Transcript Document

Manufacturing Busts, Housing Booms, and
Declining Employment
Kerwin Kofi Charles
Erik Hurst
Matthew J. Notowidigdo
University of Chicago
Harris School of Public Policy
And NBER
University of Chicago
Booth School of Business
and NBER
University of Chicago
Booth School of Business
and NBER
September 2012
This Paper

Try to explain why employment rate changed within the U.S. during
the 2000s

Focus on two prominent phenomenon:
o
Dramatic decline in manufacturing employment (secular decline)
o
Transitory housing boom followed by housing bust.

Assess how those shocks affected employment rates (and other labor
market outcomes) during 2000-2007, 2007-2010, and 2000-2010
periods .

Run counterfactuals “shutting off” the labor market effects of each of
the changes. Isolate importance of manufacturing declines.

Look at the effects of two phenomenon on human capital attainment.
Total U.S. Manufacturing Employment (in 1,000s)
21,000
19,000
17,000
15,000
13,000
11,000
9,000
Total U.S. Manufacturing Employment (in 1,000s)
21,000
~1.5 Million Jobs Lost
During 1980s and 1990s
19,000
17,000
15,000
13,000
11,000
9,000
Total U.S. Manufacturing Employment (in 1,000s)
21,000
~1.5 Million Jobs Lost
During 1980s and 1990s
19,000
17,000
15,000
13,000
11,000
9,000
~3.8 Million Jobs Lost
During 2000-2007
Total U.S. Manufacturing Employment (in 1,000s)
21,000
~1.5 Million Jobs Lost
During 1980s and 1990s
19,000
17,000
15,000
~3.8 Million Jobs Lost
During 2000-2007
13,000
Even More Jobs Lost
After 2007
11,000
9,000
7/12
This Paper: Estimate Effects on Employment Rate
2000-2007
2007-2010
2000-2010
Manufacturing
Decline
Housing Related
“Shock”
Combination of both
Phenomenon
•
•
Causally estimate effects using a local labor market strategy.
Focus on different groups: Primary focus is on non-college
men.
This Paper: Estimate Effects on Employment Rate
2000-2007
Manufacturing
Decline
Housing Related
“Shock”
Combination of both
Phenomenon
2007-2010
2000-2010
This Paper: Estimate Effects on Employment Rate
2000-2007
Manufacturing
Decline
Housing Related
“Shock”
Combination of both
Phenomenon
2007-2010
2000-2010
This Paper: Estimate Effects on Employment Rate
2000-2007
2007-2010
2000-2010
Manufacturing
Decline
Housing Related
“Shock”
Combination of both
Phenomenon
~0( )
This Paper: Estimate Effects on Employment Rate
2000-2007
2007-2010
2000-2010
Manufacturing
Decline
Housing Related
“Shock”
Combination of both
Phenomenon
•
~0( )
~0
The housing shock “masked” the labor market effects of the
manufacturing shock during the 2000-2007 period.
Summary of Main Findings
1. Manufacturing decline is important for thinking about changes in nonemployment during 2000s.
o
About 35-40% of increase in non-employment during 2000s can be
attributed to the decline in manufacturing.
2. Labor market was significantly weaker in the 2000-2007 period than
we thought.
o
Housing boom “masked” deterioration of U.S. labor market.
o
2000-2007 period marked by secular decline in one sector and a
temporary boom in another sector.
o
Implication: 2007 may not be a good benchmark to evaluate cyclical
changes in economic variables of interest.
Summary of Main Findings
3.
About one-third of the increase in non-employment during the
recession can be attributed to
o
manufacturing declines during 2007-2010 period, and
o
manufacturing declines during the 2000-2007 period that were masked
by housing boom.
4. The net effect of housing booms/busts on labor markets was small over
the entire decade.
o
5.
The bust reduced employment but the boom raised employed
Housing boom deterred college attainment during 2000-2007 period.
A Word on the “Masking” Effect

•
Masking occurred both across and within individuals.
o
Housing booms were often in places that didn’t experience the
manufacturing declines.
o
Type of workers affected differed slightly (by age, skill, and nativity).
o
However, even for a given individual, evidence that those that were
displaced from manufacturing were more likely to find employment in
places with a housing boom.
Both types of masking are interesting.
o
Implies that even though the aggregate employment rate may have
been relatively stable during 2000-2007 period, there could still have
been distributional effects (across people and locations).
Plausibility of “Masking” Effect?


For our empirical work, we are going to identify effects using cross MSA
variation.
o
Different MSAs received different combinations of manufacturing
and “housing” shocks.
o
For our aggregate calculations, need to discuss the scaling up of local
estimates (migration, etc.)
However, the potential plausibility of masking can be seen from the time
series data.
Employment Trends for Non-College Men (age 21-55)
0.40
Man + Cons Share
0.35
0.30
0.25
Manufacturing Share
0.20
0.15
0.10
Construction Share
0.05
0.00
Employment Trends for Non-College Women (age 21-55)
0.140
Manufacturing + Construction Share
0.120
0.100
Manufacturing Share
0.080
0.060
0.040
0.020
Construction Share
2011
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
0.000
Median Real Wage for Non-College Men (age 21-55)
Long-Run Increase in Non-Employment?

Results do not imply a permanent increase in non-employment
o
Workers could choose to acquire skills which could increase market
wage.
o
Workers could choose to move to different labor markets.

We think of this is as more of a medium run increase (as opposed to
being just do to cyclical fluctuations) – adjustments take time.

Our force is different from traditional mismatch stories.
o

For us, people are just moving up and down labor supply curve in
response to labor demand shocks (market wage < reservation wage).
However, our results suggest that temporary government policies to
stimulate labor demand will NOT have lasting effects on employment.
o
Only implies to the 30-40% of non-employment increase we identify.
Outline
1. Conceptual model
2. Empirical model
3. Main results
4. Counterfactual estimates
5. Examine effects on human capital attainment
6. Conclusion
Conceptual model

Purpose - To provide a simple model which highlights:
o
the interplay between shocks in different sectors
o
when those shocks will result in changes in nonemployment.
o
reasons why the response to nonemployment resulting from a shock
may change over time.
Conceptual model
•
Mass of workers have skill endowment s and reservation wage r,
distributed according to F(s,r).
•
Workers can either choose to be “employed” in either sector A or sector B
(which pay wA and wB per efficiency unit, respectively), or they can choose
to work in “home” sector H.
•
Worker of type (s,r) can either supply s efficiency units in A or (1-s) in B.
o
•
Therefore, worker chooses employment in A iff swA > (1-s)wB and
swA > r
To simplify exposition, assume aggregate production function given by the
following:
Y = αLA + βLB
so that wA = α and wB = β
r
α
LH
β
LB
LA
s*
given by
αs*=β(1-s*)
s
r
α
LH
β
A→
H
LB
LA
A→
B
s*
s'
given by
αs*=β(1-s*)
s
r
α
LH
β
H→
B
A→
H
LB
A→H→B
LA
A→
A→
A→
B
B
s*
s'
s''
given by
αs*=β(1-s*)
s
Empirical model
•
Changes in Labor Market Outcomes at Local Level (Lk )
Lk  f (DkM , DkH , DkO ,  k )
(1)
(2)
(3)
(4)
Definitions:
(1)
Effect of Manufacturing Labor Demand Shock (through all channels)
(2)
Effect of “Housing Related” Labor Demand Shock
(3)
Effect of “Other” Labor Demand Shocks (not proxied by first two)
(4)
Effect of Labor Supply Shocks
Note:
k denotes a local labor market (e.g., MSA)
Lk could be employment rate, wages, employment in a sector, etc.
Empirical model
•
Changes in Labor Market Outcomes at Local Level (Lk )
Lk  f (DkM , DkH , DkO ,  k )
Our Goal:
Estimate:
Lk / DkM and Lk / DkH
Problems:
Note:
o
We do not observe DkM and DkH
o
We ideally want proxies which are orthogonal to the labor
supply shock.
We will estimate a causal channel of manufacturing shock on labor
market outcomes (housing will be more of a catch all).
Creating a Local Manufacturing Shock
•
Instrument for the local declines in manufacturing.
•
Construct predicted change in manufacturing employment following
Bartik (1991) (  D M ).
k
•
o
interact pre-existing cross-sectional variation in industry
employment with national industry employment trends.
o
Key assumption: initial industry variation across MSAs
uncorrelated with k (local labor supply changes)
Instrument is strongly predictive of actual changes in manufacturing
employment.
Creating a “Housing Related” Labor Demand Shock
•
Use housing price growth in local area ( PkH ) as our measure of
housing related demand shock.
•
Intuition – We have two direct housing related labor demand channels
o
Wealth Effect Channel:
Wk (PkH )
(+)
o
Construction Demand Channel:
Ck (PkH )
(?)
•
The relationship between construction effect on labor demand and house
prices will be positive if variation in house prices is due to variation in
housing demand.
Relationship Between Housing Price Growth and Change in
Construction Share (Non-College Men, 25-55)
Empirical Model
Lk   0  1 DkM   2 PkH   X k  DkO   k   k
•
Note:
Housing prices are endogenous
PkH    1 DkM   g (DkH ; Z k )   X k  DkO   k  Z k   k
•
Where Z is some measure of housing supply across locations.
•
H
Where Dk is some national change in housing demand.
Note:
We do not want to take a stand on why house prices changed
during the 2000s.
 Lk dWk Lk dCk Lk d DkO Lk d k 
d Lk
2 
 h
,
,
,
H
H
H
O
H
H 
d Pk

W
d

P

C
d

P

D
d

P


d

P
k
k
k
k
k
k
k
k 

What We Estimate
Lk   0  1 DkM   2 PkH   X k  DkO   k   k
Comment 1:
dLk / d DkM   1   2  1
o
Effect of manufacturing shock on labor market outcomes includes the
direct effect and the indirect effect through house prices
-
In essence, the house price measure is residualized of
manufacturing shock.
What We Estimate
Lk   0  1 DkM   2 PkH   X k  DkO   k   k
Comment 2:
o
We estimate the above via OLS
o
We also estimate the above instrumenting for PkH to isolate a more
causal channel of house prices on labor market outcomes.
-
Use variation in Z across places (Saiz developable land measure).
-
Use temporal variation in house price movements within a city (a
new instrument).
o
Not necessarily important for our purposes to estimate a causal
relationship. Want to isolate variation orthogonal to θk.
o
OLS results and IV results are very similar in most specifications.
Data For Main Results
•
2000 Census and 2005-2007 and 2009-2010 ACS
o
Most of our analysis comes using Census/ACS data.
o
All of our analysis starts in 2000 (as a result)
o
Focus on individuals aged 21-55.
•
FHFA metro house price indexes
•
Index of Available Land (Saiz 2010)
o
Identical results if we use his housing supply elasticity measure.
Time Periods
•
•
Base estimation: 2000 – 2007 period
o
Start in 2000 because of data limitations.
o
Want to focus on pre-recessionary period to get estimated responses.
o
Interesting to focus on the boom period (highlights masking).
Follow up with estimation during the 2007-2010
o
•
Can see if the responses change in different periods.
Discuss long changes in outcomes: 2000-2010
o
Highlights the role of the temporary effects of housing booms.
A Little More on the Bartik Instruments
Bartik Shock vs. Actual Change in Manufacturing
Bartik Shock and House Price Growth, 2000-2007
Estimates from the Empirical Model:
Some Graphical Results
Change in non-employment rate for non-college men, 2000-2007
Change in non-employment rate for non-college men, 2000-2007
Change in average wage for non-college men, 2000-2007
Change in average wage for non-college men, 2000-2007
Change in construction employment share, 2000-2007
Change in construction employment share, 2000-2007
Change in manufacturing employment share, 2000-2007
Change in manufacturing employment share, 2000-2007
Estimates from the Empirical Model:
Formal Estimates
Table 4: Baseline Results
Dependent variable:
Specification:
Change in Nonemployment Rate,
2000-2007
OLS
(1)
Change in Housing Prices
[Housing Boom]
-0.034
(0.011)
Predicted Change in Share of
Non-College Men Empl. in Manuf.
[Manufacturing Bust]
-0.724
(0.245)
Housing price effect (1σ)
Manufacturing effect (1σ)
-0.011
-0.010
First stage F-statistic
N
2
R
Include baseline controls
Instrument with land availability
235
0.741
y
Table 4: Baseline Results
Dependent variable:
Specification:
Change in Average Wage,
2000-2007
OLS
(1)
Change in Housing Prices
[Housing Boom]
0.059
(0.010)
Predicted Change in Share of
Non-College Men Empl. in Manuf.
[Manufacturing Bust]
1.545
(0.369)
Housing price effect (1σ)
Manufacturing effect (1σ)
0.018
0.021
First stage F-statistic
N
2
R
Include baseline controls
Instrument with land availability
235
0.444
y
Table 4: Baseline Results
Dependent variable:
Specification:
Change in Share of NonCollege Men Employed
in Construction,
2000-2007
OLS
(5)
Change in Share of
Non-College Men
Employed in
Manufacturing,
2000-2007
OLS
(7)
Change in Housing Prices
[Housing Boom]
0.024
(0.006)
0.001
(0.004)
Predicted Change in Share of
Non-College Men Empl. in Manuf.
0.450
(0.178)
1.025
(0.074)
Housing price effect (1σ)
Manufacturing effect (1σ)
0.007
0.006
0.000
0.014
First stage F-statistic
N
2
R
235
0.492
235
0.532
Table 4: Baseline Results
Dependent variable:
Specification:
Change in Share of NonCollege Men Employed
in Construction,
2000-2007
OLS
(5)
Change in Share of
Non-College Men
Employed in
Manufacturing,
2000-2007
OLS
(7)
Change in Housing Prices
[Housing Boom]
0.024
(0.006)
0.001
(0.004)
Predicted Change in Share of
Non-College Men Empl. in Manuf.
0.450
(0.178)
1.025
(0.074)
Housing price effect (1σ)
Manufacturing effect (1σ)
0.007
0.006
0.000
0.014
First stage F-statistic
N
2
R
235
0.492
235
0.532
Instrumenting for Housing Price Changes
Temporary Nature of The Housing Boom: Booms vs. Busts
House Price Growth and Saiz Instrument
Saiz Instrument and Construction Employment, 2000-2007
Alternate Housing Instrument Identification
House Price Growth and Alternate Housing Instrument
Estimates from the Empirical Model:
Formal Estimates
Table 4: Baseline Results, With Housing Instruments
Dependent variable:
Specification:
Change in Nonemployment Rate,
2000-2007
OLS
IV-Saiz
IV-Alt
(1)
(2)
(3)
Change in Housing Prices
[Housing Boom]
-0.034
(0.011)
-0.035
(0.015)
-0.022
(0.010)
Predicted Change in Share of
Non-College Men Empl. in Manuf.
[Manufacturing Bust]
-0.724
(0.245)
-0.694
(0.220)
-0.661
(0.205)
Housing price effect (1σ)
Manufacturing effect (1σ)
-0.011
-0.010
-0.011
-0.009
-0.008
-0.009
235
0.741
y
14.290
235
0.740
y
16.90
235
0.737
y
First stage F-statistic
N
2
R
Include baseline controls
Results are Robust To Many Alternate Specifications
•
Controlling for Census regions
•
Using sub measures of the land availability index
•
Including interactions between manufacturing shocks and housing
shocks
o
None of the interaction terms were significant
Population change, non-college men, 2000-2007
o
One standard deviation decline in Bartik manufacturing shock decreases
population growth by about 3 percent (from our main empirical specification)
Non-employment Effects for Other Groups
(One Standard Deviation Effect – IV Saiz Specification)
Housing Shock
Manufacturing
Non-college Men
-0.011
-0.009
Non-college Women
-0.008
-0.007
College Men
-0.006
-0.004
College Women
-0.000
-0.003
All
-0.009
-0.007
Long Run Results
Change in non-employment rate for non-college men, 2000-2010
Change in non-employment rate for non-college men, 2000-2010
Long Run Results: IV Saiz Specification, Standardized Effects
Change in Nonemployment Over Period
2007-10
2000-10
2000-10
2000-10
Housing Change
2000-2007
Housing Change
2007-2010
0.003
-0.017
-0.004
Housing Change
2000-10
Manufacturing Shock
Relevant Period
0.005
-0.007
-0.018
-0.018
-0.018
Long Run Results: IV Saiz Specification, Standardized Effects
Change in Nonemployment Over Period
2007-10
2000-10
2000-10
2000-10
Housing Change
2000-2007
Housing Change
2007-2010
0.003
-0.017
-0.004
Housing Change
2000-10
Manufacturing Shock
Relevant Period
0.005
-0.007
-0.018
-0.018
-0.018
Long Run Results: IV Saiz Specification, Standardized Effects
Change in Nonemployment Over Period
2007-10
2000-10
2000-10
2000-10
Housing Change
2000-2007
Housing Change
2007-2010
0.003
-0.017
-0.004
Housing Change
2000-10
Manufacturing Shock
Relevant Period
0.005
-0.007
-0.018
-0.018
-0.018
Long Run Results: IV Saiz Specification, Standardized Effects
Change in Nonemployment Over Period
2007-10
2000-10
2000-10
2000-10
Housing Change
2000-2007
Housing Change
2007-2010
0.003
-0.017
-0.004
Housing Change
2000-10
Manufacturing Shock
Relevant Period
0.005
-0.007
-0.018
-0.018
-0.018
Within and Between Masking
How Much of the Masking Comes from Within Individuals?
•
Spatial correlation of shocks
o
Shocks were in Different Places
Bartik Shock and House Price Growth, 2000-2007
Manufacturing “Instrument” vs. Saiz Housing “Instrument”
How Much of the Masking Comes from Within Individuals?
•
Spatial correlation of shocks
o
•
Shocks were in Different Places
Sub-groups of the populations
o
Look at masking across broad demographic groups.
o
Focus on age and nativity.
How Much of the Masking Comes from Within Individuals?
•
Spatial correlation of shocks
o
•
•
Shocks were in Different Places
Sub-groups of the populations
o
Look at masking across broad demographic groups.
o
Focus on age and nativity.
Within Individual Results (Displaced Worker Survey)
o
Construction does not absorb lots of displaced manufacturing workers.
o
Increased some in the 2000-2007 period.
o
Exploit variation in housing market conditions.
Document Within Worker Effects: Displaced Worker Survey
•
Manufacturing workers more likely to move into construction after job
displacement during housing boom years
•
Focus on sample of displaced manufacturing workers (prime age-ish,
men and women, all education levels) and look at fraction who were
reemployed at time of survey in construction.
1992
3.0%
2002
5.3%
1994
3.8%
2004
5.7%
1996
3.9%
2006
6.6%
1998
4.1%
2008
6.5%
2000
3.6%
Average
3.7%
2010
4.2%
Average 5.9% (p-value of diff < 0.01)
Document Within Worker Effects: Displaced Worker Survey
•
Focus on people who lost jobs in manufacturing (men and women, all
education groups, broad age range, data from 2000-2006 waves).
•
Collapse the data to a state level analysis. Focus on:
o
o
o
•
Correlate fractions with state house price growth 2000-2006.
o
•
Fraction who end up out of labor force
Fraction who end up not employed
Fraction who end up employed in construction
Weight observations by number of observations from each state.
Find that the propensity for individuals to end up out of the labor force
is decreasing in house price growth!
o
o
1 standard deviation increase in house price growth reduced out of the
labor force propensity by about 1.3 p.p. (out of base of 11 p.p.).
No power to say anything about construction or employment rates.
Counterfactuals
Extrapolating Local Estimates to National Labor Market
•
•
We try to address to several concerns with this exercise:
o
Migration
o
Housing Boom → Manufacturing demand
o
Construction Boom
o
Other National GE effects (e.g., interest rates)
To the extent we can address these concerns, they seem to indicate our
results are conservative.
0.12
0.10
Model Predictions: Manufacturing Counterfactuals
Total Increase
(Raw Data)
0.08
0.06
0.04
0.02
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
0.12
0.10
Model Predictions: Manufacturing Counterfactuals
Total Increase
(Raw Data)
0.08
0.06
0.04
Predicted Effect From
Manufacturing
0.02
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
0.12
0.10
Model Predictions: Manufacturing Counterfactuals
Total Increase
(Raw Data)
0.08
~40% of
Increase
0.06
0.04
Predicted Effect From
Manufacturing
0.02
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
0.12
0.10
Model Predictions: Manufacturing Counterfactuals
Total Increase
(Raw Data)
0.08
0.06
0.04
0.02
Predicted Effect From
Manufacturing
Predicted Effect From
Manufacturing Plus Housing
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
0.12
0.10
Model Predictions: Manufacturing Counterfactuals
Total Increase
(Raw Data)
0.08
0.06
0.04
0.02
Predicted Effect From
Manufacturing
Predicted Effect From
Manufacturing Plus Housing
~ 35% during
Recession
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Education
Propensity to Have At Least One Year of College (Age: 18-29)
Did Housing Boom Delay College Attendance?

Use same local labor market design to answer this question.

The answer is YES!

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.

During the bust, this trend reversed (but, not completely).

Estimates can explain about 80% of the time series change.
Other Results
Other Results
•
Counterfactual analysis for wages for less-skilled men implies “missing”
3.3% decline, coming primarily from lack of downward wage adjustment
during bust period.
Other Results
•
Counterfactual analysis for wages for less-skilled men implies “missing”
3.3% decline, coming primarily from lack of downward wage adjustment
during bust period.
•
Decomposing non-employment results into unemployment and nonparticipation.
o
Bartik instrument primarily affects non-participation
o
Suggests much of the medium run forces we are identifying are on
non-participation.
o
Rethink earlier work on sector shifts on labor markets (Lilien (1982),
Abraham and Katz (1986)). All such tests were on unemployment –
not non-employment!
Other Results
•
Counterfactual analysis for wages for less-skilled men implies “missing”
3.3% decline, coming primarily from lack of downward wage adjustment
during bust period.
•
Decomposing non-employment results into unemployment and nonparticipation.
•
o
Bartik instrument primarily affects non-participation
o
Suggests much of the medium run forces we are identifying are on
non-participation.
o
Rethink earlier work on sector shifts on labor markets (Lilien (1982),
Abraham and Katz (1986)). All such tests were on unemployment –
not non-employment!
Other boom/bust cycle: 1980s housing boom
• Preliminary results indicate broadly similar results
Conclusions
1. Manufacturing decline is important for thinking about changes in nonemployment during 2000s (including recession).
o
About 35-40% of increase in non-employment during 2000s can be
attributed to the decline in manufacturing.
2. Labor market was significantly weaker in the 2000-2007 period than
we thought.
o
Housing boom “masked” deterioration of U.S. labor market.
o
2000-2007 period marked by secular decline in one sector and a
temporary boom in another sector.
o
Implication: 2007 may not be a good benchmark to evaluate cyclical
changes in economic variables of interest.
Conclusions
3.
About one-third of the increase in non-employment during the
recession can be attributed to
o
manufacturing declines during 2007-2010 period, and
o
manufacturing declines during the 2000-2007 period that were masked
by housing boom.
4. The net effect of housing booms/busts on labor markets was small over
the entire decade.
o
5.
The bust reduced employment but the boom raised employed
Housing boom deterred college attainment during 2000-2007 period.