Historical and Future Employmentin the United States

Robert Shimer

June 13, 2014

Outline

Past and Future Employment in the United States -p. 2

demographics and historical changes in employment

hysteresis and future changes in employment

the impact of fiscal policy on employment

Demographics

Past and Future Employment in the United States -p. 3

Age Adjustments

Past and Future Employment in the United States -p. 4

employment-population ratio varies systematically by age

Age Adjustments

Past and Future Employment in the United States -p. 4

20 25 30 35 40 45 50 55 60 65 70 75 80

−0.2

0

0.2

0.4

age

empl

oym

entp

roba

bilit

y

Age Adjustments

Past and Future Employment in the United States -p. 4

employment-population ratio varies systematically by age

standard approach: hold e-pop by age fixed, vary population shares

CBO estimates that the demographic effect has already re-duced the overall rate of participation by about 0.5 percent-age points since 2007 and that it will do so by an additional1.2 points by 2016 and by another 1.4 points between 2016and 2021 (CBOs Labor Force Projections Through 2021,2011)

Age Adjustments

Past and Future Employment in the United States -p. 4

employment-population ratio varies systematically by age

standard approach: hold e-pop by age fixed, vary population shares

CBO estimates that the demographic effect has already re-duced the overall rate of participation by about 0.5 percent-age points since 2007 and that it will do so by an additional1.2 points by 2016 and by another 1.4 points between 2016and 2021 (CBOs Labor Force Projections Through 2021,2011)

concerns with this approach

very distinct behavior of e-pop within age groups

unclear why we focus only on age decompositions

Linear Probability Model

Past and Future Employment in the United States -p. 5

outcome of interest: employment status (binary)

linear function of

age dummies

education dummies

race dummies (black, white, other)

sex dummy

marital status dummy

month-of-year dummies

year dummies

focus on behavior of year dummies

Data Source

Past and Future Employment in the United States -p. 6

Current Population Survey

January 1994 to February 2014

nearly 25 million observations over age 16

employment-population ratio averages 37.8 percent

earlier figure on e-pop by age comes from this model

Removing Demographics: All Workers

Past and Future Employment in the United States -p. 7

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controls

Removing Demographics: All Workers

Past and Future Employment in the United States -p. 7

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controlsage controls

Removing Demographics: All Workers

Past and Future Employment in the United States -p. 7

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controlsage controlsfull controls

Removing Demographics: Prime-Aged Workers

Past and Future Employment in the United States -p. 8

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controls

Removing Demographics: Prime-Aged Workers

Past and Future Employment in the United States -p. 8

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controlsage controls

Removing Demographics: Prime-Aged Workers

Past and Future Employment in the United States -p. 8

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controlsage controlseducation controls

Removing Demographics: Prime-Aged Workers

Past and Future Employment in the United States -p. 8

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.06

−0.04

−0.02

0

0.02

year

empl

oym

entp

roba

bilit

y

no controlsage controlseducation controlsfull controls

Summary

Past and Future Employment in the United States -p. 9

all workers:

raw data: 3.3 percentage point decline in e-pop since 2008

age adjustment cuts that in half, to 1.8 percentage points

other controls raise it back to 2.3 percentage points

prime-aged workers:

raw data: 2.7 percentage point decline in e-pop since 2008

age adjustment has virtually no impact

other controls increase it to 2.9 percentage points

Forecasting

Past and Future Employment in the United States -p. 10

future demographic adjustments are conceptually straightforward

what about future within-group changes?

unemployment rate is virtually the same in 1994 and 2014

e-pop is down 3 percentage points (4 for prime-aged workers)

why? job polarization phenomenon

will this continue?

Employment Probability by Education

Past and Future Employment in the United States -p. 11

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

0.04

0.06

57.5

year

empl

oym

entp

roba

bilit

y

less than high school

∗prime-aged workers, with other controls

Employment Probability by Education

Past and Future Employment in the United States -p. 11

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

0.04

0.06

57.5

71.1

year

empl

oym

entp

roba

bilit

y

less than high schoolhigh school

∗prime-aged workers, with other controls

Employment Probability by Education

Past and Future Employment in the United States -p. 11

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

0.04

0.06

57.5

71.175.9

year

empl

oym

entp

roba

bilit

y

less than high schoolhigh schoolsome college

∗prime-aged workers, with other controls

Employment Probability by Education

Past and Future Employment in the United States -p. 11

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

0.04

0.06

57.5

71.175.9

83.6

year

empl

oym

entp

roba

bilit

y

less than high schoolhigh schoolsome collegecollege

∗prime-aged workers, with other controls

Employment Probability by Education

Past and Future Employment in the United States -p. 11

1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−0.1

−0.08

−0.06

−0.04

−0.02

0

0.02

0.04

0.06

57.5

71.175.9

83.6

88.5

year

empl

oym

entp

roba

bilit

y

less than high schoolhigh schoolsome collegecollegetertiary degree

∗prime-aged workers, with other controls

Hysteresis

Past and Future Employment in the United States -p. 12

Hysteresis

Past and Future Employment in the United States -p. 13

consequences of job loss for long-tenure workers

0

–30

–40

–10

Average earnings loss as a percent of predisplacement earningsd

–2 –1–4 0 1 2 4 6 8 10 12 14 16 18

In expansions

In recessions

Percent

–20

Years

—Davis and von Wachter (2011)

Hysteresis

Past and Future Employment in the United States -p. 13

consequences of job loss for long-tenure workers

0

–30

–40

–10

Average earnings loss as a percent of predisplacement earningsd

–2 –1–4 0 1 2 4 6 8 10 12 14 16 18

In expansions

In recessions

Percent

–20

Years

—Davis and von Wachter (2011)

consequences of entering the labor market during a recession

Causes of Earnings Hysteresis

Past and Future Employment in the United States -p. 14

wages versus employment rates

Ruhm (1991): unemployment impact mostly gone in four years

Jarosch (2014): employment impact lasts ten+ years (Germany)

von Wachter: full year nonemployment elevated after 20 years

some evidence that wage declines reflect lost experience

Borovickova, Davis, and Rogerson (2014)

simplest search models miss out on much of this persistence

declining hazard of finding a job

repeated job displacement episodes

lower wage once a worker finds a job

Stopping Time Model

Past and Future Employment in the United States -p. 15

persistent state variable x(t): wage or productivity

work when state is high

stop working when state is low

recession causes a persistent decline in the distribution of x(t)

lower future wages

lower future employment

current research (with Alvarez and Borovickova):

nonparametric test and estimate of the model

decompose duration dependence into structural v. heterogeneity

Alvarez, Borovi ckov a, and Shimer (2014)

Past and Future Employment in the United States -p. 16

0 50 100 150 200 250 300 350 4000

100

200

300

400

500

total

realized duration in weeks

mea

nre

sidu

aldu

ratio

nin

wee

ks

Alvarez, Borovi ckov a, and Shimer (2014)

Past and Future Employment in the United States -p. 16

0 50 100 150 200 250 300 350 4000

100

200

300

400

500

total

portion attributed to heterogeneity

realized duration in weeks

mea

nre

sidu

aldu

ratio

nin

wee

ks

Summary of our Research

Past and Future Employment in the United States -p. 17

substantial persistence in job loss

most duration dependence is structural

caveats

Austrian data, not U.S.

different sources of shocks, not the Great Recession

no direct implications for what happens to new entrants

conjecture they start with a stochastically lower value of x(t)

similar subsequent dynamics to existing employees

The Impact of Fiscal Policy

Past and Future Employment in the United States -p. 18

What is the Elasticity of Labor Supply?

Past and Future Employment in the United States -p. 19

CBOs microsimulation model indicates the percentage change inlabor input that would be caused by the expected increase in themarginal tax rate on labor. That model incorporates estimatedlabor supply elasticities (the percentage change in total hours ofwork supplied that would result from a 1 percent increase in bothafter-tax income and the after-tax wage rate) that average 0.07for primary earners within a household and 0.40 for secondaryearners, implying an average elasticity of 0.14 for the entire laborforce. Consistent with the available evidence, CBO assumed thatabout half of the response reflects people entering and exitingthe labor force; the rest reflects decisions by working people tochange their hours on the job. (CBOs Labor Force ProjectionsThrough 2021, 2011)

What is the Elasticity of Labor Supply?

Past and Future Employment in the United States -p. 20

I estimate a Hicksian labor supply elasticity of 0.33 on the inten-sive margin and 0.25 on the extensive margin after accounting forfrictions. (Chetty “Bounds on Elasticities” Econometrica 2012)

large long-run responses to tax changes (4× CBO estimates)

wealth effects are not that important

top of the Laffer curve at around τ = 0.63

modest short-run responses to tax changes

adjustment costs are important

workers may not be able to choose labor supply

(still larger a bit larger than the long-run response)

Appendix: Laffer Curve Calculations

Past and Future Employment in the United States -p. 21

representative worker maximizes log c− γεh1/ε

budget constraint c = (1− τ)wh + T

government rebates taxes lump-sum to households

T = τwh, but each person treats T as given

solution: h = ((1− τ)/γ)ε

ε is the Frisch elasticity of labor supply

tax revenue is proportional to τ(1− τ)ε

maximized at τ = 1/(1 + ε) ≈ 0.63