1 THREE ESSAYS ON THE IMPACT OF MARKET STRUCTURE ON NETWORK INDUSTRIES A dissertation presented by Laura Wholley to The Department of Economics In partial fulfillment of the requirements for the degree of Doctor of Philosophy in the field of Economics Northeastern University Boston, MA December 2013
134
Embed
Three essays on the impact of market structure on network ...1048/fulltext.pdf · THREE ESSAYS ON THE IMPACT OF MARKET STRUCTURE ON NETWORK INDUSTRIES by Laura Wholley ABSTRACT OF
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
THREE ESSAYS ON THE IMPACT OF MARKET STRUCTURE ON NETWORK INDUSTRIES
A dissertation presented
by
Laura Wholley
to
The Department of Economics
In partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in the field of
Economics
Northeastern University Boston, MA
December 2013
2
THREE ESSAYS ON THE IMPACT OF MARKET STRUCTURE ON NETWORK INDUSTRIES
by
Laura Wholley
ABSTRACT OF DISSERTATION
Submitted in partial fulfillment of the requirements
for the degree of Doctor of Philosophy in Economics
in the College of Social Sciences and Humanities of Northeastern University
December 2013
3
Abstracts
Chapter 1: How did the Telecommunications Act of 1996 Impact Quality of Basic
Local Telephone Service?
The Telecommunications Act of 1996 changed the rules of competition in the
telecommunications industry. Regional Bell Operating Companies (RBOCs), otherwise known
as Baby Bells after the breakup of AT&T in 1984, were given the right to branch into new lines of
business, such as long-distance service. The Act also created new rules for firms to make their
networks accessible to new competitors but were also given the right to merge with each other
(subject to government approval). The years immediately after the Act saw much merger activity
which we might expect could impact service quality significantly. In this paper, I build a model
of service quality for residential and business customers using data from the FCC’s ARMIS
database to determine the impact of these mergers on service quality as measured by trouble
reports. I find that mergers tend to improve service quality while one divestiture actually may
reduce service quality.
Chapter 2: The Effect of Market Structure on Prices and Quantities in Freight Rail
Shipments
The impact of competition on a market is an issue that regulators in many industries face in the
policy making process. In the freight railroad industry, trackage rights are one of the tools used
to inject competition into a market. Trackage rights allow a firm to use another firm’s rail
infrastructure to transport goods without building its own infrastructure at a particular location.
The regulator can publicly order these trackage rights, but more often they are privately agreed
upon between firms. Often firms enter markets solely through trackage rights and this practice
4
has become increasingly prevalent in recent years. This essay estimates the impact of market
structure and trackage rights on market outcomes in freight railroad markets. Using
instrumental variables, I find that voluntary trackage rights competition tends to raise prices in
certain subpopulations of shipments.
Chapter 3: The Impact of Competition on Price Dispersion between Rail Routes
Variation in prices across firms, geography, distance, and market structure is a phenomenon
that has been widely observed and examined by economic literature. There are many theoretical
papers that point to the causes of price dispersion including price discrimination, cost variation,
search costs, and demand uncertainty. Numerous empirical studies have used these theoretical
models to determine the causes of price dispersion in industries such as airlines, life insurance,
and retail gasoline. This essay examines price discrimination as the main cause of price
dispersion across freight shipment markets. I find that there is evidence of competitive-type
price discrimination in these markets, showing that dispersion increases as concentration
decreases.
5
Dedicated to my family
6
Acknowledgments
First, I would like to extend my gratitude to my committee: John Kwoka, James Dana,
and Gustavo Vicentini. Without your tireless efforts on my behalf, I would not have been able to
accomplish this task. Thank you for all of your comments, suggestions, and critiques of all three
of my dissertation chapters. I would like to especially thank my chair, John Kwoka for getting
me in contact with several key individuals in both the telecommunications and railroads
industries to whom I also extend my thanks.
Without my parents, Maryann and Joe, none of this would have been possible. Thank
you for being there to support me without fail, cheer me on (“Yabadabadoo!”), and always wish
me “Good luck, good fortune, and break a leg!” I would also like to thank my brother, Joseph B.
Wholley III (a.k.a. Jay), for always being there for moral support and comic relief. To my aunts,
Gin and France, thank you for your advice, laughter and endless excitement.
To my husband, Brian, I cannot express my thanks enough. From the moment we met,
you have been there for me. Thank you for listening to my worrying and complaining and
assuring me that it will all turn out okay. Thank you for believing in me.
I would especially like to thank Sean Isakower. Without you, this journey would have
been very different and I am so glad that we were able to go through it together (Pretty easy,
right?). And I would also like to thank Sean Isakower and Shaun O’Brien, without whom
studying for comps would have been much more difficult and much less entertaining.
In addition, I would like to thank Neil Alper for his support and advice, especially during
my time teaching at Northeastern; for all of the moral and administrative support from Cheryl
Fonville, Kathy Downey, and Will Dirtion; and, all of the suggestions and comments from the
weekly I.O. Lunch participants.
7
TABLE OF CONTENTS
Abstract 2
Dedication 5
Acknowledgments 6
Table of Contents 7
Chapter 1: How did the Telecommunications Act of 1996 Impact Quality of Basic
Local Telephone Service?
Introduction 9
Section I: Background, Hypothesis, Literature Review 11
Section II: Data, Model & Results 22
Section III: Merger Endogeneity 41
Section IV: Conclusions 53
References 56
Chapter 2: The Effect of Market Structure on Prices and Quantities in Freight Rail
Shipments
Introduction 59
Section I: Background 60
Section II: Literature Review 62
Section III: Theoretical Model 64
Section IV: Data 70
8
Section V: First Estimation Model 74
Section VI: Second Estimation Model 85
Section VII: Conclusions 98
References 100
Chapter 3: The Impact of Competition on Price Dispersion between Rail Routes
Introduction 102
Section I: Background 103
Section II: Literature Review 106
Section III: Model 115
Section IV: Conclusions 132
References 133
9
Chapter 1: How did the Telecommunications Act of 1996 Impact Quality of Basic
Local Telephone Service?
In the telecommunications industry, other than price, quality of service is probably the
most important factor that consumers consider when evaluating a provider. Other factors such
as availability of new products and price are also extremely important, but if the phone
connection is constantly full of static, or installation wait-times are prohibitively long, then a
consumer might choose to change providers quite quickly.
Service quality in an experience good such as telephone access is a very interesting area
to study. In the realm of telecommunications, it is particularly interesting because of the myriad
of variables and constant change that may have an impact on service quality. Those variables
include innovation and changes in regulation. In addition to both of these variables, we also see
constant change in market structure due to merger activity that has been rampant in the last 15
years. The industry has been regulated partially to ensure that consumers receive fair prices and
that firms are able to maintain provision of service through adequate revenues. However, in
addition to the maintenance of fair prices, the industry has been regulated to ensure that
consumers receive the best possible level of service.
In this paper firstly, I document that service quality as measured by trouble reports has
increased since the Telecommunications Act of 1996 was implemented. Secondly, I construct a
model of service quality in local telephone service to measure the mechanism through which
quality changed due to the Act. This change in the federal regulatory structure is discussed in
detail in the Background section of the paper but in short, the Act allowed the Baby Bells to
branch out into businesses other than local telephone service including Internet provision and
video transmission. Because of this new ability to enter different lines of business, it is possible
that the level of service quality for telephone service declined in the years after the Act was
10
implemented. The 1996 Act also allowed new entrants into the local telephone service market
and as such it is possible that service quality increased after the Act due to increased
competition. Another important aspect of deregulation is that it allowed the Baby Bells to merge
with each other and other firms. Diversification of the product line by firms could be a means to
take advantage of economies of scope, while mergers in the industry may have enabled firms to
utilize cost efficiencies, know-how of other firms, and economies of scale. These three
mechanisms: competition, diversification, and mergers, will be the focus of the service quality
model. This paper focuses on residential service quality following the literature on this topic
The second stage results provide a somewhat odd result for MAINEX. Maintenance
expenditures in time t-1 have a positive (only statistically significant for business customers)
impact on both Initial and Repeat Trouble reports. For business customers, this significant
coefficient implies that when we increase maintenance expenditures by $1 per line, the number
of initial trouble reports increases by about 232 holding all else constant. I believe this to be an
endogeneity issue, and therefore using a one-year lag of maintenance expenditures is
inappropriate in the case of basic local telephone service.
For operating expenditures in both customer groups and each measure of quality, the
resulting coefficient is negative and statistically significant. Therefore, operating expenditures in
time t-1 improve quality in time t.
This exercise has provided some evidence that mergers impact both maintenance and
operating expenditure for firms. However, the results of the regressions showing the impact of
those maintenance and operating expenditures on quality is somewhat mixed. The second stage
regressions show that last year’s operating expenditures have a significant impact on this year’s
quality except for in the case of business customers. Despite some ambiguity in the results, there
does appear to be this two-stage relationship in the impact of mergers on quality.
53
Section IV
Economic and Policy Implications
The Telecom Act of 1996 was intended to deregulate the industry from a federal
standpoint. States still retained control over pricing regulation. However, the federal standpoint
on entry, diversification, and mergers changed with the passing of the Act. The goal of this paper
is to measure how those three aspects of the industry impacted residential service quality.
According to the results, the ease of entry created by the Telecommunications Act did not
significantly change service quality in either direction for basic local service. The
Telecommunications Act was intended to allow and encourage competition into the local service
market. Prior to 1996, CLECs were merely a fringe segment of the market, but the number of
competitors skyrockets after 1996. Unfortunately, this competition had not real impact on either
the infrastructure quality of the local telephone system, nor the customer service that a firm
provided consumers. Clearly, there were much more important factors than competition that
influenced the increase in service quality for this industry.
The implication of this result is not that competition does not matter for improving
quality. The implication is that this regulation did not influence competition perhaps in the way
in which it might be expected for quality. Regulators may have only been attempting to create
competition in terms of prices. While price and quality can be thought of as two sides of the
same coin, in this case we do not see an impact on the quality provided by firms.
Diversification should increase quality of service if there are economies of scope amongst
different products. The telecommunications system can support several different product
offerings using the same infrastructure. Traditional copper lines to provide phone service can
54
also provide DSL for Internet access. Fiber optic cables use light to transmit voice, data, and
video information. For example in the 1996 Act, the government required that firms prove that
their local service markets were competitive before being allowed to diversify into the long-
distance market. Diversification into other lines of business such as Internet and wireless service
started becoming popular around the same time. At the very least, there should be some
economies of scope in terms of managing customers and their various product purchases from
diversification. However, the results indicate that basic local service does not benefit from
diversification of products. A more specialized firm will provide higher quality infrastructure
and customer service. The argument for economies of scope between various product offerings
does not hold water in this case. While firms may decide to diversify, the results of this paper
imply that the reason to become less specialized has virtually nothing to do with improving the
quality of the baseline product offering.
Conclusions and Future Research
The most important mechanism through which the Telecommunications Act of 1996
impacted service quality in basic local service is the ability for firms to merge. Competition from
CLECs had almost no impact on quality of infrastructure or customer-oriented service for either
consumer group. The results of this paper find that a more specialized firm will have better
quality infrastructure, but the magnitude of the coefficients is relatively small compared to the
coefficients on merger activity. Converting to fiber optic technology does not appear to improve
or reduce the quality of basic local service, which is unsurprising as this technology is more
important for broadening the range of products provided through one system.
Mergers are clearly the most important factor influencing quality of service for the Baby
Bells. The main impact of the Telecommunications Act of 1996 is the newfound ability of firms
55
to merge with each other to take advantage of economies of scale, other firm’s know how, and
better ability to test and implement new strategies and technologies. Mergers do not seem to be
based on an endogenous quality consideration, but instead are based on proximity of service
territories. The question of what factors influence the decision to merge with another firm is
something that should be addressed in future research.
Future research may also want to examine why there is a difference of service quality
between business and residential customers in the first place. Another question that should be
addressed by future research is why are firms merging so much in this industry? A final question
that I find to be compelling is to see ask how the Telecommunications Act of 1996 impacted
well-established Incumbent Local Exchange Carriers that were not part of the Bell System, such
as GTE and Southern New England Telephone.
56
References
Abel, Jaison R. and Michael E. Clements (1998). “A Time Series and Cross-Sectional Classification of State Regulatory Policy Adopted for Local Exchange Carriers” for the National Regulatory Research Institute. http://nrri.org/pubs/telecommunications/98-25.pdf Accessed 1 April 2011.
Ai, C. and D. Sappington (2002). “The Impact of State Incentive Regulation on the U.S.
Telecommunications Industry,” Journal of Regulatory Economics, 22: 133-159.
Aron, Debra J. “Ability, Moral Hazard, Firm Size, and Diversification” in The RAND Journal of Economics, 19:1:72-87.
Banerjee, A. (2003) “Does Incentive Regulation “Cause” Degradation of Retail Telephone Service Quality?” Information Economics and Policy, 15: 243-269.
Banker, Rajiv, H. Chang, and S. Majumdar (1998) “Economies of scope in the U.S. telecommunications industry,” Information Economics and Policy, 10: 253-272.
Brenner, Steven R. (1999). “Potential Competition in Local Telephone Service: Bell-Atlantic – NYNEX (1997)” in The Antitrust Revolution, 3rd edition, J. Kwoka and L. White, eds. Oxford University Press.
Chang, Hsihui and Raj Mashruwala (2006) “Was the Bell System a natural monopoly? An application of data envelopment analysis” in Annals of Operation Research 145: 251-263.
Clements, M. (2001) “Local Telephone Quality of Service: The Impact of Regulation and
Competition,” Ohio State University Ph.D. Dissertation.
Clements, Michael E. (2004). Local telephone quality-of-service: a framework and empirical evidence,” Telecommunications Policy 28, 413-426.
Crandall, Robert and Jerry Hausman (2000). “Competition in US Telecommunications,” in Deregulation of Network Industries, Peltzman and Winston, eds. P. 77-89.
Evans, D. and J. Heckman. (1984). “A Test for Subadditivity of the Cost Function with an Application to the Bell System.” American Economic Review September, 615–623.
FCC (2011a). Wireline Competition Bureau ARMIS Database “Customized Reports 43-05 Service Quality Report,” http://fjallfoss.fcc.gov/eafs7/MainMenu.cfm Accessed 1 March 2011.
FCC (2011b). Wireline Competition Bureau ARMIS Database “Customized Reports 43-02 USOA Report,” http://fjallfoss.fcc.gov/eafs7/MainMenu.cfm Accessed 1 March 2011.
FCC (2011d). Wireline Competition Bureau ARMIS Database “Customized Reports 43-07 Infrastructure Report,” http://fjallfoss.fcc.gov/eafs7/MainMenu.cfm Accessed 1 March 2011.
FCC (2011e). Wireline Competition Bureau Statistical Reports “Local Telephone Competition Reports 1994-2008,” http://transition.fcc.gov/wcb/iatd/comp.html Accessed 1 March 2011.
Gabel, David and D. Mark Kennet (1994). “Economies of Scope in the Local Exhange Market,” in Journal of Regulatory Economics, 6: 381-398.
Kahn, Alfred E, Timothy J. Tardiff, and Dennis L. Weisman (1999). “The Teleommunications Act at three years: an economic evaluation of it implementation by the Federal Communications Commission” Information Economics and Policy, 11 p. 319-365.
Kidokoro, Yukihiro (2002). “The Effects of Regulatory Reform on Quality” in Journal of the Japanese and International Economies. 16, p.135-146.
Kittl, Jorg, Martin Lundborg, and Ernst-Olav Ruhle (2006). “Infrastructure-Based Versus Service-Based Competition in Telecommunications” Communications and Strategies 64 p. 67-87.
Noll, Roger G. and Bruce M. Owen (1994). “The Anticompetitive Uses of Regulation: United States v. AT&T (1982) in The Antitrust Revolution, 2nd edition, J. Kwoka and L. White, eds. Oxford University Press.
Norsworthy, J.R. and Diana H. Tsai (1999). The role of service quality and capital technology in telecommunication regulation, ” Telecommunications Policy 11. 127-145.
Pérez-Chavolla, Lilia (2004). “State Retail Rate Regulation of Local Exchange Providers as of September 2004” for the National Regulatory Research Institute. http://nrri.org/pubs/telecommunications/04-13.pdf Accessed 1 April 2011
Pérez-Chavolla, Lilia (2006). “State Retail Rate Regulation of Local Exchange Providers as of September 2005” for the National Regulatory Research Institute. http://nrri.org/pubs/telecommunications/06-05.pdf Accessed 1 April 2011
Pérez-Chavolla, Lilia (2007). “State Retail Rate Regulation of Local Exchange Providers as of December 2006” for the National Regulatory Research Institute. http://nrri.org/pubs/telecommunications/07-04.pdf Accessed 1 April 2011.
Roycroft, Trevor R. and Martha Garcia-Murrilo (2000). “Trouble reports as an indicator of service quality: the influence of competition, technology, and regulation” in Telecommunications Policy 24, 947-967.
Rubin, Jonathan L. (2005). “Local Telecommunications,” in Network Access, Regulation and Antitrust, D. Moss, Ed.
58
Sappington, David E. M. (2005). “Regulating Service Quality: A Survey” in Journal of Regulatory Economics 27:2, 123–154.
Securities and Exchange Commission (1994-2008). Holding Company Annual Reports from Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system. http://www.sec.gov/edgar.shtml Accessed December 2011.
Ter-Martirosyan, Anna and John Kwoka (2010). “Incentive regulation, service quality, and standards in U.S. electricity distribution” in Journal of Regulatory Economics 38: 258–273.
Tirole, Jean. (1988). The Theory of Industrial Organization. MIT Press. Cambridge, MA.
59
Chapter 2: The Effect of Market Structure on Prices and Quantities in Freight Rail
Shipments
The impact of competition on a market is an issue that regulators in many industries face
in the policy making process. Regulators often create policies to foster competition in markets so
that consumers can reap the benefits in the form of lower prices. In the freight railroad industry,
trackage rights are one of the tools that can be used to inject competition into a market.
Trackage rights allow a firm to use another firm’s rail infrastructure to transport goods without
building its own infrastructure at a particular location. The regulator can publicly order these
trackage rights or they can be privately agreed upon between firms. Often firms enter markets
solely through trackage rights and this practice has become increasingly prevalent in recent
years. The question that this paper aims to answer is whether or not entry through trackage
rights provides meaningful competition in a market.
Section I provides a brief background of the freight railroad industry in the United States
as it stands today. Section II is a review of the literature relevant to this research while Section
III presents the model used to examine the research question and Section IV identifies the data
used. Section V presents an estimation model used to examine the impact of market structure in
markets that do not have firms competing through trackage rights as well as the results of that
estimation. Section VI presents the model and results of the main research question, looking at
various subsets of freight shipments to examine the impact of trackage rights competition on
market outcomes. Section VII concludes.
60
Section I: Background
The means by which goods are transported to various locations is a very important issue,
especially for a country as large as the United States. Commodities are typically transported via
truck or rail, though some may be transported via air. About forty percent of all freight is
transported by railroads. In particular commodities such as minerals, metallic ores, and
petroleum are large portions of the total shipped. Seventy percent of all coal is transported by
railroads and coal makes up more than forty percent of all rail shipments (AAR 2010b).
In 2010, there were 566 freight railroads in the United States with mileage of 138,623.
Freight railroads are divided up into different classes based on the amount of annual revenue
that each firm makes. Defined by having annual revenue of $250 million or more, in 2010 Class
I railroads were actually characterized by revenue greater than $398.7 million (AAR 2010b).
There are currently seven Class I railroads operating in the United States: Burlington National
Santa Fe (BNSF), Canadian National (CN), Canadian Pacific (CP), Chessie and Seaboard System
Railroad (CSX), Kansas City Southern (KCS), Norfolk Southern (NS), and Union Pacific (UP).
The Surface Transportation Board (STB), an agency that is part of the Department of
Transportation (DOT), regulates freight railroads. The STB is in charge of resolving any rate and
service disputes between railroads as well as between railroads and shippers. It is also charged
with the responsibility to approve or deny railroad mergers. The STB has not approved a merger
between any Class I firms since the 1996 merger of Union Pacific and Southern Pacific.
Additionally, there has been no entry to or exit from the Class I firm classification since this
merger. Market structure as described in this manner has remained constant for the last 17
years.
61
Figure 1 above shows a map depicting the Class I firms’ ownership of track throughout
the United States. Regional and short-line railroads are indicated in grey. These smaller carriers
make up the vast majority of the number of railroads in the United States, but account for very
little of the trackage ownership and revenue. Class I firms account for about 69% of mileage,
94% of revenue, and 90% of employees in the freight rail industry (AAR 2010b). Most areas of
the country are served by only one or two Class I carriers while there are some markets that are
served by three or more of these firms. However, there are no markets that are served by all
seven Class I firms through track ownership alone. In those cases where 7 Class I railroads are
present, at least one is present due to trackage rights.
Figure 3: Association of American Railroads Map of Class I Track Ownership, 2010
62
Section II: Literature Review
In 1976, the Railroad Revitalization and Regulatory Reform Act was enacted to
encourage competition and reliance on cost-based rate making. And then in 1980, the Staggers
Act deregulated railroad rates in an effort to help railroads increase profitability. In their 1987
study, Lee, Baumel, & Harris use a structure, conduct, performance analysis of Class I railroads
from 1971 to 1984 finding that these changes in regulation had no long-term impact on the trend
for larger firms seen in the Class I railroad group. The biggest impact of the Staggers Act was
further concentration as measured by the four-firm concentration ratio. One limitation of this
study is that it uses aggregate Class I industry data and due to this constraint, cannot look at the
impact on specific commodities, geographies or regional railroads. Using data envelopment
analysis to answer a similar question, Chapin & Schmidt (1993) try to determine if there has
been an increase in efficiency since the Staggers Act and if so, if that efficiency increases can be
attributed to the mergers since deregulation. The authors find that deregulation did in fact
increase efficiency in the market but that many firms are larger than the efficient scale. The
authors conclude that mergers have reduced scale economies and any efficiency gains cannot be
attributed to mergers.
The last major merger in the railroad industry was between two Class I firms: Union
Pacific and Southern Pacific in 1996. Kwoka & White (2004) and Breen (2004) both provide
case study analyses of the UP/SP merger. The initial years after the merger were very difficult
for the newly integrated firm and shippers were subjected to poor service quality as the firms
tried to coordinate activities. Due to the extreme difficulties of the merger, the Surface
Transportation Board has essentially placed a moratorium on mergers and there have not been
any allowed between Class I firms since 1996.
63
To allay some of the fears and anxiety of shippers prior to the merger of UP and SP, the
Surface Transportation Board created some conditions with which the companies needed to
comply in order for the merger to be consummated. The most important condition of the merger
between UP and SP was the trackage rights granted to BNSF – the only other Class I track-
owner in the west. This merger significantly impacted many markets in the west because many
of those markets went from having two competitors to only one competitor (2-to-1 markets) or
from three competitors to only two competitors (3-to-2 markets). Those shippers who only have
access to one carrier are called captive shippers and as Pittman (2010) argues, need to be
protected by the government due to this status.
Pittman (2010) discusses a few means by which captive shippers may be protected. The
most direct path, Pittman argues, would be tighter regulation of rates charged to captive
shippers, though this would require significant time and effort in streamlining the process of
rate setting and regulation. Another route would be to introduce legislation that would place
more railroad behavior under the jurisdiction of the antitrust laws (there is currently an
exemption for railroads), which could limit the ability of a railroad to exploit its market power at
the expense of captive shippers. Mandatory switching, which would need to be regulated or
legislated, is a third way of protecting captive shippers from high rail rates (Pittman 2010).
Trackage rights are another way to try to accomplish this task.
While Christensen (2009) finds that trackage rights may be difficult to implement and
sustain because of coordination issues with track-owners, studies of the trackage rights imposed
in the west find that they can be effective. Karikari, Brown, & Nadji (2002) look specifically at
the impact of the merger on rail rates for those shippers in potential 2-to-1 markets after the
merger of UP and SP. They find that the competition of BNSF through government imposed
64
trackage rights was more effective in a specific economic area– kept downward pressure on
prices – than SP’s pressure on UP prices prior to the merger. However, these results were only
relevant to the Salt Lake City economic area and varied by commodity type, direction of traffic
and shipper type. Winston, Maheshri, & Dennis (2011) examine the long-run effect of the
mergers of UP/SP and BN/SF, focusing specifically on grain shipments. The authors find that in
the long run, these mergers have not raised rates and have had negligible impact on consumer
welfare.
The impact of market structure on prices and quantities is an important question both
theoretically and practically. Schmidt (2001) attempts to answer this question for the freight
railroad shipment market. His paper provides the theoretical background and method for
measuring market structure for this research. The theoretical model is presented next and the
method of measurement is discussed in Section V.
Section III: Theoretical Model
The theoretical model is based on the one presented in Schmidt (2001). Schmidt
estimates the reduced form of a structural model of supply and demand for railway shipment of
a commodity. The definition of the market in this model is a shipment of a particular commodity
from an origin location to a destination location. Schmidt is attempting to answer the following
question: is shipment using only one firm (a single line shipment) cheaper than shipping a
commodity using two or more carriers (an interline shipment)? To answer this question,
Schmidt analyzes markets using both regional and Class I firms that own trackage in the origin
and/or destination location. While Schmidt does not include firms with trackage rights as
competitors in the market, the theoretical model remains the same.
65
On the demand side of the market, there are assumed to be a large number of shippers
who are price takers with the options of shipping via rail or truck:
Qr = fr (Pr, Pt, X, b) (1)
Qt= ft (Pr, Pt, X, b) (2)
where Qr is the quantity of the good shipped via rail, Qt is the quantity of the good shipped via
truck, Pr is the price of the shipment via rail, Pt is the price of the shipment by truck, X is a
vector of exogenous variables that impact demand for shipment, and b is a vector of structural
parameters.
The trucking sector is competitive and therefore behavior of suppliers can be described
as:
Qt= gt (Pt, Zt, g) (3)
where Zt is a vector of exogenous variables that affect the cost and supply of transportation via
truck, and g is a vector of parameters.
Equilibrium in the trucking sector can therefore be solved by setting supply equal to
demand:
ft (Pr, Pt, X, b)= Qt= gt (Pt, Zt, g) (4)
and when solving for equilibrium price, Pt, we find:
Pt=h (Pr, X, Zt, b, g) (5)
Substitution of the price of shipments by truck into the demand equation for railroad
shipments provides the inverse demand function for freight as a function of its own prices:
66
Pr=fr’ (Pr, X, Zt, b, g) (6)
Due to the fact that the rail industry has a small number of firms, solving for price and
quantity is more difficult because the market is not competitive. Schmidt (2001) follows
Bresnahan (1989) by assuming that the ith rail firm maximizes profit where marginal revenue
equals marginal cost:
MC (Qri, Zri) = Pr + Qri*θi(N)* dfr/dPr (7)
which can be rewritten as:
MC (Qri, Zri) = Pr*(1 + Sri(N)* θi(N)/εr) (8)
where Qri is the quantity of output produced by the ith rail firm, MC (Qri) is the marginal cost of
the ith rail firm of carrying another unit of this commodity from the origin location to the
destination location, Sri is the market share of the ith rail firm, Zri is a vector of exogenous
variables that impact firm i’s costs of providing rail freight transport, εr is the elasticity of market
demand, and θi is a parameter to describe the competitiveness of firm behavior.
Equation (8) implies that a firm’s demand elasticity is equal to its market share divided
by the market’s demand elasticity. This result may not always come about. For instance, suppose
a firm has a very high market share and the demand elasticity of the market is very elastic. This
would imply a high elasticity of demand for that firm which in turn implies that there are a
significant amount of close substitutes for the good. Therefore, the market power of the firm
may be overestimated by this model. It is important to keep this in mind when interpreting the
67
results of this exercise as the impact of another competitor on market outcomes may be
overestimated by the model.
θi may depend on the number of rail competitors serving the market. θi may also vary
between firms because some firms may be more aggressive than others in competition for
shippers’ products. If we hold demand constant, Qri will change as the number of firms changes
and MC(Qri) may also change with the number of firms.
Since the sum of the outputs of individual firms equals the market output, we have the
following identity:
Qr = ∑ Qri (9)
If the market has N firms, there are therefore N + 2 endogenous variables: Qri for each of
the N firms, Qr, and Pr; and, there are N + 2 equations relating those variables (equation (8) for
each of the N firms, equation (9), and the demand curve equation (6)). If we assume that the
functional forms of these equations are well behaved, we can solve the system for each of the
endogenous variables as a function of only the exogenous variables and parameters:
Pr = k(X, Zt, Zr, b, g, θ, N) (10)
Qr = m(X, Zt, Zr, b, g, θ, N) (11)
Qri = m(X, Zt, Zr, b, g, θ, N) (12)
where Zr and θ are vectors composed of the individual Zri and θi for each firm.
Schmidt (2001) does not estimate the full structural model for three reasons. First, firm-
specific information on the exogenous variables is needed to be able to identify the model.
68
Second, data about firm-specific quantities is needed and is something that is not publicly
available. And third, one must make an assumption about the behavior of the firms, but there is
little empirical evidence to provide guidance. For the same reasons, this paper will first estimate
a reduced form model of the market to look at the impact of market structure on prices and
quantities.
To analyze the results, Schmidt considers four theories of market structure which have
implications for the magnitudes and significance of the coefficients in the model:
(1) Constant returns to scale and competitive (Bertrand) behavior. If this is
the case, then firm marginal cost does not depend on the number of firms
in the market and therefore θ, the parameter to describe the level of
competitiveness of firms will be equal to zero. Therefore, price will be
equal to marginal cost, quantity and price in the market will not depend
on the number of firms and will not be statistically significant in the
estimation.
(2) Decreasing returns to scale and competitive (Bertrand) behavior. In this
situation, as the number of firms in the market increases, quantity per
firm falls and therefore firm costs fall which would cause increasing
market quantities and falling prices.
(3) Increasing returns to scale and competitive (Bertrand) behavior. Here,
as firms enter, quantity per firm would decline and therefore costs of
individual firms will rise. We would expect to see falling market quantities
and increasing prices as the number of firms goes up.
69
(4) Constant returns to scale and imperfect competition between firms. As
the number of firms increases, market shares of individual firms fall. θ,
the parameter to describe the level of competitiveness of firms may also
fall if competition is characterized as Cournot. As market share falls, each
firm has less ability to hold up price and increases quantity. If θ falls,
increasing the number of firms reduces a single firm’s ability to raise
price.
An additional theory of industry structure is described below:
(5) N is correlated with unobserved factors in the vector of exogenous
variables that influence demand for shipment of products. If this case is
true, a market would exhibit higher prices and quantities.
In this model, the number of firms is taken to be exogenous but it is possible (if not
likely) that the number of firms is dependent on unobservable demand characteristics. If firms
are more likely to enter markets that have high demand for transportation of a specific
commodity, then market structure would be dependent on N. The results in Schmidt (2001) are
therefore short-run results because of this characteristic as well as the fact that entry is costly in
this industry and exit requires a firm to abandon assets which are sunk. Exit of the market also
requires regulatory approval. The assumption means that the number of firms is fixed in the
short-run.
For the purposes of this paper, this assumption only holds true if we do not include firms
with trackage rights as competitors. As discussed earlier, market structure for the Class I firms
has remained constant since 1996. However, it is very likely that trackage rights are not created
70
randomly or exogenously. Therefore, this possibility will become very important for my
estimation methodology. The next sections describe the data used to estimate the impact of
market structure and trackage rights on prices and quantities in these markets.
Section IV: Data
The main source of data used in this paper is the 2010 Public Use Waybill Sample from
the STB. This dataset is a sample of waybills from Class I, II, and III railroads in the United
States in the year 2010. An observation is defined as a shipment of commodity i from origin
Bureau of Economic Analysis (BEA) economic area to destination BEA economic area. Figure 2
provides a map depicting the definitions of BEA economic areas used by the STB in the Public
Use Waybill Sample. Note from Figure 2 that a BEA economic area can be a fairly large
geographic area. The use of BEA economic areas as the geographic unit is not ideal because a
firm present at the north end of an economic area may have little influence on a firm at the
southern tip of that same area if the area is particularly large. This may lead to the belief that
shippers have more options in choosing a carrier than they do in actuality. Additionally, a firm
in one BEA economic area may have a significant impact on prices in its neighboring economic
area, which by construct cannot be captured in this model. However despite these potential
issues, this research is constrained to using BEA economic areas because that is what is
provided in the STB’s Public Use Waybill Sample. The Confidential Waybill Sample provides
much more detailed geographic information and would provide a more detailed analysis that fits
the reality of railroad location better than the Public Use Waybill Sample.
71
Figure 4: Map of BEA Economic Areas (1995)
The sample provides information on many aspects of a shipment including but not
limited to the weight of the shipment in tons, the shortest distance via rail track from origin to
destination, and the revenue of the shipment. A market is defined as a commodity shipped
within an origin-destination pair. Data will be aggregated to the origin-destination market level
for each type of good.
The Public Waybill Sample does not identify the carrier of the shipment, this is only
included in the Confidential Waybill Sample. When looking at pure market structure, the
identity of the firm may not matter. However, to determine the impact of a particular firm on
prices and quantities in a market, identity must be known. Data has been collected from the
Association of American Railroad’s publication Railroads and States to supplement the Waybill
72
sample with carrier identification. From the information provided in Railroads and States, I
was able to match railroad locations as identified by the AAR with BEA economic areas across
the country for each of the seven Class I railroads and 17 regional railroads. With this
information, it is not only possible to count the number of firms at an origin and destination, but
to also identify track owners at both ends of a shipment.
Table 1 below describes the market structure in the 2010 Public Use Waybill Sample. The
first column indicates the number of single-line carriers on a route (firm is located at both ends
of a shipment). The second column indicates the percent of shipments in the 2010 Public Use
Waybill Sample that exhibit this market structure. As discussed above, it is important to note
that the data used in this research does not identify carrier location beyond BEA economic area.
The reader should keep in mind that the results may not be capturing the exact market structure
at the station level. However, the BEA economic area provides an approximation of location
which is correlated with the presence of a carrier at a particular station; if a carrier is not located
in the BEA economic area, it cannot be present at a particular station in that BEA economic
area.
73
Table 6: Market Structure of Single Line Firms by Number of Shipments in the
2010 Public Use Waybill Sample
This table includes all seven Class I railroads as well as the 17 largest regional railroads
as defined by the Association of American Railroads Railroads and States publication. A vast
majority of the shipments in the sample are shipped along a duopoly route (73.53%). In second
place are monopoly routes with 14.55% of the shipments. A little over five percent of the
shipments in the sample do not have any single line firms on the route. These shipments are
either interline shipments, where the shipment must be transferred to another carrier mid-
route, or they are carried by a railroad not included in this research.
Additional supplementary data sources for control variables include the U.S. Census
Bureau, the Bureau of Economic Analysis, the United States Geological Sample, the Annual
Survey for Manufactures, and the United States Department of Agriculture. The variables
obtained from these resources are discussed in the next section and outlined in Table 2.
Number of
Single Line
Firms
Percent of
Shipments
None 5.40%
One 14.55%
Two 73.53%
Three 3.82%
Four 1.65%
Five 0.21%
Six
Seven 0.84%
Total 100.00%
74
Table 7: Variable Definitions, Sources, & Units
Section V: First Estimation Model
The first estimation model looks only at shipments between markets where firms do not
compete through trackage rights in which the only competitors are trackage owners. The market
structure of trackage-owners for Class I firms has remained unchanged for the last seventeen
years (since the merger of UP and SP in 1996). Thus, the rationale for discarding trackage rights
competitors is to attempt to only look at the effects of what may be called “exogenous” market
structure on market outcomes in the short run. Section VI will address the approach to and
results of the second estimation model dealing with the impact of trackage rights on market
Variable Source Description/Units
Shipm ent STB Public Carload Way bill Sample Shipment from j to k of commodity i
Price Created from STB Carload Waybill
Av erage Rev enue per ton-mile,
weighted by the number of carloads
in the market for commodity i
Quantity Created from STB Carload WaybillNumber of carloads in a market for
commodity i
Structure AAR - Railroads and StatesNumber of class I RR operating in
BEA Economic Area
Origin T rackage Rights Firm s DOT RITA NTAD DatabaseNumber of Class I RR with trackage
rights in Origin BEA
Destination T rackage Rights Firms DOT RITA NTAD DatabaseNumber of Class I RR with trackage
rights in Destination BEA
State Fuel Price (FORG,FDEST )
EIA 2010 Transportation Sector
Distillate Fuel Prices, 2010 Dollars per million BTU
City Population (PORG, PDEST ) U.S. Census Bureau 2010, Number of persons
State GDP (GDPORG, GDPDEST ) Bureau of Economic Analy sis 2010, Millions of current dollars
State T otal Value of Agri. Sector (AGORG,
AGDEST )
USDA Economic Research Serv ice:
U.S. and State Farm Income and
Wealth Statistics
2010, Thousands of dollars
State Manufacturing Value Added
(MANUFORG, MANUFDEST )Annual Surv ey of Manufactures 2010, Thousands of dollars
State T otal Value of Minerals Extracted
(MINORG, MINDEST )USGS Minerals Handbook 2010, Thousands of dollars
Short Line Distance (SHIP) STB Public Carload Way bill SampleShortest distance in miles of track
from Origin BEA to Dest. BEA
WAT ER STB Public Carload Way bill SampleDummy =1 if shipment travels v ia
water, zero otherwise
Comm odity STB Public Carload Way bill SampleDummy =1 for commodity i , zero
otherwise
Unit of Observation
Independent Variables
Dependent Variables
75
outcomes using instrumental variable regression analysis. One important limitation of this first
exercise is that the markets examined may not include firms with trackage rights because they
are not as profitable and thus endogeneity may still be present. All of the models presented are
necessarily static models and do not account for any dynamic interactions.
However, it is important to note that track-ownership could be endogenous to the
market itself as well, particularly if there are bottleneck issues present. If it is very difficult for a
new track-owner to build into a market simply due to space constraints, then the number of
competitors is restricted. Therefore, it is important to keep in mind that the results presented
may face some issues of endogeneity and may only be relevant in the short-run as discussed in
Section III.
A very general functional from is used by Schmidt (2001) and the same technique is
adopted here for markets that do not have trackage rights competitors. For a particular
shipment, a firm will be designated as “single-line” firm if it serves both ends of a particular
route. A firm will be considered an “origin-only” firm if it only serves the origin BEA economic
area and a firm will be designated “destination-only” if is only serves the destination BEA
economic area. A dummy is created for each possible combination of the numbers of each type
of firm. For example, the base case will be a monopoly single-line firm with no origin-only or
destination-only firms in the market (1-0-0). When there are two single-line firms, the dummy
for (2-0-0) will be equal to 1, zero otherwise. For a market with a single-line firm, an origin-only
firm, and a destination-only firm, the dummy for (1-1-1) will be equal to 1, zero otherwise. There
are over one hundred different combinations of single-line, origin-only, and destination-only
firms in the data.
76
Market price will be measured as the weighted average revenue per ton-mile in the
market of a particular commodity. Using price per ton-mile purges the effects of weight and
distance from the price as these characteristics clearly influence how much a shipment will cost
to transport. The average price in a market is weighted by the number of carloads on a
shipment. Quantity will be measured as the number carloads of commodity i in an origin-
destination pair. I will include variables to control for fuel prices, population, and economic
well-being in the origin and destination cities or states. I also include dummies for each
commodity type in addition to the market structure dummies. The base commodity will be STCC
1132, which is the category for barley shipments. The estimation model is defined as follows:
From Table 4 we see that the coefficients on all of the control variables are statistically
significant. The signs of the control variable coefficients are fairly consistent with the results
presented in Schmidt (2001), however the statistical significance is different in some cases. The
magnitude of the results presented here is larger than those of the results presented by Schmidt.
The results show that fuel cost increases prices and quantities at the origin. Fuel costs at
the destination point tend to decrease prices and quantities. Fuel costs are also large in
magnitude as compared to the other variables indicating, as one might expect, that fuel costs are
a large component of rail costs. The extremely large coefficients on fuel costs imply that there
may be some residual multicollinearity in the estimation. However, at the very least it can be
said that fuel costs are a significant component of price and the cost of fuel at the origin has a
much larger impact than the price of fuel at the destination point of a shipment. Other cost
characteristics such as the distance of the shipment (SHIP) and the weight of the shipment
(TONS) are also all statistically significant at the 0.001 level, even when using price per ton-mile
as is used here. The negative signs of these coefficients indicate that heavier shipments or those
that must travel further have lower average price per ton-mile. A shipment traveling via water
has prices approximately 12% higher than those that do not.
Included demand characteristics are also statistically significant at least at the 0.001
level. The results for population show that a one percent increase in the population at the origin
is associated with a 6.09% increase in prices and an 18.6% increase in quantity. As population at
the destination point increase by one percent, there will be a 4.32% decrease in price and a
26.6% increase in quantity holding all else equal. Population at both ends of a route is clearly a
significant factor in determining shipment prices. Prices to ship goods tend to be higher at the
originating location if there is a larger population. It may be more difficult to run track or
83
schedule times to run trains in very populated areas than in less densely populated areas, thus
increasing the price of shipping from these locations.
The total agricultural value of the origin and the destination states also has significant
impacts on the price of shipments. Holding all else equal, a one percent increase in the total
value of the agricultural sector in the origin state will increase the price of the shipment by
3.72%. A one percent increase in the value of the agricultural sector in the destination state
decreases the price of shipments by 14.7% and increases quantities by about 22.4% holding all
else equal. The agricultural value of an origin or destination is important because those areas
whose business is primarily farming will be shipping products to areas where there is less
agricultural activity. Agricultural goods are less likely to be shipped by rail because they typically
are goods with limited shelf life, thus the minimal (though statistically significant) impact on
price.
Minerals are a group of products more likely to be transported via rail because they
typically do not have a limited shelf life and are often very heavy. A one percent increase in the
value of the mineral sector in the origin state decreases shipment price by about 2.9% and
increases quantity by about 2.4% holding all else equal. A one percent increase in the value of
the mineral sector in the destination state will increase prices by about 10% and decrease
quantity by about 15.3% holding all else equal. Again, there is a small but statistically significant
impact on prices from this variable. From the results, it appears that if a destination starts to
produce more minerals itself, prices will increase significantly.
The coefficients for the market structure dummies are presented in Tables 4 and 5. These
coefficients should be interpreted in relation to the price in a monopoly single-line market, that
is the market structure 1-0-0. This type of market would occur when the only option for freight
84
shipment via rail is service from one Class I or regional carrier. It is important for the reader to
note again that there is a high likelihood of market structure endogeneity and that the results
presented in this section must be considered to be short-run with no change in market structure.
These results are static rather than dynamic.
As can be seen in the results, prices (Table 4) are generally higher and quantities (Table
5) lower when there is no firm that serves both BEA economic areas in the market pair directly
(0 single line firms). This result, consistent with Schmidt (2001), indicates that interline service
is more costly than shipping a good using only one carrier for the duration. For example, if we
compare the situation where there is no single-line carrier and exactly one carrier in the origin
and one carrier in the destination (0-1-1), we see that prices are 32.6% higher and quantities
131% lower than the base case of the monopoly single-line shipper. The coefficient on quantity (-
1.31) is clearly confusing because it implies negative shipment, so clearly the multicollinearity or
some other issue is present in this estimate. However, we can say that prices are higher but
cannot definitively state if quantities are lower than the base case when interline shipment is the
only option. Schmidt (2001) finds that interline shipments tend to be more expensive than
single-line shipments and the results presented here coincide with that finding. Those markets
without any single-line firms have generally higher prices – and coefficients of large magnitude
– as well as significantly lower quantities.
Analysis of the coefficients presented in Tables 4 and 5 reveals some strange results.
While the results for markets without any single line carriers seem to make some sense
(interline shipment is more costly than single-line shipment), as the number of single line firms
increases, the results show that prices continue to increase and quantities fall. Unfortunately,
these results do not coincide with those found by Schmidt (2001) nor do they follow traditional
85
economic theory. However, there are a couple of possible explanations. The first is that there is
still an endogeneity problem present. Track ownership and the number of firms is endogenous
to the prices and quantities of shipments (and therefore profitability) in a market. The other
possibility is that there is some oligopolistic behavior occurring that the model is capturing.
However, there is no way to tease this out of the model with the available data.
It is important to note that the results presented in this section do not consider the
impacts of firms competing through trackage rights on market outcomes. Competition via
trackage rights is a very important aspect of this industry that should not be excluded from a
contemporary exploration of the freight railroad shipment market. Section VI presents the
methodology and results of such an exploration.
Section VI: Second Estimation Model
A firm may enter a market without building its own infrastructure by obtaining the right
to use the track of existing firms. These trackage rights may be negotiated between firms using
contracts or may be imposed by the regulator, the Surface Transportation Board. As noted by
Kwoka & White (2004):
Sometimes a railroad will extend “trackage rights” to a second railroad, so that the latter can run its trains over the former’s track (typically for a relatively short distance) and thereby connect shippers/recipients to the second railroad’s track network. In principle this can result in competition between the two carriers despite the single track. However, the relationship between the two railroads is that of landlord and renter, and there are many ways that the landlord can use its position to mute the competitive threat from the tenant. For example, the fee for use of the tracks may be set so high that the second railroad has to price its service noncompetitively. In addition, the landlord railroad can use its train scheduling (“dispatching”) prerogatives, track maintenance routines, and longer-run investments affecting the route to favor itself and raise its rival’s costs or degrade the latter’s service. And the extensive and close contact between the two railroads, especially on routes where they are the only providers of rail service, may provide the basis for oligopolistic coordination.
86
If railroads are voluntarily extending trackage rights to competitors, there must be some
benefit to doing so. Therefore, these voluntary trackage rights may not provide the expected
downward pressure on prices that should emerge from meaningful competition. This caveat is
very important to keep in mind, as the data used in this estimation uses trackage rights that are
voluntarily agreed upon as well as those mandated by the government.
As discussed earlier, the regulator may impose trackage rights as part of the conditions
of a merger approval as was done in the UP/SP merger in 1996. The STB made these trackage
rights the centerpiece of the merger agreement and awarded approximately 4,000 track-miles of
rights (Kwoka & White 2004). These trackage rights were seen by the industry as likely
ineffective because these they were imposed only in places where the shipper and the recipient
were directly connected to the UP and SP (and only UP and SP) track. In doing so, the STB
neglected to include instances where either UP or SP might be close enough to the shipper or
recipient such that the threat of using another means of shipment or the possibility of build out
to the competitor led to effective competition between UP and SP. These instances were counted
as pre-merger monopoly that would not change post-merger. Furthermore, the STB ignored
instances where UP or SP may be a monopoly at one end of a shipment, but the other carrier
was competing at the opposite end of a shipment. Merging would create a single-line service
monopoly. It was estimated that there would be freight price increases of about 20% post-
merger (Kwoka & White 2004).
Some additional data is required to obtain the identities of firms with trackage rights in
particular BEA economic areas. The Department of Transportation’s Research and Innovative
Technology Administration (RITA) publishes the National Transportation Atlas Databases
(NTAD) on an annual basis. Using 2010’s NTAD, I am able to obtain information on the
87
identities of firms with trackage rights (voluntary or mandated) in an economic area. Even
though I do know the identity of each firm in each BEA economic area, I do not present results
looking at the impact of a particular firm’s use of trackage rights because of endogeneity issues.
The presence of a particular competitor may be endogenous and I do not have a good
instrument to mitigate the issue at this time. As noted, the trackage rights included in this
dataset include both those that have been mandated by the STB through merger conditions as
well as trackage rights privately agreed upon between firms themselves. Mixing of these
trackage rights is not strictly appropriate as they differ in fundamental ways, but the data does
not provide information on the reason that trackage rights exist on a particular route. In one of
the estimations, I attempt to separate the two types of trackage rights by looking only at those
routes for which trackage rights were granted in the UP/SP merger approval. However, the
other estimations do not differentiate between voluntary and mandated trackage rights.
About 52% of the shipments in the sample are not served by any firms that are present
through trackage rights only. About three percent of the sample is comprised of shipments for
which one trackage rights competitor is present at both the origin and the destination point,
while about 0.14% of shipments are in markets where two trackage rights competitors are
present at both the origin and the destination point of a shipment. There are, however, many
routes where there may not be a single-line trackage rights competitor, but there are trackage
rights competitors at either the origin or destination point of a shipment. Table 6 below
describes how trackage rights are mixed in with the market structure of single-line track-
owners.
88
Table 11: Shipments Distribution of Trackage Rights by Track Ownership Market
Structure in the 2010 Public Use Waybill Sample
From Table 6 it can be seen that when there is a duopoly in track-ownership, most of the
time there will not be any trackage rights competitors but it is more likely that there will be one
trackage rights competitor at both ends of a shipment than two. Additionally, it is far more
likely for there to be just one trackage rights competitor at one end of a shipment when the
track-ownership market structure is duopolistic than two or more trackage rights competitors.
These tables provide some insight into the distribution of trackage rights across market
structures in the sample which leads to the question of how these trackage right competitors
impact market outcomes.
Table 7 below describes the number of markets (origin, destination, commodity
combination) in which firms have trackage rights in the sample. Trackage rights are typically
granted in miles along a route. Table 8 simply reports the number of markets not the number of
Table 12: Number of Markets in which Firms have Trackage Rights
Of the almost 8,000 markets in the sample, less than 350 markets have firms competing
through trackage rights. CP – at 130 markets – is by far the largest trackage rights competitor in
terms of its presence in the most markets. KCS follows with 71 markets and next are CN and
BNSF with 36 and 32 markets respectively. CSX with 25 markets, UP with 13 markets, and NS
with 6 markets rounds out the competition from Class I railroads in terms of trackage rights.
FEC and WE are regional carriers but both have a fair presence in trackage rights competition.
The existence of trackage rights in a particular market is most likely not exogenous.
Firms are probably much more likely to negotiate trackage rights in a market that would be
profitable than one that is not. Additionally, the regulator is much more likely to try to impose
competition in a market where either there are already high prices or a merger will reduce
competition. To deal with the issue of endogeneity, I employ two-stage least-squares.
To instrument for the number of single-line track owners and the number of single-line
trackage rights competitors, I use county-level data on agricultural production, manufacturing,
Firm Number of Markets
BNSF 32
CN 36
CP 130
CSX 25
FEC 22
KCS 71
NS 6
UP 13
WE 8
Total 343
Total Markets 7,991
90
and population from the 1900 U.S. Census of Population and Housing. These variables were
successfully used as instruments for the number of firms at the origin and destination of a
shipment in Hughes (2011). Because total size of the U.S. railroad network peaked in the early
1900s and existing track operates on historical right-of-way, it is expected that historical rail
service will be correlated with present railroad participation. As actual historical railroad
participation is not observed, county level population, value of agricultural goods, livestock, and
manufacturing at the origin and destination are used to predict railroad participation along a
route (Hughes 2011).
Tests for relevance in the following regressions show that these instruments are valid.
The model is over-identified, in that there are more instruments than endogenous regressors. As
such, exogeneity can be tested statistically. Unfortunately, the results of these tests reject the
null hypothesis indicating that they are in fact endogenous. However, the qualitative reasoning
above and the successful use of these instruments by Hughes (2011) inspires some confidence in
their usefulness for this exercise. First stage results and test statistics are available from the
author upon request.
The rest of this section discusses the second-stage results of four explorations of the
impact of trackage rights on freight shipment prices and quantities. I have chosen these four
ways to examine the data because of their relevance to the sector. The first set of results
presented look at all shipments whether they occur in markets with or without trackage rights.
The second results presented look specifically at the impact of government imposed trackage
rights. Third, shipments of coal are examined as they make up a large portion of freight rail
traffic. The fourth and fifth regressions look at the impact of trackage rights on short-haul versus
long-haul shipments. A market continues to be defined as a shipment from origin j to
destination k of commodity i.
91
Table 13: All Shipments
(Standard errors in parentheses) ***p<0.001; **p<0.01; *p<0.05
Table 8 presents the results using the instruments from the 1900 U.S. Census of
Population and Housing described above for all of the shipments in the 2010 Public Use Waybill
Sample. The results show that increasing the number of single-line track owners along a route
will decrease prices by about 30% and increase quantities by about 179%, holding all else equal.
The statistical significance of this sign indicates, as would be expected, that the addition of an
additional competitor – that owns track – will create downward pressure on market prices.
A negative sign is also present for Single TR, indicating that the addition of a trackage
rights competitor will also exert downward price pressure. However, the coefficient of 323% is
very large in magnitude making the result and interpretation questionable. The railroad must be
paying the shipper to use its services in order for this result to hold. This result suggests that
ln (Price) ln (Quantity)
Single Lines -0.305*** 1 .7 87 ***
(0.01 ) (0.01 )
Single TR -3 .230*** 3.47 5***
(0.07 ) (0.1 5)
ln (SHIP) -0.821 *** 1 .1 23***
(0.01 ) (0.01 )
ln (TONS) -0.423*** 0.31 6***
(0.002) (0.005)
WATER -0.325 -1 .041 **
(0.1 7 ) (0.35)
Constant 3.37 4*** -8.055***
(0.09) (0.1 9)
Commodity Dummies? YES YES
N 257 ,840 257 ,840
92
trackage rights are still endogenous, which is confirmed with statistical testing. It is also
important to remember that this dataset includes voluntary trackage rights which are clearly
endogenous.
Overall, shipments that weigh more and are transported over longer distances have
lower prices and higher quantities. Long-distance shipments tend to be less costly because they
involve fewer switches and since only single-line carriers are included here, the negative sign is
expected. The dummy for water transport has no significant effect on price. The next set of
results looks only at routes on which BNSF was granted trackage rights by the merger of UP and
SP.
Table 14: UP/SP Merger Condition Routes - Government Mandated Trackage
Rights
(Standard errors in parentheses) ***p<0.001; **p<0.01; *p<0.05
Table 9 presents the results exclusively for those markets in which trackage rights were
government imposed in the UP/SP merger approval in 1996. There are 16 routes of this list in
ln (Price) ln (Quantity)
Single Lines 0.552*** 0.648***
(0.02) (0.03)
Single TR 0.292*** 0.7 97 ***
(0.03) (0.05)
ln (SHIP) -0.1 62*** -0.1 7 3***
(0.01 ) (0.01 )
ln (TONS) -0.1 83*** 0.1 48***
(0.01 ) (0.02)
Constant -0.927 * -1 .1 47
(0.37 ) (0.62)
Commodity Dummies? YES YES
N 4,985 4,985
93
the 2010 Public Waybill Sample and 74 markets. The government imposed trackage rights at the
route level. The Surface Transportation Board mandated trackage rights along these routes to
mitigate the impact of a lost competitor due to the merging of UP and SP in 1996. The routes
identified by the STB are primarily in Arkansas, California, Colorado, Louisiana, Nevada, Texas,
and Utah with the majority of routes in the state of Texas. These trackage rights may be
considered as exogenous (as in comparison to voluntary trackage rights agreements) because
they were imposed in an effort to mitigate predicted effects of the loss of a competitor along
these routes.
As discussed above, these trackage rights were seen by the industry as likely ineffective
because these they were imposed only in places where the shipper and the recipient were
directly connected to the UP and SP (and only UP and SP) track. The STB did not include
situations in which either UP or SP might be close enough to the shipper or recipient such that
the threat potential entry of the other or shipper build out to the competitor led to effective
competition. Additionally, the STB ignored instances where UP or SP may be a monopoly at one
end of a shipment, but the other carrier was competing at the opposite end of a shipment, such
that interline alternatives had previously been available to shippers. 20 percent price increases
were estimated (Kwoka & White 2004). Karikari, Brown, & Nadji (2002) found that these
trackage rights were effective in decreasing prices in the Salt Lake City economic area.
These results look only at the routes on which the STB imposed trackage rights in 1996.
Holding all else equal, the addition of one more single line track owner would increase prices by
55.2% and increase quantities by 64.8% in these markets. In contrast, to the results obtained by
Karikari, Brown, & Nadji (2002) trackage rights competition has a positive and statistically
significant impact on price in the results presented in Table 9. The addition of a trackage rights
competitor increases price by 29.2% and increases quantity by about 80%, ceteris paribus.
94
While the endogeneity problem may not be entirely mitigated by only looking at government
mandated trackage rights (the sign and magnitude of the coefficient on Single TR for quantity is
positive and fairly large), the sign and coefficient of Single TR in the price regression coincides
with the estimated impact described in the case study by Kwoka & White (2004). Even
seventeen years after the merger was approved and trackage rights were mandated, higher
prices are still present. The next section exclusively examines coal shipments in the 2010 Public
Use Waybill Sample.
Table 15: Shipments of Coal, STCC 11
(Standard errors in parentheses) ***p<0.001; **p<0.01; *p<0.05
ln (Price) ln (Quantity)
Single Lines 0.662*** -0.824***
(0.03) (0.06)
Single TR 1 .960*** -5.468***
(0.38) (0.69)
ln (SHIP) -0.339*** -0.347 ***
(0.02) (0.03)
WATER 0.1 85 -1 .835
(1 .65) (2.96)
Constant -6.7 7 1 *** 1 3.37 ***
(0.1 9) (0.34)
N 25,866 25,866
95
The results presented in Table 8 are exclusively for shipments of coal – Standard
Transportation Commodity Code (STCC) 11. Coal has been singled out for two reasons. The first
reason to look at coal separately from other commodities is that it makes up about 40% of all rail
shipments. For this reason, coal is clearly an important product for railroads and thus may be
impacted by trackage rights differently from other commodities. The second reason coal has
been singled out is the captive shipper problem. As discussed in Pittman (2010), captive
shippers are those shippers that are served by only one firm and have very little potential to
attract an additional carrier. Coalmines may be particularly subject to these restrictions if they
are located in remote areas to which it may be very difficult and costly to build additional track
to serve. It is important to note that coal is not the only product for which the captive shipper
problem is present. Other shippers such as petrochemical plants among others may be in similar
situations. Coal was primarily chosen because of its prevalence as a commodity shipped via rail.
The results in Table 8 indicate that adding an additional track owner at both ends of a
coal shipment will increase prices by about 66%. And, holding all else equal, one more single-
line trackage rights competitor will increase the price of a coal shipment by 196%.
The positive sign on these coefficients could be for a few reasons. It could be that firms
are making agreements with each other for trackage rights because they provide some benefit –
profit or otherwise – and this is at the expense of the shipper. Thus, voluntary trackage rights do
not benefit shippers.
Econometrically, there may still be some endogeneity that is not being taken care of by
the instrumental variable method. Though there is statistical evidence that the instruments are
endogenous, though valid, I do not believe this to be the primary issue. Additionally, the
unavoidable use of BEA economic area as origin and destination location is correlated with the
96
existence of the railroad in a place where the shipper may use its services, but it is not the best
measure. A coal mine may be located in an area with only one carrier choice (captive shipper)
but the data (wrongly) categorizes that shipment in a market that has more than one shipper.
However, it is impossible to avoid this problem with the available data.
As the first estimation results show, the distance of a shipment has a statistically
significant impact on both prices and quantities. It is possible that the impact of a trackage
rights competitor may be different depending on the distances that the shipment must travel.
The results presented in Table 11 are for short haul shipments (traveling less than 250 miles),
while Table 12 presents the results for long haul shipments (a distance greater than 250 miles).
Table 16: Short Haul Shipments (Less than 250 Miles)
ln (Price) ln (Quantity)
Single Lines 0.01 25* 0.07 41 ***
(0.01 ) (0.01 )
Single TR 0.41 1 *** 0.464***
(0.03) (0.05)
ln (SHIP) -0.3 7 5*** -0.229***
(0.01 ) (0.01 )
ln (TONS) -0.57 9*** 0.427 ***
(0.01 ) (0.01 )
Constant 0.31 8* 3.91 5***
(0.1 4) (0.25)
Commodity Dummies? YES YES
N 21 ,449 21 ,449
97
Table 17: Long Haul Shipments (Greater than 250 Miles)
(Standard errors in parentheses) ***p<0.001; **p<0.01; *p<0.05
Short-haul and shipments experience higher prices and higher quantities when an
additional single-line track owner enters the market. Holding all else equal, short haul
shipments experience 1.25% higher prices and 7.4% higher quantities. In contrast, an additional
single-line track owner for long-haul shipments decreases prices 46.7% and increases quantities
by 211%. The impact of a firm entering through trackage rights has a statistically significant and
positive impact on price. For example, a competitor entering along an entire route will increase
prices by 41.1% and increase quantity by 46.4% holding all else equal.
In contrast, a firm competing through track ownership and trackage rights at both ends
of a long haul shipment will decrease price and increase quantity. This sign is consistent with
the results in the first regression (Table 8), where long-haul shipments are less costly than
short-haul shipments. The magnitude of the impact on price and quantity for both track owners
ln (Price) ln (Quantity)
Single Lines -0.467 *** 2 .1 1 5***
(0.01 ) (0.02)
Single TR -3 .359*** 1 .57 3***
(0.08) (0.1 8)
ln (SHIP) -0.625*** 1 .1 91 ***
(0.004) (0.01 )
ln (TONS) -0.396*** 0.281 ***
(0.002) (0.004)
WATER -0.524*** -0.447
(0.1 5) (0.3 2)
Constant 2.37 5*** -8.950***
(0.08) (0.1 8)
Commodity Dummies? YES YES
N 236,391 236,391
98
and trackage rights competitors is much larger than the magnitude of the coefficients for short-
haul shipments. As the number of observations used in this regression is almost the entirety of
the shipment data, the same issues of endogeneity discussed above likely apply. However, the
results do provide some evidence that long-distance shipments tend to be less costly because
they involve fewer switches and since only single-line carriers are included here, the negative
sign is expected.
Section VII: Conclusion
This paper has used publically available cross-sectional data to examine the impact of
market structure on market outcomes in the freight railroad shipment industry in the United
States. It has confirmed previous findings that prices and quantities do vary with the number of
firms competing through track-ownership.
This paper has also taken the analysis one step further by investigating the impact of
firms competing through trackage rights without building their own infrastructure. The trackage
rights used here are both voluntarily agreed upon between firms and imposed by the
government. Trackage rights are clearly not exogenous and while instrumental variables have
been employed, there is still some level of endogeneity present.
If firms agree to grant trackage rights to each other, there must be some positive benefit
to both parties. As Kwoka & White (2004) points out, in principle this should provide
competition and result in lower prices for shippers. However, the “landlord” is easily able to
impose costs on the renting railroad and use its position to mitigate the competitive threat from
the renter. These techniques could be used to raise rivals costs, thus increasing the costs of a
99
shipment along a route where trackage rights are present. Additionally, close contact between
railroads may facilitate collusive behavior.
The results of this exercise imply that these voluntary trackage rights are actually not
beneficial to shippers and in fact do create higher market prices per ton-mile. Even looking
specifically at trackage rights imposed by the government in the UP/SP merger approval, prices
increase with the entrance of a trackage rights competitor. This result is likely due to the way in
which the Surface Transportation Board applied trackage rights.
Trackage rights may be a viable way to create competition in a market, but only through
careful economic analysis and appropriate application will this result in lower prices. Further
research into this topic is clearly needed, and this paper is just a first step. In future, the
Confidential Waybill Sample should be used to provide a more detailed analysis with better
geographic and price data. Additionally, instruments that can be proven to be both statistically
valid and exogenous would also help to improve this analysis. Furthermore, it is imperative to
be able to separate voluntary and government imposed trackage rights to determine empirically
if trackage rights are a good policy alternative to increase consumer welfare.
100
References
Association of American Railroads (2010a). Railroads and States. https://www.aar.org/keyissues/Pages/Railroads-And-States.aspx#.UkQ1YnflfTo. Accessed 20 December 2012.
Association of American Railroads (2010b). Freight Railroads in United States. https://www.aar.org/StatisticsAndPublications/Pages/default.aspx. Accessed 20 December 2012.
Borenstein, Severin (1990). “Airline Mergers, Airport Dominance, and Market Power” in American Economic Review 80:2, 400-404.
Breen, Denis A. (2004). “The Union Pacific/Southern Pacific Rail Merger: A Retrospective on Merger Benefits” in Federal Trade Commission Working Papers Number 269. http://www.ftc.gov/be/workpapers/wp269.pdf. Accessed 15 October 2012.
Bureau of Transportation Statistics (2010). National Transportation Atlas Database DVD from RITA. http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_atlas_database/index.html
Burton, Mark and Wesley W. Wilson (2006). “Network Pricing: Service Differentials, Scale Economies, and Vertical Exclusion in Railroad Markets” in Journal of Transport Economics and Policy 40:2, 255-277.
Cambridge Systematics (2010). Regional Freight System Planning Recommendations Study for Chicago Metropolitan Agency for Planning. http://www.cmap.illinois.gov/freight-system-planning. Accessed 5 November 2012.
Chapin, Alison and Stephen Schmidt (1999). “Do Mergers Improve Efficiency? Evidence from Deregulated Rail Freight” in Journal of Transport Economics and Policy 33:2, 147-162.
Farrell, Joseph, Paul A. Pautler, and Michael G. Vita (2009). “Economics at the FTC: Retrospective Merger Analysis with a Focus on Hospitals” in Review of Industrial Economics 35, 369-385.
Hughes, Jonathan E. (2011). “The higher price of cleaner fuels: Market power in the rail transport of fuel ethanol” in Journal of Environmental Economics and Management 62, 123-139.
Interstate Commerce Commission (1920). Thirty-Fourth Annual Report on the Statistics of Railways in the United States for 1920. Google Books. Accessed 10 September 2013.
Ivaldi, Marc and Gerard McCullough (2007). “Railroad Pricing and Revenue-to-Cost Margins in the Post-Staggers Era” in Research in Transportation Economics 20, 153-178.
Karikari, John, Stephen Brown, and Mehrzad Nadji (2002). “The Union Pacific/Southern Pacific Railroads Merger: Effect of Trackage Rights on Rates” in Journal of Regulatory Economics 22:3, 271-285.
101
Krishnan, Ranjani (2001). “Market restructuring and pricing in the hospital industry” in Journal of Health Economics 20, 213-237.
Kwoka, John and Lawrence White (2004). “Manifest Destiny? The Union Pacific and Southern Pacific Railroad Merger (1996)” in The Antitrust Revolution 4th Edition, 27-51.
Lee, Tenpao, C. Phillip Baumel, and Patricia Harris (1987). “Market Structure, Conduct, and Performance of the Class I Railroad Industry, 1971-1984” in Transportation Journal 26:4, 54-66.
Pittman, Russell (1990). “Railroads and Competition: The Santa Fe/Southern Pacific Merger Proposal” in The Journal of Industrial Economics 39:1, 25-46.
Pittman, Russell (2009). “Railway Mergers and Railway Alliances: Competition Issues and Lessons for Other Network Industries” in Department of Justice Economic Analysis Group Papers. http://www.justice.gov/atr/public/eag/246372a.html. Accessed 15 October 2012.
Pittman, Russell (2010). “The Economics of Railroad ‘Captive Shipper’ Legislation” in Administrative Law Review 62:3, 919-936.
Schmidt, Stephen (2001). “Market structure and market outcomes in deregulated rail freight markets” in International Journal of Industrial Organization 19, 99-131.
Surface Transportation Board (2008). “Decision No. 16” in STB Finance Docket No. 35087, http://www.stbfinancedocket35087.com/html/bkgrndoverview.html. Accessed 15 October 2012.
Surface Transportation Board (2009). Statement No. 81-1: Procedure for Sampling Waybill Records by Computer. http://www.stb.dot.gov/stb/industry/econ_waybill.html. Accessed 15 October 2012.
Surface Transportation Board (2010). Public Use Waybill Sample. http://www.stb.dot.gov/stb/industry/econ_waybill.html. Accessed 15 October 2012.
Werden, Gregory J., Andrew S. Joskow, and Richard L. Johnson (1991). “The Effects of Mergers on Price and Output: Two Case Studies from the Airline Industry” in Managerial and Decision Economics 12, 341-352.
Winston, Clifford, Vikram Maheshri, and Scott M. Dennis (2011). “Long-run Effects of Mergers: The Case of US Western Railroads” in Journal of Law and Economics 54:2, 275-304.
102
Chapter 3: The Impact of Competition on Price Dispersion between Rail Routes
Variation in prices across firms, geography, distance, and market structure is a
phenomenon that has been widely observed and examined in economic literature by authors
such as Pratt, Wise, & Zeckhauser (1979), Carlson & McAfee (1983), Stahl (1989), Shepard
(1991), Borenstein & Rose (1994), Dana (1999), Gerardi & Shapiro (2009), and Orlov (2011)
among others . There are many theoretical papers that point to the causes of price dispersion
including price discrimination (Borenstein & Rose 1994; Gerardi & Shapiro 2009), cost