“Masking and Aggregate Employment Changes: Housing Booms and Manufacturing Decline During the 2000s” “Housing Booms, Labor Market Outcomes and Educational.

“Masking and Aggregate Employment Changes: Housing Booms and Manufacturing Decline During the 2000s”

“Housing Booms, Labor Market Outcomes and Educational Attainment”

Kerwin CharlesErik Hurst

Matt Notowidigdo

Research Design

• A local labor market approach

o Identify a “manufacturing” labor demand shifter

o Identify a “housing boom” labor demand shifter

• Some towns experienced larger manufacturing declines than others

o Detroit vs. Orlando

• Some towns experienced larger “housing” demand shocks than others

o Las Vegas vs. Dallas

• Adjust for migration responses

Simple Labor Market Model

ln

ln

D D Dkt kt l kt

S S Skt kt l kt

N w

N w

D M M H H O O Xkt kt kt kt ktX

Estimating Equation

0 1 2

1

2

N N M N H N Nkt kt kt kt kt

SN Ml

D Sl l

SN Hl

D Sl l

S DN O O Sl lkt ktD S D S

l l l l

N X

The Manufacturing “Instrument”: Shift Share (Bartik)

• Identifying Assumption:

o Manufacturing composition in location k in period t is orthogonal to local supply shocks and local changes in demand of other sectors.

• Highly Predictive “First Stage”:

o Shift share measure strongly predicts actual manufacturing employment changes within the MSA.

,~ , 1 ,~ ,( )Mkt ikt i k t i k t

i

s m m

Predicted vs. Actual Change in Manufacturing

Inferring Housing Demand Changes

,

,

log log

log log

D H D Hk k k k

S H S Hk k k k

H P

H P

, , ,H D H S D H S Hk k k k k k kP H P

• Assuming no local housing supply shocks

• Housing demand changes are potentially correlated with other labor demandchanges and labor supply changes.

• Need an instrument.

Estimating Equations

Effects of interest:

o β1 + δ1β2 (Total effect of predicted manufacturing decline)

o β2 (Effect of predicted housing demand change)

Key Assumption: Housing demand change does not affect predicted manufacturing decline in location

(Data strongly support this assumption)

0 1 2

0 1

(1)

( ) (2)

N N M N H N Nkt kt kt kt kt

H M H Okt kt kt kt kt kt kt

N X

f Z X D

Estimating Equations

• Motivation for using an instrument for housing demand change:

o Housing demand change measured with error (e.g., housing supply shocks are possible, measurement error in supply elasticity estimate).

o Housing demand change may be result of other labor demand shocks or labor supply shocks (omitted variables bias)

• Instrument using sharp, structural break in quarterly house price series that occurred in some MSAs during mid-2000s.

o Isolate the “Bubble” component of housing demand change (wish test)o Look for “structural breaks” in housing demand series.

0 1 2

0 1

(1)

( ) (2)

N N M N H N Nkt kt kt kt kt

H M H Okt kt kt kt kt kt kt

N X

f Z X D

Identifying Assumptions

Trying to capture housing markets during the 2000s.

Evidence that national/local house prices changed in part because of speculative behavior and changes in lending technology

o As opposed to traditional housing demand factors (e.g., income growth, population growth, etc.)

o Speculative behavior may differ spatially.

o Lending technology changes may not differ spatially.

Our structural break measure is uncorrelated with all traditional labor market variables (lagged population growth, lagged employment growth, composition of workforce, etc.).

Our structural break measure is highly correlated with changes in Price-to-Rent ratios and share of out-of-town home owners in MSA.

Our New Housing “Instrument”: Structural Breaks

Relationship Between Instrument and Housing Demand Change

Relationship Between Instrument and Lagged Housing Change

Relationship Between Instrument and Supply Elasticity

Relationship Between Instrument and Lagged Non-Employment

Relationship Between Instrument and Lagged Wages

Instrument vs. “Out of Town” Buyers (21 MSAs)

Instrument vs. “Out of Town” Buyers (21 MSAs)

Effects on Employment: Manufacturing Decline

Manufacturing declines depress employment

o A one standard deviation manufacturing decline reduced employment by 0.7 percentage points between 2000 and 2007.

o A one standard deviation manufacturing decline between 2000 and 2007 reduced employment by 1.1 percentage points between 2000 and 2011 (suggesting persistence in manufacturing declines).

Manufacturing declines also reduced wage growth 2000-2007 (but not between 2007 and 2011).

Manufacturing declines caused an in migration of workers (but employment propensities of the migrants were similar to natives).

Manufacturing declines hit older workers harder than younger workers (and also resulted in higher disability take ups).

Effects on Employment: Housing Boom

Housing boom lifted employment

o A one standard deviation housing demand change increased employment by about 1 percentage points between

2000 and 2007.

o A one standard deviation housing boom between 2000 and 2007 had essentially no effect on employment between 2000 and 2011 (the booms were followed by busts – different interpretation of Mian and Sufi results.)

Housing booms increased wage growth between 2000-2007 and 2000-2011 (wags declines during bust didn’t offset the boom).

Housing boom caused an in migration of workers (but employment propensities of the migrants were similar to natives).

For men, employment response concentrated in construction (90%); For women concentrated in FIRE (about 50%). Housing boom hit younger workers more than older workers.

Within Individual Masking: Re-Employment Rate

A Simple Counterfactual

Estimated Effect of Manufacturing Decline on Non-Employment

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-0.020

-0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

0.070

0.080

Estimated Effect of Manufacturing Decline on Non-Employment

~42% Explained

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-0.020

-0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

0.070

0.080

Estimated Effect of Housing Cycle on Non-Employment

Housing Cycle (Construction and Other)

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-0.020

-0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

0.070

0.080

Manufacturing Decline

The Housing Boom Masked The Manufacturing Decline in 2000s

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-0.020

-0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

0.070

0.080

Manufacturing

Data

Housing + Manufacturing

The Housing Boom Masked The Manufacturing Decline in 2000s

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-0.020

-0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

0.070

0.080

Manufacturing

Data

Housing + Manufacturing

34% DuringRecession

Housing Boom and Human Capital Attainment

Propensity to Have At Least One Year of College (Age: 18-29)

1979 1984 1989 1994 1999 2004 20090.40

0.45

0.50

0.55

0.60

0.65

Did Housing Boom Delay College Attendance?

Use same local labor market design to answer this question.

The answer is YES – in both survey and administrative data

Places that had large housing booms had a large reduction in the propensity to attend at least one year of college.

o Nearly all the action was on two year colleges (community colleges, technical schools, trade schools, etc.).

o Found effects for both men and women.

o Effect only present among “lost generation”; those who were young in the early 2000s in boom markets.

For this “lost generation”, the effect was persistent through 2013.

Estimates can explain about 40% of the time series change.

Summary: Our Interpretation

Interpretation

Housing boom “masked” some of the labor market effects of declining manufacturing during the early 2000s.

o Cross-MSA masking (Detroit vs. Las Vegas)

o Cross individual masking (Old hurt by manufacturing decline while young lifted by housing boom)

o Within individual masking (Displaced manufacturing workers are more likely to be reemployed in a MSA that experienced a housing boom).

Is the 2007 labor market the right benchmark to assess cyclical fluctuations?

o Our results suggest no

o Large temporary housing boom lifted labor markets during early 2000s and then brought them back to trend (particularly for low skilled).

Interpretation

We are predicting a period of a “medium run” decline in employment to population decline (relative to pre-recession period)

o Some displaced middle age and older workers in manufacturing decline MSAs have taken up disability (Sloane, 2014).

o Younger workers will slowly adjust to new labor market conditions (process was delayed because of housing boom).

Is this transition from manufacturing (routine) jobs to non-routine services different from the transition from agriculture to manufacturing?

o We think so. We are working on estimating the transition rate across sectors for different types of workers.

Policy Thoughts

Temporary policy stimulus (either monetary or fiscal) will:

o Only have temporary effects on labor market outcomes

o Potentially slow down the human capital accumulation process

For example, another temporary housing boom could temporarily improve labor markets and again deter schooling choices.

How do we train workers displaced by manufacturing (routing jobs) to move to non-routine services?.

O Are those workers willing/able to work at service job wages?

O Will those policies only work for younger workers – or can they lift the employment propensities of older workers.

o Not likely something influenced by Fed policy.