Countercyclical Restructuring and Jobless Recoveries David Berger Northwestern David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Jun 16, 2015
Countercyclical Restructuring and Jobless Recoveries
David Berger
Northwestern
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Motivation: understanding jobless recoveries
The Great Recession1 Hours and employment still 7% below pre-recession high2 Average labor productivity was back above trend by 2009q2
.9.9
51
1.05
1.1
2007q1 2008q1 2009q1 2010q1 2011q1
Total hoursLabor productivity
Hours and Average Labor Productivity in the Great Recession
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Stylized facts
Until mid-1980s
1 Employment rebounded quickly during recovery2 Average labor productivity (ALP) was strongly procyclical
Since mid-1980s
1 Last three recoveries have been jobless2 Low or negative correlation between ALP and output
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
New stylized facts
1 Jobless recoveries2 Less cyclical ALP
20
24
6%
chg
0 2 4 6 8Quarters since NBER trough
Avg pre911990912001200709
Total Hours
.50
.51
1947q3 1960q1 1972q3 1985q1 1997q3 2010q1
Rolling correlation: output and ALP
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Results
Model can match new business cycle facts
Restructuring e¤ect on the intensive margin
Basic mechanism
Firms get "fat" during booms
learning takes time and adjustment is costly
Restructuring is concentrated in recessions
Employment adjustment cost is procyclical and costly in terms ofcurrent output
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Intuition
Acyclical ALP
Selective �ring ) match quality is countercyclical
Jobless recoveries
Fire low quality workers and hire average quality workers
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Results
Provide evidence on what changed in the 1980s
Large decline in union power ) more scope for selective �ringFirms more able to learn about worker quality
Model explains change by increased ability to selectively �re
Worker heterogeneity model without selective �ring explains pre-1984Worker heterogeneity model with selective �ring explains post-1984
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Fact 1: Jobless recoveries
20
24
6%
chg
0 2 4 6 8Quarters since NBER trough
Avg pre911990912001200709
Total Hours
20
24
6%
chg
0 2 4 6 8Quarters since NBER trough
Employment
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Fact 2: Correlation of output and ALP
.50
.51
1947q3 1960q1 1972q3 1985q1 1997q3 2010q1
Rolling correlation: output and ALP
Decline is robust to di¤erent choices of labor and output measure
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Two period model
Goal:
1 Illustrate how heterogeneity, learning and selective �ring interact togenerate restructuring margin
2 Show how model can generate both countercyclical ALP and joblessrecoveries
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Setup
Production
One homogeneous good is produced and sold in competitive markets
2 types of matches:
E¢ ciency units θH and θL with θH � θL
Total e¢ ciency units: θ(α) = αθH + (1� α)θL
Production function of establishment i : yi = z(θ(α)Li )γ
Learning
Match speci�c productivity of all new workers is unknown ex-ante
Average match quality of a new hire is exogenous and equal to p
Learn quality of match through production after one period (Jovanovic 1979)
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Timing
First periodFirms endowed with (z1, L1, α1)
Second periodLearn quality of all employeesFirms hire or �re and produce output
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Firms Problem
Given α1, L1 and z2, the �rm�s optimal employment in the second period, L2,solves:
maxL2z2�θ(α2)L2
�γ � L2
where γ 2 (0, 1) and α2 is given by the following weighted average:
α2 =
8>><>>:α1�L1L2
�+p
�L2�L1L2
�if hiring
α1�L1L2
�if �ring fewer than (1� a1)L employees
1 if �ring more than (1� a1)L employees
9>>=>>;
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Implication 1: ALP
Consider what happens after a fall in TFP: z2 = (1� τ)z1Aggregate output and labor: Y =
Ryi and L =
RLi
ALP: Y /L
Selective �ring causes ALP to increase:
(Y /L)f > Y /L
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Implication 2: Jobless recoveries
Consider what happens after an increase in TFP: z2 = (1+ τ)z1Employment growth rate if hiring:
Homogeneous �rm:
ge =h(1+ τ)
11�γ � 1
i=) hire when τ > 0
Heterogenous �rm:
ge =�
θ(α1)
θ(p)
� "�(1+τ)θ(p)
θ(α1)
� 11�γ � 1
#De�ne τ� = θ(α1)
θ(p)� 1
=) hire when τ > τ�
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Implication 2: Jobless recoveries
We learn two things from : τ� = θ(α1)
θ(p)� 1
1 τ� > 0
Firms need larger TFP draws to begin hiring than in the homogenousworker model
2 τ� is increasing in α1
Firms with a better matched workforce in period one need larger TFPto begin hiring
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
What changed in the 1980s
Large decline in union power in the U.S. during the 1980s
Unions: layo¤s done by seniority rather than by quality
Union power # leads to more scope for selective �ring
Decline in union power:
2000
4000
6000
8000
1000
0N
umbe
r of e
lect
ions
1960 1965 1970 1975 1980 1985 1990 1995 2000
Source: Farber and Western (2004)
Number of NLRB Certification Elections
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
What changed in the 1980s
Large decline in union power in the U.S. during the 1980s
Unions: layo¤s done by seniority rather than by quality
Union power # leads to more scope for selective �ring
Decline in union power:
2000
4000
6000
8000
1000
0N
umbe
r of e
lect
ions
1960 1965 1970 1975 1980 1985 1990 1995 2000
Source: Farber and Western (2004)
Number of NLRB Certification Elections
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
What changed in the 1980s
Large decline in union power in the U.S. during the 1980s
Unions: layo¤s done by seniority rather than by quality
Union power # leads to more scope for selective �ring
Decline in union power:
2000
4000
6000
8000
1000
0N
umbe
r of e
lect
ions
1960 1965 1970 1975 1980 1985 1990 1995 2000
Source: Farber and Western (2004)
Number of NLRB Certification Elections
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
What changed in the 1980s
Large decline in union power in the U.S. during the 1980s
Unions: layo¤s done by seniority rather than by quality
Union power # leads to more scope for selective �ring
Decline in union power:20
0040
0060
0080
0010
000
Num
ber o
f ele
ctio
ns
1960 1965 1970 1975 1980 1985 1990 1995 2000
Source: Farber and Western (2004)
Number of NLRB Certification Elections
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Testing the union power hypothesis
1 Idea: states with larger declines in union coverage rates should have largedeclines in corr(Y , YE )
2 Data: Annual data on output, employment and % of workers covered bycollective bargaining aggreements
3 Estimating Equation:
(ρi ,post95 � ρi ,pre85) = α+ β log(UC i ,post95/ UC i ,pre85) + ε
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
State level results
Change in corr(Y,Y/E) % change in corr(Y,Y/E)bα 0.082 0.182(0.183) (0.291)bβ 0.565* 0.878**(0.292) (0.431)
Observations 51 51R-squared 0.079 0.083
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
State level results: right-to-work states
Idea: If union power hypothesis is correct, results should be stronger in nonright-to-work states where unions were historically more powerful
Right-to-work states Non right-to-work statesChange in corr(Y,Y/E) Change in corr(Y,Y/E)bα -0.238 0.222
(0.393) (0.300)bβ 0.067 0.894**(0.470) (0.322)
Observations 21 29R-squared 0.006 0.128
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries
Conclusion
Since mid-1980s:
Employment recoveries have been joblessALP has been acyclical
Standard models cannot match these facts
Model with countercyclical restructuring margin:
1 Generates a decline in the procyclicality of ALP2 Generates jobless recoveres after a large recessions3 Both output growth and changing productivity dynamics important
Provided evidence that structural change related to decline in union power
David Berger (Northwestern) Countercyclical Restructuring and Jobless Recoveries