Discussion of Chen, Kannan, Loungani, and Trehan

Daniel Aaronson

Chicago Fed

April 2011

Summary of paper

• Estimate shocks to sectoral reallocation using dispersion in industry stock market returns.

• Interesting result for at least two reasons:

1. Explains ½ (!!) of the historic rise in LTU (Figure 10). a. Nice check -- No impact on ST unemployment, which is less likely to be

“structural.”

2. Potential estimate of “structural unemployment” (Fig 11). Very large effects here too.

• Key assumption:

– Assumes movements in stock prices within an industry reflect permanent (structural) shifts, rather than temporary (cyclical) changes.

– But fair amount of evidence that “structural change” coincides with cyclical downturns. Makes identification difficult.

– Particularly tricky when industry cyclicality varies.

Some reasons to be skeptical of magnitudes

1. Using the current period in the estimation.

• Concern: Big outlier (dispersion, LTU) in 2008/09 much of the variation used to predict rise in LTU and structural unemployment recently.

y = 0.9199x + 0.0647R² = 0.1469

y = 0.4755x + 0.0965R² = 0.0477

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0 0.05 0.1 0.15 0.2 0.25

Shar

e o

f u

ne

mp

loye

d >

26

we

eks

Chen et al Dispersion index, lagged 8 quarters

Why is this a potential concern?

• Lots going on recently. Control for overall stock market dispersion is nice but not far enough. Much more of this would be great (broader financial conditions, within-industry dispersion).

• Need out of sample tests. Effects almost surely smaller.

-4

-3

-2

-1

0

1

2

3

4

5

6

Jan

-73

Jan

-75

Jan

-77

Jan

-79

Jan

-81

Jan

-83

Jan

-85

Jan

-87

Jan

-89

Jan

-91

Jan

-93

Jan

-95

Jan

-97

Jan

-99

Jan

-01

Jan

-03

Jan

-05

Jan

-07

Jan

-09

Jan

-11

Stan

dar

d d

ev

un

its

Chicago Fed National Financial Conditions Index

Removes variation due to economic activity (CFNAI) and inflation (PCE)

http://chicagofed.org/webpages/publications/nfci/index.cfm

Long lags between business cycle shock and the rise in

LTU are standard parts of recovery.

• But do we think that recessions always involve sectoral reallocation? Not clear (some evidence in a few slides).

0

5

10

15

20

25

30

35

40

45

50

0 2 4 6 8 10 12

Frac

tio

n o

f La

bo

r Fo

rce

Un

em

plo

yed

27

w

ee

ks +

Unemployment rate

Unemployment Rate vs share long-term unemployed, selected cycles

12/07-3/1211/73-10/777/81-6/857/90-6/943/2001-2/2005

2. Suggests alot more structural unemployment (8.5-

9%) than other approaches

Shock: 16% reduction in match efficiency (BF 2011)New Beveridge Curve

With free entrycondition (and otherassumptions),u/v must rise by 45%.

Shock to matchefficiency can only leadto an upper bound of7% SS unemployment

Barlevy (2011, forthcoming)

And its probably even lower: 1. Davis et al (2010) – Firm search intensity; 2. UI – worker search intensity

(Aaronson/Mazumder, Mazumder, Valetta, etc.)So can’t all be about mismatch (within DMP framework).

2 pp

And a lot more sector reallocation

(e.g. Sahin et al (2011))

• Economy is a large number of distinct labor markets (industries, occupations, etc) with unemployed resources and vacancies.– Kind of like a industry-specific (or occupation-specific, etc) Beveridge Curve

analysis

• Question: how much unemployment can be “fixed” by freely moving resources to job openings in other sectors. – i.e. the share of unemployment due to workers being in the “wrong” labor

market (the one without openings).

• By industry: explains 0.4 to 0.7 perc points of 5 point rise during recession and early recovery. Most of this is construction (and a little manufacturing).

• Index has a little trouble with timing because it reverts back pretty quickly in 2009-10 when UR and LTU are particularly high (unless there are long lags like in Prakash’s VARs).

-30

-20

-10

0

10

20

30

Construction

-30-20-10

010203040

Durable manufacturing

-20-15-10

-505

1015

Nondurable manufacturing

-15

-10

-5

0

5

10

15

Transportation and utilities

-10

-5

0

5

10

15

Wholesale Trade

-10

-5

0

5

10

15

Retail Trade

-10

-5

0

5

10

Finance, insurance, and real estate

-6-4-202468

10

Services

-4-202468

10

Government

Employment has fallen broadly

-30

-20

-10

0

10

20

30

Construction

-30-20-10

010203040

Durable manufacturing

-20-15-10

-505

1015

Nondurable manufacturing

-15

-10

-5

0

5

10

15

Transportation and utilities

-10

-5

0

5

10

15

Wholesale Trade

-10

-5

0

5

10

15

Retail Trade

-10

-5

0

5

10

Finance, insurance, and real estate

-6-4-202468

10

Services

-4-202468

10

Government

Employment has fallen broadly (and fairly cyclically)

Statistical model of Rissman (2010): Noncyclical reallocation

barely budged

0

1

2

3

4

5

6

7

8

1950 1954 1958 1962 1966 1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 2010

Important caveat: Limited categorization (9 industries). Mismatch may be within industry or across other labor market segmentations.

Rissman, Ellen, 2010, Economic Perspectives.

Aaronson et al (2010)

• Oaxaca-Blinder decomposition of long-term unemployment changes over last 30 years.

• Virtually no evidence that across-industry (again, broad) changes can explain much.

• Within-industry changes are actually much more useful. Maybe this is picking up some of Prakash’s results. – Easy test: play with the level of aggregation.

• Rob V has better evidence on how unusual the current period is, after adjusting for demographics.

0.00

0.01

0.02

0.03

0.04

0.05

1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009

3. Is there Unusual Demand in Certain Sectors? Standard

wage data doesn’t suggest it.Dispersion of private industry hourly earnings(year-over-year)

Highest industry less averageStandard deviation of industries

Average hourly earnings, by private industry (CES). Excludes mining and logging. Red line = highest industry less average industry. Blue line = standard deviation of industry average hourly earnings.

Industry wage dispersion in the SIPP. Mixed evidence?

• Pro: Can deal with some compositional biases (not completely done yet…).

• Con: Very Preliminary, not many industries (yet), smallish samples

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

19

84

19

85

19

86

19

87

19

88

19

89

19

90

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

20

07

20

08

20

09

20

10

Dispersion of private industry wage growth , SIPP

Job switchers

Spell of U

Job stayers

*Thru Mar 2010

Real wage growth among hires from U is falling in

tandem

-0.10

-0.05

0.00

0.05

0.10

0.15

19

84

19

85

19

86

19

87

19

88

19

89

19

90

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

20

07

20

08

20

09

20

10

Log

wag

e c

han

ge

Annual Real Wage Growth, Individuals experiencing E-U-E

Goods-producing

Finance

Construction

Change, 2008-2010 vs 1984-2007 growth

Finance -0.099Construction -0.044Goods prod -0.037Other serv -0.018Gov +0.029

*Thru Mar 2010

Hiring is broadly bad. Some sectoral reallocation likely

but evidence is far from clear that its substantial.

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

40.0%

45.0%

Jan

-05

Jan

-06

Jan

-07

Jan

-08

Jan

-09

Jan

-10

0-4 weeks

5-14 weeks

15-26 weeks

27 plus weeks

0.0

2.0

4.0

6.0

8.0

10.0

12.0

1976 '81 '86 '91 '96 '01 '06

TotalCollege*

Unemployment rates(total; college degree or more, aged 16-24, SA percent)

* Recent college grads = No UI, no homes, highly mobile (geographically, sector). Yet UR rose to similar heights.

Fraction that move from U to E, by initial unemployment duration(total unemployed in t-1; SA percent)

Hiring bad at every duration spell.

Extra slides

Wage changes among LTU (SIPP)

-0.15

-0.1

-0.05

0

0.05

0.1

19

84

19

85

19

86

19

87

19

88

19

89

19

90

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

20

07

20

08

20

09

20

10

Log

Wag

e C

han

ge

Annual Wage Growth by Spell Length (with 90% confidence intervals)

1 to 6 month spell

7 to 12 month spell

*Thru Mar 2010