The Ins and Outs of Unemployment: AConditional Analysis

Fabio Canova, ICREA-UPF, AMeN and CEPR

Claudio Michelacci, CEMFI and CEPR

David Lopez Salido, Federal Reserve Board and CEPR

December 2008

Motivation

- Conventional wisdom: recessions (high unemployment) begin with layoffs

and persist because unemployed can’t find jobs.

- Hall (2005) and Shimer (2007): job finding rate fluctuates; job separation

rate acyclical.

- Problems:

i) What drives fluctuations in finding rates?

ii) What is the direction of causality? Could movements in separation rates

drive fluctuations in finding rates?

iii) Do conclusions hold true for important business cycle shocks?

Contribution

- Analyze the dynamics of unemployment in technology induced recessions

(consider investment-neutral and investment-specific shocks).

- Look both the intensive margin (hours per employee) and the extensive

margin (number of employed workers) of the labor market.

- Characterize unemployment dynamics in terms of the job separation and

the job finding rates.

Results 1

- Investment-neutral technology shocks increase unemployment in the short

run and affect labor market variables along the extensive margin.

- Investment-specific technology shocks expand aggregate hours (both

hours per worker increase and unemployment falls), but the intensive mar-

gin more important.

- Impact response of unemployment is almost entirely due to a jump in the

separation rate. Dynamics due to movements in the finding rate.

- Neutral shocks important sources of cyclical fluctuations in unemploy-

ment. They explain the recession of the late 80’s and the subsequent

(jobless) recovery of the early 90’s.

Results 2

Results challenge sticky price explanation of technology shock/hours rela-

tionship (Galı (1999)).

- When technology improves and monetary policy is not accommodating

enough, demand is sluggish to respond and firms take advantage of tech-

nology improvements to economize on labor input.

- Naturally applies to the intensive margin - displacing workers is more

costly than changing prices.

- Extensive margin plays a key role for neutral shocks. Fall in hours is

related to the reallocation of workers across jobs. Consistent with the

Schumpeterian creative destruction (see model in Canova, et. al. (2007)).

The empirical model

- Basic specification: X = (∆q, ∆yn, h, u, s, f)0.

- q and GDP scaled by output deflator.

- Identification of shocks: long run restrictions as in Fisher (2006) or

Michelacci and Lopez Salido (2007).

- Use a VAR with 8 lags, a constant and a decay restriction.

- f and s: both approximate/exact flow rates as in Shimer (2005).

Problem: low frequency movements

- There is a bias in full sample results because of low frequency comove-

ments

- Taking subsamples is inefficient and probably biased - LR restrictions

used on a small sample problematic.

- Subsample dynamics roughly unchanged; VAR constant is affected. Break

the constant at 1973:2 and 1997:1.

Neutral Shock55:I-00:IV (continuous), 55:I-73:I (dotted), 73:II-97:I (dash-dotted)

Relative Price of Investment

5 10 15 20 25-0.08

0.00

0.08

0.16

0.24

Labor Productivity

5 10 15 20 250.18

0.36

0.54

0.72

0.90

Unemployment

5 10 15 20 25-1.8

0.0

1.8

3.6

5.4

Hours

5 10 15 20 25-0.6

-0.4

-0.2

-0.0

0.2

Finding Rate

5 10 15 20 25-4.2

-2.8

-1.4

0.0

1.4

Separation Rate

5 10 15 20 25-2

-1

0

1

2

Hours per Employee

5 10 15 20 25-0.32

-0.24

-0.16

-0.08

0.00

0.08

Output

5 10 15 20 25-0.25

0.00

0.25

0.50

0.75

1.00

Investment Specific Shock55:I-00:IV (continuous), 55:I-73:I (dotted), 73:II-97:I (dash-dotted)

Relative Price of Investment

5 10 15 20 25-1.25

-1.00

-0.75

-0.50

-0.25

Labor Productivity

5 10 15 20 25-0.4

-0.3

-0.2

-0.1

-0.0

0.1

Unemployment

5 10 15 20 25-1.8

-1.2

-0.6

0.0

0.6

1.2

Hours

5 10 15 20 25-0.2

0.0

0.2

0.4

0.6

Finding Rate

5 10 15 20 25-3

-2

-1

0

1

Separation Rate

5 10 15 20 25-2.0

-1.5

-1.0

-0.5

0.0

0.5

Hours per Employee

5 10 15 20 25-0.16

0.00

0.16

0.32

0.48

Output

5 10 15 20 25-0.36

-0.18

0.00

0.18

0.36

0.54

Neutral shock Investment specific shock

Evidence

- (Positive) Neutral shock increases unemployment; hours fall. Change in

hours-per-employee small and insignificant.

- Unemployment increases on impact due to the rise in the separation rate

and of the fall in the job finding rate. The separation rate returns quickly

to normal; the job finding rate takes up to fifteen quarters to recover.

- Output takes about 5 quarters to significantly respond but then gradually

increases until it reaches its new higher long-run value.

- Positive investment specific shock increases output and hours per capita;

unemployment fall (insignificant).

- The fall of unemployment on impact is due to a drop in the separation

rate. This effect is partly compensated by a fall in the job finding rate.

- The increase in hours is due to the sharp and persistent increase in the

number of hours-per-employee.

- Labor market adjustments to neutral technology shocks along the ex-

tensive margin; those in response to an investment specific technology

shock along the intensive margin.

Neutral ShockRelative Price of Investment

5 10 15 20 250.00

0.05

0.10

0.15

0.20

Labor Productivity

5 10 15 20 250.18

0.36

0.54

0.72

0.90

Unemployment

5 10 15 20 250.0

1.6

3.2

4.8

Hours

5 10 15 20 25-0.6

-0.4

-0.2

-0.0

0.2

Finding Rate

5 10 15 20 25-4.8

-3.2

-1.6

-0.0

Separation Rate

5 10 15 20 25-0.5

0.0

0.5

1.0

1.5

2.0

Hours per Employee

5 10 15 20 25-0.36

-0.27

-0.18

-0.09

-0.00

0.09

Output

5 10 15 20 25-0.25

0.00

0.25

0.50

0.75

1.00

Investment Specific ShockRelative Price of Investment

5 10 15 20 25-1.12

-0.96

-0.80

-0.64

-0.48

-0.32

Labor Productivity

5 10 15 20 25-0.32

-0.24

-0.16

-0.08

0.00

0.08

Unemployment

5 10 15 20 25-2.4

-1.6

-0.8

-0.0

0.8

Hours

5 10 15 20 250.00

0.25

0.50

0.75

Finding Rate

5 10 15 20 25-1.8

-0.9

0.0

0.9

1.8

Separation Rate

5 10 15 20 25-2.5

-2.0

-1.5

-1.0

-0.5

0.0

Hours per Employee

5 10 15 20 250.0

0.2

0.4

0.6

Output

5 10 15 20 25-0.35

0.00

0.35

0.70

Neutral shock Investment specific shockApproximate rates.

Neutral ShockRelative Price of Investment

5 10 15 20 25-0.025

0.000

0.025

0.050

0.075

0.100

Labor Productivity

5 10 15 20 250.4

0.5

0.6

0.7

0.8

0.9

Unemployment

5 10 15 20 250

1

2

3

4

5

Hours

5 10 15 20 25-0.64

-0.48

-0.32

-0.16

0.00

Finding Rate

5 10 15 20 25-4.2

-2.8

-1.4

0.0

Separation Rate

5 10 15 20 250.0

1.2

2.4

3.6

Hours per Employee

5 10 15 20 25-0.36

-0.24

-0.12

0.00

Output

5 10 15 20 25-0.25

0.00

0.25

0.50

0.75

1.00

Investment Specific ShockRelative Price of Investment

5 10 15 20 25-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

Labor Productivity

5 10 15 20 25-0.36

-0.24

-0.12

0.00

Unemployment

5 10 15 20 25-2.1

-1.4

-0.7

0.0

Hours

5 10 15 20 250.0

0.2

0.4

0.6

Finding Rate

5 10 15 20 25-1.0

-0.5

0.0

0.5

1.0

1.5

Separation Rate

5 10 15 20 25-3

-2

-1

0

Hours per Employee

5 10 15 20 250.00

0.16

0.32

0.48

Output

5 10 15 20 25-0.32

-0.16

0.00

0.16

0.32

0.48

(a) Neutral technology shock (b) Investment specific technology shock

Exact rates (dotted lines) and approximated rates (solid lines).

Comparison exact/approximate rates

- The sign/shape of responses are similar.

- For neutral shocks: with exact rates, dynamics of the separation rate

more accentuated.

- For investment shocks with exact rates, the separation rate falls more on

impact. The fall in the unemployment rate is more pronounced (extensive

margin more important).

Difference with Hall (2005) and Shimer (2007)

- Analysis is conditional on technology shocks, rather than unconditional.

- Measures the ins and outs of unemployment on impact and over the

adjustment path, rather that at generic business cycle frequencies.

- Permits feedbacks in response to technology shocks.

Fujita and Ramey (2006) and Yashiv (2007) look unconditionally to the

same data as Hall and Shimer, but results differ.

How important is the separation rate?

- Stock of unemployment evolves as:

ut = S(lt − ut)− Fut (1)

lt =labor force and ut = unemployment, S and F are the separation and

finding rates in levels. The unemployment rate converges to the following

fictional rate:

u =S

S + F≡ exp(s)

exp(s) + exp(f).

- Shimer (2007): u close to actual unemployment rate.

- Can calculate the contribution of f and s to the cyclical fluctuations in

u and evaluate how accurately u approximates u in technology induced

recessions (misses movements in and out of the labor force).

Contribution of Separation rateu (continuous), u-fictional (dotted), Separation only (dash-dotted)

Neutral Shock

5 10 15 20 25-0.9

0.0

0.9

1.8

2.7

Investment Specific Shock

5 10 15 20 25-0.96

-0.80

-0.64

-0.48

-0.32

-0.16

Contribution of Separation rateu (continuous), u-fictional (dotted), Separation only (dash-dotted)

Neutral Shock

5 10 15 20 25-0.9

0.0

0.9

1.8

2.7

Investment Specific Shock

5 10 15 20 25-1.25

-1.00

-0.75

-0.50

-0.25

0.00

Approximated rates Exact rates

True unemployment rate: solid line; u:dotted line. Dash-dotted is u|f fixed

- Impact effect of u after a neutral shock explained by s, especially with

exact rates (90 per cent). Contribution falls to 40 per cent after one

quarter and to 20 per cent one year after the shock.

- Impact response of actual and fictional unemployment different: flow in

and out of the labor force important.

- Following an investment specific shock, unemployment falls little on im-

pact with approximate rates(the fall in s and the fall in f compensate each

other) and a lot with exact rates (mainly due to the fall in s).

- Impact response of actual and fictional unemployment similar: others

labor market flows small.

The contribution of technology shocks

Variable Neutral Investment specificHorizon (quarters) Horizon (quarters)1 8 16 32 1 8 16 32

A. Approximated ratesOutput 1 6 30 55 3 5 5 4Hours 8 9 8 7 14 16 21 22Hours per Worker 5 5 4 4 17 23 29 29Unemployment 23 21 21 21 3 3 6 6Finding Rate 17 17 17 17 0 1 2 2Separation Rate 10 8 7 6 5 8 12 14B. Exact ratesOutput 8 4 17 37 14 8 6 6Hours 22 19 18 16 24 15 14 14Hours per Worker 14 12 11 10 35 27 28 28Unemployment 34 30 29 27 3 1 1 1Finding Rate 1 25 24 24 0 1 2 3Separation Rate 34 34 30 26 0 1 1 1

- Neutral shocks explain a substantial proportion of the volatility of unem-

ployment (20 per cent with approximate rates), little of hours (5 percent).

- Investment specific shocks account for a substantial proportion of the

volatility of hours (20 per cent of hours per capita, 30 per cent of hours

per worker) and little of unemployment (less than 10 per cent).

- With exact rates the numbers are larger.

Data and Technology componentUnemployment

1974 1978 1982 1986 1990 1994 1998-32

-16

0

16

32

Finding

1974 1978 1982 1986 1990 1994 1998-24

-12

0

12

24

Hours

1974 1978 1982 1986 1990 1994 1998-5.0

-2.5

0.0

2.5

5.0

Separation

1974 1978 1982 1986 1990 1994 1998-16

0

16

32

Original solid; component due to technology dotted

All filtered with HP(1600). Grey areas are NBER recessions.

- Technology shocks drive cyclical fluctuations in labor market variables

(more for unemployment than for hours).

- They account for several important business cycle episodes, including the

recession of the late 80’s and the subsequent remarkably slow labor market

recovery of the early 90’s.

Original solid, component due to technology shocks dotted

Exact rates. All HP(1600) filtered. The vertical lines comprise the NBER recession.

- Recession of late 1980 generated a jobless recovery (see Bernanke (2003)).

- Downturn in employment severe; peak in unemployment lags by two years

the trough in output - different from previous episodes.

- The technology component of the data tracks quite closely the evolution

of the raw data.

- Mainly due to neutral shocks that naturally induce jobless recoveries: fol-

lowing the initial rise in job separation and unemployment, output increases

to its new higher long run value, while unemployment remains above trend

because of the low job finding rate, which induces a remarkably slow re-

covery in the labor market.

Robustness

Results robust to

- omitted variables/omitted shocks.

- VAR lag length.

- Alternative treatment of low frequency components.

- Medium, sign or long-run identifying restrictions.

- Choice of Price deflators for q and GDP.

- Alternative data sets (Elsby, et. al (2007)).

Conclusions

- Neutral/investment specific shocks produce different labor market re-

sponses. VAR with just one technology shock is not enough.

- Distinction extensive vs. intensive margin important to understand prop-

agation of technology shocks.

- Unemployment rate moved both by separation and finding rate - roughly

as conventional wisdom would suggest. On impact separation rate crucial.

- Neutral shocks important sources of cyclical fluctuations in unemploy-

ment. They explain the recession of the late 80’s and the subsequent

(jobless) recovery of the early 90’s.