production worker
hour, averaged $126.12 in the U.S. manufacturing sector in
2007. But Table 1 shows that it varied widely across the
states, from more than double the national average in
Louisiana ($259.90) to just two-thirds of the average in
neighboring Arkansas ($84.71). The most productive state
had a value more than three times that of the least productive.
In their 1992 study at the state level Carlino and Voith noted
that the most productive state’s aggregate productivity was
2.3 times that of the least productive state’s, and the range
was only 1.2 for manufacturing. Although their method for
measuring productivity was quite different from that used in
this study, their productivity index varied significantly less
across states than the measure used in this study.
So how is Pennsylvania doing? Unfortunately, the state’s
productivity performance is a bit disappointing. Despite its
strong manufacturing tradition, Pennsylvania’s productivity
is $118.13, 6.3% below the national average, and ranking #27
of the 51 “states”. Of PA’s neighbors, Delaware, New
Jersey, Maryland, and New York all beat the national
average, while Ohio and West Virginia did worse than
Pennsylvania. While Delaware’s productivity is about 54%
higher than PA’s, West Virginia’s is about 10% less.
The Metro Level
Productivity varies even more widely at the metro level.
Table 2 shows productivity in the sixteen metro areas that lie
wholly or partially in Pennsylvania, as well as in selected
other MSAs. Perhaps the most notable fact from Table 2 is
the extremely wide range of metro productivity values, from
$637 an hour in Cheyenne WY to just $43 an hour in El
Centro CA. This 15-fold range across MSAs is much wider
than across the states.
Only four of PA’s metro areas surpass the national average
for productivity, with Philadelphia at the forefront beating the
national average by nearly 25%. Laggard Johnstown’s
productivity is only 51% of the national average, and only
41% of Philly’s. More detail about productivity at the metro
level will be presented in a later section. But first, it’s
necessary to discuss industry disaggregation.
The Industry Level
Of course, a key question is “why?” The answer must start
with the fact that “manufacturing” is not the same in all of
these areas. As mentioned above, not all industries exist in
all areas. “Manufacturing” is hardly a homogeneous industry.
In Erie it includes locomotives and plastics, but in Pittsburgh
it is robotics and medical devices. Of course, these
individual industries can all have very different levels of
productivity. This means we need to disaggregate the
manufacturing supersector to see what industries comprise
“manufacturing” in each area.
But does productivity vary much across industries? If not,
perhaps this is a moot question. Table 3 shows the range of
productivity across 3-digit NAICS industries for the U.S. and
Pennsylvania. In fact, the range across industries is even
greater that it is across metro areas—from a low of $48.86 in
apparel manufacturing (NAICS 315) to $828.16 in petroleum
and coal products (324), a 17-fold range. Clearly, the
industry mix of any economy is going to have a major impact
on its overall manufacturing productivity.
At the 3-digit level, Pennsylvania’s productivity patterns
follow those of the nation for the most part; their correlation
coefficient is 0.85. But that leaves room for some significant
differences, too. The range across 3-digit industries is from
$38.64 in leather products (316) to $409.66 in chemical
manufacturing (325), an 11-fold range—compared with the
17-fold range nationally.
On the high end, PA’s productivity exceeds national levels in
ten of the 21 3-digit industries. Relative to the national
industry levels, PA’s most productive industry is primary
metal manufacturing (NAICS 331), with a productivity that is
about 29% higher than the national level in that industry.
This is followed by food manufacturing (311), with a 25%
premium, and textile mills (313), with a 22% premium. But
the PA industry with the highest absolute level of
productivity, in dollar terms, is chemical products (325). At
$409.66 per hour, it is 3½ times the average PA productivity,
and is about 6% higher than the level of the industry in the
nation as a whole. Clearly a winner for the state!
Proceedings of the Pennsylvania Economic Association 100
On the other end of the scale relative to the U.S., PA’s
productivity in petroleum and coal products (324) is only
39% of the national level. But since this is the industry with
the highest productivity at the national level, PA’s value still
registers $324.46 per hour, about 2.7 times the state average,
despite the 61% reduction from this industry’s national
average. Relative to its competition in other states, it appears
that this industry doesn’t fare so well, but compared to other
industries in the state, it makes a major contribution to
incomes.
More of the story is revealed if we delve further into industry
disaggregation. The striking difference between U.S. and PA
productivity levels in petroleum/coal products (324) is
explained partially by digging down to the 5-digit NAICS
level. Table 4 shows the 5-digit components of that industry
and its sister industry, chemicals (325). It is clear that there
is some dramatic variation within these 3-digit industries. At
the national level, within NAICS 324 the productivity of the
petroleum refining industry (32411) is more than six times as
high as that of the paving and roofing materials industry
(32412), and more than three times as high as the remainder
of the 3-digit industry (32419). In productivity terms, these
industries are clearly quite different. Similarly in the
chemicals industry (325), petrochemical manufacturing
(32511) with productivity at a phenomenal $2,144 per hour is
3.7 times the next closest industry, pharmaceutical and
medicine manufacturing (32541) which itself has a notably
high productivity at over $582 per hour. Both of these are far
greater than the $97 per hour of the lowly synthetic fibers and
filaments industry (52522). Clearly, the 3-digit industries are
far from homogeneous when it comes to productivity.
All of this implies that industrial productivity analysis should
be done at the greatest level of industrial detail possible. But
there’s the rub. The greater the level of industry detail, the
more likely it is that the data will be unavailable due to the
legislation-imposed confidentiality responsibilities of the
Census Bureau. Table 5 shows the actual number of NAICS
industries in each level of industrial detail, as well as the
number of industries for which productivity data are
available at the national and state level. At the 3-digit level,
data are available for all 21 industries for both the nation and
PA. But once we get down to the 6-digit level, even at the
national level it is necessary to conceal data for a few
industries. And for the state, data are available for only 277
of the 472 officially defined 6-digit industries. It appears that
105 of the missing values are due to nondisclosure of the data
by the Census Bureau, and the other 90 are cases where the
industry simply doesn’t exist in the state. If we go to smaller
levels of geography, such as metro areas, the nondisclosure
problem becomes considerably more problematic, as we’ll
see below. At the micropolitan or county level, it is rare to
have good data for detailed industries. This clearly makes
the analysis more challenging.
Productivity by Metro Area, for Disaggregated Industries
As mentioned above, Pennsylvania’s metro areas are a
diverse lot, with quite different industry mixes. Table 6
shows productivity data for each MSA at the 3-digit level, as
well as for the manufacturing supersector. The number of
empty cells in this table clearly shows how difficult it is to
get comprehensive metro data that are industrially detailed,
even at the relatively aggregated 3-digit level. Of the 21 3-
digit industries, Lancaster had data for 12 industries and
Scranton for 11. All the other MSAs had data for
significantly fewer industries. And none of the PA MSAs
had data for four of the 21 3-digit industries.
At the supersector level, only four of the sixteen metro areas
had higher productivity than the U.S. average. Philadelphia’s
$157 an hour was 24% higher than the national average.
Allentown was 15% higher than the U.S. average, Pittsburgh
13%, and New York 12%. All of the other twelve had less
than average productivity, with Johnstown pulling up the rear
with productivity 49% less than the nation’s.
Given this, it is not surprising that most of PA’s metro areas
had lower productivity than the national average in each
industry, too. But in food manufacturing (211), five of the
MSAs surpassed the industry average. In fact Harrisburg’s
$198 an hour was 87% higher than the national average for
that industry. Another notable PA industry is machinery
manufacturing (333) where four of nine PA MSAs had higher
than average productivity, with Lebanon’s productivity 77%
higher than the nation’s. Lebanon also surpassed the national
average in fabricated metal products (332), where it had more
than double the national average productivity. And given its
primary industrial cluster, perhaps it is not surprising that
State College surpassed the national average in the printing
and related support activities (323) by nearly 90%. Erie,
unfortunately, did not surpass the national average in any of
the seven industries for which data are available.
Which Industries Are Pennsylvania’s Productivity Stars?
Table 7 shows industries that have productivity higher than
the state’s average of $118.13 per production worker hour in
manufacturing. The industry with the highest value is
biological product manufacturing (325414) at $786.64 per
hour. There were twenty establishments in this industry
employing 2,533 production workers in Pennsylvania in
2007, and they generated over $4.4 billion of value added.
Six other industries from the chemical products (325)
category are in the top eleven highest productivity six-digit
industries. So it’s no surprise that chemical products (325)
tops the three-digit list with a productivity of $409.66.
Chemical products industries also make up three of PA’s top
eight four-digits, and six of the top ten five-digits.
Proceedings of the Pennsylvania Economic Association 101
In 2007 the state had 624 establishments in the chemical
products (325) category, employing over 20,000 production
workers who produced over $17.3 billion of value added.
That made it the largest 3-digit industry in the state in terms
of value added, although food manufacturing (311) had over
2.5 times as many production workers and fabricated metals
(332) had more than 3.3 times as many.
Does PA Specialize in High Productivity Industries?
One question that arises from these data is whether the state
specializes in the manufacturing industries where it has high
productivity.
Table 8 shows productivity per production worker hour in 3-
digit manufacturing industries in PA and the U.S., and
relative to U.S. productivity in each industry. The final
column shows the manufacturing location quotient (LQ) for
each industry. This is calculated as the industry’s percent of
total manufacturing production worker employment in PA
compared to its percent in the U.S.:
LQi = % of PA’s mfg prdn workers in industry i . (1)
% of U.S.’s mfg prdn workers in industry i
An LQ greater than one implies that PA has a greater share of
its manufacturing activity in this industry than does the
nation, and is one of the state’s manufacturing specialties.
Eight of the 21 3-digit industries fall into this category, led
by primary metal manufacturing (331) which has nearly
double the national concentration.
The question above asks about “high productivity”, but this
can be interpreted in two ways. First, there is the absolute
sense; which industries have productivity values greater than
the PA average? Table 9 presents the same data as in the
previous table, but sorted by PA productivity, in dollars. Of
the seven industries with productivity above the PA average,
only three (43%) have LQs greater than one. Of the fourteen
industries with less than average productivity, five (36%)
have LQ’s greater than one. Alternatively, of the eight
industries for which PA has an LQ greater than one, only
three (38%) have productivity greater than the average. Of
the thirteen industries for which PA has an LQ less than one,
four (31%) have productivity greater than the average. These
numbers imply a slightly greater than average concentration
in higher productivity industries. In fact, the LQ for the
seven higher-than-average productivity industries taken
together is 1.05 (with 33.0% of manufacturing production
workers in PA, and 31.3% in the U.S. in these seven
industries.)
So the answer to the question posed above is “yes”, but only
a little bit.
A second way to answer this question would be to look at
industries in which PA’s productivity is higher than the
nation’s. Table 10 presents the same data as the previous two
tables, but this time ranked by PA’s productivity as a percent
of U.S. in each industry. Of the ten industries which had
higher productivity than the nation in their respective
industries, four of them (40%) had LQs greater than one. For
the eleven industries which had lower than national
productivity, we also find that four (36%) had LQs greater
than one. As with the absolute productivity measure, these
numbers imply a slightly greater than average concentration
in higher productivity industries. In fact, the LQ for the ten
higher-than-national-average productivity industries taken
together is 1.07 (with 47.2% of manufacturing production
workers in PA, and 44.3% in US in these ten industries.)
Again the answer to the question posed above is “yes”,
Pennsylvania does tend to specialize in its high relative
productivity industries, but only a bit.
CONCLUSIONS
This paper has intentionally been descriptive, and did not
seek to explain high or low productivity levels. It has
presented productivity data, in the form of value added per
production worker hour, for the state of Pennsylvania, both
for the manufacturing supersector and for various levels of
industry disaggregation. It also presented data for
Pennsylvania metro areas, both at the supersector level and
for 3-digit industries.
Some key findings are:
► Productivity varies dramatically across states and metro
areas of the nation.
► Productivity varies dramatically across industries within
manufacturing.
► Because of this, it is necessary to perform productivity
analysis at the greatest level of industry detail possible.
► Pennsylvania is “middle of the pack” compared to other
states in manufacturing productivity. Many of our near
neighbors are doing better, some significantly better.
► A handful of PA’s metro areas have higher productivity
than the national average, but ¾ of them lag the national
average, by significant amounts in some cases.
► PA and its metro areas are also home to some industries
with quite high levels of productivity, both in absolute and
relative terms.
► Pennsylvania tends to specialize in its high productivity
industries, but only to a small extent.
Proceedings of the Pennsylvania Economic Association 102
► Data problems become more severe the greater the level
of industrial detail, and the finer the level of geographical
detail.
Proceedings of the Pennsylvania Economic Association 103
Table 1
Manufacturing Productivity* by State, 2007
Table 2
Manufacturing Productivity* for PA and Selected Other Metro Areas, 2007
*Value added per production worker hour, in dollars.
Rank State Productivity % of U.S. Rank State Productivity % of U.S.
1 Louisiana $259.90 206.1 26 Illinois 118.44 93.9
2 Wyoming 210.60 167.0 27 Pennsylvania 118.13 93.7
3 Delaware 181.46 143.9 28 Ohio 115.45 91.5
4 Hawaii 172.31 136.6 29 Kentucky 113.58 90.1
5 Texas 171.65 136.1 30 Minnesota 112.92 89.5
6 New Mexico 158.77 125.9 31 Michigan 112.41 89.1
7 Oregon 154.11 122.2 32 Tennessee 112.10 88.9
8 Arizona 153.74 121.9 33 Oklahoma 110.35 87.5
9 Washington 153.14 121.4 34 Kansas 109.42 86.8
10 Connecticut 148.06 117.4 35 Wisconsin 107.80 85.5
11 Montana 146.44 116.1 36 North Dakota 107.00 84.8
12 Massachusetts 141.71 112.4 37 West Virginia 105.97 84.0
13 New Jersey 140.28 111.2 38 South Carolina 104.03 82.5
14 North Carolina 138.44 109.8 39 Maine 102.85 81.6
15 California 137.40 108.9 40 Georgia 101.11 80.2
16 Maryland 136.32 108.1 41 Alabama 100.84 80.0
17 Nevada 130.85 103.8 42 New Hampshire 95.18 75.5
18 Colorado 130.42 103.4 43 Idaho 94.45 74.9
19 Indiana 129.98 103.1 44 South Dakota 93.90 74.5
20 New York 127.69 101.2 45 Rhode Island 93.71 74.3
UNITED STATES 126.12 100.0 46 Alaska 92.93 73.7
21 Iowa 125.69 99.7 47 Mississippi 90.84 72.0
22 Utah 123.51 97.9 48 Vermont 90.07 71.4
23 Missouri 121.45 96.3 49 District of Columbia 86.76 68.8
24 Florida 120.99 95.9 50 Nebraska 85.09 67.5
25 Virginia 120.66 95.7 51 Arkansas 84.71 67.2
*Value added per production worker hour, in dollars.
Rank of
16 PA
MSAs
Rank of
332
MSAs
Metro Area Productivity % of U.S.
1 Cheyenne, WY $637.64 505.6
2 Lake Charles, LA 455.52 361.2
3 Alexandria, LA 426.35 338.1
1 57 Philadelphia-Camden-Wilmington, PA-NJ-DE-MD 157.41 124.8
2 75 Allentown-Bethlehem-Easton, PA-NJ 145.52 115.4
3 83 Pittsburgh, PA 142.18 112.7
4 85 New York-Northern New Jersey-Long Island, NY-NJ-PA 140.58 111.5
UNITED STATES 126.12 100.0
5 120 Scranton--Wilkes-Barre, PA 121.39 96.3
6 124 Williamsport, PA 119.79 95.0
7 128 Harrisburg-Carlisle, PA 118.96 94.3
8 129 Lancaster, PA 118.49 94.0
Pennsylvania 118.13 93.7
9 130 York-Hanover, PA 118.00 93.6
10 191 State College, PA 101.18 80.2
11 208 Altoona, PA 96.79 76.7
12 258 Youngstown-Warren-Boardman, OH-PA 86.49 68.6
13 264 Erie, PA 84.35 66.9
14 286 Lebanon, PA 78.36 62.1
15 308 Reading, PA 68.69 54.5
16 318 Johnstown, PA 64.26 50.9
330 Gadsden, AL 46.65 37.0
331 Kokomo, IN 46.48 36.9
332 El Centro, CA 42.58 33.8
Proceedings of the Pennsylvania Economic Association 104
Table 3
Manufacturing Productivity* for U.S. and PA, for 3-digit Manufacturing Industries, 2007
*Value added per production worker hour, in dollars.
Table 4
5-digit Detail for the Petroleum/Coal Products and Chemicals Industries
Table 5
Availability of Productivity Data by Level of Industry Detail
NAICS Industry US PA PA % of U.S.
31-33 Manufacturing $126.12 $118.13 93.7
311 Food manufacturing 105.60 131.45 124.5
312 Beverage and tobacco product manufacturing 458.20 279.08 60.9
313 Textile mills 60.86 73.95 121.5
314 Textile product mills 60.16 63.23 105.1
315 Apparel manufacturing 48.86 55.61 113.8
316 Leather and allied product manufacturing 53.91 38.64 71.7
321 Wood product manufacturing 49.18 48.16 97.9
322 Paper manufacturing 120.05 137.04 114.2
323 Printing and related support activities 71.94 72.16 100.3
324 Petroleum and coal products manufacturing 828.16 324.46 39.2
325 Chemical manufacturing 384.93 409.66 106.4
326 Plastics and rubber products manufacturing 73.75 73.99 100.3
327 Nonmetallic mineral product manufacturing 94.78 92.06 97.1
331 Primary metal manufacturing 120.40 154.71 128.5
332 Fabricated metal product manufacturing 77.07 75.85 98.4
333 Machinery manufacturing 110.04 103.81 94.3
334 Computer and electronic product manufacturing 230.30 149.76 65.0
335 Electrical equipment, appliance, and component manufacturing 104.23 88.21 84.6
336 Transportation equipment manufacturing 131.12 112.22 85.6
337 Furniture and related product manufacturing 60.60 69.09 114.0
339 Miscellaneous manufacturing 118.06 114.81 97.3
NAICS Industry U.S. PA PA % of U.S.
324 Petroleum and coal products mfg $828.16 $324.46 39.2
32411 Petroleum refineries 1,211.49 509.06 42.0
32412 Asphalt paving, roofing, and saturated materials mfg 190.69 136.30 71.5
32419 Other petroleum and coal products mfg 378.60 141.00 37.2
325 Chemical mfg 384.93 409.66 106.4
32511 Petrochemical mfg 2,144.42
32512 Industrial gas mfg 402.78 321.26 79.8
32513 Synthetic dye and pigment mfg 253.18 439.53 173.6
32518 Other basic inorganic chemical mfg 334.07 172.21 51.5
32519 Other basic organic chemical mfg 288.59 607.00 210.3
32521 Resin and synthetic rubber mfg 279.89 168.92 60.4
32522 Artificial and synthetic fibers and filaments mfg 96.60
32531 Fertilizer mfg 276.87 175.10 63.2
32532 Pesticide and other agricultural chemical mfg 522.81 165.37 31.6
32541 Pharmaceutical and medicine mfg 582.36 619.85 106.4
32551 Paint and coating mfg 237.33 292.40 123.2
32552 Adhesive mfg 162.03 179.01 110.5
32561 Soap and cleaning compound mfg 491.01 143.16 29.2
32562 Toilet preparation mfg 335.96 130.01 38.7
32591 Printing ink mfg 161.64
32592 Explosives mfg 109.81 227.01 206.7
32599 All other chemical product and preparation mfg 197.19
Empty cells mean either that industry does not exist in that area or the data are nondisclosed.
"2-digit" 3-digit 4-digit 5-digit 6-digit
Max number of NAICS industries: 1 21 86 184 472
United States 1 21 86 184 469
Pennsylvania 1 21 79 145 277
Nondisclosed 0 0 6 29 105
Nonexistent in PA 0 0 1 10 90
Proceedings of the Pennsylvania Economic Association 105
Table 6
Productivity in Pennsylvania’s Metro Areas
NAICS Industry
U.S
.
PA
All
en
tow
n
Alt
oo
na
Erie
Ha
rris
bu
rg
Jo
hn
sto
wn
La
nca
ster
Leb
an
on
New
Yo
rk
Ph
ila
delp
hia
Pit
tsb
urg
h
Rea
din
g
Scra
nto
n
Sta
te C
oll
eg
e
Wil
lia
msp
ort
Yo
rk
Yo
un
gst
ow
n
Nu
mb
er
of
MS
As
wit
h d
ata
Nu
mb
er
of
MS
As
gre
ate
r th
an
US
valu
e:
31-33 Manufacturing $126.12 $118.13 $145.52 $96.79 $84.35 $118.96 $64.26 $118.49 $78.36 $140.58 $157.41 $142.18 $68.69 $121.39 $101.18 $119.79 $118.00 $86.49 16 4
311 Food mfg 105.60 131.45 53.97 78.11 197.61 111.26 183.30 54.44 102.34 121.72 95.53 136.88 10 5
312 Beverage and tobacco product mfg 458.20 279.08 0 0
313 Textile mills 60.86 73.95 93.96 1 1
314 Textile product mills 60.16 63.23 0 0
315 Apparel mfg 48.86 55.61 0 0
316 Leather and allied product mfg 53.91 38.64 0 0
321 Wood product mfg 49.18 48.16 46.41 47.99 59.23 3 1
322 Paper mfg 120.05 137.04 65.30 74.19 2 0
323 Printing and related support activities 71.94 72.16 53.14 59.64 136.56 3 1
324 Petroleum and coal products mfg 828.16 324.46 0 0
325 Chemical mfg 384.93 409.66 304.69 1 0
326 Plastics and rubber products mfg 73.75 73.99 67.91 54.62 69.46 61.70 73.12 100.25 115.97 7 2
327 Nonmetallic mineral product mfg 94.78 92.06 129.10 138.50 83.26 70.82 4 2
331 Primary metal mfg 120.40 154.71 113.30 81.25 2 0
332 Fabricated metal product mfg 77.07 75.85 74.94 65.18 69.16 48.98 62.72 155.83 71.80 75.19 71.96 78.53 10 2
333 Machinery mfg 110.04 103.81 89.47 83.99 81.73 125.86 128.72 194.24 98.05 122.97 78.08 9 4
334 Computer and electronic product mfg 230.30 149.76 84.02 112.88 238.90 121.97 4 1
335 Elect eqpt, appliance, & component mfg 104.23 88.21 79.34 120.60 161.33 3 2
336 Transportation equipment mfg 131.12 112.22 85.83 48.02 44.17 88.22 4 0
337 Furniture and related product mfg 60.60 69.09 57.30 87.50 50.33 3 1
339 Miscellaneous mfg 118.06 114.81 52.77 120.39 118.89 3 2
Number of 3-digit industries with data 21 21 2 6 7 5 3 12 5 2 1 5 1 11 2 3 3 1
Number of 3-digits > US value 10 1 1 0 1 2 4 2 0 0 2 1 5 1 2 1 1
Boldface means a value greater than the U.S. value.
Proceedings of the Pennsylvania Economic Association 106
Table 7
Pennsylvania’s High Productivity Industries
NAICS Industry Productivity NAICS Industry Productivity
31-33 Manufacturing $118.13 SIX-DIGIT
THREE-DIGIT 325414 Biological product (except diagnostic) mfg $786.64
325 Chemical mfg 409.66 336391 Motor vehicle air-conditioning mfg 553.70
324 Petroleum & coal products mfg 324.46 325131 Inorganic dye & pigment mfg 551.48
312 Beverage & tobacco product mfg 279.08 325412 Pharmaceutical preparation mfg 526.27
331 Primary metal mfg 154.71 324110 Petroleum refineries 509.06
334 Computer & electronic product mfg 149.76 334510 Electromedical & electrotherapeutic apparatus mfg 495.74
322 Paper mfg 137.04 312111 Soft drink mfg 332.74
311 Food mfg 131.45 333295 Semiconductor machinery mfg 327.59
FOUR-DIGIT 325120 Industrial gas mfg 321.26
3254 Pharmaceutical & medicine mfg 619.85 325510 Paint & coating mfg 292.40
3251 Basic chemical mfg 445.28 325998 All other miscellaneous chemical product & preparation mfg 254.06
3241 Petroleum & coal products mfg 324.46 311613 Rendering & meat byproduct processing 249.98
3112 Grain & oilseed milling 317.13 311320 Chocolate & confectionery mfg from cacao beans 236.84
3221 Pulp, paper, & paperboard mills 314.51 311821 Cookie & cracker mfg 229.42
3121 Beverage mfg 279.35 334514 Totalizing fluid meter & counting device mfg 228.20
3122 Tobacco mfg 277.96 331111 Iron & steel mills 223.73
3255 Paint, coating, & adhesive mfg 252.36 327310 Cement mfg 219.98
3345 Navigational, measuring, electromedical, & control instruments mfg 232.50 337125 Household furniture (except wood & metal) mfg 215.92
3119 Other food mfg 230.58 334290 Other communications eqpt mfg 211.67
3259 Other chemical product & preparation mfg 229.11 312112 Bottled water mfg 211.21
3311 Iron & steel mills & ferroalloy mfg 222.64 334515 Instrument mfg for measuring & testing electricity & electrical systems 203.68
3342 Communications eqpt mfg 189.73 334517 Irradiation apparatus mfg 202.77
3113 Sugar & confectionery product mfg 188.49 334516 Analytical laboratory instrument mfg 202.18
3253 Pesticide, fertilizer, & other agricultural chemical mfg 173.14 311111 Dog & cat food mfg 197.03
3111 Animal food mfg 168.76 325132 Synthetic organic dye & pigment mfg 192.34
3252 Resin, synthetic rubber, & artificial synthetic fibers & filaments mfg 168.49 332913 Plumbing fixture fitting & trim mfg 191.86
3314 Nonferrous metal (except aluminum) production & processing 148.59 333921 Elevator & moving stairway mfg 190.06
3313 Alumina & aluminum production & processing 143.03 334220 Radio & TV broadcasting & wireless communications eqpt mfg 188.69
3333 Commercial & service industry machinery mfg 138.01 311822 Flour mixes & dough mfg from purchased flour 187.55
3274 Lime & gypsum product mfg 137.65 339114 Dental eqpt & supplies mfg 186.05
3256 Soap, cleaning compound, & toilet preparation mfg 136.70 327999 All other miscellaneous nonmetallic mineral product mfg 184.88
3391 Medical eqpt & supplies mfg 128.73 334210 Telephone apparatus mfg 183.28
3118 Bakeries & tortilla mfg
128.06
334513 Instruments & related prdcts mfg for measuring, displaying, &
controlling industrial process variables 181.55
3332 Industrial machinery mfg 126.78 325520 Adhesive mfg 179.01
3353 Electrical eqpt mfg 122.46 311330 Confectionery mfg from purchased chocolate 176.23
3312 Steel product mfg from purchased steel 119.81 311520 Ice cream & frozen dessert mfg 176.05
3339 Other general purpose machinery mfg 119.04 333516 Rolling mill machinery & eqpt mfg 174.04
FIVE-DIGIT 337214 Office furniture (except wood) mfg 173.46
32541 Pharmaceutical & medicine mfg 619.85 325314 Fertilizer (mixing only) mfg 172.59
32519 Other basic organic chemical mfg 607.00 325188 All other basic inorganic chemical mfg 172.21
32411 Petroleum refineries 509.06 333512 Machine tool (metal cutting types) mfg 172.19
32513 Synthetic dye & pigment mfg 439.53 331315 Aluminum sheet, plate, & foil mfg 166.31
32512 Industrial gas mfg 321.26 324122 Asphalt shingle & coating materials mfg 166.06
32551 Paint & coating mfg 292.40 325320 Pesticide & other agricultural chemical mfg 165.37
31211 Soft drink & ice mfg 287.39 339112 Surgical & medical instrument mfg 162.25
31132 Chocolate & confectionery mfg from cacao beans 236.84 324191 Petroleum lubricating oil & grease mfg 154.98
33451 Navigational, measuring, electromedical, & control instruments mfg 232.50 333319 Other commercial & service industry machinery mfg 153.68
32599 All other chemical product & preparation mfg 227.01 335314 Relay & industrial control mfg 152.72
33141 Nonferrous metal (except aluminum) smelting & refining 223.53 333618 Other engine eqpt mfg 150.41
33111 Iron & steel mills & ferroalloy mfg 222.64 334519 Other measuring & controlling device mfg 149.62
32731 Cement mfg 219.98 334417 Electronic connector mfg 149.18
33994 Office supplies (except paper) mfg 214.85 327420 Gypsum product mfg 146.96
33429 Other communications eqpt mfg 211.67 331210 Iron & steel pipe & tube mfg from purchased steel 146.44
33142 Copper rolling, drawing, extruding, & alloying 201.36 334512 Auto. environmental control mfg for resdntl, comrcl, & appliance use 144.38
33422 Radio & television broadcasting & wireless communications eqpt mfg 188.69 333414 Heating eqpt (except warm air furnaces) mfg 139.72
33391 Pump & compressor mfg 185.96 335313 Switchgear & switchboard apparatus mfg 137.09
33421 Telephone apparatus mfg 183.28 331314 Secondary smelting & alloying of aluminum 136.86
32552 Adhesive mfg 179.01 336413 Other aircraft parts & auxiliary eqpt mfg 136.42
31133 Confectionery mfg from purchased chocolate 176.23 333220 Plastics & rubber industry machinery mfg 136.16
31152 Ice cream & frozen dessert mfg 176.05 311119 Other animal food mfg 134.50
32531 Fertilizer mfg 175.10 331112 Electrometallurgical ferroalloy product mfg 133.76
32518 Other basic inorganic chemical mfg 172.21 311911 Roasted nuts & peanut butter mfg 133.75
32521 Resin & synthetic rubber mfg 168.92 333992 Welding & soldering eqpt mfg 133.39
31111 Animal food mfg 168.76 339920 Sporting & athletic goods mfg 133.22
32532 Pesticide & other agricultural chemical mfg 165.37 327410 Lime mfg 133.04
32742 Gypsum product mfg 146.96 326122 Plastics pipe & pipe fitting mfg 132.68
33121 Iron & steel pipe & tube mfg from purchased steel 146.44 333518 Other metalworking machinery mfg 132.49
32561 Soap & cleaning compound mfg 143.16 326130 Laminated plastics plate, sheet (except packaging), & shape mfg 132.13
33131 Alumina & aluminum production & processing 143.03 333131 Mining machinery & eqpt mfg 130.68
32419 Other petroleum & coal products mfg 141.00 325620 Toilet preparation mfg 130.01
33331 Commercial & service industry machinery mfg 138.01 311812 Commercial bakeries 129.96
32412 Asphalt paving, roofing, & saturated materials mfg 136.30 324199 All other petroleum & coal products mfg 129.23
33322 Plastics & rubber industry machinery mfg 136.16 339113 Surgical appliance & supplies mfg 128.02
33992 Sporting & athletic goods mfg 133.22 311340 Nonchocolate confectionery mfg 126.33
32741 Lime mfg 133.04 333298 All other industrial machinery mfg 124.53
32613 Laminated plastics plate, sheet (except packaging), & shape mfg 132.13 333111 Farm machinery & eqpt mfg 124.24
32562 Toilet preparation mfg
130.01
331491 Nonferrous metal (exc copper & aluminum) rolling, drawing, extruding,
& alloying 124.06
33911 Medical eqpt & supplies mfg 128.73 335932 Noncurrent-carrying wiring device mfg 123.91
33329 Other industrial machinery mfg 126.82 333294 Food product machinery mfg 122.10
31134 Nonchocolate confectionery mfg 126.33 333314 Optical instrument & lens mfg 121.65
33531 Electrical eqpt mfg 122.46 313230 Nonwoven fabric mills 121.43
33313 Mining & oil & gas field machinery mfg 121.45 335121 Residential electric lighting fixture mfg 120.54
31323 Nonwoven fabric mills 121.43 315232 Women's & girls' cut & sew blouse & shirt mfg 120.24
33639 Other motor vehicle parts mfg 118.44 322291 Sanitary paper product mfg 120.08
327993 Mineral wool mfg 119.20
324121 Asphalt paving mixture & block mfg 118.39
327125 Nonclay refractory mfg 118.20
Proceedings of the Pennsylvania Economic Association 107
Table 8
Productivity and Location Quotients in 3-digit Industries
Table 9
Productivity and Location Quotients in 3-digit Industries, Ranked by PA Productivity
PA U.S.
31-33 Manufacturing $118.13 $126.12 93.7
311 Food mfg 131.45 105.60 124.5 0.92
312 Beverage and tobacco product mfg 279.08 458.20 60.9 0.92
313 Textile mills 73.95 60.86 121.5 0.83
314 Textile product mills 63.23 60.16 105.1 0.77
315 Apparel mfg 55.61 48.86 113.8 0.85
316 Leather and allied product mfg 38.64 53.91 71.7 0.69
321 Wood product mfg 48.16 49.18 97.9 1.09
322 Paper mfg 137.04 120.05 114.2 1.19
323 Printing and related support activities 72.16 71.94 100.3 1.32
324 Petroleum and coal products mfg 324.46 828.16 39.2 1.33
325 Chemical mfg 409.66 384.93 106.4 0.90
326 Plastics and rubber products mfg 73.99 73.75 100.3 1.04
327 Nonmetallic mineral product mfg 92.06 94.78 97.1 0.99
331 Primary metal mfg 154.71 120.40 128.5 1.94
332 Fabricated metal product mfg 75.85 77.07 98.4 1.17
333 Machinery mfg 103.81 110.04 94.3 0.98
334 Computer and electronic product mfg 149.76 230.30 65.0 0.78
335 Electrical equipment, appliance, and component mfg 88.21 104.23 84.6 1.30
336 Transportation equipment mfg 112.22 131.12 85.6 0.59
337 Furniture and related product mfg 69.09 60.60 114.0 0.82
339 Miscellaneous mfg 114.81 118.06 97.3 0.96
Mfg LQNAICSProductivity*
PA % of US
*Value added per production worker hour, in dollars.
Industry
PA U.S.
31-33 Manufacturing $118.13 $126.12 93.7
325 Chemical mfg 409.66 384.93 106.4 0.90
324 Petroleum and coal products mfg 324.46 828.16 39.2 1.33
312 Beverage and tobacco product mfg 279.08 458.20 60.9 0.92
331 Primary metal mfg 154.71 120.40 128.5 1.94
334 Computer and electronic product mfg 149.76 230.30 65.0 0.78
322 Paper mfg 137.04 120.05 114.2 1.19
311 Food mfg 131.45 105.60 124.5 0.92
339 Miscellaneous mfg 114.81 118.06 97.3 0.96
336 Transportation equipment mfg 112.22 131.12 85.6 0.59
333 Machinery mfg 103.81 110.04 94.3 0.98
327 Nonmetallic mineral product mfg 92.06 94.78 97.1 0.99
335 Electrical equipment, appliance, and component mfg 88.21 104.23 84.6 1.30
332 Fabricated metal product mfg 75.85 77.07 98.4 1.17
326 Plastics and rubber products mfg 73.99 73.75 100.3 1.04
313 Textile mills 73.95 60.86 121.5 0.83
323 Printing and related support activities 72.16 71.94 100.3 1.32
337 Furniture and related product mfg 69.09 60.60 114.0 0.82
314 Textile product mills 63.23 60.16 105.1 0.77
315 Apparel mfg 55.61 48.86 113.8 0.85
321 Wood product mfg 48.16 49.18 97.9 1.09
316 Leather and allied product mfg 38.64 53.91 71.7 0.69
*Value added per production worker hour, in dollars.
NAICS IndustryProductivity*
PA % of US Mfg LQ
Proceedings of the Pennsylvania Economic Association 108
Table 10
Productivity and Location Quotients in 3-digit Industries, Ranked by PA Productivity as % of U.S.
1Thanks to Economic Research Institute of Erie Research Assistant Brittany Martinelli for able assistance in preparation of the
databases for this research.
PA U.S.
31-33 Manufacturing $118.13 $126.12 93.7
331 Primary metal mfg 154.71 120.40 128.5 1.94
311 Food mfg 131.45 105.60 124.5 0.92
313 Textile mills 73.95 60.86 121.5 0.83
322 Paper mfg 137.04 120.05 114.2 1.19
337 Furniture and related product mfg 69.09 60.60 114.0 0.82
315 Apparel mfg 55.61 48.86 113.8 0.85
325 Chemical mfg 409.66 384.93 106.4 0.90
314 Textile product mills 63.23 60.16 105.1 0.77
326 Plastics and rubber products mfg 73.99 73.75 100.3 1.04
323 Printing and related support activities 72.16 71.94 100.3 1.32
332 Fabricated metal product mfg 75.85 77.07 98.4 1.17
321 Wood product mfg 48.16 49.18 97.9 1.09
339 Miscellaneous mfg 114.81 118.06 97.3 0.96
327 Nonmetallic mineral product mfg 92.06 94.78 97.1 0.99
333 Machinery mfg 103.81 110.04 94.3 0.98
336 Transportation equipment mfg 112.22 131.12 85.6 0.59
335 Electrical equipment, appliance, and component mfg 88.21 104.23 84.6 1.30
316 Leather and allied product mfg 38.64 53.91 71.7 0.69
334 Computer and electronic product mfg 149.76 230.30 65.0 0.78
312 Beverage and tobacco product mfg 279.08 458.20 60.9 0.92
324 Petroleum and coal products mfg 324.46 828.16 39.2 1.33
Mfg LQ
*Value added per production worker hour, in dollars.
NAICS IndustryProductivity*
PA % of US
Proceedings of the Pennsylvania Economic Association 109
REFERENCES
Brunot, Justin A and James A. Kurre. 2012. Manufacturing
Productivity: How Much Does It Vary Across Metro Areas
and Why? Paper presented at the 66th Annual Conference of
the Association for University Business and Economic
Research, Honolulu HI, Oct 2012.
Carlino, Gerald A. and Richard Voith. 1992. Accounting for
Differences in Aggregate State Productivity. Regional
Science and Urban Economics, v. 22, #4, pp 597-617.
Dall’erba, Sandy, Yiannis Kamarianakis, Julie LeGallo, and
Maria Plotnikova. 2005. Regional Productivity Differentials
in Three New Member Countries. What Can We Learn from
the 1986 Enlargement to the South? The Review of Regional
Studies, v. 35, #1, pp. 97-116.
Executive Office of the President, Office of Management and
Budget. 2006. Update of Statistical Area Definitions and
Guidance on Their Uses OMB Bulletin 07-01. December
18, 2006.
Hammill, Michael L. 2002. Productivity Growth in Erie
Manufacturing Through Time. Erie PA: Economic Research
Institute of Erie, Penn State Erie. December 2002. Available
online at www.ERIEdata.org.
Hill, Edward W. and John F. Brennan. 2000. A
Methodology for Identifying the Drivers of Industrial
Clusters: The Foundation of Regional Competitive
Advantage. Economic Development Quarterly, v. 14, #1, pp.
65-96.
Iranzo, Susana and Giovanni Peri. 2006. Schooling
Externalities, Technology And Productivity: Theory And
Evidence From U.S. States. NBER Working Paper 12440,
August 2006.
Kurre, James A. 2004. Determinants of Productivity
Differences Across American Metro Areas. Presented at the
58th Annual AUBER Fall Conference, Tucson AZ , October
17-19, 2004.
Kurre, James A. and Edward R. Miseta. 2008 Determinants
of Productivity Differences across Metro Areas for
Manufacturing Industries. Presented at the 47th Annual
Meeting of the Southern Regional Science Association,
Arlington, VA, March 2008.
National Bureau of Economic Research. 2013. Productivity,
Innovation and Entrepreneurship Program. Online at:
http://www.nber.org/programs/pr/
Primont, Diane F. and Bruce R. Domazlicky. 2005. Which
Matters Most to the Estimation of Efficiency and
Productivity Growth in State Manufacturing: Method or
Measurement? Review of Regional Studies, v. 35, #2, pp.
117-138.
U.S. Bureau of Economic Analysis. 2013. NIPA Table
1.12. National Income by Type of Income.
U.S. Census Bureau. 2007A. 2007 Economic Census. Table
ECO731SG1: Manufacturing: Summary Series: General
Summary: Industry Statistics for Industry Groups and
Industries: 2007.
U.S. Census Bureau. 2007B. 2007 Economic Census
Methodology: Manufacturing (sector 31-33), page 6.
Available online at:
http://www.census.gov/econ/census07/pdf/meth/meth_31.pdf
.
Proceedings of the Pennsylvania Economic Association 110
FORCE-USING FIRMS IN THE COMPETITIVE EQUILIBRIUM
WHEN THE ONLY PUBLICLY PROVIDED GOOD IS ENVY GRATIFICATION
Johnnie B. Linn III
Division of Business
Concord University
Athens, WV, 24712
ABSTRACT
Envy is a public bad because if another’s possessions
engender envy in one person, they will engender envy in all
other persons similarly situated. Gratification of envy is
therefore a public good. Government maximizes solely this
public good subject to a budget constraint in extracting
capital from firms. In the equilibrium solution, the tax on
firm size is regressive. Firms have an incentive to become
larger. As a consequence, the tax base out of which envy is
gratified, and the amount of envy gratified, becomes larger.
INTRODUCTION
Envy is unhappiness that arises from the well-being of others.
It can augment a utility function otherwise based purely on
self-interest as an extra argument, denoting the well-being of
someone else, making a negative contribution to utility
(Zizzo, 2007). Envy thus implies more than the mere desire
to take from another, because it is gratified even if that which
is taken is not received by the person having the envious
desire. Envy thus has the attribute of a public bad, for if
particular assets exist that engender envy, they engender envy
in all persons that are envious. Gratification of envy, or
schadenfreude, is therefore a public good. A traditional
function of government is to provide public goods. This
paper explores a competitive equilibrium in the presence of a
government whose only function is the gratification of envy.
Frank (1997) coined the term positional externality to denote
the dependence of utility on relative consumption. Positional
externalities are analytically no different than environmental
pollution, and Frank proposes a tax on consumption
analogous to a tax on pollution (p. 1843). It is the relative
consumption, not consumption per se, that is the target of the
tax, so the tax would necessarily be progressive. Frank
further likens the optimization of consumption as having the
same public-good attribute as military preparedness:
Because each individual's consumption affects the
frame of reference within which others evaluate
their own consumption, this frame of reference
becomes, in effect, a public good. The
uncoordinated consumption decisions of individuals
are not more likely to result in the optimal level of
this public good than the uncoordinated actions of
individuals are likely to result in an optimal level of
military preparedness. The progressive consumption
tax is a simple policy measure that can help mould
the frame of reference in mutually beneficial ways
(p. 1844).
This paper takes as given that the function of government is
to forcibly provide public goods that the people want without
making a moral judgment as to whether what the people want
is morally appropriate, but also that those being taxed will
use force to ameliorate the government’s taxing power. The
resultant tax regime may therefore not be progressive.
Envy can take various forms in what is envied. The object of
envy can be all the possessions of all others, whether or not
the others are of higher or lower station than the envier, or all
the possessions of only those who are of higher station than
the envier, or only the possessions among those of higher
station than the envier that exceed the level of possessions of
the envier. The third form will be used in this paper to avoid
the complications of change in rank of individuals in regard
to their station.
THE MODEL
Labor is assumed to be homogenous in the population and
each individual is endowed with one unit of labor. Capital is
the object of envy and the envy engendered by each
individual is a function of that individual’s holding of capital.
Individuals may hire capital and guards and operate their own
firms, or may work as a guard, or may work as a tax
collector. Production requires only capital and one
entrepreneur, and force requires only labor. Following Linn
(2007), a ratio rule will be used for the allocation of winnings
by force.
We consider a horizon encompassing n individuals embedded
in a world market in which labor and capital can be
purchased at constant wages. Individuals are given a
designation j according to their rank in holdings of capital.
Envy extends only to the holdings of others who can be
“seen”, that is, those within the horizon. The amount of envy
engendered by individual j is
∑
∑
(1)
Proceedings of the Pennsylvania Economic Association 111
where the wage for capital is r. The amount of envy
engendered by a unit of capital is proportional to its value to
its soon-to-be dispossessed owner, or r. The jth
set of terms
in the first summation is individual j’s self-envy which will
be zero.
The Government’s Optimization Problem
The government’s objective function is to maximize the
value of a quantity of capital extracted from all individuals in
proportion to the envy engendered by them, less outlays for
tax collectors, thus:
∑ ( ∑
)
∑
(2)
where the government deploys Tj tax collectors paid a wage
w into arena j occupied by firm j and extracts a proportion tj
of the value of the firm’s capital. The taxes are weighted by
the amount of envy engendered in Equation (1) by each firm.
Taxes are levied only in arenas that have output. The other
arenas are empty, having been vacated by individuals who
have been hired as guards or as tax collectors. The
government expects the firms to anticipate its actions and
regards their variables as fixed. The government also regards
the number of individuals in the horizon to be unaffected by
the deployment of tax collectors.
Under a ratio rule, the government’s own-force elasticity for
its winnings is
(3)
The government’s force function in elasticized form for its
tax collectors is
(4)
The first order condition for the government in a given arena
is, employing the chain rule with Equations (3) and (4),
( ) ( ∑
)
(5)
If we sum the government’s first-order conditions over all n
arenas, we obtain
∑ ( ) ( ∑
)
∑
(6)
At equilibrium, setting government outlays for tax collectors
equal to the value of capital extracted, we obtain
∑ ( ) ( ∑
)
∑
(7)
or
∑ ( )( ∑
)
∑
(8)
For the government to balance its budget, it must break even
in at least one arena, so
|
(9)
The Firm’s Optimization Problem
The objective function for the firm in a given arena is
(10)
where Kj is the amount of capital hired and Gj is the number
of guards hired. In calculating the firm’s first-order
conditions, the firm expects the government to anticipate all
the firms’ actions, so the government’s variables, and the
amount of envy gratification to be generated for each
individual, are regarded as fixed. The firm’s production
function in elasticized form for its output is
(11)
The firm’s first-order condition for its capital is
( )
(12)
The force technologies for the firm and the government are
assumed to be the same, so the firm’s force elasticity with
respect to its guards is . Under the ratio rule, the firm’s
cross-force elasticity with respect to a tax rate is, when the
firm and the government are the only users of force,
Proceedings of the Pennsylvania Economic Association 112
(13)
The first-order condition for the firm’s guards is
( )
(14)
As in the competitive equilibrium in the absence of
government as set out in Linn (2007), we expect to find
stable competitive equilibria here only if is less than unity
and is greater than unity.
The Second-Order Conditions
The government’s matrix of second derivatives for its tax
collectors is a diagonal matrix each of whose diagonal
elements is the second derivative for a particular arena. The
second-order sufficiency condition is met if each of the
diagonal elements is negative.
The firm’s matrix of second derivatives for its capital and
guards is
(
( )
( )
)
where
(15)
The second-order sufficiency condition is met when the
diagonal elements are negative and the determinant
( )
|
( )
( )
|
(16)
is greater than zero. For r we may substitute in its value from
the first-order condition for Kj to get a further simplification
( )
( )
| ( ) ( )
( ) |
(17)
The second-order condition for the firm will not be met if the
tax rate is too small or or are excessively large.
We will return to the second-order conditions when we
consider simplified solutions in the neighborhood of the
long-run competitive equilibrium.
The Long-Run Competitive Equilibrium
In the long-run competitive equilibrium, the surviving firms
will earn a normal profit w*, the amount that its proprietor
could earn not only as a guard or a tax collector, but also the
amount of additional envy gratification that the former
capitalist would receive by entering the ranks of the non-
holders of capital. In the long-run competitive equilibrium,
labor is fully employed, so each person hired as a guard
supplies one unit of labor to one firm. After substituting for
the outlays to capital and guards in the profit function, we
obtain
( ) [ ( ) ( )]
(18)
The government must be breaking even in at least one arena.
For that arena, making use of Equation (9) and the guards’
first-order condition, we obtain
(19)
Let us suppose that in this arena, the number of guards and
number of tax collectors is the same. Since they have the
same force technology, the resultant tax rate would be 0.5.
Only for a value of unity for is such a solution possible.
For all values of greater than unity, only values of tj
greater than 0.5 are possible in arenas where the government
breaks even.
Progressivity or Regressivity of the Tax Regime
We now consider whether the tax regime across firms of
different sizes is progressive or regressive from the short-run
perspective of the firms. We take the derivative of equation
(18) with respect to tj and obtain
[ ( ) ]
( )
(20)
We set to zero and back-substitute for w
* from Equation
(18):
( ) ( )
( )
(21)
After cancellation of the expressions involving (1 - ), the
quantity can be factored out of the remaining terms
Proceedings of the Pennsylvania Economic Association 113
leaving a quadratic relation involving only tj. Its positive root
is
√
(22)
The value of Yj is maximized at this tax rate, so the firm size
and tax rate are directly correlated (progressive) as lower
levels of the tax and the firm size and tax rate are inversely
correlated (regressive) at higher levels of the tax.
For a tax rate other than 0.4142, there are two tax rates
compatible with a given size. The government will not allow
all firms to select a tax rate less than 0.4142, because for the
government to balance its budget in an arena the tax rate
must be at least 0.5. Firms that are sitting on the right hand
side of the firm size curve as a function of the tax rate, as
shown in Figure 1, will have an incentive to move up and left
on the curve, increasing their size and reducing their tax rate.
The government will lose surplus funds from the arenas of
such firms, so it will be forced to balance its budget in all
arenas, and since all firms employ identical factors, they will
all be the same size. The tax rate faced by all firms will be
regressive.
THE SOLUTION WHEN ALL FIRMS ARE ALIKE
Let the lowest-ranking, or only, firm in the horizon have the
designation m. The government’s objective function from
Equation (2) is
∑
(23)
Its first-order condition is
(24)
Setting the government’s revenue in arena m equal to its
outlays there,
(25)
The second derivative for the government in each arena is
identical to the firms’ second derivative for guards, as seen in
the lower right element of H in Equation (15), except that the
G’s are replaced with T’s. The second-order sufficiency
condition for the government is
(26)
or, from Equation (25),
(27)
which is met for all value of greater than unity.
The second order sufficiency condition for the firm will be
met if
|
|
(28)
In the long-run equilibrium, each firm defending against the
tax collectors will have one proprietor and an integral number
of guards, and there will be n – m + 1 such firms. Firms with
larger numbers of guards will be favored in the competitive
equilibrium because, from Equation (14), capital and guards
are positively correlated, and more capital means a lower tax
rate, since all firms are sitting on the right hand side of their
firm size curves. For a horizon of a given size, larger firms
means fewer firms. In the long run, there will be only one
firm in the horizon, and in Equation (28), the m’s will be
replaced with n’s. Figure 2 shows how the zone where
Equation (28) is satisfied varies with horizon size.
A greater horizon size selects for greater robustness of the
firm and a larger range of and in which the competitive
equilibrium is stable.
IMPLICATIONS
A stable competitive equilibrium where envy gratification is
maximized has been found. Contrary to what might be
wished for, the associated tax regime is regressive. The
absence of a progressive tax regime does not imply that
maximum envy gratification has not been achieved. This
model is built on the basest of human emotions, thus is likely
to be empirically robust. A government that wants to provide
more to the people than envy gratification can use this model
to improve upon.
Proceedings of the Pennsylvania Economic Association 114
Figure 1. Short-run firm size as a function of the tax rate on capital.
Figure 2. Zone of long-run stability as a function of horizon size.
REFERENCES
Frank, R.H. 1997. The Frame of Reference as a Public Good,
The Economic Journal, 107 (November), 1832-1847.
Linn, J. B. 2007. Force-Using Firms in the Competitive
Equilibrium when Public Goods are Absent, Pennsylvania
Economic Review, 15(1): 21-34.
Zizzo, D.J. 2007. The Cognitive and Behavioral Economics
of Envy, in Richard H. Smith (ed.), “Envy: Theory and
Practice”, Oxford University Press (Affective Science
Series), forthcoming.
0
0.4
0.8
1.2
1.6
2
0 0.2 0.4 0.6 0.8 1
Y
t
1
1.2
1.4
1.6
1.8
2
0 0.2 0.4 0.6 0.8 1
n 2
n 3
n
Proceedings of the Pennsylvania Economic Association 115
THE EFFECTS OF INCREASES IN CIGARETTE PRICES ON SMOKING BEHAVIORS:
ESTIMATES USING MSA AS A NATURAL EXPERIMENT
Zhen Ma
Department of Business
Misericordia University
301 Lake Street
Dallas, PA 18612
ABSTRACT
This paper examines the impact of the large increases in
cigarette prices after the Master Settlement Agreement
(MSA) on cigarette consumption for smokers by estimating
dynamic panel data models. I use system GMM estimator
with fixed effects to address the unobserved heterogeneity
and initial conditions issues. I find that older smokers are
virtually nonresponsive to price changes, and younger
smokers are more price sensitive; women smokers might be
more price responsive; and the overall estimated price
elasticity for the full sample of smokers is -0.26. In addition,
smoke-free air laws do not show significant effects.
I. INTRODUCTION
Since the release of the Surgeon General’s report on smoking
and health in 1964, the health hazard caused by smoking has
been more and more recognized. Smoking harms almost
every organ of the body; and it remains the single largest
preventable cause of death in the United States. Each year,
there are 393,600 estimated deaths caused by direct
smoking1. Secondhand smoking is no less harmful.
Secondhand smoke contains more toxic chemicals than does
smoke inhaled from a cigarette. Currently, there are 49,400
estimated deaths of nonsmokers annually from secondhand
smoking2. For half a century, the federal and state
governments have been engaged in anti-smoking campaigns.
Major policy instruments include widespread dissemination
of information on the health consequences of cigarette
smoking, ban of broadcast advertising of cigarettes,
restrictions on smoking in public places and private
workplaces, increased cigarette excise taxes and more
(Chaloupka and Grossman 1996). Over the past decade or
so, America has seen the historically largest increases in
cigarette excise taxes. Since January 1 1997, with the
exception of Florida, Mississippi, Missouri, North Dakota,
South Carolina and Virginia, all other states and the District
of Columbia have increased their taxes on cigarettes at least
once. The following are the number of states as of 2009 that
raised cigarette taxes by 25 cents or more per pack in
nominal terms from the previous year: 3 states in 1997 and
1998, 1 in 1999, 2 in 2000, 7 in 2002, 18 in 2003, 8 in 2004,
9 in 2005 and 6 in 2006 through 2009. Among these, 13
states had one-time increases of a dollar or more at least
once. As of 2009, Rhode Island has the highest state excise
tax of $3.46 per pack, followed by $2.75 in New York
(Orzechowski and Walker 2009). Also over the past decade,
an increasing number of states have passed stronger smoke-
free air laws. As of March 2011, 27 states have enacted
statewide bans on smoking in all enclosed public places,
including private workplaces, restaurants and bars, whereas
there was none prior to the year 2002.
In November 1998, 46 states and the four major tobacco
companies reached the Master Settlement Agreement which
stipulated that the tobacco companies pay the states $206
billion over the next 25 years to compensate most states for
Medicaid expenses for treating tobacco-related illnesses. The
other four states (Florida, Mississippi, Texas and Minnesota)
settled with the tobacco companies individually. The tobacco
companies are financing these payments by increasing
cigarette prices. As a result, the cigarette prices went up by
45 cents per pack3, or 19.5%, nationwide immediately after
the settlement, and continued to rise over the following five
years in many states due to substantial increases in their
excise taxes on cigarettes.
In this paper, I investigate how cigarette consumption among
smokers has been affected by these unprecedented increases
in cigarette prices and the changes in anti-smoking
regulations (the full prices of smoking)4. I estimate system
GMM estimator with fixed effects to address the unobserved
heterogeneity and initial conditions issues associated with
estimation of dynamic panel data models.
II. BACKGROUND AND LITERATURE
Raising cigarette taxes has been one of the most effective
means to prevent and reduce smoking. There is an extensive
literature documenting the impact of changes in cigarette
taxes or prices on the demand for cigarettes. Estimates of
price elasticity of demand for cigarettes vary from -0.05 to -
0.60.
Lewit and Coate (1982) is one of the earliest works that
examine the impact of cigarette excise taxes on reducing
smoking. The unique contribution of their study is that it
controls for possible cigarette bootlegging by eliminating the
individuals that reside in an area where the price of cigarettes
is higher than another price found within a 20-mile-wide
Proceedings of the Pennsylvania Economic Association 116
band around it. Using 1976 Health Interview Survey, they
estimate a price elasticity of -0.16 for adults, -0.15 for the 20-
25 age group, -0.03 for the 26-35 age group, and -0.30 for 36
years of age and older. For smokers, price elasticity is -0.10
for all adults, -0.20 for the 20-25 age group, -0.04 for the 26-
35 age group, and -0.15 for adults over 35 years of age.
Following the idea of Lewit and Coate (1982) on accounting
for cigarette bootlegging, Wasserman et al (1991) uses a
sample from the National Health Interview Survey (NHIS)
between 1970 and 1985. They dummy out the individuals
who boarder on an area that has lower-priced cigarettes,
rather than deleting them. Their estimates of price elasticities
of demand for cigarettes among adults range from 0.06 in
1970 to -0.23 in 1985, with the elasticities becoming
increasingly negative over time, but relatively lower than the
results from Lewit and Coate (1982).
Using repeated cross sections from Second National Health
and Nutrition Examination Survey (NHANES2) 1976-1980,
Chaloupka (1990) uses two staged least squares (2SLS) to
estimate demand equations derived from the Becker-Murphy
rational addiction model. He finds that, for men, the price
elasticity of demand for cigarettes centers on -0.60, while
women are virtually unresponsive to changes in cigarette
prices. Chaloupka (1991), using the same data, empirical
models, and estimation methods, finds that the price elasticity
of demand for the full sample ranges from -0.36 to -0.27.
The price elasticity of demand for cigarettes by only current
smokers ranges from -046 to -0.30.
Townsend et al. (1994) employ the British data General
Household Survey 1972-1990 and find that women are more
responsive to changes in cigarette prices. Their estimates of
price elasticity of demand for cigarettes for men and for
women are -0.08 and -0.23, respectively.
Hersch (2000), using the Current Population Survey (CPS)
Tobacco Use Supplements from September 1992, January
1993, and May 1993, finds that price elasticity of demand for
cigarettes to be -0.46 for men, and -0.38 for women.
Farrelly et al. (2001) estimate the two-part model with
National Health Interview Survey (NHIS) 1976-1993. They
find that the overall price elasticity for the full sample is -
0.15. They also estimate the price elasticities across gender
and age groups. They find that the demand elasticities for
cigarettes are -0.21 and -0.32 for men and for women,
respectively, -0.55 for ages 18-24, -0.53 for ages 25-39, and -
0.08 for ages 40 and older.
Stehr (2007) uses a sample of 1.3 million observations drawn
from the 1985-2000 Behavioral Risk Factor Surveillance
System (BRFSS), and finds that, by interacting dummy
variables for sex, race, ethnicity, age and income quartile
with cigarette taxes and state fixed effects, women are more
responsive to cigarette taxes than men, as opposed to what
previous studies have indicated. Specifically, the demand
elasticities are -0.09 for men and -0.12 for women.
All these studies use repeated cross-sectional data and fail to
control for reinforcement effects. Reinforcement of an
addictive good means that past consumption increases current
consumption by raising the marginal utility of current
consumption. Due to the addictive nature of cigarettes, past
consumption will reinforce its current consumption. This
paper advances the literature by using panel data to
consistently estimate the price elasticity of demand for
cigarettes allowing for reinforcement.
III. CONCEPTUAL FRAMEWORK AND EMPIRICAL
MODEL
The model starts with an individual’s utility in period t as a
function of the consumption of cigarettes )( tC and the
composite good )( tG given consumption of cigarettes from
last period )( 1tC . The individual maximizes utility:
)|,( 1 ttt CGCuU
subject to a budget constraint:
t
G
tt
C
tt GpCpY
where tY is income in period t , C
tp and G
tp are prices for
cigarettes and the composite good, respectively. Prices of
cigarettes are the full prices that include monetary prices and
policies that regulate the consumption of cigarettes.
Maximizing utility subject to the budget constraint yields the
demand function for the individual. The dependent variable
can be either amount of cigarettes consumed or a
dichotomous indicator for smoking. The following empirical
model for the probability of smoking for individual i in
period t will be estimated
titi
c
ti
c
tititi xpolpcc ,,3,2,11,1, (1)
where the error term ti , consists of the unobserved individual
fixed effects iu and the idiosyncratic disturbances tiv ,
( ),, tiiti vu . tic , is an indicator for smoking
participation or the amount of cigarettes consumed, 1, tic is
smoking status for the previous period, c
tip , is the money
prices for cigarettes, c
tipol , is the policy variable for
smoking, tix , is a vector of exogenous social-demographic
variables including family income, age, gender, race,
Proceedings of the Pennsylvania Economic Association 117
ethnicity, education, employment and marital status. 1
measures the reinforcement effect of past smoking behavior
on current smoking, and it is expected to be positive. 1 and
2 , the coefficients on the money price and policy variables,
should be negative based on the law of demand.
Estimation of equation (1) requires longitudinal data.
Difficulty arises in obtaining unbiased and consistent
estimates of the state dependence due to the correlation
between the time invariant unobserved individual
heterogeneity iu and the lagged smoking behaviors. Serial
correlation in the idiosyncratic disturbances tiv , also
complicates the estimation. Even when coefficients on
lagged dependent variables are not of direct interest,
estimating them consistently may be crucial for recovering
consistent estimates of other parameters (Bond 2002).
IV. DATA
The data are from the 1999, 2001, 2003, 2005, 2007 and
2009 waves of the Panel Study of Income Dynamics (PSID).
Conducted by the Survey Research Center, Institute for
Social Research, University of Michigan, the PSID is a
longitudinal study of a nationally representative sample of
U.S. individuals and the family units in which they reside.
The initial wave of the PSID was administered in 1968.
Follow-up interviews were conducted annually until 1996
and biennially thereafter. The health behavior measures such
as alcohol and cigarette consumption have been collected
since 1999.
4.1 Dependent Variables
The research objective is to find out whether smoking
behavior is significantly affected by changes in cigarette
prices. Two dependent variables are used: smoking
participation and the average number of cigarettes per day the
individual consumes. The values of the two dependent
variables are based on the answers from the respondents to
the questions “Do you smoke cigarettes?” and “On the
average, how many cigarettes per day do you usually
smoke?” Figure 1 shows the trends in cigarette consumption
among smokers.
4.2 Independent Variables
To control for the reinforcement effects of past smoking,
variables for lagged smoking participation and lagged
cigarettes consumption are included.
The PSID provides a wide range of demographic and
socioeconomic variables. The estimation of the model
controls for age, sex, race, ethnicity, educational attainment,
family income, household size, marital and employment
status. Three categories of race are used: white, black and
other races. White is the omitted group. Four categories of
educational attainment are created: college, some college,
high school and less than high school. College is defined as
16 or more years of schooling; some college is defined as 13
to 15 years of schooling; high school is 12 years of schooling
and less than high school is 11 or fewer years of schooling.
College is chosen to be the omitted group. Family income is
the total family income in thousands of dollars deflated to
1999 dollars from the previous year of interview. Household
size is the number of persons living in the same family unit.
For marital status, a dichotomous indicator for marriage is
constructed. Similarly for employment status, a dichotomous
indicator is used which is equal to one if the individual is
working now or only temporarily laid off or on
sick/maternity leave, and zero otherwise.
The key variables of the model are the costs of consuming
cigarettes. The costs consist of monetary costs and non-
monetary costs. Cigarette prices come from the state level
weighted average prices per pack in the Tax Burden on
Tobacco (Orzechowski and Walker 2009). All prices are
deflated to 1999 dollars. Non-monetary cost refers to
policies and regulations that increase the degree of
inconvenience for consuming cigarettes. The estimation uses
a smoking ban index that is constructed based on the
smoking restriction decisions of the smoke-free air laws in
the following 12 locations: Government worksites, private
worksites, child care centers, health care facilities,
restaurants, recreational facilities, cultural facilities, public
transit, shopping malls, public schools, private schools, and
free standing bars. Smoking restrictions for some locations
are coded 0, 1,2 and 3, with 0 being no restrictions against
smoking and 3 being smoking banned at all times; other
locations are coded from 1 to 5 in the similar fashion. The
smoking ban index for each state of each year is the sum of
the numerical codes. The price and policy variables are
merged with the PSID data based on the respondent’s state of
residence and the year of interview. The law of demand
predicts negative signs cigarette prices and the smoking ban
index. Table 1 shows the summary statistics of the
independent variables.
V. ESTIMATION
Equation (1) will be estimated as a linear dynamic panel data
model with fixed effects using the two-step system
Generalized Method of Moment (GMM).
Due to the addictive nature of cigarettes, 1, tic is positively
correlated with time invariant preferences for smoking, tiu , .
Therefore, treating all other variables as strictly exogenous,
Proceedings of the Pennsylvania Economic Association 118
OLS estimation of equation (1) will yield upward biased and
inconsistent estimates.
A system of the first differenced equation and the original
equation in levels are estimated
titi
c
ti
c
tititi vxpolpcc ,,3,2,11,1, (2)
titi
c
ti
c
tititi xpolpcc ,,3,2,11,1,
tic , is smoking participation or number of cigarettes
consumed; 1, tic is a predetermined variable: past
participation in smoking 1, tic is not correlated with current
shocks to smoking tiv , , but is affected by the past ones.
System GMM uses the lagged levels of the predetermined
variable as instruments for their first-differenced form and
uses the first differenced predetermined variable (now
exogenous to the fixed effects) to instrument them in levels.
Let tiX , be a vector of all exogenous variables in first
differences ),,,( ,,,, ti
c
ti
c
titi xpolpX . I use tiX ,
to
instrument tiX , in the difference equation and to
instrument themselves in the level equation. One period
lagged price and policy variables can serve as the additional
instruments for the predetermined variables5. Define
),( 1,1,,
c
ti
c
titi polpLP . Instrument matrices are given by6
6,64321
5,5321
4,421
3,31
000
000
000
000
iiiiii
iiiii
iiii
iii
di
LPXcccc
LPXccc
LPXcc
LPXc
Z (3)
6,64321
5,5321
4,421
3,31
000
000
000
000
iiiiii
iiiii
iiii
iii
li
LPXcccc
LPXccc
LPXcc
LPXc
Z (4)
li
di
siZ
ZZ
0
0 (5)
The moment conditions are7
0)( QZE s (6)
where
U
UQ ,
iT
i
i
i
v
v
v
U
4
3
and
iT
i
i
i
v
v
v
U
4
3
VI. RESULTS
This section reports the estimation results. Separate
regressions are done for the full sample, as well as for sub-
samples stratified by age group and sex8. Table 2 reports
OLS parameter estimates. Table 3 contains the GMM
estimates. The p-values associated with the AR (1) and AR
(2) tests are also reported Table 3. Both test the validity of
the instruments. To take into account of any time-specific
common trends, year fixed effects are included in all
regressions.
In Table 2, lagged cigarette consumptions are positive and
significant at the 1% level for all sample specifications.
Since this lagged term is positively correlated with the
unobserved individual fixed effects in the error terms, these
estimates might be upward biased and inconsistent. As such,
estimates of other key parameters might also be biased and
inconsistent. Indeed, the coefficients on cigarette prices have
unstable signs across columns, and are all insignificant.
In Table 3, all coefficients on the lagged cigarette
consumptions are lower in magnitude than those in Table 1.
This is in compliance with the expectation that OLS
overestimates the state dependence. Coefficients on cigarette
price are all negative, as expected, and are significant for
younger adults and females. Smoke-free air laws do not
show any significant effects.
Table 4 reports estimates of price elasticity of demand for
cigarettes. The elasticity is -0.43 for the age group of 50 and
younger, and -0.04 for 51 year-olds and older. For males and
females, their elasticities of demand are -0.27 and -0.49,
respectively. The price elasticity for the full sample of
smokers is -0.26.
The validity of the instruments depends on the assumption of
no autocorrelation (AR) in the error terms. Table 4 also
reports the P-values of the AR tests. The null hypothesis of
the AR test is no autocorrelation. Due to first-differencing in
the GMM estimator, there will be first-order autocorrelation
in the errors. Thus, the AR tests essentially test for second-
order autocorrelations. The P-values indicate that the sub-
sample of female smokers fails the autocorrelation test.
VII. CONCLUSIONS
The MSA has led to unprecedented increases in cigarette
prices in the U.S. since 1998. In this paper, I use data from
Proceedings of the Pennsylvania Economic Association 119
the PSID to examine how large increases in cigarette prices
affect cigarette consumption by smokers. I estimate the price
elasticity of demand for cigarettes for the full sample of
smokers, and for the subsamples stratified by age group and
sex. To my knowledge, all prior studies utilize repeated
cross-sectional data. I take advantage of the panel features of
the PSID, and apply the consistent system GMM estimator to
estimate a dynamic panel data model that accounts for the
reinforcement effects of past consumption on current
consumption. There are three major findings. (1) Older
smokers are virtually nonresponsive to price changes, and
younger smokers are more price sensitive. This may be
because younger smokers have been smoking for a shorter
time and are less addicted to smoking; they are less
financially stable and are affected more by higher prices. (2)
There is limited evidence in this paper that women smokers
might be more price responsive. However, the validity of
this claim is weakened by the failure in the autocorrelation
test. (3) The overall estimated price elasticity for the full
sample is -0.26, lower than the -0.40 to -0.60 range in the
literature. One explanation for this reduced price elasticity
could be that smokers who do not quit when facing
substantially higher prices are likely to be more addicted.
The regression analyses seem to indicate that raising price or
taxes is a more effective policy instrument for reducing
demand for cigarettes among smokers than smoke-free air
laws. This does not mean, however, that smoke-free air laws
are not effective in altering smoking behaviors. They appear
insignificant in the regressions because there is not much
variation in the smoke-free air laws within states during the
time of this analysis. These laws would induce occasional
smokers to quit, or deter nonsmokers from starting smoking
by increasing nonmonetary costs of smoking or by
influencing preferences for smoking.
A shortcoming of this research is that it does not take into
consideration of the effects of future prices on current
cigarette consumption. According to the rational addiction
framework developed by Becker and Murphy (1988),
consumers are forward-looking, and current consumption is
also affected by the anticipated future prices. However,
including a lead price will leave only three waves of data to
use. It would be more feasible to test rational addiction as
more data become available.
11
12
13
14
15
16
17
18
19
1999 2001 2003 2005 2007 2009
Fig.1: Trends in Cigarette Consumption
all age 50 & younger age 51 & older males females
Proceedings of the Pennsylvania Economic Association 120
Table 1. Independent Variables Descriptive Statistics
Variables Definition Mean
Standard
Deviation
Cigarette Price
Price of a pack of 20 cigarettes adjusted by consumer
price index to 1999 dollars
3.42 0.53
Smoking Ban
Index
Numerical values created from the smoking restriction of
the smoke-free air laws
17.87 13.66
Age Age of respondent
43.94 11.61
Male Dichotomous indicator for male
0.47 0.50
White Dichotomous indicator for white
0.61 0.49
Black Dichotomous indicator for black
0.30 0.46
Other Race Dichotomous indicator for race other than white or black
0.10 0.29
Hispanic Dichotomous indicator for Hispanic
0.05 0.21
College Dichotomous indicator for having college degree or higher
0.23 0.42
Some College Dichotomous indicator for having some college
0.14 0.35
High School
Dichotomous indicator for having high school diploma or
equivalent
0.35 0.48
Less than High
School
Dichotomous indicator for not having completed high
school
0.28 0.48
Family Income
Total family income of previous year in thousands of
dollars discounted 1999 dollars
68.24 53.69
Household
Size Number of persons residing in the household
2.92 1.51
Married Dichotomous indicator for being married
0.62 0.48
Employed Dichotomous indicator for employed 0.70 0.46
The maximum sample size is 8394. Some variables may have fewer observations due to non-
response.
Table 2: OLS Parameter Estimates for Smokers Only
Full
Aged 50 and
younger
Aged 51 and
older Males
Females
Independent Variable (1) (2) (3) (4)
(5)
Lagged Cigarette
Consumption 0.571***
0.558***
0.572***
0.592***
0.534***
(0.009)
(0.011)
(0.017)
(0.013)
(0.012)
Cigarette price -0.115
0.096
-0.516
-0.364
0.092
(0.195)
(0.229)
(0.001)
(0.330)
(0.228)
Smoking ban index -0.005
-0.009
-0.001
0.006
-0.015*
(0.039)
(0.009)
(0.014)
(0.013)
(0.009)
Number of observations 5684 3939 1745 2658 3026
All regressions include year fixed effects.
*significant at 10% level, **significant at 5% level, ***significant at 1% level
Proceedings of the Pennsylvania Economic Association 121
Table 3: GMM Parameter Estimates for Smokers Only
Full
Aged 50 and
younger
Aged 51
and older Males
Females
Independent Variable (1) (2) (3) (4)
(5)
Lagged Cigarette
Consumption 0.175***
0.153***
0.206***
0.081
0.356***
(0.045)
(0.045)
(0.080)
(0.056)
(0.063)
Cigarette price -1.059
-1.728**
-0.158
-1.252
-1.798*
(0.895)
(0.781)
(1.485)
(1.458)
(1.012)
Smoking ban index .0023
-0.032
0.024
0.023
-0.007
(0.022)
(0.027)
(0.005)
(0.039)
(0.0037)
Age
-0.317
-0.173
0.071
-0.730*
0.924***
(0.425)
(0.323)
(0.79)1
(0.438)
(0.281)
Male
6.381
2.403
-22.264
-
-
(11.010)
(8.980)
(31.779)
-
-
Black
-2.903
3.117
19.362
0.447
-0.146
(8.536)
(7.912)
(23.856)
(54.767)
(9.563)
Other race
5.99
1.681
11.681
-104.335
2.375
(8.692)
(7.541)
(27.340)
(125.755)
(6.667)
Hispanic
28.839
27.035
-15.157
-32.684
13.059
(22.41)
(20.967)
(55.462)
(235.539)
(19.747)
Some college -4.071
-3.505
12.530
9.555
12.304*
(8.33)
(6.308)
(17.453)
(18.834)
(6.49)
High school
-1.278
4.726
-45.671*
-15.656
-5.915
(6.018)
(4.753)
(23.365)
(31.434)
(4.814)
Less than high school 13.46**
4.587
11.211
6.197
0.074
(5.561)
(6.125)
(15.285)
(38.800)
(5.102)
Family income 0.009*
0.008
0.004
0.005
0.011
(0.005)
(0.008)
(0.005)
(0.011)
(0.008)
Household size 0.817
0.771*
-0.336
0.445
1.463
(0.499)
(0.428)
(1.415)
(0.801)
(0.737)
Married
-1.043
0.281
-0.439
0.358
-0.527
(1.615)
(1.553)
(7.930)
(2.540)
(2.010)
Employed
0.317
0.924
-1.307
1.273
-0.899
(0.614)
(0.661)
(1.316)
(0.906)
(0.809)
p-value AR(1) test 0.000
0.000
0.000
0.000
0.000
p-value AR(2) test 0.443
0.359
0.972
0.903
0.038
Number of observations 5685 3943 1742 2657 3028
Windmeijer finite-sample corrected standard errors are in parentheses.
All regressions include year fixed effects.
*significant at 10% level, **significant at 5% level, ***significant at 1% level
Table 4: Estimates of Price Elasticity of Demand for Cigarettes for Smokers Only
Full Aged 50 and younger
Aged 51 and
older Males Females
Price Elasticity of Demand -0.26 -0.43 -0.04 -0.27 -0.49
The elasticity is calculated at the mean of cigarette prices of 3.42.
Proceedings of the Pennsylvania Economic Association 122
ENDNOTES
1 Source: Center for Disease Control and Prevention (CDC).
“Tobacco Use: Targeting the nation’s leading killer.”
http://www.cdc.gov/chronicdisease/resources/publications/A
AG/osh.htm. Accessed on July 11, 2011.
2 Source: Facts about secondhand smoke.
http://www.co.dakota.mn.us/HealthFamily/HealthyLiving/S
moking/ShSfacts.htm. Accessed on July 15, 2011.
3 Adjusted to 1999 dollars.
4 Because non-smokers should not be affected by increases in
cigarette prices, they are excluded in this study.
5 These are used as IV style instruments in addition to the
GMM style instruments.
6
diZ is the instrument matrix for the differenced equation
for each individual; liZ is the instrument matrix for the level
equation for each individual; siZ is the instrument matrix for
the system of equations for each individual.
7
sZ ,iU and
iU are stacked matrices across individuals.
8 Balanced panels are used.
REFERENCES
Arellano, M., & Bond S. “Some tests of specification for
panel data: Monte Carlo evidence and an application to
employment equations.” The Review of Economic Studies,
April 1991, 58, pp. 277 – 297.
Arellano, M., & Bover, O. “Another look at the instrumental-
variable estimation of error- components models.” Journal of
Econometrics, 1995, 68, 29-52.
Becker, G., Grossman, M., & Murphy K. “An empirical
analysis of cigarette addiction.” The American Economic
Review, 1994, 84(3): 396–418.
Becker, G., & Murphy, K. “A theory of rational addiction.”
The Journal of Political Economy, 1998, 96(4): 675-700.
Blundell, R., & Bond, S. “Initial conditions and moment
restrictions in dynamic panel data models.” Journal of
Econometrics, 1998, 87: 115–143.
Bond, S. “Dynamic panel data models: a guide to micro data
methods and practice.” Institute for Fiscal Studies, WP09/02,
2002.
Bond, S., & Windmeijer, F. “Finite sample inference for
GMM estimators in linear panel data models.” Institute of
Fiscal Studies, WP02/x, 2002.
Cameron, C., & Trivedi, P. Microeconometrics: Methods and
Applications. New York, NY: Cambridge University Press,
2005
Chaloupka, F. “Men, Women, and Addition: the Case of
Cigarette Smoking.” NBER working paper #3267, 1990.
Chaloupka, F. “Rational Addictive Behavior and Cigarette
Smoking.” The Journal of Political Economy, 1991, 99(4):
722-742.
Chaloupka, F.J., & Grossman, M. “Price, tobacco control
policies and youth smoking.” National Bureau of Economic
Research, Inc, NBER Working Papers #5740, 1996.
Chaloupka, F.J., & Warner, K.E. “The Economics of
Smoking.” In The Handbook of Health Economics, ed.
Joseph P. Newhouse and Anthony J. Cuyler, 2000.
Chamberlain, G. “Panel data.” In Handbook of Econometrics,
Griliches Z, Intriligator M (eds), vol. 2. North Holland:
Amsterdam, 1984, 1247–1318.
Chay, Kenneth Y. and Dean Hyslop (1998), “Identification
and Estimation of Dynamic Binary Response Panel Data
Models: Empirical Evidence using Alternative Approaches.”
Center for Labor Economics, Working Paper No. 5, UC
Berkeley.
Farrelly, M., Brady, J., Pechacek. T., Woollery, T.,
“Response by Adults to Increases in Cigarette Prices by
Sociodemographic Characteristics.” Southern Economic
Journal, 2001, 68(1): 156-165.
Greene, W. Econometric Analysis (5th ed). Upper Saddle
River, NJ: Pearson Education, Inc. 2003.
Grossman, M., et al. “Policy watch: Alcohol and cigarette
taxes.” The Journal of Economic Perspective, autumn 1993,
Vol. 7, No. 4, pp. 211-222.
Hajivassiliou, V. “Smooth simulation estimation of panel
data LDV models.” Unpublished Manuscript, Yale
University, 1990.
Hayashi, F. Econometrics. Princeton, NJ: Princeton
University Press, 2000.
Heckman, J. 1981. “The incidental parameters problem and
the problem of initial conditions in estimating a discrete
time–discrete data stochastic process.” In Structural Analysis
of Discrete Data and Econometric Applications, ed. C. F.
Proceedings of the Pennsylvania Economic Association 123
Manski and D. L. McFadden, 178–195. Cambridge, MA:
MIT Press.
Hersch, J., “Gender, Income, Levels, and the Demand for
Cigarettes.” Journal of Risk and Uncertainty, 2000, 21(2/3):
263-282
Holtz-Eakin, D., Newey, W., & ROSEN, H. “Estimating
Vector Autoregressions with Panel Data.” Econometrica,
1988, 56, 1371-1395.
Hsiao, C. Analysis of Panel Data. New York, NY:
Cambridge University Press, 2003.
Keane M. “A computationally practical simulation estimator
for panel data.” Econometrica, 1994, 62, 95-116.
Kenkel, D., Mathios, A., & Pacula, R. “Economics of youth
drug use, addition and gateway effects.” Addiction, 2001, 96,
151-164.
Markowitz, S., & Tauras, J. “Substance use among
adolescent students with consideration of budget constraints.”
Review of Economics of the Household, 2009, 7:423–446.
Nickell, S. “Biases in dynamic models with fixed effects.”
Econometrica, 1981, 49(6): 1417-26.
Orzechowski, W. & Walker R. The tax burden on tobacco:
Historical compilation. Arlington, Virginia, 2009.
Pacula, R. “Economic modeling of the gateway effect.”
Health Economics, 1997, 6: 521–524.
Picone, G., Sloan, F., & Trogdon, J. G. “The effect of the
tobacco settlement and smoking bans on alcohol
consumption.” Health Economics, 2004, 13(10), 1063–1080.
Roodman, D. “How to do xtabond2: an introduction to
“Difference” and “System” GMM in Stata.” Center for
Global Development Working Paper Number 103, December
2006.
Stewart, M. “The Inter-related Dynamics of Unemployment
and Low-Wage Employment.” Journal of Applied
Econometrics, April 2007, Vol.22, No. 3, pp. 511-531.
Townsend, J., Roderick, P., Cooper, J., “Cigarette Smoking
by Socioeconomic Group, Sex, and Age: Effects of Price,
Income, and Health Publicity.” British Medical Journal,
1994, 309: 923-927.
Windmeijer, F. “A finite sample correction for the variance
of linear two-step GMM Estimators.” Journal of
Econometrics, 2005, 126: 25-51.
Wooldridge, J. Econometric Analysis of Cross Section and
Panel Data. MIT Press: Cambridge, MA, 2002.Wooldridge,
J. “Simple solutions to the initial conditions problem in
dynamic, nonlinear panel data models with unobserved
heterogeneity.” Journal of Applied Econometrics, 2005, 20:
39–54.
Proceedings of the Pennsylvania Economic Association 124
MORTGAGE CASH FLOW ANALYSIS AND PRICING USING CAS (COLLATERAL ANALYSIS SYSTEM)
Stephen M. Mansour and Riaz Hussain
The University of Scranton
Kania School of Management
Monroe and Linden Sts.
Scranton, PA 18510
ABSTRACT
The Collateral Analysis System (CAS) was developed as an
analytic database software product primarily serving the
mortgage industry. In this paper we will examine the cash
flow analysis and pricing tools. CAS uses an object-oriented
scripting language to generate cash flows which can be
discounted and priced accordingly. We will look at several
simple examples which can be built upon to create more
complex models.
HISTORY
During the Savings and Loan Crisis of the 1980’s mortgage
companies often acquired loans from several different
sources due to bankruptcies and reorganizations. The data
were stored in differing formats on computer tapes. Many
of these companies hired “tape-crackers” to extract and
reformat the data to company specifications. CAS was
originally developed to facilitate this process. Later on
database queries, reporting and analytical tools were
developed in CAS. The first commercial release of CAS
was in 1996. Currently there are over 40 companies using
the product.
CAS is involved in all facets of the mortgage industry,
including investment and commercial banking, origination,
servicing, finance, accounting and bond insurance. CAS is
used for tape-cracking and importing data, reporting, edit and
error analysis, exporting data, underwriting and automated
data scrubbing, inventory monitoring, secondary marketing,
shadow servicing, pricing, research and analysis, and cash
flow analysis. CAS is written in Dyalog APL to take
advantage of its array-processing capability, graphical user
interface, object-oriented model and quick development time.
CASH FLOW ANALYSIS
CAS generates cash flows from various user inputs including
balance, interest rate, term, prepayment rate, day-count
method, rate type (Fixed or Adjustable), amortization term
and/or payment. These inputs allow the CAS user to analyze
fixed-rate, adjustable-rate, balloon and interest-only
mortgages using various prepayment scenarios including the
CPR (Constant Prepayment Rate) and PSA (Public Securities
Administration) models. Default models including MDR
(Monthly Default Rate) and SDA (Standard Default
Assumption) can also be included; users can also assign loss
severity and recovery periods. Cash Flows generated
include Remaining Balance, Scheduled Principal, Actual
Interest, Voluntary Prepayments, Servicing Fees, Principal
Recoveries and Losses, New Defaults, Loans in Foreclosure
and Lost Interest. Cash flows can be generated for individual
mortgages or for a mortgage pool.
From previously generated cash flows, CAS will calculate
price, yield, modified and Macaulay duration average Life as
well as other industry-specific measurements. Various
prepayment and default assumptions can be set up
dynamically allowing the user to observe the price sensitivity
as well as the sensitivity of other measurements to changes in
the interest rate, prepayment rate, and default assumptions.
Error! Reference source not found. shows the basic logic
flow from mortgage-level data to pool sensitivity analysis.
GENERATING CASH FLOWS
To Generate Cash Flows using the object model, you must
first create a CashFlow object:
CF <- CashFlow.New ''
The left arrow (<-) evaluates the expression on its right and
assigns the result to the name on the left. Thus A<-2+2
assigns the value to the name A.
The Generate method takes two inputs, the coupon and the
balance. It assumes a term of 360 months (30 years). Thus
to amortize $100,000 AT 5.25% enter:
D <- CF.Generate 5.25 100000
The result of Generate is a multi-dimensional object, which
contains the cash flows. To make it readable we can convert
it to a table which is known as a RecordSet. The rows of the
table will be time periods, and the columns will be cash
flows.
R <- D.MakeRecordSet 'Time'
Proceedings of the Pennsylvania Economic Association 125
Since the result is a RecordSet, we can use the Show method
to display the cash flows
R.Show ''
The results are shown in Error! Reference source not
found.
The rows of the RecordSet are the 360 time periods, and the
columns are the following Cash Flows:
PERFBAL – Performing Balance (Balance at beginning of
Period)
ACTAM – Actual Amortization (Principal)
EXPINT – Expected Interest
If we wanted to use model a 15-year mortgage, we could
reset the Periods property to a different term by inserting the
following statement before the Generate method:
CF.Periods <- 180
The record set would now have only 180 records representing
time periods.
Prepayments and Defaults
Most mortgages do not pay as scheduled – Some property
owners pay off early due to sale of the property or
refinancing; others may default on the mortgage. To handle
these situations we introduce the SetPrepaymentRate and
SetRateOfDefault methods.
The simplest type of PrepaymentRate is the CPR – Constant
Prepayment Rate. We can generate CashFlows with an
annualized prepayment rate of 6% by inserting the following
statement immediately before the Generate method:
CF.SetPrepaymentRate 'CPR' 6
After running the script, the display will have a new column:
VOLPREPAY – Voluntary Prepayments
If we are dealing with non-Agency mortgages, we need to
concern ourselves with default rates. The simplest type of
default rate is CDR – Constant Default Rate. To model an
annualized default rate of 0.6% we insert the following
statement:
CF.SetRateOfDefault 'CDR' 0.6
Loss severity is the proportion of the balance which is lost to
the lender upon foreclosure. We will set the Loss Severity at
20%.
CF.LossSeverity <- 20
Now if we run the script, the display will have eight new
columns:
NEWDEF – New Defaults
ADB – Amortized Default Balance
FCL – Foreclosures
AMDEF – Amortization from Defaults
ACTINT – Actual Interest (This will differ from Expected
interest)
LOSTINT – Lost Interest (ACTINT – EXPINT This value
will be negative)
PRINLOST – Principal Lost (This value will be negative)
PRINRECOV – Recovered Principal
Variable rate (ARM)
Generating Cash Flows for an adjustable-rate mortgage
(ARM) requires that we set up the floating rates ahead of
time. The Floating Rates vary over time are depend on the
value of an underlying index. To do this we need to create a
TimeSeries Object containing the index.
IDX <- TimeSeries.New 'Index'
Since the index values are time-dependent we need to set the
Start Date for January 1, 2005:
IDX.StartDateTime <- 20050101 IDX.Periods
<- 360
To simulate rates between 4 and 6% set the Floor and Ceiling
properties:
IDX.Floor <- 4
IDX.Ceiling <- 6
Now generate “LIBOR” rates starting at 5.25%
IDX.Generate 'LIBOR' 5.25
Alternatively, we can simply copy in rates from a Data Table
using the Import method:
IDX.Import 'ARMINDEX'
Now we need to create a second TimeSeries object to reflect
the Floating Rates used in the calculation:
C<-TimeSeries.New 'FloatingRate'
Now set the contract provisions of the loan. The next reset
date is June 1, 2006, reset frequency is six months, reset cap
is 2% and margin is 3%:
Proceedings of the Pennsylvania Economic Association 126
C.ResetDate <- 20050601 C.ResetFrequency
<- 6
C.ResetCap <- 2
C.Margin <- 3
Next, we need to link the floating rate time series to the Index
time series we created previously.
C.Indextable <- IDX
C.Index <- 'LIBOR'
C.StartDate <- 20050101
C.Generate 'ARM1' 4.39
Next, create a Cash Flow object.
CF.CashFlow.New ''
Since the monthly payments will vary with an ARM, add a
new Cash Flow, PANDI
(Principal and Interest)
CF.AddFlow 'PANDI'
Instead of using a constant rate, substitute the floating-rate
time-series object for the coupon rate to generate cashflows
for a $100,000 mortgage:
D <- CF.Generate C 100000
The following two statements will allow us to view the cash
flows:
R <- D.MakeRecordSet 'TIME'
R.Show ''
The results are shown in Error! Reference source not
found.. Observe that the monthly payment (PANDI) changes
periodically in response to the changes in coupon payments
due to the variable rate.
Commercial Mortgages
Some Commercial mortgages have terms which are more
complex. These include interest-only mortgages whose cash
flows are similar to corporate bonds, and balloon mortgages
which amortize like traditional fixed-rate mortgages for a set
period, then require the balance to be paid off at a specific
time. Commercial mortgages often use different day count
methods instead of the standard 30/360 day count that is used
for residential mortgages. Hybrid mortgages combine
interest only, fixed-rate and balloon mortgages. A script to
generate a hybrid mortgage is displayed below, showing the
initial interest-only period of 3 months, the amortization
period of 30 years which determines the monthly payment
and the remaining term of 10 years after which the balloon
payment is due:
CF <- CashFlow.New ''
CF.Periods <- 360
CF.DayCount <- 'ACT/360'
CF.RemainingTerm <- 120
CF.StartDate <- 20050101
CF.IOPeriod <- 3
CF.Payment <-12000
D <- CF.Generate 5.25 2000000
R <- D.MakeRecordSet 'TIME'
R.Show ''
Aggregation of Cash Flows
If we want to generate cash flows for a pool of mortgages, we
need to link the Cash Flow object to a RecordSet, which
contains the individual loans. First we must create the
RecordSet: RS <- RecordSet.New ''
RS.GetCurrentFile ''
Error! Reference source not found. shows the first few
records of a RecordSet containing loan-level data
Now create a new cash flow object, using the record set as
the argument. CF <- CashFlow.New RS
Individual properties can be modified using the SetProp
method; the right argument has two elements; the name of
the property, and the name of a field in the record set:
CF.SetProp 'LoanID' 'LOANID'
CF.SetProp 'Coupon' 'RATE'
CF.SetProp 'Balance' 'CURBAL'
CF.SetProp'Periods' 'CALCRTM'
To generate total cash flows use field names for the coupon
and balance:
D<-CF.Generate'RATE' 'CURBAL'
To generate cash flows for individual loans you must include
the field containing the Loan ID. This is not recommended
for large RecordSets because of memory constraints.
D <- CF.Generate 'LOANID' 'RATE'
'CURBAL'
PRICING
In order to price individual loans, one must first create a
pricing object from the same record set used to create the
cash flows:
Proceedings of the Pennsylvania Economic Association 127
PX <- Pricing.New RS
The pricing object allows you to calculate the price from the
yield or the yield from the price. To calculate the price, you
first need to determine the yield. For example, to create a
field containing the yield assuming a base index of 4.25,
enter the following statements:
YLD<-RS.Evaluate'4.25+Margin'
YLD.Name <- 'YIELD'
YLD.Format <- 'F9.5'
RS.AddField YLD
Now set the yield and other properties for the pricing object:
PX.SetYield 'YIELD'
PX.SettlementDate <- 20050413
To calculate the Price, apply the Analyze method to the Cash
Flow object.
PD <- PX.Analyze CF
To display the pricing information, select the bond
measurements of interest:
PD <- PD.Select 'BONDMEAS =
PRICE,YIELD,AVGLIFE,MODDUR'
PR <- PD.MakeRecordset 'COLL'
PR.Show ''
To update the database with loan level pricing information:
PR.Key <- 'YIELD'
RS.Key <- 'LOANID'
RS.JoinInPlace PR
RS.SetCurrentFile ''
Price Sensitivity
In addition to loan-level pricing, CAS enables us to price
mortgage pools which are the basis for mortgage-backed
securities. We can analyze how prices of mortgage-backed
securities respond to changes in interest rates. Instead of
using static analysis, we observe that the prepayment rate
changes with respect to yield. Suppose the current yield is
5.25%. Let us assume that if the yield increases 25 basis
points, the prepayment rate will decrease from 6% CPR to
3% CPR, and if the yield decreases 25 basis points, the
prepayment rate will increase to 12%. The script to
calculate the price and duration is listed below: CF <- CashFlow.New ''
CF.SetPrepaymentRate 'CPR' (12 6 3)
DC<-CF.Generate 5.25
P<-Pricing.New ''
P.SettlmentDate<-20050101
P.SetYield 5 5.25 5.5
DP <- P.Analyze DC
R<- DP.MakeRecordSet 'PRE'
R<-R.Select '' 'PRICE,YIELD,MODDUR'
R.Show ''
CAS will produce the following output:
PRICE YIELD MODDUR
101.261 5.000 4.938
100.000 5.250 6.970
97.837 5.500 8.488
To find the effective duration, calculate
= 6.849 Note that the effective duration is smaller than the modified
duration of 6.970 because of the embedded call option. (See
Fabozzi et. al. 1994 pp. 170-171).
CONCLUSION
Thus, CAS can take loan level data, convert it into aggregate
cash flows for the pool with various prepayment and default
assumptions, and price the pool accordingly. In addition to
pricing, CAS can perform other calculations such as average
life and duration. If the price of a pool is known, then CAS
can figure out the yield.
The recent mortgage crisis has resulted in the sale of many
mortgage portfolios, often at a discount. The cash flow and
pricing tools in CAS have been used extensively by some of
our clients as an aid in pricing deals based on the valuation of
mortgage pools.
Proceedings of the Pennsylvania Economic Association 128
Figure 1: Pricing Object Model Flow Chart
Figure 2: The first few rows of the mortgage amortization table
TIME PERFBAL ACTAM EXPINT
20130523 100,000.00$ 114.70$ 437.50$
20130623 99,885.30$ 115.21$ 437.00$
20130723 99,770.09$ 115.71$ 436.49$
20130823 99,654.38$ 116.22$ 435.99$
20130923 99,538.17$ 116.72$ 435.48$
20131023 99,421.44$ 117.23$ 434.97$
20131123 99,304.21$ 117.75$ 434.46$
20131223 99,186.46$ 118.26$ 433.94$
20140123 99,068.20$ 118.78$ 433.42$
Proceedings of the Pennsylvania Economic Association 129
Figure 3 - Cash Flows for an ARM
TIME PERFBAL ACTAM EXPINT PANDI
20130523 100,000.00$ 134.34$ 365.83$ 500.17$
20130623 99,865.66$ 134.83$ 365.34$ 500.17$
20130723 99,730.83$ 135.32$ 364.85$ 500.17$
20130823 99,595.51$ 135.82$ 364.35$ 500.17$
20130923 99,459.70$ 136.31$ 363.86$ 500.17$
20131023 99,323.38$ 94.64$ 528.90$ 623.53$
20131123 99,228.75$ 95.14$ 528.39$ 623.53$
20131223 99,133.61$ 95.65$ 527.89$ 623.53$
20140123 99,037.96$ 96.16$ 527.38$ 623.53$
20140223 98,941.80$ 96.67$ 526.87$ 623.53$
20140323 98,845.13$ 97.18$ 526.35$ 623.53$
20140423 98,747.95$ 68.39$ 678.89$ 747.29$ Figure 4 - Record Set containing loan level data
LOANID NAME CITY STATE CURBAL RATE RTERM PANDI
10066 PAUL MANSOUR CHARLOTTESV VA 41764.73 7.000 333 500.00
10322 WILLIAM A BLOCK HADDON HEIG PA 86432.00 7.000 332 467.04
11098 ROBERT A. LOWRY SARASOTA FL 66469.19 6.875 333 446.71
12161 TYRONE R SMITH CHARLESTON SC 55716.77 6.875 334 374.45
13250 HENRY G PROSACK HARRISONBUR VA 75973.15 6.875 333 511.09
13268 ALBERT V. GIAMBALVOHARRISONBUR VA 76175.39 7.500 333 543.99
Figure 5 – Pricing Information at the Loan Level
COLL PRICE YIELD AVGLIFE MODDUR
10066 -$ 7.20 0 0.0000
10322 97.64$ 7.25 18.03953 9.1741
11098 96.92$ 7.20 18.10344 9.2350
12161 96.46$ 7.25 18.10344 9.2071
13250 96.47$ 7.25 18.03814 9.1906
13268 115.78$ 6.00 18.36883 9.9074
13854 96.47$ 7.25 18.03814 9.1906
13862 96.46$ 7.25 18.10344 9.2071
14191 96.47$ 7.25 18.03814 9.1906
REFERENCES
The Bond Market Association, 1990 Standard
Formulas for the Analysis of Mortgage-Backed
Securities and Other Related Securities
Fabozzi, 1997 Handbook of Fixed Income Securities,
Fifth Edition , JohnWiley & Sons
Fabozzi, 1994 The Handbook of Mortgage-Backed
Securities , Sixth Edition, McGraw-Hill
Fabozzi, Ramsey and Ramirez 1994 Collateralized
Mortgage Obligations, Structures and Analysis, 2nd
ed., FJF Associates
Mansour, P. and Mansour S., Using APL Expressions
in Database Operations, APL 1998 Conference
Proceeedngs, 22-26
Mansour, S., 2000 Houses, Windows and DOHRs
ACM SIGAPL APL Quote Quad, Vol. 30 Issue 4,
June 2000, 145-152
Proceedings of the Pennsylvania Economic Association 130
DISCUSSANT COMMENTS
THE GEOGRAPHIC CONCENTRATION OF ECONOMIC ACTIVITY
Stephen M. Mansour
The University of Scranton
Kania School of Management
Monroe and Linden Sts.
Scranton, PA 18510
The concepts in this paper are somewhat esoteric so it is
somewhat difficult to follow what the author is trying to
accomplish. There are some interesting historical data on
land use; mainly that manufacturing and industry has become
more dispersed and that farmland has become more
concentrated. However, the paper fails to discuss in detail
the consequences of these trends.
Thomas Hylton mentions in his book “Save Our Land, Save
Our Towns”, that since the 1950s, Pennsylvania has lost in
excess of 4 million acres of farmland. He claims this is due
mainly to misguided public policies that have encouraged
urban sprawl. The economic consequences of this include
excessive infrastructure costs for roads, bridges, water and
sewer lines to suburban and exurban communities. These
costs are often subsidized by the government. Social
consequences include include a loss of community and the
devastation of inner cities in the 1970’s.
Ironically, Detroit, which is in a manufacturing-dense county,
was destroyed by the very industry which made it an
industrial power. The automotive industry allowed people to
live farther from the industrial centers where they worked, so
the middle class taxpayers were able to move out.
Another issue is the author’s use of the Gini coefficient
which measures inequality. This can be misleading because
it looks at relative wealth rather than absolute wealth (or
concentration in this case). To illustrate this point, the late
Prime Minister Margaret Thatcher responded to Parliament
member Simon Hughes’ question in 1990 about the wealth
gap in Britain:
“What the honorable member is saying is that he would
rather that the poor were poorer, provided that the rich
were less rich. So long as the gap is smaller, they would
rather have the poor poorer. You do not create wealth
and opportunity that way. You do not create a property-
owning democracy that way.”
Is increased concentration of agriculture good or bad? The
“locavore” movement encourages the sale and consumption
of food produced within a 100 mile radius. Farmer’s markets
are a primary example of this. This would suggest that the
concentration of agriculture is not a good thing. On the
other hand more efficient farm yields per acre allow us to use
land more efficiently and for purposes other than farming.
Proceedings of the Pennsylvania Economic Association 131
CREDENCE GOODS AND STATE MANDATED VEHICLE SAFETY INSPECTIONS:
HOW NON-PROFIT INSPECTION SERVICES CAN CORRECT FOR MARKET FAILURE
John McCollough,
Department of Economics
Lamar University
Beaumont, TX 77710
ABSTRACT
The problems associated with asymmetric information and
credence goods are a common worry to consumers who
require the services of technicians with expert knowledge. A
study was designed to see if the concerns of consumers were
justified. This paper reports the empirical results of this
study. Specifically, the test looks at two different samples of
vehicle owners and the repair costs associated with a vehicle
state safety inspection. In one sample the vehicles were
inspected by a ‘non profit’, state affiliated inspection station
while in the second sample the vehicles were inspected by a
‘for profit’ vehicle inspection station. The results suggest that
those vehicles inspected at a ‘for profit’ inspection station
had higher repair costs than those vehicles inspected by a
‘non profit' vehicle’ station.
INTRODUCTION
The issue of credence goods is a special case of asymmetric
information and as such, it can lead to market failure. More
specifically, credence goods deal with service goods
provided by a technician with expert knowledge and,
furthermore, the consumer’s knowledge is much less than the
technician’s (Rasch and Waibel, 2012). There are many
examples of this and it is quite common. Because of this
asymmetric knowledge between the technician and the
consumer, there can be an incentive for the service provider
to exploit the consumer, and as a result of this exploitation a
market failure can arise. With respect to vehicle repairs,
Schneider, (2012) estimates the welfare loss of this market
failure at $8.2 billion. It could be that a majority of
technicians within any one particular trade are honest, but if
perceptions spread among consumers that this trade group
has a proclivity toward exploitation then less service will be
demanded by consumers than is socially optimal (Dulleck
and Kerschbamer, 2006). As a result, this market failure may
even result in environmental damage as consumers decide to
forgo the chance of being exploited and dispose of a product
that could have been repaired for further reuse (McCollough,
2010).
There are many common examples of credence goods. As
suggested above, one example would be services provided by
an auto mechanic. Indeed, vehicle repairs rank first in
customer complaints. However, there are many other
examples which range from services provided by your local
roofer or plumber or even services provided by the medical
profession. Because of the consumer’s reliance on the
mechanic’s expert knowledge, the consumers are at an
information disadvantage and can, therefore, be easily
exploited. Exploitation can take the form of overcharging for
service, providing more service than is needed, or perhaps
even charging for services that never took place (Darby and
Karni, 1973,Webbink,1978).
The objective of this paper is to find evidence of this type of
exploitation and to find out if the consumers’ fears and
suspicions are justified. This paper also attempts to quantify
the exploitation. In addition, this paper will also show how
state-affiliated agencies, acting in the role of safety inspectors
without profit motives can play an important role in
correcting the market failure associated with credence goods.
The hypothesis set out in this paper is that repair costs
associated with vehicle safety inspections provided by quasi-
governmental agencies will be statistically less than repair
costs associated with inspections provided by privately
owned, ‘for-profit’, service stations. The reason for this is
that for- profit service stations have an incentive to provide
more service than is required. For example, a ‘for profit’
service station might require brake work or perhaps a tire
replacement when, in fact, these services are not really
needed in order for the vehicle to pass inspection. Worse yet,
the ‘for-profit’ service station might require certain repair
work before the vehicle can pass the safety inspection, but
then never provide the service, charging the customer for
work that never took place. On average, any difference in
repair costs should represent the cost of the market failure.
A test was designed which compares the repair costs
associated with vehicle safety inspection for residents from
the state of Pennsylvania and for residents from New Jersey.
In Pennsylvania, vehicle owners must have their vehicle
inspected by a ‘for profit’ service station, while in New
Jersey the residents can choose to have their vehicle
inspected by either a ‘for profit’ service station or a ‘not for
profit’ vehicle inspection station. When New Jersey
residents have their vehicle inspected by a ‘non profit’, state-
affiliated, vehicle inspection station, the vehicle is actually
inspected by a private firm that has been contracted out to
Proceedings of the Pennsylvania Economic Association 132
perform all state safety vehicle inspections. Safety inspectors
do not work as state employees, rather they work for the firm
which provides the inspections. Neither the firm nor the
inspectors have a profit motive. These inspectors tell the
vehicle owner what needs to be fixed before the vehicle can
pass inspection. The inspection stations are prohibited from
performing any repairs. The vehicle owner will then fix the
problem at a service station of his or her choosing and then
come back to the state-affiliated inspection station for an
inspection sticker as proof that the vehicle passed inspection.
LITERATURE REVIEW
Due to the nature of credence goods, and the fact that the
expert knowledge required to perform the service is
asymmetric, the services provided can often times be price
insensitive with low price elasticity’s (Peppers and Rogers,
2006). The lower the price elasticity for the service, the
easier it is for disreputable service providers to take
advantage of the consumer. The literature typically cites lack
of competition as the cause for price insensitivity.
Geographic locations are a prime determinant in how
competitive vehicle repairs and vehicle inspection services
are. Typically, the denser the geographic location, the more
competition there is and, hence, the ‘switching costs’ are low.
In other words if it is easy for consumers to find other service
providers then this makes it more difficult for service
providers to overcharge customers.
Rasch and Waibel (2012) state that overcharging for vehicle
repairs occur more frequently in less densely populated, non-
competitive locations. They find that non-competitive, low
density locations just off the interstate overcharge since there
is less chance of repeat business. Customers at these
locations are mainly one time customers just passing through.
They conclude that in more dense geographic locations
where competition is higher, service providers are dependent
on repeat business.
As service providers seek repeat customers, protection of
their reputation can “discipline” service providers, especially
when there is a possibility of repeat business by customers
(Schneider, 2012). However, Hubbard (2002, pg 466) warned
that “Incentives are weaker when consumers are naive about
sellers’ private objectives, believe that sellers are
homogeneous, or when switching costs are high.” As an
example of this, Hubbard (1998) finds that independently
owned service stations are more likely to pass vehicles for
inspection than chain store service stations, new car
dealerships, and tune-up shops, because the latter work on
commission whereas the independent shop is motivated by
repeat business. He also finds that the more inspectors there
are at a service station, the more likely a vehicle is to fail. In
addition, Schneider (2012) states that higher quality service
can be provided by those technicians looking for repeat
business, but the prospect for repeat business must be likely.
In a follow-up study, Hubbard (2002) finds that the
reputation effect does pay off. More specifically, he finds that
consumers are 30% more likely to utilize a service station in
the future if that service station had passed the vehicle for
inspection in the recent past. Biehal (1983) also finds that
consumers make choices with respect to auto repair services
based on previous experiences with repair facilities.
However, with respect to annual state vehicle safety
inspections, the desire for repeat business can actually create
a moral hazard problem. For example, Hubbard (1998) found
that in California private inspection facilities pass vehicles at
twice the rate of state inspection facilities, except in cases
when the emission repairs are covered under a warranty for
late model, low mileage vehicles that are being inspected at
new car dealerships. Interestingly, Hubbard also found that
the inspection failure rate was even slightly lower when the
service provider was located in a more competitive location,
and this he attributes to ‘low switching costs’ and the ease
with which to obtain a second opinion.
Schneider (2012) finds that initial diagnostic fees are lower
for possible repeat customers. This suggests that when
reputation was important, the service provider charged a
lower up front diagnostic fee but Schneider (2012) found no
difference in repair recommendations, repair prices or the
number of legitimate repairs when the service mechanic was
trying to protect his reputation. Schneider concludes that the
ability of consumers to discipline service providers with the
possibility of repeat business is ‘fruitless’.
So, how can consumers protect themselves from
unscrupulous service providers? One way is to obtain a
second opinion. However, second opinions are usually
expensive with respect to either money or time (perhaps
both) for the consumer and the service provider, particularly
when it is cheaper to provide the diagnosis and the repair
service together as opposed to the repair service and
diagnosis taking place separately (Emmons, 1997). In
addition, it is unclear to the consumer if a proper diagnosis
was even performed. Pesendorfer and Wolinsky (2003)
suggest that in competitive markets with competitive prices
efforts by the service provider to provide proper diagnosis
might be sub-optimal.
Therefore, barring a second opinion it is difficult for
consumers to determine if they actually were taken advantage
of because who, other than the service provider, can really
judge if service was required or not. With respect to vehicle
repairs mandated by an annual vehicle inspection, second
opinions can be costly to the consumer, particularly with
respect to time. Typically a consumer must leave their
vehicle for half a day or more with the service provider
giving the second opinion. Typically, this means alternative
transportation must be arranged. Customers then find
themselves in a dilemma. If the vehicle inspection station
Proceedings of the Pennsylvania Economic Association 133
does repair work themselves, which is usually the case in
Pennsylvania, then the customer must decide to either go
ahead or trust the inspector to do the repairs while the vehicle
is still queued up. Or, does the customer take the vehicle in
for a second opinion, requiring additional time and expense
on the part of the customer.
Another common strategy is that the consumers can ask for
the old part back when a part was replaced. This helps to
prevent fraudulent billing and overcharging for work that was
not performed. But it still does not prevent ‘over-treatment’,
which is providing more repairs than necessary (Dulleck and
Kerschbamer, 2006)
States can help to protect consumers from fraudulent repairs
by requiring licensing or certification of service providers.
Unfortunately, this can increase barriers to competition
which reduces competition when more competition should be
encouraged. Often time consumer publications and rating
agencies such as AAA membership clubs or Angie’s list
allow consumers to undertake information searches to help
sort out reliable service providers from the unreliable ones.
However, Biehal (1983) suggests that consumers are not as
proactive in their information search as they need to be. For
example, from a survey of customers who recently had their
vehicles repaired Biehal found that 31.7% of the respondents
felt as if their bills were unreasonable and one-fourth were
dissatisfied with their service. But, the level of customer
dissatisfaction decreased as the amount of customer external
information for auto repair services increased.
Whether or not a repair will fall into the hands of a reputable
or disreputable mechanic, the consumer needs to weigh the
expected benefits against the cost of a repair. The expected
benefits of a repair includes both an expectation that a repair
will be completed correctly as well as an expectation that the
product will have its useful life extended. The more trust a
consumer has in a mechanic (either because the mechanic is
certified, or has an excellent “word of mouth” reputation, or
has been endorsed by a rating agency) the higher the
expectation that a repair will be made correctly. If this
expectation is low then it is more likely that the consumer
will forgo the repair in favor of choosing to replace the
product (McCollough, 2010, pg 189)
EMPIRICAL TEST
An empirical test is designed to see if the repair bills
associated with vehicle state inspections are statistically
different for vehicle owners who go to a ‘not-profit’, state-
affiliated inspection station as opposed to those who go to a
‘for-profit’, private inspection station. Obviously, if the
repair bills at the private inspection stations are statistically
higher than this would suggest evidence of market failures
from an asymmetric information and credence goods point of
view. In other words, according to the literature, technicians
with asymmetric and expert information regarding the repair
and maintenance of a product are thought to have an
incentive to cheat the customer and are in fact doing so. This
then would suggest that there is role for government with
respect to vehicle inspection services because the government
would be able to cut back on unnecessary repair bills by
providing an initial diagnostic and inspection service as in the
case of New Jersey’s vehicle safety inspection program. On
the other hand, if repair bills associated with vehicle
inspections from a private vehicle inspection station are not
statistically different from the state-affiliated vehicle
inspection station, then this might suggest that there is no
need for governments to be involved in the vehicle inspection
business. Finally, if the repair charges associated with vehicle
inspections are statistically less at private vehicle inspection
stations than for the state-affiliated vehicle inspection
stations, then one might conclude that the reputation effect is
at work and that private inspection stations are working hard
to keep customers satisfied. However, it is very troubling to
think that private inspection stations could possibly be
overlooking necessary and important repairs at inspection
time because they are afraid of losing potential long term
clients. On the other hand, it is just as unsettling to think that
the state-affiliated inspectors could be overlooking necessary
and important repairs which are being caught by private
sector inspectors.
The data for the empirical test was taken from the 2005 BLS
annual consumer expenditure survey. In this data set
households are chosen at random from around the country
and the head of the household keeps a bi-weekly diary on
their day to day expenditures. In addition, the head of
household responds to a detailed monthly survey with respect
to purchases that are not routine and do not occur on a daily
or weekly basis. During the interview the respondents are
asked to list their vehicle repair expenditures as well as
annual vehicle registration fees and vehicle inspection fees.
Additional information is also collected on the vehicle’s
make, age and mileage, as well as to whether or not the
vehicle was purchased new or used.
Respondents to the survey from both New Jersey and
Pennsylvania were chosen for the empirical test. Both New
Jersey and Pennsylvania have a vehicle inspection program.
However, the major difference between the two states is that
Pennsylvania requires its residents to have their vehicle
inspected once a year by a private inspection station. These
privately owned service stations could be a large chain of
service stations, or proprietarily owned service station, or
maybe even a car dealership. In this case, the vehicle owner
pays the ‘for-profit’ vehicle inspection station a fee for the
state inspection. If a repair is required to pass the inspection,
the vehicle owner can then opt to have the inspection station
do the repairs or have a different service station do the work.
The car is then re-inspected, and if it passes, the vehicle gets
its annual inspection sticker. In general, many vehicle owners
Proceedings of the Pennsylvania Economic Association 134
simply choose to have the original inspection station perform
the repairs since it more convenient and usually will save
time.
New Jersey residents, on the other hand, can opt to have their
vehicle inspected by one of many state-affiliated vehicle
inspection stations located around the state or by a privately
run inspection station. In the past the state of New Jersey
would actually provide inspection services as an alternative
to the privately run inspection stations. However, at the time
of this data set, New Jersey no longer provided the inspection
service themselves. Instead, they sub-contract this service out
to a private firm. This firm is prohibited from performing any
repairs; they only provide the inspection service. Therefore,
in the case of New Jersey, there is no incentive to cheat the
vehicle owner at a state-affiliated inspection station by
requiring unnecessary repairs. If the inspectors at the state-
affiliated stations find a problem with the vehicle, then the
owner must go to a service station of his or her choice, have
the problem fixed and return to the inspection station for a
final inspection and inspection sticker. There is no charge to
New Jersey residents who use this inspection service, but
there is an inspection fee for those New Jersey residents who
opt to use the inspection service of a private inspection
station in New Jersey.
Other than New Jersey residents having the option to go to a
state-affiliated inspection station, there are two other
important differences between New Jersey and Pennsylvania.
First, in New Jersey residents are required to have their
vehicle inspected only once every 2 years as opposed to
Pennsylvania where vehicles are inspected annually.
Secondly, New Jersey does not mandate vehicle inspections
for vehicles that are five years old or newer.
Only New Jersey and Pennsylvania respondents who had
reported owning only one vehicle were selected for the
empirical test. The reason for this is that from the data set it
is impossible to determine which vehicle was being inspected
for households with two or more vehicles. New Jersey
respondents were also deselected from the data set if their
vehicle was newer than 5 years old since, as previously
stated, those vehicles are not required to have an inspection.
It is important to keep in mind that New Jersey residents are
only required to have their vehicles inspected every other
year, while in Pennsylvania the vehicle must get inspected
once a year. Therefore, the only difference, for example,
between a 2005 Honda CRV owned by a New Jersey resident
and a 2005 Honda CRV owned by a Pennsylvania resident is
that the New Jersey vehicle had not been inspected in two
years while the Pennsylvania vehicle was just inspected the
year before. Because of this fact one would expect that repair
bills for Pennsylvania vehicles to be half the amount of repair
bills for New Jersey vehicles. To correct for this discrepancy
the repair costs reported by Pennsylvania residents were
doubled.
A total of 28 vehicles from New Jersey and 120 vehicles
from Pennsylvania were selected for the empirical test. The
difference in the number of observations between the two
states resulted from the facts stated above, which are vehicles
newer than 5 years old do not need to be inspected in New
Jersey and vehicles in Pennsylvania are inspected twice as
often as those vehicles in New Jersey.
Various vehicle repair bills were cumulated and totaled per
survey respondent. The total of the repair bills per vehicle
constitutes the explanatory variable. The following types of
repairs were included as the explanatory variable; brake
work, tire repair, tire purchases and mounting, front end
alignment, wheel balancing and wheel rotation, steering or
front end work, electrical system work, engine repair or
replacement, exhaust system work, engine cooling system
work, clutch or transmission work, motor tune-up, battery
purchase and installation, and finally other vehicle services,
parts, and equipment. Survey respondents reported other
types of vehicle repairs. However, these repairs were most
likely not associated with passing a vehicle inspection, such
as air conditioning repair, tune-up, body work, or radio
repair.
The empirical test is modeled as follows.
C = b1(S) + b2(F/D) + b3(N) + b4(R) + b5(MSRP) + b6(Y) +
b7(M*A)
The variables are defined as follows:
C = This is the total of repair bills during the month of the
vehicle owner’s annual state inspection as in the case of
Pennsylvania residents or during the month of the annual car
registration for New Jersey residents.
S = A value of ‘0’ was assigned to vehicles owned by New
Jersey residents and a value of ‘1’ was assigned to vehicles
owned by Pennsylvania residents.
F/D = A value of ‘0’ was assigned to a vehicle if it was
manufactured by a foreign car manufacturer such as Toyota
or Volvo. A value of ‘1’ was assigned to the vehicle if it was
manufactured by a domestic car manufacturer such as Ford.
These values were assigned regardless of where the vehicle
was manufactured.
N = A value of ‘0’ was assigned if the vehicle was purchased
by the owner as new and a value of ‘1’ was assigned to the
vehicle if the owner purchased it as used.
R = This is a general vehicle reliability index found on the
website ‘autos.msn.com/research/vip/Reliability.aspx?year’
for that specific make, model, and year.
Proceedings of the Pennsylvania Economic Association 135
MSRP = This is the manufacturer’s suggested retail price
found on the website
‘autos.msn.com/research/vip/Reliability.aspx?year’ for that
specific make, model, and year
Y = This represents the average number of miles driven per
year.
M = This represents the total number of miles on the vehicle.
The results of the regression analysis are exhibited below.
The purpose of this paper is to find empirical evidence in
support of problems associated with asymmetric information
and credence goods. Therefore, determining the factors that
explain vehicle repair costs associated with state inspections
is not the reason for running the empirical tests. Rather the
empirical test is more narrowly focused than that, and that is
to find out whether or not the cost of vehicle repairs
associated with a state inspection is statistically different
depending on whether the vehicle is inspected by a private
inspection station or a state-affiliated inspection station.
Therefore, in this regression the explanatory variable of
interest is the state variable.
Table 1 reports the regression results for the empirical model.
The regression results show that the state variable is positive
and significant at the 6.8% level. The positive coefficient of
$166.17 suggests that those vehicles inspected by privately
owned service station can expect to have, on average, an
additional $166.17 in repair bills during the month of their
state inspection. This amount is relevant on a bi-annual basis
since, as stated above, repair bills for Pennsylvania residents
were doubled to account for the fact that they are inspected
twice as often as New Jersey vehicles Since New Jersey
residents are having their vehicles inspected every other year,
then they can expect to save $166.17 every time they go in
for a state inspection.
The finding from the regression analysis supports the
literature on credence goods and asymmetric information,
meaning that service technicians with superior and expert
knowledge over the customer have an incentive to cheat the
customer, and they are, in fact, doing so. This ‘cheating’ or
even the belief that vehicle owners will be cheated is what
creates the market failure.
It should be pointed out that Poitras and Sutter (2002) have
reported the results of a similar study which looks to see if
vehicle inspections can increase vehicle repair cost. They use
a dataset of 733 vehicle inspections for vehicles that were 12
years or older in 50 different states between the years of 1953
– 1967. They find that state inspections do not increase
repair costs (ie, repair revenue for the inspecting facility).
The difference in these results and the results reported here
are most likely attributed to the time period used in both
studies, the type of vehicles used in the study, and the fact
that this study categorizes vehicles inspected into those
inspected by a for-profit or a non-profit inspection facility.
The regression results also show vehicles produced by
foreign manufacturers have statistically higher repair costs
during the month of their state inspection at the 9.3% level of
significance. This result suggests that repair costs associated
with vehicles from foreign manufacturers is $129.63 higher
than for vehicles from domestic manufacturers. There could
be a number of reasons for this result. First, it could be the
case that parts cost more for vehicles from foreign
manufacturers as opposed to domestic manufacturers.
Second, for whatever reason, it could be that vehicles from
foreign manufacturers are slightly more complicated for
American mechanics to work on. Perhaps American
mechanics have a good deal more experience and training
with vehicles made by domestic manufacturers.
The only other variable that was statistically significant was
total mileage. It was positive and significant at the 5.2%
level. The value of the coefficient suggests that for each
addition mile on the vehicle, owners can expect to pay an
additional $.002. This result should be the most intuitive of
all the explanatory variables. The higher the mileage on the
vehicle, the more it cost to maintain
The remaining variables all turned out to be insignificant,
including the number of miles driven over the most recent
year. The coefficient of determination was .229. The
consumer expenditure survey data lacked one or two other
relevant pieces of information which could have increased
the coefficient of determination. This would be the labor
rates charged per vehicle repair shop and information
regarding each shop’s productivity. However, as stated
above, the primary focus of this paper is to find empirical
evidence on the problems associated with asymmetric
information and credence goods.
The data set yields socio-economic characteristics on the
survey respondents. However, it is interesting to point out
that each of these socio-economic characteristics was highly
insignificant. Meaning that, on average, a survey
respondent’s race, gender, age, education level or income
level was insignificant in determining how much they paid
with respect to vehicle repair bills in the month of their state
inspection. Suspicions that a vehicle owner was being taken
advantage of based on their gender, race, age, etc. were
unfounded in this empirical test.
CONCLUSION
Credence good types of services provided by technicians that
are characterized as yielding asymmetric information leave
the consumers at an information disadvantage. This creates
an opportunity for unscrupulous service providers to take
Proceedings of the Pennsylvania Economic Association 136
advantage of the consumer. Indeed, from time to time one
hears stories in the news media of consumers being taken
advantage of. As a result, a market failure arises. An
empirical test was designed and reported on in this paper to
see if the consumer’s fears are warranted.
The empirical test in this paper looks at vehicle owners who
have had their vehicles inspected by either a for-profit
inspection service station or a state affiliated, ‘non-profit’
inspection station. The results from this test indicate that if
your vehicle is inspected by a state-affiliated, ‘non-profit’,
inspection station rather than a ‘for-profit; inspection station,
the vehicle repair bills will be less. The amount as reported
from the regression results is $166.17 on a bi-annual basis.
This is a meaningful savings for owners who utilize the ‘non-
profit’, state affiliated inspection stations. However, there are
a number of factors to consider when interpreting these
results. First, there is the cost of providing vehicle inspection
services. Those vehicle owners in New Jersey do not directly
pay an inspection fee whereas a direct fee is paid to the ‘for
profit’ inspection station in Pennsylvania (when
accumulating repair costs, the cost of the vehicle inspection
fee was not included). The bi-annual savings of $166.17 in
repair bills from New Jersey residents have to be compared
to the cost of running the vehicle inspection stations in New
Jersey. The costs of running the state-affiliated’ inspection
stations are funded by the New Jersey taxpayers.
Secondly, we do not know for sure if vehicle owners in
Pennsylvania are getting more thorough and higher quality
inspections. Perhaps the state-affiliated inspectors from New
Jersey are simply shirking their duties and passing vehicles
that, in reality, do require some repair and maintenance.
There is no way to tell for certain. Although it is beyond the
scope of this paper, it might be possible to look at highway
traffic accident data and see if there is a correlation between
increased accidents and vehicles inspected by state-affiliated
inspection stations. However, the literature with respect to
this topic is inconclusive and shows conflicting results
between vehicle safety inspections and their effectiveness at
preventing accidents (For example, see Fosser 1992, White
1985, or Merrell, D., Poitras, M. & Sutter, D. 1999)
Finally, it need not be that inspections stations that provide
diagnostic inspection services only and none of the required
repairs do not have to be state-affiliated or state operated.
This type of service could just as easily be provided by the
private sector, and perhaps the private sector could perform
the services more efficiently than the state affiliated or state
inspection station. And then it might just be possible that
these private sector companies could run their services more
efficiently than the state affiliated inspection company.
Tables
Table 1 – Regression Results for the Empirical Model
Variable Coefficient “t” – value Significance
State 166.17 1.84 .068
Foreign or domestic 129.63 1.692 .093
Purchase new or used 29.19 .394 .694
Reliability index -21.50 -1.154 .251
MSRP .005 1.009 .315
Annual mileage -.005 -1.225 .223
Total mileage .002 1.956 .052
R-sq .229
REFERENCES
Biehal, G. J. 1983. Consumers' prior experiences and
perceptions in auto repair choice. The Journal of Marketing.
3, summer 1983: 82-91.
Darby, M. & Karni, E, 1973. Free Competition and the
Optimal Amount of Fraud. Journal of Law and Economics.
16, April 1973: 67-88.
Dulleck, U., & Kerschbamer, R. 2006. On doctors,
mechanics, and computer specialists: The economics of
credence goods. Journal of Economic Literature. 44, March
2006: 5-42.
Emons, W. 1997. Credence goods and fraudulent experts.
The Rand Journal of Economics. 28, Spring 1997: 107-119.
Fosser, S. 1992. An experimental evaluation of the effects of
periodic motor vehicle inspection on accident rates. Accident
Analysis & Prevention. 24 December 1992: 599-612.
Proceedings of the Pennsylvania Economic Association 137
Hubbard, Thomas 1998, An Empirical Examination of Moral
Hazard in the Vehicle Inspection Market. Journal of
Economics. 29 Summer 1998: pg 406 – 426.
Hubbard, Thomas 2002. Ho Do Consumers Motivate
Experts? Reputational Incentives in an Auto Repair Market.
Journal of Law and Economics. 45. No. 2, Part 1. Oct. 2002:
437-468
McCollough, J. 2010. Consumer Discount Rates and the
Decision to Repair or Replace a Durable Product: A
Sustainable Consumption Issue. Journal of Economic Issues.
44, March 2010: 183-204.
Merrell, D., Poitras, M., & Sutter, D. 1999. The effectiveness
of vehicle safety inspections: An analysis using panel data.
Southern Economic Journal. 65, January 1999: 571-583.
Peppers, D., & Rogers, M.2006. CUSTOMER
SEGMENTATION STRATEGIES. Pricing on Purpose:
Creating and Capturing Value, 197.
Pesendorfer, W., & Wolinsky, A. 2003. Second opinions and
price competition: Inefficiency in the market for expert
advice. The Review of Economic Studies. 70(2), 2003: 417-
437.
Poitras, M., & Sutter, D. 2002. Policy ineffectiveness or
offsetting behavior? An analysis of vehicle safety
inspections. Southern Economic Journal. 68, April 2002:
922-934.
Rasch, A., & Waibel, C. 2012. What drives fraud in a
credence goods market?-Evidence from a field experiment
(No. 03-07). Cologne Graduate School in Management,
Economics and Social Sciences.
Schneider, H. S. 2012. Agency problems and reputation in
expert services: Evidence from auto repair. Journal of
Industrial Economics. 60, September 2012: 406-433
Webbink, D. W. 1978. Automobile Repair: Does Regulation
or Consumer Information Matter?. The Journal of Consumer
Research. 5, December 1978: 206-209.
White, W. T. 1986. Does periodic vehicle inspection prevent
accidents?. Accident Analysis and Prevention.
18, February 1986: 51-62
Proceedings of the Pennsylvania Economic Association 138
SOME POSSIBLE REASONS FOR THE IRRATIONAL CHOICE OF GIFT CARDS
David Nugent
Robert Morris University
Moon Township, PA 15108
ABSTRACT
This paper is a proposal for a study that will address the
purchase of gift cards. Topics include economic theory that
suggests that the utility of a gift card is less than the utility of
an equal amount of cash. Reasons for the choice of cash gifts,
gift cards and tangible gifts are addressed, with a focus on
gift cards. A number of gift-card related hypotheses are
presented. A questionnaire is proposed that will include
questions related to the hypotheses. Analyses of the
questionnaire data will determine if data is consistent with
the hypotheses.
INTRODUCTION
This paper is a proposal for a study that will address the
question of why people purchase gift cards even though an
equal gift of cash would provide greater utility to most
recipients. Types of gifts given include cash, gift cards and
tangible gifts. In recent years, the purchase of gift cards has
become popular. Card Hub (2013) reports that gift card sales
in 2011 were approximately $99 billion.
Economic theory suggests that a gift of cash should provide
more utility to a recipient than a gift card. Economic
textbooks (Samuelson, 1980; Schiller, 2008) suggest that if a
person had some amount of cash, that person would use the
money to purchase a “market basket” of products and
services that would maximize utility. The market basket
would consist of a combination of housing, food,
transportation, clothing, appliances, etc. that is consistent
with the person’s tastes and preferences. Different people
have different tastes and preferences. Accordingly, market
baskets of purchases also differ. A set of purchases that
maximizes utility for one person will likely not maximize
utility for most other people. If a person has cash, that person
has the choice of buying whatever will maximize utility. But,
if that same person has a gift card for a particular store or
restaurant, it is likely that utility will be less than if the
person had cash that could be spent anywhere. To give a gift
card that that is likely to provide the recipient with less utility
than an equal gift of cash seems to be irrational.
Evidence that at least some recipients would prefer cash to
gift cards includes the market for gift cards. ABC News
(2012) reported that there are several gift card exchange
websites where holders of gift cards can sell their gifts cards
for amounts less than face value. Offenberg (2007) conducted
a study of gift card auctions on eBay for 25 merchants.
Results showed that average sales prices were approximately
15% less than the face value of the gift cards. When fees and
other expenses are considered, the effective discount was
approximately 20%.
REASONS FOR GIVING TANGIBLE GIFTS
Before addressing reasons for choosing gift cards, consider
the decision to give either cash or tangible gifts. For reasons
similar to the preceding argument that a gift card has less
value to a recipient than cash, a tangible gift is likely to have
less value to a typical recipient than the cash spent to
purchase the gift. Waldfogel (1993) conducted a study in
which gift recipients were asked to estimate the cost of gifts
received, and to estimate the value to them, measured as
either the maximum that the recipient would pay for the gifts,
or the minimum that the recipient would accept in lieu of the
gifts. The results indicated that value fell short of cost by
between ten percent and one-third.
Although the monetary value to recipients may be less than
the costs of gifts, it is commonly recognized that the
significance of a gift can be more than simply the money
spent. Studies that address gift-giving (Belk, 1976; Caplow,
1982; Camerer, 1988; Davies, Whelan, Foley and Walsh,
2010; Sherry, 1983; Solnick and Hemenway, 1996; Webley
and Wilson, 1989 ) suggest that the topic is complicated.
Reasons may include expressing affection, reinforcing and
improving personal relationships, facilitating social bonding,
and provide a long-lasting, tangible reminder of the gift-
giver. In choosing gifts, a gift-giver will expend time and
effort to select and purchase an appropriate, thoughtful
expression of an understanding of a recipient’s likes and
dislikes. A resulting gift may be appreciated by a recipient
not because the gift has monetary value, but instead because
of the symbolism that it represents, or because it has
Proceedings of the Pennsylvania Economic Association 139
sentimental value or because it serves as a remembrance of
the giver. In some cases, a gift may have little or no monetary
value. For example, suppose that a 5-year old girl were to
give her grandmother a home-made piece of pottery. A piece
of pottery made by a 5-year old probably has zero market
value, but to the grandmother it could be a prized possession
prominently displayed in her china cabinet.
In contrast to tangible gifts, a gift of cash will likely not
provide a lasting reminder of the gift-giver. A gift of cash
may be viewed as impersonal and as not entailing time and
effort and thought. Imagine if a group of people were to
gather to exchange gifts on a gift-related holiday. If each
person were to give each other person a $50 bill, the result
would be each person ending the day where they started.
Alternately, if some people gave $50 bills and others gave
$20 bills, the inequality of gifts would be immediately
apparent. If the goal of gift exchange were a reciprocal
exchange of equal value, those giving the smaller cash gifts
would be embarrassed and possibly viewed as not fulfilling
the expectation of reciprocity.
The exchange of tangible gifts may give rise to uncertainty
regarding cost that may obscure differences between gifts. If
two people exchange gifts that may or may not have cost $50
each, the potential embarrassment of unequal gifts is
diminished. If each gift entailed time and effort and
thoughtfulness to select and purchase, the significance of
unequal cost would be further diminished.
In comparing cash gifts, gift cards, and tangible gifts, gift
cards have many of the same shortcomings as cash. The cost
of the gift card is explicitly stated. If people were to
exchange gift cards, any inequality of gifts would be
apparent. The recipients of a gift card may perceive that the
purchase of a gift card required little time and effort.
If a person is considering the purchase of a tangible gift, the
non-monetary aspects of the gift may make it an appropriate
choice. However, if a person is considering the purchase of a
gift card, economic theory, as addressed in the preceding,
suggests that a gift card is an irrational choice. A gift of cash
would be more likely to maximize recipient utility.
In some circumstances cash gifts are made. For example,
parents sometimes give substantial amounts of cash to
children and grandchildren with no expectation of equal
reciprocity of cash gifts. In such cases, cash gifts are
appropriate. Imagine, in contrast, if a parent were to give
each child and grandchild a $10,000 gift card to a national
restaurant chain. It seems likely that each recipient would
prefer the cash.
HYPOTHESES
Consider next possible reasons why a gift-giver may make
the seemingly irrational choice of gift cards.
Discounts and Rebates for Purchasers of Gift Cards
The argument that a gift card purchase is irrational is based
on the assumption that the cost of a gift card is the same as
the face value of the gift card. If a purchaser were to receive a
rebate, or if a gift card could be purchased at a discount, or if
some other benefit were available to purchasers, then the
purchase of a gift card may provide the purchaser an
incentive to give a gift card rather than cash. An example of
such an incentive would be the discounts on gasoline
provided by the Giant Eagle supermarket chain. Giant Eagle
(2013) has an incentive program that allows customers 10
cents per gallon discount for up to 30 gallons on Giant Eagle
gasoline for each $50 that a customer spends at Giant Eagle.
Qualified purchases include the purchase of gift cards for
more than 150 stores and restaurants.
If a customer had a gas tank large enough to hold 30 gallons,
the potential discount would be $3.00 ($0.10 X 30 = $3.00).
Accordingly, the purchaser of a $50 gift card could
potentially save on gasoline an amount equal to 6% ($3.00 /
$50.00) of the cost of the gift card.
If the equivalent monetary value to the recipient falls short by
more than the discount or other benefit received by the
purchaser, the net benefit of a gift card would be negative.
If modest incentives influence purchasers of gift cards, the
following can be hypothesized:
H1: Gift-givers are more inclined to purchase
gifts cards if an incentive such as a
discount or rebate is available.
Reduction of Social Risk
The complicated nature of selecting an appropriate tangible
gift may make gift selection difficult. Austin and Huang
(2012) suggest that if a gift-giver lacks knowledge of a
recipient’s needs and preferences, the choice of a gift may
Proceedings of the Pennsylvania Economic Association 140
expose a gift-giver to social risk. A gift-giver may feel
anxiety about choosing the wrong gift, and may be inclined
to reduce exposure to social risk by choosing a gift card.
The preceding leads to the following hypothesis:
H2: Gift-givers are more inclined to purchase
gift cards if their knowledge of the needs
and preferences of recipients is limited.
Possible Preference For Gift Card Characteristics
Although social risk may make a gift-giver inclined to choose
a gift card instead of a tangible gift, it could similarly be
argued that social risk could make a gift-giver inclined to
give cash instead of a tangible gift. Reasons for a preference
for gift cards may be characteristics that make gift cards
seem more appropriate than cash.
A recipient may perceive the selection of a gift card as
entailing some thoughtfulness by the gift-giver. The purchase
of a gift card may entail some time and effort, particularly if
it is for a store or restaurant that the buyer knows that the
recipient regularly patronizes.
Valentin and Allred (2012) suggest that liquidity may
influence the selection of a gift card. If a gift-giver wants to
provide a recipient with a gift that can be readily used to
acquire products or services that the recipient values, but the
gift-giver does not want to simply give cash, a gift card may
be the preferred choice.
This leads to the following hypothesis:
H3: Gift-givers are more inclined to purchase
gift cards if their objective is to provide
recipients with a wide range of product
choices.
Reduction of Recipient Guilt and Regret
Studies that address the emotional aspects of purchases
(Burnett and Lunsford, 1994; Keinan and Kivetz, 2008;
Kivetz and Keinan, 2006) suggest that unneeded or
extravagant purchases may cause a person to feel guilt. A
prudent person may refrain from extravagant purchases in
favor of more responsible purchases. However, over time,
people who have denied themselves luxuries and extravagant
purchases may feel regret because of what they missed.
If a person were to give a gift card for luxury products and
services, the gift card would provide the recipient with an
excuse to indulge in something that ordinarily would be an
unnecessary extravagance. By giving such a gift card, the
gift-giver would be relieving the recipient of the guilt and
regret that otherwise would arise.
This leads to the following hypothesis:
H4: Gift-givers are more inclined to purchase
gift cards if their objective is to provide
recipients with excuses to indulge in
extravagant luxury products.
RESEARCH DESIGN
At this time the design of the research instrument is not
complete. The general approach to data collection will entail
the completion of questionnaires by subjects who have
knowledge of buying and receiving gift cards. Potential
subjects may include college students.
Questions may include:
Have you received gifts cards?
Have you purchased gift cards to give to others?
In choosing gift cards, have you chosen a particular card
because the purchase resulted in a discount, rebate or other
incentive?
Have you given gift cards instead of tangible gifts because
you did not know the kind of tangible gifts that recipients
would prefer?
Have you given gift cards that you felt would provide the
recipients with a wide range of redemption choices?
Have you given gift cards for extravagant luxury products
because you wanted the recipient to have an excuse to
indulge?
Do you consider a gift card to be a more appropriate gift than
cash?
Questionnaire items may also include responses on a scale
that could be quantified for analysis purposes. For example, a
Proceedings of the Pennsylvania Economic Association 141
question might ask subjects to state the degree to which they
feel that cash is an appropriate or inappropriate gift. Another
question might ask subjects to state the degree of inclination
to either purchase a gift card or to give cash.
Analysis of results will entail determining whether
questionnaire responses are consistent with the hypotheses.
CONCLUSIONS
This paper is a proposal for a study that will address the
question of why people purchase gift cards even though an
equal gift of cash would provide greater utility to most
recipients. Economic theory suggests that if a person has
cash, a person will purchase a mixture of products and
services that maximizes utility. A gift card for a specific store
or restaurant limits a person’s choices and likely would result
in less utility than if the person had received cash.
The study addresses three categories of gifts: cash, gift cards
and tangible gifts. For reasons similar to the argument against
gift cards, tangible gifts are likely to provide less utility than
cash. However, for a variety of reasons such as reinforcing
personal relationships, social bonding and other social
factors, tangible gifts may be viewed as more appropriate
than either gift cards or cash. Gift cards generally do not
provide for the social facilitation that tangible gifts provide.
A number of hypotheses are presented that address the topic
of the choice of gift cards. Topics addressed include
discounts and rebates for purchasers of gift cards, reduction
of social risk, possible preference for gift card characteristics,
and reduction of recipient guilt and regret.
This paper proposes that data will be gathered through a
questionnaire that will include questions related to choice of
gift cards. Analyses of the data will determine if data is
consistent with the hypotheses.
REFERENCES
ABC News Website, Hate That Gift Card? Trade It, Sell It,
www.abcnews.go.com/Business/hate-gift-care-trade-
sell/story?id=18083809 retrieved 5/27/2013.
Austin, Caroline Graham and Lei Huang, 2012, First Choice?
Last Resort? Social Risks and Gift Card Selection, Journal of
Marketing Theory and Practice Volume 20, No 3, pp. 293-
305.
Belk, Russell W., 1976, It’s the Thought that Counts: A
Signed Digraph Analysis of Gift Giving, Journal of
Consumer Research Volume 3, pp. 155-162.
Burnett, Melissa S. and Dale A. Lunsford, 1994,
Conceptualizing Guilt in the Consumer Decision-Making
Process, The Journal of Consumer Marketing” Volume 11,
No 3, pp. 33-43.
Camerer, Colin, 1988, Gifts as Economic Signals and Social
Symbols, American Journal of Sociology Volume 94, pp.
S180-S214.
Caplow, Theodore, 1982, Christmas Gifts and Kin Networks,
American Sociological Review Volume 47, No 3, pp. 383-
392.
Card Hub Website, Gift Card Market Size,
www.cardhub.com/edu/gift-card-market-size/ retrieved
5/27/2013.
Davies, Gary, Susan Whelan, Anthony Foley and Margaret
Walsh, 2010, Gifts and Giving, International Journal of
Management Reviews Volume 12, pp. 413-434.
Giant Eagle Website,
www.gianteagle.com/Save.fuelperks/fuelperks-Rules-
Regulations/ retrieved 5/22/2013.
Keinan, Anat and Ran Kivetz, 2008, Remedying Hyperopia:
The effects of Self-Control Regret on Consumer Behavior,
Journal of Marketing Research Volume XLV, pp. 676-689.
Kivetz, Ran and Anat Keinan, 2006, Repenting Hyperopia:
An Analysis of Self-Control Regrets, Journal of Consumer
Research Volume 33, Issue 2, pp. 273-282.
Offenberg, Jennifer Pate, 2007, Markets: Gift Cards, Journal
of Economic Perspectives Volume 21, No 2, pp. 227-238.
Samuelson, Paul A., 1980, Economics, New York: McGraw-
Hill.
Schiller, Bradley R., The Economy Today, New York:
McGraw-Hill Irwin.
Proceedings of the Pennsylvania Economic Association 142
Sherry, John F., 1983, Gift Giving in Anthropological
Perspective, Journal of Consumer Research Volume 10, pp.
157-168.
Solnick, Sara J. and David Hemenway, 1996, The
Deadweight Loss of Christmas: Comment, The American
Economic Review Volume 86, No 5, pp. 1299-1305.
Valentin, Erhard K. and Anthony T. Allred, 2012, Giving and
getting Gift Cards, Journal of Consumer Marketing Volume
29, No 4, pp. 271-279.
Waldfofgel, Joel, 1993, The Deadweight Loss of Christmas,
The American Economic Review Volume 83, No 5, pp.
1328-1336.
Webley, Paul and Richenda Wilson, 1989, Social
Relationships and the Unacceptability of Money as a Gift,
Journal of Social Psychology Volume 129, Issue 1, pp. 85-
91.
Proceedings of the Pennsylvania Economic Association 143
SERVICE QUALITY IN THE U.S. AIRLINE INDUSTRY: FACTORS AFFECTING CUSTOMER SATISFACTION
Logyn Pezak
Rose Sebastianelli
Department of Operations & Information Management
Kania School of Management
University of Scranton, Scranton, PA 18510
ABSTRACT
Airlines have consistently received low customer satisfaction
ratings, with the industry ranking third from the bottom in
terms of the American Customer Satisfaction Index (ACSI).
Using data obtained from the U.S. Department of
Transportation’s Air Travel Consumer Report, we examine
the relationship between ACSI ratings and a number of
explanatory variables related to service quality. Regression
results indicate that in addition to time period and percentage
of on-time arrivals, ACSI ratings are significantly impacted
by four specific complaint categories. These four categories
are complaints related to (1) reservations, ticketing and
boarding, (2) baggage, (3) customer service, and (4)
disability.
INTRODUCTION
Airlines are notoriously noted for poor service quality. In
2012, the airline industry had the third lowest American
Customer Satisfaction Index (ACSI) rating among all
industries (American Customer Satisfaction Index, 2012). Its
ACSI rating actually decreased 6.9% from 1995, the first
year it was measured. While technological advances over this
period would presumably have led to improved service
quality and higher levels of customer satisfaction, this has not
been the case for most U.S. airlines. Airlines are
simultaneously faced with satisfying customers that expect
lower prices, better service quality, and higher comfort levels
while operating profitably in a competitive business
environment with a number of economic challenges (J.D.
Power and Associates, 2012). In order to be successful,
however, airlines must understand that financial well-being
depends on sales, and sales depend on meeting and/or
exceeding customer expectations. Airlines that recognize the
importance of satisfying customers are those that will
continue to be financially viable and profitable in the long
term.
This paper reports the findings of an empirical study
designed to determine what factors are significant in
explaining customer satisfaction in the U.S. airline industry.
We use ACSI ratings for U.S. airline companies as the
measure of customer satisfaction. While a number of factors
affect customer satisfaction, our focus is on the service
quality provided by airlines. By doing so, we limit attention
to factors over which airline companies have some control.
Most of the independent variables included in our study
involve specific categories of airline customer complaints.
Complaints play a large part in any satisfaction rating system.
For example, The Wall Street Journal annually ranks airlines.
Its rankings showed that Southwest is at No. 1 with the
lowest number of customer complaints among U.S. airlines
in 2012. Consequently, determining which customer
complaint issues have the most impact on customer
satisfaction should help airlines to better manage service
quality and meet the expectations of its customers.
RELEVANT LITERAURE
A number of studies have examined service quality within
the airline industry. Most recently, Degirmenci et al. (2012)
used the SERVQUAL scale to measure customer satisfaction
levels among Turkish Airline passengers. The surveys used
were designed according to Skytrax, the most accepted and
prestigious official airline quality 5-star rating system. The
results of this study showed that meals and passenger
transferring services have the highest impact on customer
satisfaction. Huang (2009) also employed a marketing
approach to examine service quality in the airline industry.
Based on a sample of airline passengers in Taiwan, the study
explored how companies’ responses to “other-customer
failures” influenced customer satisfaction. Customers may
become dissatisfied with a company because of how other
customers act or behave. Employee efforts and
compensation were considered as contributing factors in this
study. An important finding was that companies can benefit
from encouraging dissatisfied customers to voice their
complaints as it gives them the opportunity to make amends
and identify root causes responsible for service failures.
The terrorist attacks of September 11, 2001, are perhaps the
most important singular event to impact the U.S. airline
industry. Cunningham, Young, and Lee (2004) examined its
effect on customer perceptions of airline service quality using
the SERVPERF scale. They found that although the number
Proceedings of the Pennsylvania Economic Association 144
of airline trips declined following 9/11, there was no
statistically significant change in overall satisfaction levels
among airline passengers. These results suggest that events
perceived as being “outside” the direct control of airlines
may have limited to no effect on customer satisfaction levels.
Our study most closely resembles that of Rhodes and
Waguespack (2000) in that they too examined complaint data
(obtained from the Department of Transportation’s Air
Travel Consumer Report). Their goal was to use these data
to evaluate the level of service quality both within the airline
industry and between airline companies. Based on data
collected for the period from 1987 to 1996, they found that
airline quality was improving, with Southwest having the
lowest complaint rate (3 complaints per 10,000 departures) of
all airlines. Their study also found that many airline
passengers associate service quality with safety. They did
not link complaint data, however, to any measure of customer
satisfaction.
RESEARCH STUDY
Objectives
The main objective of this empirical study is to identify
factors that have a significant impact on customer satisfaction
in the U.S. airline industry. Primarily, we examine the
relationship between customer satisfaction ratings (ACSI
rating) of U.S. airlines and a set of explanatory variables,
most representing the specific categories of complaint data
compiled by the U.S. Department of Transportation. In
addition, we also examine the differences in mean ACSI
ratings among the U.S. airlines included in our study.
Variables and Data Collection
The secondary data for this study were obtained from two
sources: The American Consumer Satisfaction Index (ACSI)
and the Department of Transportation’s Air Travel Consumer
Report. Data were collected for seven U.S. airlines
(American, Continental, Delta, Northwest, Southwest,
United, and US Airways) that were in operation during the
14-year period from 1998 to 2011. Two airlines (Northwest
and Continental) were not in operation during the entire
period as a result of mergers.
The dependent variable is the annual ACSI rating for each
airline. The ACSI rating is a national benchmark of customer
satisfaction produced by a private company based in Ann
Arbor, Michigan. It aims to quantify customer satisfaction
and quality perceptions and relate them to firm financial
performance (Brecka, 1994). It is based on a sample of 250
customer interviews, with more than 70,000 interviews
conducted annually using random samples via telephone and
email (Carey, 2012). The ACSI quantifies how perceived
quality, perceived value, and customer expectations affect
each other in addition to how they affect customer
satisfaction.
Data for the independent (explanatory) variables were
obtained from the Air Travel Consumer Report, a report
published monthly by the U.S. Department of
Transportation. With the exception of Percentage of On-
Time Arrivals, all independent variables represent specific
passenger complaint categories. Please refer to Table 1 for a
list of the variables used in this study and a brief description
of each.
Several adjustments to the data were required in order to
make them comparable across airlines and with the time
frame of the ACSI rating. To make the complaint data
comparable across airlines, it was necessary to account for
differences in company size and number of enplaned
passengers. This was done by computing an annual index for
each complaint type per airline per year. This was achieved
by dividing the number of complaints in a particular category
in a given year by the total number of complaints for that
airline in the same year. This was then multiplied by the
complaints per 100,000 enplanements (given in the report)
for that airline and year. These indexes made it possible to
compare specific complaint categories across airlines as the
index now represented the number of a specific type of
complaint per 100,000 enplaned passengers.
The ACSI ratings for each airline are published in June and
reflect the previous 12 months (i.e., a particular year’s ACSI
rating reflects data from June of the previous year to May of
the current year). However, the annual complaint data
(independent variables) obtained from the Air Travel
Consumer Report represent calendar years (i.e., January
through December). In order to make the time periods
comparable between the dependent variable (ACSI rating)
and the independent variables, we adjusted the ACSI ratings
so they too represented a calendar year. To do this,
proportions were taken to combine two years of ACSI data
into one representing a standard calendar year. For example,
the ACSI rating for the year 2000 in this study is actually
five-twelfths of the ACSI rating for 2000 and seven-twelfths
of the ACSI rating for 2001.
Proceedings of the Pennsylvania Economic Association 145
Data Analysis and Results
Before determining which explanatory variables have a
significant impact on the ACSI ratings of U.S. airlines, we
first examine if (and how) the mean ACSI ratings might
differ among the seven airline companies in our study. Given
that the annual airline ACSI ratings may be influenced by
economic and/or other factors related to time period, we use
year as the blocking variable. Consequently, we perform an
analysis of variance (ANOVA) using one factor (airline) and
one block (year), the response variable being annual ACSI
rating. The ANOVA results appear in Table 2.
Taking into account the variation in ACSI ratings explained
by time period (the blocking variable) we can reject the null
hypothesis that the mean ACSI ratings are the same across all
seven airlines. Further analysis reveals that the mean ACSI
rating for Southwest is significantly higher than for all other
airlines included in our study. Moreover, Continental was
found to have a significantly higher mean ACSI rating than
the remaining six airlines. In addition, American and Delta,
while not significantly different from each other, had
significantly higher mean ACSI ratings compared with
Northwest, United and US Airways.
Stepwise regression was used to fit a multiple regression
model with annual airline company ACSI rating as the
dependent variable and thirteen potential independent
variables (the twelve listed in Table 1 and Year). Stepwise
regression is an automatic model building procedure that
employs a forward selection method for adding variables to
the model with a “backward” look to drop any that become
insignificant. The data used to fit the model spanned the 14-
year period from 1998 to 2011. However, as a result of
mergers, data for Northwest and Continental were available
only through 2009 and 2010, respectively. The data were
found to satisfy the basic assumptions for regression.
Regression results are reported in Table 3.
Of the thirteen potential independent variables presumed to
be related to ACSI ratings, six are found to be statistically
significant (at α = 0.05). These are the four complaint
categories of Reservations/Ticketing/Boarding, Baggage,
Customer Service, and Disability as well as Year and
Percentage of On-Time Arrivals. The model explains almost
65% of the variability in ACSI ratings as indicated by the
resulting R2 value of .648. As expected, the coefficient
associated with the percentage of on-time arrivals is positive.
Also as expected, the coefficients associated with the various
complaint categories are negative, with the exception of
Customer Service. The correlation matrix of all variables
shows that Customer Service complaints are indeed
negatively associated with ACSI ratings (r = -0.395), but it
also shows that Customer Service complaints are positively
correlated with all three other complaint categories included
in the regression model (the correlations are 0.641 with
Reservations, Ticketing, Boarding, 0.688 with Baggage, and
0.444 with Disability). Consequently, in the presence of the
other variables in the model, Customer Service complaints
have the effect of increasing ACSI ratings. This is because
the decrease in ACSI ratings caused by Customer Service is
already being accounted for by the other complaint categories
included the model.
DISCUSSION, LIMITATIONS, IMPLICATIONS
Our study shows that in addition to the percentage of on-time
arrivals, complaints concerning reservations / ticketing /
boarding, baggage, disability, and customer service have a
statistically significant impact on ACSI ratings of airlines.
Year was also found to be significant, indicating that factors
related to time period, perhaps economic or environmental in
nature, also affect customer satisfaction ratings. These
findings suggest that airline companies can improve
satisfaction among its customers by offering superior service
in these particular areas. Reducing the number of customer
complaints in these specific categories airlines improves the
ACSI rating. However, this does not mean that airlines
should discourage customers from complaining. As
previously mentioned, studies suggest that airlines can
benefit from encouraging dissatisfied customers to complain
because it gives them an opportunity to make amends and
identify root causes (Huang 2009). It is important, however,
for them to act on the complaints and work to eliminate the
root causes for problems that result in dissatisfaction. This is
what leads to fewer complaints, improved service quality,
and increased levels of customer satisfaction.
Another finding from this study is that Southwest has a
statistically significantly higher mean ACSI rating than all
other airlines. This corresponds to Southwest having
substantially fewer complaints per 100,000 customers.
Boxplots of ACSI ratings for specific airline companies over
time show increased variability, suggesting that airline
companies are increasingly differentiating themselves in
terms of meeting and/or exceeding customer expectations
(see Figure 1). This leads to the question: What makes
Southwest so much better than its competitors? Southwest is
the only so called “low-cost carrier” to be included in this
Proceedings of the Pennsylvania Economic Association 146
study. In the past, a low-cost carrier was defined as having a
one-class cabin, high aircraft utilization, all internet sales and
low fares (Low-cost Carriers Become Harder to Define,
2008). That definition, however, is beginning to become
obsolete with the influx of so many low-cost carriers
worldwide. Today it is harder to classify airlines as such
because of the narrowing gap in cost between the “low-cost”
companies and the legacy companies. However, as this study
dates back to 1998, the fact that Southwest was originally
perceived as a “low-cost” carrier may help explain why its
customer satisfaction is higher than other airlines. Since
customer satisfaction depends on the fulfillment of customer
expectations, perhaps its customers’ expectations are lower
than those of the more “expensive” carriers. Southwest is
also recognized as having a very pleasant workforce (The
Rise of Southwest Airlines, Advance! Business Consulting).
“Pleasant” is not always a term used to describe any airport
experience, but Southwest aims to keep its employees happy
in order to keep its customers happy. Its mission places a
high value on customer satisfaction (The Mission of
Southwest Airlines, Southwest.com.). It is evident that this
customer-driven strategy has been quite successful for
Southwest Airlines.
As with all research studies, ours is not without limitations.
First, surely there are other factors not included here that
impact customer satisfaction in the airline industry. Second,
our study relies on complaint data. It is likely that complaint
data seriously underestimates the true level of customer
dissatisfaction. Many customers simply do not complain.
For example, in 2011 there were 16.2 mishandled baggage
reports per 100,000 customers yet there were only 1.13
baggage complaints per 100,000 customers. The difference in
these numbers shows that either airline passengers were
unaware that they could file a complaint with the Department
of Transportation or just did not bother to do so. Finally, we
use the ACSI rating as a proxy for customer satisfaction
rather than measuring it directly.
CONCLUDING REMARKS
Even with its limitations, our study supports the notion that
airlines can (and should) improve service quality in an effort
to increase customer satisfaction. While a number of factors
affecting customer satisfaction may not be under the control
of airlines (e.g., airport issues, weather, fuel prices, and the
economy), our results suggest that by paying attention to a
few specific service areas that are under their control, namely
those related to reservations, ticketing and boarding,
baggage, customer service, and disability, may help airlines
keep more of its customers satisfied. By offering improved
service in these key areas, and making customer satisfaction
a high priority, an airline may be able to gain the competitive
edge needed to remain viable and profitable well into the
future.
Proceedings of the Pennsylvania Economic Association 147
Table 1: Independent (Explanatory) Variables
Variables Descriptions
Flight Problems Cancellations, delays, or any other deviations from schedule, whether planned or
unplanned.
Reservations, Ticketing,
Boarding
Airline or travel agent mistakes made in reservations and ticketing; problems in
making reservations and obtaining tickets due to busy telephone lines or waiting in
line, or delays in mailing tickets; problems boarding the aircraft (except oversales).
Fares Incorrect or incomplete information about fares, discount fare conditions and
availability, overcharges, fare increases and level of fares in general.
Refunds Problems in obtaining refunds for unused or lost tickets, fare adjustments, or
bankruptcies.
Baggage Claims for lost, damaged or delayed baggage, charges for excess baggage, carry-on
problems, and difficulties with airline claims procedures.
Customer Service Rude or unhelpful employees, inadequate meals or cabin service, treatment of
delayed passengers.
Disability Civil rights complaints by air travelers with disabilities.
Advertising Advertising that is unfair, misleading or offensive to consumers.
Discrimination Civil rights complaints by air travelers (other than disability); for example,
complaints based on race, national origin, religion, etc.
Animals Loss, injury or death of an animal during air transport provided by an air carrier.
Other Frequent flyer, smoking, tours credit, cargo problems, security, airport facilities,
claims for bodily injury, and others not classified above.
Percentage of On-Time
Arrivals
A flight is counted as "on time" if it operated less than 15 minutes after the
scheduled time shown in the carriers' Computerized Reservations Systems (CRS).
Proceedings of the Pennsylvania Economic Association 148
Table 2: ANOVA Results
Source Degrees of Freedom
Sums of Squares
Mean Square F Statistic P-value
Airline 6 1843.90 307.316 60.95 < 0.0001
Year (Block)* 11 151.34 13.758 2.73 0.0060
Error 66 32.77 5.042
Total 83 2328.00 *Data used in ANOVA are for the years 1998-2009. Because of mergers, data were not available for Northwest
after 2009 and for Continental after 2010.
Table 3: Multiple Regression Results
Model Coefficient T-ratio P-value
Constant -709.2 -2.80 0.006
Reservations, Ticketing, Boarding
-45.680 -6.57 0.000
Baggage -11.787 -2.50 0.014
Customer Service 16.159 4.50 0.000
Percentage of On-Time Arrivals
0.2824 2.58 0.012
Disability -45.11 -3.03 0.003
Year 0.3796 3.01 0.003
Proceedings of the Pennsylvania Economic Association 149
Figure 1: Boxplots of ACSI Ratings by Airline and Over Time
US AirwaysUnitedSouthwestNorthwestDeltaContinentalAmerican
80
75
70
65
60
55
Da
ta
Boxplot of American, Continental, Delta, Northwest, Southwest, ...
20112010200920082007200620052004200320022001200019991998
80
75
70
65
60
55
Da
ta
Boxplot of 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, ...
Proceedings of the Pennsylvania Economic Association 150
REFERENCES
American customer satisfaction index. 2012. Benchmarks by
Industry: Ann Arbor, MI.
Carey, S. 2013. U.S. Airlines scored poorly in consumer
survey. The Wall Street Journal. January.
http://online.wsj.com/article/SB10001424052702303703004
577475002478830004.html
Cunningham, L.F. & Young, C.E. 2004. Perceptions of
airline service quality: pre and post 9/11. Public Works
Management and Policy. 9(10): 10-25.
Degirmenci, E., Basligil, H., Bolat, A., & Ozdemir, Y. 2012.
Customer satisfaction measurement in the airline services
using SERVQUAL. Open Access Scientific Reports. 1(5): 1-
9.
Air travel consumer report. 1998-2011. Department of
Transportation Office of Aviation Enforcement and
Proceedings, Washington, D.C.
Huang, W. 2010. Other-customer failure. Journal of Service
Management. 21(2): 191-211.
2012 North America airline satisfaction study.
.http://www.jdpower.com/content/press-
release/aOGunkG/2012-north-america-airline-satisfaction-
study.htm
Low-cost carriers become harder to define. Flight Global.
Airline Business. 19, May 2008. Retrieved from
http://www.flightglobal.com/news/articles/low-cost-carriers-
become-harder-to-define-223826/
McCartney, S. 2013. Believe it or not, flying is improving.
The Wall Street Journal.
Rhoades, D.L. & Waguespack, B. 2000. Service quality in
the U.S. airline industry: variations in performance within
and between airlines and the industry. Journal of Air
Transportation World Wide. 5(1): 60-77.
The Rise of Southwest Airlines. Advance! Business
Consulting.
http://www.advancebusinessconsulting.com/advance!/strategi
c-alignment/strategic-alignment-business-cases/the-rise-of-
southwest-airlines.aspx
Proceedings of the Pennsylvania Economic Association 151
WHAT AFFECTS NEW ZEALAND WINE PRICES? ESTIMATION OF THE EFFECTS OF SENSORIAL,
REPUTATIONAL, OBJECTIVE, AND QUALITY FACTORS IN THE HEDONIC PRICE MODEL
Angela M. Rowland
Indiana University of Pennsylvania
ABSTRACT
This paper applies the hedonic price model to determine the
factors that influence the bottle price of New Zealand wines.
This study primarily focuses on sensorial, reputational,
objective, and quality factors and how they influence the
bottle price. Included are wine vintages from 1995 to 2012
from every main wine-producing region of New Zealand. All
data and ratings were taken from wine expert Robert Parker’s
website. Results indicate that though the Parker Rating is
significant in determining bottle price in other countries, it
only explains a small percentage of the variation in bottle
price for New Zealand wines. Various sensorial and objective
characteristics are more statistically significant to bottle price
than others when regional and quality attributes are
controlled for, regardless of Parker Ratings. This suggests
that, though the Parker Rating does have a significant effect
on bottle prices of New Zealand wines, objective,
reputational, and sensorial elements are also significantly
correlated.
INTRODUCTION
Background
A generally unknown fact about wine is that it is the most
traded liquid commodity in the world, second only to the
global oil trade (Sechrist, 2012). A social lubricant that has
withstood the test of time, it actually may have been one of
the catalysts for the creation of civilization. While many
theories attribute the birth of civilization to the growing of
grains and the need to gather in large groups to protect
immobile resources, there are some that claim grapes and
wine began this process instead. While this will most likely
always be a debate for speculation, the fact that wine has
been around since the beginning of civilization remains. Not
only is it used leisurely for pleasure in social situations, but it
also can have significant ritualistic and spiritual connotations
in various cultures and religions, particularly those found in
western societies. A commonly known use of wine in a
religion is the Eucharist in Christianity, but many other
religions use it as well. Various sects of Judaism, for
instance, celebrate with Kosher wines such as Manischewitz
(Kramer, 2010).
Wine Consumption
Over the past 10 years global wine-consumption has risen
10%, roughly 1% per year, which fell short of the predicted
20% due to the deteriorating economic climate (AFP with
Wine-Searcher Staff, 2013). According to the Beverage
Information Group, wine-consumption has consistently
increased every year for the past 16 years in the United States
alone, which is the predominant wine-consuming nation by
both value and volume. China’s wine-consumption is on
track to becoming second globally by 2016 and already has a
significant impact on global wine-consumption (AFP with
Wine-Searcher Staff, 2013). Wine-consumption in traditional
wine-producing areas like France, Italy, Spain, and the
United Kingdom, however, has dramatically decreased due to
the recent financial crises. Despite these decreases and the
recent recessions faced around the world, global consumption
is expected to grow 5% between 2012 and 2016, compared to
the 3% increase from 2007 to 2011 (AFP with Wine-Searcher
Staff, 2013). United States wine consumption is expected to
rise at least 40% between 2012 and 2016, and Chinese wine-
consumption is expected to rise at least 12% in the same
period (AFP with Wine-Searcher Staff, 2013). These
increases are largely due to a progressively more competitive
marketplace with new brands and competitive prices. In
contrast, European wine demand is expected to continue its
downward trend, partially due to the economic crisis and
partially due to changing drinking preferences (AFP with
Wine-Searcher Staff, 2013). Today, wine demand is so high
that nations around the world have begun to challenge the
traditional wine-growing countries like France, Italy, and
Spain and have begun developing their own reputations as
competitive wine-makers. One of the most recent of these
upstarts in the wine industry is New Zealand, whose wines
have become particularly popular in Australia, China, and the
United States.
New Zealand Wines
Prior to the 1960s, New Zealand’s grape cultivation on any
scale was nonexistent. Early grape-growers produced low-
quality grapes that made even worse wines, leading to a
dead-end for New Zealand wineries (Sechrist, 2012). The
technology and skills to grow quality grapes in the New
Zealand terroir simply was not available. Persistence
prevailed, however, and by the 1980s around 14,000 acres of
land were being used to grow grapes for the purpose of wine-
making (Kramer, 2010). With new developments in grape-
growing knowledge and technology across the world,
however, New Zealand grape-growers began experimenting
and updating their methods of both growing grapes and
making wine. Early attempts ended poorly with prices for
New Zealand wine plummeting because of the terrible
quality of its wines (Sechrist, 2012). White wines, such as
Proceedings of the Pennsylvania Economic Association 152
Sauvignon Blanc and Chardonnay, were the most successful
until the 1990s when the South Island’s red wines started to
meet globally competitive standards (Sechrist, 2012). Today,
New Zealand’s Pinot Noir, Cabernet Sauvignon, and Merlot
hold their own even against traditional French Bordeaux
wines (Kramer, 2010).
The issue in question, however, is what characteristics make
a wine desirable to consumers, in this case from New
Zealand. The topic examined in this paper is the determinants
of bottle price using the hedonic price model for New
Zealand wines in the global market for 2012. Various studies,
such as Bicknell and MacDonald (2012), Oczkowski (2001)
and Anderson and Schamel (2003), analyze the effects of
ratings and reputation on wine prices in New Zealand. There
has also been some research into the effects of sensorial
characteristics on New Zealand wines, such as that conducted
by Combris, Lecocq, and Visser (1997). Studies such as
Bicknell and MacDonald (2012) have been conducted on the
influences of a particular type of grape variety or wine on
bottle price.
This study will broaden the research to include reputational,
sensorial, quality, and objective variables to create a more
complete breakdown of influences on New Zealand wine
prices. As of yet, no other study to my knowledge has
analyzed New Zealand wines for years later than 2010, and
no paper later than 2003 conducts a comprehensive study for
New Zealand such as this one. Much of the information I
gathered came from world-renowned wine critic Robert
Parker’s website and database, Robert Parker’s Wine
Advocates. Robert Parker is famous for his ability to
influence bottle prices with his ratings and reviews
(Anderson & Schamel, 2003).
This study seeks to provide better insight into what
characteristics make a bottle of wine desirable to consumers,
particularly in determining how significant Parker Ratings
are to New Zealand bottle prices. This could in turn be used
by individuals and businesses to make wiser purchases when
buying New Zealand wines and by New Zealand vineyards to
produce more wines with those particular desired attributes.
If the Parker Rating were significant for New Zealand wines,
then wineries would be able to determine what characteristics
are correlated with the highest Parker Ratings and target
those characteristics. If the Parker Rating were not significant
for New Zealand wines, then wineries could still use the
information from this study to see which characteristics are
the most significant influences on bottle price.
Outline of Paper
This study attempts to determine the significant objective,
sensorial, reputational, and quality factors that influence the
bottle price of New Zealand wines in the hedonic price
model. The second section will review the literature used in
this study to identify the factors associated with bottle price.
The descriptive statistics, data, and variables used will be
discussed in the third section, and the fourth will detail the
econometric model and methodology utilized in this study.
The fifth section will discuss the regression equations and
results of the various models. Conclusions and implications
of the findings will be examined in the final section.
LITERATURE REVIEW
Various studies have undertaken the task of analyzing the
factors that influence bottle prices of wine. Oczkowski
(1994) conducted one of the first empirical wine studies,
adopting the hedonic price model estimating the log-linear
function of Australian premium wines. A similar method was
also used by Combris et al. (1997) in analyzing the hedonic
prices of Bordeaux wines of France. They explained that the
determinants of wine prices are not necessarily easy to define
because of their qualitative nature. Results indicated that the
market prices for Bordeaux wines were primarily determined
by characteristics found right on the bottle. Combris et al.
(1997) focused on the impact of both reputational and
sensorial characteristics on bottle prices.
Oczkowski (2001) claimed that when single indicators are
used for reputation and quality they ultimately contain
measurement error because, as shown by Cliff and King
(1996), expert wine tasters’ evaluations differ. This problem
can be averted if quality and reputation are treated as latent
constructs, which can only be reflected by multiple observed
indicators. Thus, Oczkowski (2001) used factor analysis to
consistently estimate hedonic prices in the presence of
attributes measured with error by using expert ratings by
Robert Parker. He found that theoretical and empirical
evidence both indicated that wine prices depend upon quality,
reputation, and objective characteristics. In the application to
Australian premium wines, reputational factors proved
significant to bottle price while quality factors were
insignificant. This theory was fortified by Bicknell and
MacDonald (2012), who noted that wine is a good with
highly differentiated experiences with limited availability of
information prior to consumption. Consumers are forced to
use reputation and expert ratings as proxies of a wine’s value
to make purchasing decisions. Thus, Bicknell and
MacDonald (2012) used hedonic price analysis to generate
implicit prices for a sensorial quality rating and regional
reputation for premium New Zealand wine.
Benfratello, Piacenza, and Sacchetto (2004) were the first to
use a combination of the different groups of variables to
estimate the market price of Italian premium wines. As
suggested by Oczkowski (2001), consumers of wine
frequently face imperfect information, especially concerning
sensorial characteristics. One cannot determine the sensorial
characteristics prior to purchasing the wine. Thus, other types
of information provided by quality, reputation, and objective
Proceedings of the Pennsylvania Economic Association 153
data fill the void. Using price as the dependent variable, and
objective, sensory, reputational, and quality characteristics
for the independent variables they develop new methods of
explaining wine price using a single producer’s reputation.
They found that reputation outperformed taste significantly in
estimating price, but both were significant.
Arguea and Hsiao (1993) examined the econometric issues
found in estimating hedonic price functions by analyzing the
U.S. market for automobiles. They found that consumers are
ambivalent to focusing on every minute detail. This can also
be applied to the wine market, as the average consumer of
wine likely would not notice the small details in a particular
bottle. For instance, consumers are more likely to think a
wine tastes fruity as opposed to associating a flavor with a
particular spice like black pepper or anise. Dummy variables
were used by Arguea and Hsiao (1993) and were found to be
highly significant. Not including them would have created
substantial omitted variable bias. There was some degree of
multicollinearity; however, they claimed that
multicollinearity is to be expected with dummy variable
groups and as long as the equations are homoskedastic it is
not a problem. Therefore, this study also uses dummy
variables for sensorial characteristics, such as flavors from
the wine wheel and body style, along with other variables
such as growing region and grape variety.
Similarly, Anderson and Schamel (2003) conducted a
combined study of changes in hedonic price functions of
Australian and New Zealand wines from 1992 to 2000 with a
focus on regional reputations and expert ratings. The
dependent variable was the logged price of a bottle of wine.
They also tested a log-log form and a linear form with the
log-linear model giving the best results. For their independent
variables they included dummies for red varieties, white
varietes, growing region, and whether a wine is a classic or
not. They also included Parker Rating, star rating, vintage
rating, winery rating, and point rating. Anderson and
Schamel (2003) concluded that in New Zealand, vintage
ratings were significant and fairly constant over time.
Varietal and regional differences were far less significant
from 1992 to 2000.
DATA
This study is a cross-sectional analysis that uses the hedonic
price model to test for a relationship between the bottle price
and a wine’s objective, quality, reputation, and sensorial
characteristics. All data was taken from Robert Parker’s
website, Robert Parker’s Wine Advocates. Anderson and
Schamel (2003) use this website as the main source for their
data as well. Robert Parker is a renowned wine enthusiast
whose critiques are internationally valued and trusted by
wineries and consumers alike (Oczkowski, 2001). The data
pertains to all major wine-producing regions in New Zealand,
including vintages ranging from 1995 to 2012. A map of
these regions, along with common grape varieties found in
each region, can be found in Figure 1 of the Appendix. Only
red and white wines are included, as including rose would
create bias and be inconclusive because only 3 or 4 bottles
were available for the sample.
Data taken from Robert Parker’s website includes wine
ratings (PARKRAT), which range from 0 to 100; 0 being the
worst and 100 being the best. Descriptions of these ratings
can be seen in Table 1 of the Appendix. Any rating below a
60 is deemed an unacceptable wine and is not included on his
website, and thus not in this study. Also taken from his
website are the sensorial characteristics, which are used to
distinguish between different flavors and textures present in
different wines. Arguea and Hsiao (1993) state that it is
unlikely that the average consumer will distinguish between
flavors on a more specific scale than the basic categories of
the aromatic wine wheel, thus this study limits the sensorial
variables to the broadest flavor categories consumers
typically detect. These categories are taken from the Davis
Wine Aroma Wheel, which can be found at The Wine Cellar
Insider (Leve, 2010).
Combris et al. (1997) found that the quality measures used by
wine tasters are primarily explained by sensorial
characteristics. Since wine tasters are essential in determining
the value of a particular wine, sensorial variables are clearly
important in determining the market price of wine and I will
use them in this study. Bicknell and MacDonald (2012)
indicated that both regional reputation and quality ratings
vary between grape varieties, using three different varieties.
Therefore, I will be including grape varieties to account for
this variation, although I include six varieties.
Oczkowski (2001) revealed that quality, reputational and
objective variables are potentially significant to the bottle
price of wine; therefore, they will be included in the sample.
Though they focused on Australian wines, the same can be
applied to New Zealand wines. Benfratello et al. (2004) used
a similar approach but included sensorial characteristics as
well. They also included variables such as alcohol gradation,
which was unavailable for the majority of the wines in my
sample and thus excluded from this study. Producer
reputation was also incorporated, which is unavailable for
this dataset.
Anderson and Schamel (2003) also incorporated reputational,
quality, and objective variables, but for both Australian and
New Zealand wines and excluding any sensorial variables.
They created an overall examination of what influences
bottle prices of New Zealand wine rather than isolating a
particular characteristic that only influences a part of the
bottle price. This study concentrates most heavily on quality
and reputational variables, using objective variables mostly
as controls to reduce omitted variable bias. Prior to this
study, however, no study to my knowledge has included
Proceedings of the Pennsylvania Economic Association 154
reputational, quality, and objective characteristics along with
sensorial characteristics such as flavor for New Zealand
wines. Thus, this is a broadly inclusive study of New Zealand
wines for 2012 prices.
Sensorial Variables
Sensorial variables used in this study include flavors, color,
and body. Flavors are divided into dummy variables for the
general categories taken from the wine wheel: fruity,
herbal/vegetal, nutty, caramel, woody, earthy, chemical,
pungent, oxidized, microbiological, floral, and spicy. These
are FRUITY, HERBVEG, NUTTY, CARAM, WOODY,
EARTHY, CHEM, PUNG, OXID, MICROB, FLORAL, and
SPICY, respectively. Though PUNG, CHEM, and OXID are
primary categories on the aromatic wine wheel, they were
condensed into one category OTHERFLV. Wine color is also
dummied for red and white; COLORR and COLORW,
respectively. Rose wines were not included due to lack of
data availability. Similarly, dummy variables were used for
body, which is divided into full, medium, and light:
BODYFULL, BODYMED, and BODYLGT, respectively.
For the same reasons that PUNG, CHEM, and OXID were
condensed, so were medium-full and medium light.
Medium-light was combined with light, and medium-full was
combined with full. Most wines fell between full and
medium, very few falling into the medium-light and light
categories.
Reputational Variables
Reputational variables include grape variety and wine region.
Grape varieties were dummied, including Sauvignon Blanc,
Chardonnay, Riesling, Syrah, Pinot Gris, Cabernet
Sauvignon, Pinot Noir, Merlot, Proprietary Blends (usually
similar to Bordeaux blends), Cabernet Franc, Viognier,
Montepulciano, Gewurztraminer, Nebbiolo, and Malbec.
Some of these were condensed for not having a large enough
share in the sample. Thus, the variables included are only
PINNOIR, SYRAH, PROPBLE, CHARD, and SAUVBLA
with all other varieties represented by OTHERVAR.
Dummy variables were created including Auckland,
Gisborne, Hawkes Bay, Wairarapa, Nelson, Waipara, Otago,
and Marlborough. These are represented by AUCKLE,
GISBOR, HAWKESB, WAIRAR, NELSON, WAIPAR,
OTAGO, and MARLB, respectively. They are the main
wine-producing regions; however, there are some specific
regions within these that have been incorporated into the
larger, inclusive regions. For instance, Central Otago is a part
of Otago, but some wines from Central Otago are specifically
labeled as such rather than just being labeled as being from
Otago. The same conditions apply to Clevedon, Kumue,
Ponui, Waiheke Island, and Matakana in Auckland, Esk
Valley in Hawkes Bay, Prophets Rock, Waitaki, Felton Road,
Surveyor Thompson, and Bannockburn in Otago, Canterbury
and Hakataramea Valley in Waipara, Solstone and
Martinborough in Wairarapa, and Moutere in Nelson. Some
of these, such as Waiheke Island, may soon be classified as
their own wine-producing regions, but currently they are still
included under the main regions listed. Again, because of
small sample sizes, NELSON and GISBOR are condensed
into OTHERREG.
Quality Variables
The main quality variable included in this study is the Parker
Rating (PARKRAT). Quality ranges are 96 to 100 which are
extraordinary wines, 90 to 95 are outstanding, 80 to 89 are
barely above average to good, 70 to 79 are average, 60 to 69
are below average, and any wine rated below 60 is
unacceptable. Table wines are not included in this study as
they are typically not of a high enough quality to be rated.
Objective Variables
Objective variables included in this study are vintage and
shelf life. Vintage (VINTAGE) describes the year a particular
wine was produced, subtracting the vintage year from 2012.
For example, if a wine was bottled in 1992, the vintage
would be 20. Shelf life (SHELF) states how many years the
wine can be stored as of 2012 before it is no longer a
desirable buy. Negative numbers for this variable indicate the
wine is past its drinking time. Not all bottles have data for
this variable. A distinction between classic and not could not
be found for New Zealand wines, and thus this variable was
excluded.
Expected Signs
A table of all expected signs can be found in Table 2 of the
Appendix. There are several variables that one would expect
to have a positive sign in relation to bottle price, including
Parker Rating and shelf life. The higher the Parker Rating,
the better the wine, and the higher the price is likely to be.
The longer the bottle will last indicates a higher quality wine
and thus would also be expected to be positive. Similarly, the
longer a wine can be stored, the more willing a buyer may be
to spend more money on it.
Many of the variables used in this study, however, have
ambiguous signs, including body type, grape variety, wine
region, and flavors. These could be either positive or negative
depending on the preferences of the consumers, which will
be reflected in the bottle price. For instance, Pinot Noir is one
of the most well-known grape varieties in New Zealand, and
therefore buyers may be more willing to pay higher prices for
a Pinot Noir than a Nebbiolo. Benfratello, et al. (2004) used
the most well-known grape variety as the omitted condition
in their study, which in this case would be Pinot Noir.
According to the literature available, there is no particular
established method for choosing the omitted condition for
Proceedings of the Pennsylvania Economic Association 155
grape variety. Therefore, I simply used OTHERVAR as the
omitted condition and left the main specific grape variety
variables in the model. I used a similar method for the region
and flavor dummy variables, using OTHERREG and
OTHERFLV as the omitted conditions. BODYLGT was the
omitted condition for body style variables, and COLORR for
wine color.
Descriptive Statistics
Complete descriptive statistics for all variables can be found
in Table 3. The most interesting statistics are as follows. The
average bottle price of New Zealand wine in this sample was
$57.08, with a maximum of $275.00 and a minimum of
$38.00. The average vintage age in this sample was 4.5 years,
with a maximum of 17 years and a minimum of 1 year. The
average shelf life of the wines included in this sample was
4.2 years, with a maximum of 13.0 years and a minimum of -
10.0 years, meaning that bottle was past its drinking prime by
approximately 10 years. In this sample, 92% of the wines
have a fruity flavor and none had an oxidized flavor. The
region from which the largest proportion of the sample
originates was Auckland with 22.5% of the sample. The
region with the smallest proportion was Nelson with 2.0%
ECONOMETRIC MODEL
This study uses ordinary least squares (OLS) regression
analysis as done in the similarly comprehensive study by
Anderson and Schamel (2003). The original hypothesized
equation is as follows:
This model represents a combination of Anderson and
Schamel (2003) and Oczkowski (2001). Table 1 exhibits the
variables included in each category in the equation above.
Anderson and Schamel (2003) particularly influenced the
form of the dependent variable (which is logged), dummy
variables for region, grape variety, flavors, body, stage, class,
and color. Unlike Anderson and Schamel (2003), this study is
cross-sectional rather than a panel, which is more similar to
Oczkowski (2001). I conduct a more focused analysis,
however, emphasizing the effects of the chosen independent
variables on only New Zealand wines, intending to update
the research for present-day wines. I also include sensorial
characteristics with the other dummy groups, unlike past
studies concerning New Zealand wines.
Econometric Issues
The first and most prevalent issue in this study was the
availability of data. Other studies have included winery
ratings or regional ratings, which were unattainable within
the constraints of this study. Furthermore, some of the bottles
used in this sample did not have data for each of the
variables, decreasing the sample size depending upon the
model. This was particularly prevalent with the body
variables. Additionally, the expectation that the stochastic
error term does not have a constant variance across
observations is confirmed by the results of White tests. The
null hypothesis of homoskedasticity is rejected for all
regressions. As a result, all regressions are corrected for
heteroskedasticity.
RESULTS
The results for all models appear in Table 4 of the Appendix.
As expected, PARKRAT remained positive and significant
across all models. Model 1 was a simple regression using
only PARKRAT in the estimation of LNPRICE. PARKRAT
is significant at the 1% level in this model, with an adjusted
R-squared of 0.037. Though this model obviously contains
omitted variable bias, it does suggest that the Parker Rating is
significantly correlated with the bottle price of New Zealand
wines. This is consistent with the findings in Anderson and
Schamel (2003). However, it clearly explains only a small
portion of the story in understanding the influences on bottle
prices for New Zealand wines. Anderson and Schamel (2003)
relied more heavily upon Parker Rating. This may have been
necessary for the combined examination of Australian and
New Zealand wines, but was less useful for New Zealand
alone for 2012 wines. Anderson and Schamel (2003) also
used winery rating and regional rating, however, which may
have an influence on the significance of the Parker Rating in
their models. The significance of the Parker Rating to New
Zealand wine prices is also consistent with the results
Oczkowski (2001). Though Oczkowski (2001) used different
sources for the various ratings he used, all ratings proved
significant in the various models. Thus, all subsequent
models contain PARKRAT.
Model 2 includes PARKRAT as well as VINTAGE and
SHELF. This functioned as the base equation for all future
models. Also as expected, the coefficients on VINTAGE and
SHELF remain positive and significant across all models
after Model 1. Both VINTAGE and SHELF are significant at
the 1% level, while PARKRAT is significant at the 5% level.
The inclusion of VINTAGE and SHELF caused the adjusted
R-squared to increase to 0.165. Oczkowski (2001) also used
variables equivalent to vintage and shelf life, which proved
significant in all models.
Each of the following models contains a different group of
dummy variables. Several models containing multiple
dummy groups were tested, but were excluded due to
multicollinearity between groups. Model 3 includes wine
regions along with the variables from Model 2. For this
model, the adjusted R-squared is 0.175. Interestingly,
Proceedings of the Pennsylvania Economic Association 156
PARKRAT’s significance falls to the 10% level, but
VINTAGE and SHELF remain significant at the 1% level. Of
the region variables, only AUCKLE and OTAGO are
statistically significant at the 10% level. Despite these low
individual significance levels, F-test results indicate that as a
group the region variables are significant to the model.
AUCKLE, MARLB, OTAGO, WAIPAR, and WAIRAR
tend to see higher bottle prices compared to OTHERREG,
while HAWKESB is correlated with lower bottle prices than
OTHERREG. Though AUCKLE and OTAGO are
statistically significant for New Zealand wine prices in 2012,
this is no indication that they would be significant for other
years. Anderson and Schamel (2003) conducted a panel
study, which showed the change in significance of the
various regions over time. They found that Auckland most
frequently was significantly correlated with higher bottle
prices, however. This indicates that Auckland tends to
produce wines for which consumers tend to be willing to pay
higher prices.
Model 4 replaces the region variables with the flavor
variables, resulting in an adjusted R-squared of 0.192.
NUTTY is significant at the 10% level, while EARTHY and
WOODY are significant at the 5% level. Though none of the
other flavors are statistically significant, as a group they are
significant to the model based on F-test results. CARAM,
EARTHY, FLORAL, FRUITY, HERBVEG, SPICY, and
WOODY are all positive, while MICROB and NUTTY are
negative. This reveals that consumers tend to pay higher
prices for the former flavors than they would for
OTHERFLV, which includes pungent and oxidized. On the
other hand, they are more willing to pay higher prices for
OTHERFLV than they are for MICROB and NUTTY. Some
flavor preferences may change over time with wine
“fashion,” while others may remain consistent over time. A
panel study delving into the changes in preferences for
particular sensorial characteristics could be a point of interest
for a future study.
Flavors are exchanged for grape varieties in Model 5, with an
adjusted R-squared of 0.206. The only variety not significant
is SUAVBLA, which is also negative in sign. This initially
seemed odd because New Zealand is known for its
Sauvignon Blanc. However, this actually makes sense if
approached from the perspective that New Zealand wineries
produce more Sauvignon Blanc than most other varieties
(probably because of its popularity), which would eventually
trigger downward pressure on the bottle price in comparison
with OTHERVAR. CHARD is also negative, but it is
significant. Sauvignon Blanc and Chardonnay also had a
consistently negative signs in Anderson and Schamel (2003).
This suggests that consumers are inclined to pay lower prices
for these two white New Zealand wines compared to the
wines included in OTHERVAR. PROPBLE is positive and
significant at the 10% level, and PINNOIR and SYRAH are
positive and significant at the 1% level. This indicates that
consumers are willing to pay higher prices for these three red
New Zealand wines than for those other wines included in
OTHERVAR. Grape varieties changed in significance over
time Anderson and Schamel (2003), but Pinot Noir and Syrah
were significant at the 5% level in most years.
Model 6 contains the variables for body style, including
BODYFULL and BODYMED in the regression with
BODYLGT as the omitted condition. With an adjusted R-
squared of 0.223, BODYFULL is significant at the 5% level,
while BODYMED is insignificant in the model. Though
there is obviously some level of omitted variable bias, F-test
results indicate that the body style variables are jointly
significant to the model overall and are thus included. The
coefficients of BODYFULL and BODYMED are both
negative, suggesting that consumers will pay higher prices
for wines with light bodies than for full or medium-bodied
wines. The last model, Model 7, includes only COLORW
with COLORR being the omitted condition. With an adjusted
R-squared of 0.200, COLORW is negative. This suggests that
consumers tend to pay higher prices for red wines rather than
white wines.
Adjusted R2 values ranged from 0.23 to 0.69 in the literature,
the panel studies typically yielding higher values than the
cross-sectional studies. Since this study is a cross-sectional
analysis, the adjusted R2 values obtained of roughly 0.17 to
0.22 fall closer to the lower end of this range. When
considered in groups rather than individual characteristics,
the flavor variables are less statistically significant compared
to region, grape variety, body, and color. This may be due to
the fact that flavors cannot be experienced until after the
bottle is already purchased. Combris et al. (1997) argued that
most consumers base their purchases upon the information
found on the bottle. They must then rely heavily upon
reputational, quality, and objective variables to attempt to
make wise purchasing decisions. Flavors, on the other hand,
are used to determine the rating of a wine such as those by
wine tasters like Robert Parker. Therefore, the flavor
variables are significantly correlated with wine prices, but
other characteristics take precedence in determining bottle
price because they are tangible at the time of purchase.
CONCLUSION
This study suggests that sensorial, objective, reputational,
and quality variables are all correlated with bottle price. Most
of the variables in this study were included because they
were found to be important the existing literature and were
necessary to control for different conditions that could
influence individual bottle prices, even if they do not affect
all bottles. This study found that while Parker Rating is
significant in determining bottle prices for New Zealand
wines, it explains only a small portion of the variation in
prices. Vintage, shelf life, region, flavors, grape variety,
body, and color are all significantly correlated with New
Proceedings of the Pennsylvania Economic Association 157
Zealand wine prices. The particular regions, varieties, body
styles, flavors, and colors of significance depend upon the
sample taken and year.
In the case of the time period considered in this study,
Auckland and Otago had a statistically significant and
positive correlation with higher bottle prices compared to
Nelson and Gisborne. Similarly, earthy, nutty, and woody
flavors were significantly correlated with bottle price,
whether positive or negative, in relation to oxidized and
pungent flavors. Pinot Noir, Proprietary Blends, and Syrah
were significantly correlated with higher prices as opposed to
the many varieties included in OTHERVAR, while
Chardonnay and Sauvignon Blanc were negatively
correlated. For instance, a sturdy Syrah would most likely
fetch a higher price than a more temperamental Nebbiolo,
which usually requires significant periods of aging. In the
case of body style, light-bodied wines were correlated with
higher prices than those for medium and full body styles.
White wines also seemed to receive lower prices than red
wines in this time period. The significance of the particular
variables within the dummy groups may change over time or
with different samples.
Implications of Results
Wineries in regions that tend to receive lower prices for their
wines can try to adopt some of the characteristics found in
the regions that tend to command higher prices to increase
their own profits. These results indicate the flavors, body
styles, colors, and grape varieties for which consumers seem
to be the most willing to pay more money. Wine producers
could also try to increase the shelf life of their wines, which
would subsequently increase the probability of being able to
sell older vintages. Any of these methods would be viable
options for increasing profits and improving customer
satisfaction. Combris et al. (1997) indicated that consumers
frequently choose their purchases based upon information
found directly on the bottle. Therefore, wineries could use the
information in this study to better understand what
information they should include on their bottle labels and
descriptions of their wines. Essentially, while the Parker
Rating is important, it is not the be-all-end-all for New
Zealand wine sellers when setting bottle prices. They would
do better focusing on producing wines that last for long
periods of time with flavors consumers prefer as opposed to
catering to wine tasters’ preferences.
(I would like to thank Dr. James Jozefowicz, for all the time
and effort you put in to helping me with my paper. You will
never know how much your constant support is valued and
appreciated. Thank you also to Dr. Yaya Sissoko, my
discussant, Brian Foster-Pegg, for your input, and my former
partner, Grey Berrier, for the moral support and answering
random questions at all hours of the night)
Proceedings of the Pennsylvania Economic Association 158
APPENDIX
Figure 1 Sechrist (2012)’s Map of New Zealand Wine Regions
Table 1 Parker Rating Ranges and Descriptions
Rating Descriptions
96-100 An extraordinary wine of profound and complex character displaying all the attributes expected
of a classic wine of its variety. Wines of this caliber are worth a special effort to find, purchase,
and consume.
90-95 An outstanding wine of exceptional complexity and character. In short, these are terrific wines.
80-89 A barely above average to very good wine displaying various degrees of finesse and flavor as
well as character with no noticeable flaws.
70-79 An average wine with little distinction except that it is a soundly made. In essence, a
straightforward, innocuous wine.
60-69 A below average wine containing noticeable deficiencies, such as excessive acidity and/or tannin,
an absence of flavor, or possibly dirty aromas or flavors.
50-59 A wine deemed to be unacceptable.
Proceedings of the Pennsylvania Economic Association 159
Table 2 Expected Signs and Definitions for all Variables
Dependent Variable: Logged Bottle Price of New Zealand Wines
Independent Variable Definition Expected Sign
PARKRAT Parker Rating from eRobertParker.com +
Objective Variables
VINTAGE 2012 minus the year the bottle was produced +
SHELF Shelf life +
Regions
AUCKLE Dummy variable for the region of Auckland +/-
HAWKESB Dummy variable for the region of Hawkes Bay +
MARLB Dummy variable for the region of Marlborough +
OTAGO Dummy variable for the region of Otago +/-
WAIPAR Dummy variable for the region of Waipara +/-
WAIRAR Dummy variable for the region of Wairarapa +/-
OTHERREG†
Omitted condition for region variables – includes Nelson and Gisborne N/A
Flavors
CARAM Dummy variable for the flavor caramel +/-
EARTHY Dummy variable for the flavor earthy +/-
FLORAL Dummy variable for the flavor floral +/-
FRUITY Dummy variable for the flavor fruity +/-
HERBVEG Dummy variable for the flavor herbaceous/vegetal +/-
MICROB Dummy variable for the flavor microbiological +/-
NUTTY Dummy variable for the flavor nutty +/-
SPICY Dummy variable for the flavor spicy +/-
WOODY Dummy variable for the flavor woody +/-
OTHERFLV† Omitted condition for flavor variables – includes oxidized and pungent N/A
Grape Varieties
CHARD Dummy variable for the grape variety Chardonnay +/-
PINNOIR Dummy variable for the grape variety Pinot Noir +/-
PROPBLE Dummy variable for the grape variety Proprietary Blend +/-
SAUVBLA Dummy variable for the grape variety Sauvignon Blanc +/-
SYRAH Dummy variable for the grape variety Syrah +/-
OTHERVAR† Omitted condition for grape variety variables N/A
Body Style
BODYFULL Dummy variable for full body style +/-
BODYMED Dummy variable for medium body style +/-
BODYLGT† Omitted condition for body style variables N/A
Colors
COLORW Dummy variable for white wine +/-
COLORR† Omitted condition for color variables N/A
† indicates the omitted conditions for each dummy group, which were dropped from the regressions
N/A: Not Applicable because they are the omitted conditions
Proceedings of the Pennsylvania Economic Association 160
Table 3 Descriptive Statistics
Mean Median Maximum Minimum Std. Dev. Observations
PRICE 57.08333 50.00000 275.0000 38.00000 24.93101 492
PARKRAT 90.54158 91.00000 97.00000 82.00000 2.067164 493
VINTAGE 4.488844 4.000000 17.00000 1.000000 2.643903 493
SHELF 4.241279 4.000000 13.00000 -10.00000 3.044363 344
AUCKLE 0.083164 0.000000 1.000000 0.000000 0.276411 493
HAWKESB 0.225152 0.000000 1.000000 0.000000 0.418107 493
MARLB 0.199187 0.000000 1.000000 0.000000 0.399795 492
OTAGO 0.249493 0.000000 1.000000 0.000000 0.433159 493
WAIPAR 0.064909 0.000000 1.000000 0.000000 0.246615 493
WAIRAR 0.129817 0.000000 1.000000 0.000000 0.336444 493
OTHERREG 0.048682 0.000000 1.000000 0.000000 0.215420 493
CARAM 0.237323 0.000000 1.000000 0.000000 0.425874 493
EARTHY 0.375254 0.000000 1.000000 0.000000 0.484680 493
FLORAL 0.225152 0.000000 1.000000 0.000000 0.418107 493
FRUITY 0.929006 0.000000 1.000000 0.000000 0.257076 493
HERBVEG 0.326572 0.000000 1.000000 0.000000 0.469436 493
MICROB 0.267748 0.000000 1.000000 0.000000 0.443236 493
NUTTY 0.260437 0.000000 1.000000 0.000000 0.260437 493
SPICY 0.432049 0.000000 1.000000 0.000000 0.495864 493
WOODY 0.423935 0.000000 1.000000 0.000000 0.494682 493
OTHERFLV 0.139959 0.000000 1.000000 0.000000 0.353101 493
CHARD 0.107505 0.000000 1.000000 0.000000 0.310069 493
PINNOIR 0.519270 0.000000 1.000000 0.000000 0.500136 493
PROPBLE 0.152130 0.000000 1.000000 0.000000 0.359511 493
SAUVBLA 0.029398 0.000000 1.000000 0.000000 0.166275 493
SYRAH 0.093306 0.000000 1.000000 0.000000 0.291157 493
OTHERVAR 0.093306 0.000000 1.000000 0.000000 0.291157 493
BODYFULL 0.393258 0.000000 1.000000 0.000000 0.489161 356
BODYMED 0.568627 0.000000 1.000000 0.000000 0.495963 357
BODYLGT 0.036415 0.000000 1.000000 0.000000 0.187582 357
COLORW 0.196755 0.000000 1.000000 0.000000 0.397949 493
COLORR 0.801217 1.000000 1.000000 0.000000 0.399490 493
Proceedings of the Pennsylvania Economic Association 161
Table 4 OLS Regressions of Hedonic Price Model
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
CONSTANT 1.212
(1.753)
1.250
(1.351)
1.627
(1.626)
1.353
(1.396)
0.677
(0.686)
0.545
(0.572)
0.612
(0.649)
PARKRAT 0.031***
(3.999)
0.027** (2.592) 0.0221* (1.954) 0.024** (2.232) 0.033***
(2.933)
0.036***
(3.264)
0.035***
(3.273)
VINTAGE 0.024***
(2.751)
0.0260***
(2.798)
0.032***
(3.388)
0.019** (2.089) 0.026***
(2.657)
0.017* (1.936)
SHELF 0.038***
(4.269)
0.040***
(4.221)
0.037***
(4.082)
0.031***
(3.305)
0.044***
(4.266)
0.030***
(3.227)
AUCKLE 0.176* (1.881)
HAWKESB -0.068 (0.102)
MARLB 0.068
(1.085)
OTAGO 0.113* (1.867)
WAIPAR 0.029
(0.416)
WAIRAR 0.091
(1.355)
CARAM 0.048
(1.311)
EARTHY 0.088** (2.565)
FLORAL 0.059
(1.548)
FRUITY 0.045
(0.411)
HERBVEG 0.048
(1.403)
MICROB -0.001
(0.033)
NUTTY -0.106* (1.932)
SPICY 0.049 (1.590)
WOODY 0.070** (2.234)
CHARD -0.109**
(2.522)
PINNOIR 0.100***
(2.840)
PROPBLE 0.103* (1.798)
SAUVBLA -0.027
(0.327)
SYRAH 0.148*** (2.626)
BODYFULL -0.151**
(2.107)
BODYMED -0.038
(0.525)
COLORW -0.160***
(4.893)
R-Squared 0.039 0.172 0.197 0.220 0.224 0.236 0.209
Adjusted
R-Squared 0.037 0.165 0.175 0.192 0.206 0.223 0.200
F-statistic for
Group Joint
Significance
NA NA 5.831 2.257 4.448 36.708 14.974
N 492 343 342 343 343 296 343
Note: Calculated t-statistics in parentheses are based on White heteroskedasticity-consistent standard errors.
* = significant at the 10% level
** = significant at the 5% level
*** = significant at the 1% level
Proceedings of the Pennsylvania Economic Association 162
References available upon request from Angela M. Rowland.
Proceedings of the Pennsylvania Economic Association 163
Author Index
Alt, Ashley M. 16
Armstrong, Thomas O. 28
Balagyozyan, Aram 34
Byun, Chong Hyun C. 41
D’Angelo, Dana 48
Epstein, Susan 48
Gargone, David 78
Ghosh, Soma 52
Giannikos, Christos 34
Hussain, Riaz 123
Kallianiotis, Ioannis N. 54
Kara, Orhan 67
Kearney, Timothy F. 78
Kohli, Tanu 87
Kurre, James A. 97
Linn III, Johnnie B. 109
Ma, Zhen 114
Mansour, Stephen M. 123, 129
McCollough, John 130
Mona, Kyoko 34
Nugent, David 137
Oberkofler, Daniel R. 16
Pezak, Logyn 142
Rowland, Angela M. 150
Sebastianelli, Rose 142
Tesfu, Solomon T. 52
top related