Does Extending Benefits Improve Job Quality?

Arash Nekoei Andrea Weber IIES Stockholm CEU − Folk wisdom: more generous UI → higher job quality

− Mixed Empirical Findings

Card, Chetty, Weber 2007; Lalive 2007; Schmieder, Von Wachter, Bender 2013

Question

Unemployment (UI) increases unemployment duration

Does UI also affect job quality? Question

Unemployment insurance (UI) increases unemployment duration

Does UI also affect job quality?

− Folk wisdom: more generous UI → higher job quality

− Mixed Empirical Findings

Card, Chetty, Weber 2007; Lalive 2007; Schmieder, Von Wachter, Bender 2013 Overview

I Postitive UI effect on post- Age-based discontinuity in UI rules in Austria 9 extra weeks of UI cause 0.5% wage increase Overview

I Postitive UI effect on post-employment wage Age-based discontinuity in UI rules in Austria 9 extra weeks of UI cause 0.5% wage increase II Directed-search model Two off-setting forces: Subsidized search causes unemployed to look for better job, but longer unemployment deteriorates job opportunities Reconcile our result with previous estimates Negative correlation bw wage and duration effects of UI Overview

I Postitive UI effect on post-employment wage Age-based discontinuity in UI rules in Austria 9 extra weeks of UI cause 0.5% wage increase II Directed-search model Two off-setting forces: Subsidized search causes unemployed to look for better job, but longer unemployment deteriorates job opportunities Reconcile our result with previous estimates Negative correlation bw wage and duration effects of UI III Mechanisms underlying the positive UI wage effect IV implications: UI wage effect as fiscal externality Data/Institutional background

Austrian Unemployment and Social Security Registries 1980-2011 Daily records of employment status and UI receipt

Austrian UI system Benefit level: Replacement rate 55%, Benefit duration:

Experience over UI duration Age the last n years in weeks 1 over 2 - 20 3 over 5 - 30 6 over 10 >=40 39 Empirical Research Design

Identification: Discontinuity at age 40 Laid-off workers eligible for 30 or 39 weeks of UI as age ≶ 40 Sample: 1,738,789 job separations Presentation: 10-year Bandwidth and polynomial degree = 2 Robustness checks Optimal bandwidth and bias-corrected RD estimates

Imbens and Kalyanaraman (2011) and Calonico et al. (2014)

Experience over UI duration Age the last n years in weeks 1 over 2 - 20 3 over 5 - 30 6 over 10 >=40 39 Density .04 .045 .05 .055 30 35 RDD ValidityTestI ofAge Age at layoff 40 45 50 Log previous wage 7.48 7.49 7.5 7.51 7.52 7.53 30 Covariates AroundUIExtensionThreshold 35 RDD ValidityTestII Age at layoff 40 45 50 Log previous wage −.055 −.05 −.045 −.04 −.035 30 Predicted Wagechangebetweenjobs 35 RDD ValidityTestIII Age at layoff 40 45 50 Non−employment duration (days) 105 110 115 120 125 30 UI ExtensionEffectonNon−Employment Duration 35 Age at layoff 40 45 50 Prob. of finding job within 39 weeks .75 .8 .85 .9 30 UI ExtensionEffectonNon−Employment Duration 35 Age at layoff 40 45 50 Wage change between jobs −.06 −.055 −.05 −.045 −.04 30 UI ExtensionEffectonWage 35 Age at layoff 40 45 50 Prob(New wage > benefit level) .956 .958 .96 .962 .964 .966 30 UI ExtensionEffectonWage 35 Age at layoff 40 45 50 Effect of UI Extension from 30 to 39 Weeks Discontinuity at age 40 Dependent variable Non- Find job Wage employment within 39 change New wage > duration weeks between jobs UI benefit Covariates (1) (2) (3) (4)

No 1.932*** -0.0131*** 0.00449*** 0.00388*** (0.526) (0.00164) (0.00170) (0.00105)

Yes 1.918*** -0.0119*** 0.00455*** 0.00385*** (0.466) (0.00146) (0.00146) (0.00102)

Mean dep. var. 114.7 0.842 -0.0440 0.962 Observations 1,589,178 1,738,787 1,187,476 1,187,476 Wage change between jobs −.025 −.02 −.015 −.01 −.005 0 30 35 Pre−Reform Discontinuity Placebo Test Age at layoff 40 45 50 Benchmark: Risk-neutral agent, no leisure/search

2 (1 − τ) w e = 2b + L (1 − τ) ∆w e | {z } | {z } MC MB

∆w e 2 ⇒ = (1 − α) = 0.14% w e L

where

L : Duration of new job = 567 days b α : replacement rate = = 55% (1 − τ) w e

Is the Order of Magnitude Reasonable?

Estimated effect of UI extension: 0.5% wage increase and 2 extra days of search Is the Order of Magnitude Reasonable?

Estimated effect of UI extension: 0.5% wage increase and 2 extra days of search Benchmark: Risk-neutral agent, no leisure/search

2 (1 − τ) w e = 2b + L (1 − τ) ∆w e | {z } | {z } MC MB

∆w e 2 ⇒ = (1 − α) = 0.14% w e L

where

L : Duration of new job = 567 days b α : Net replacement rate = = 55% (1 − τ) w e Previous Literature

2.0%

1.5% Lalive 2007 of UI ExtensionUIof 1.0% NW Card, et al 2007

0.5% Wage EffectWage UI Duration (weeks) 0.0% 20 30 39 52 78 104

-0.5%

-1.0% Schmieder, et al 2013

Schmieder, et al 2013 -1.5%

-2.0%

Without UI

wage

Target Target

w τ

Time Layoff

UI Effect

wage

Target Target

Δw τ w τ

Time Layoff Benefit Exhaustion

UI Effect

wage

Target Target

In admin data, we observe target wage only when search is successful (The starting wage of the new job)

We can identify UI effect on average wage, we Δw τ w τ

Time Layoff Benefit Exhaustion UI Effect

e τ

wage Δw = λ .Δw ≥ 0

Wage Effect Selectiveness

Target Target

e w : Average wage λ : Probability of finding a job before benefit exhaustion

Δw τ w τ

Time Layoff Benefit Exhaustion

With Negative Duration Dependence

Target wage Target

w τ

Time Layoff

With Negative Duration Dependence

Duration Dependence implies w τ (0) ≥ we

Target wage Target

w τ

Time Layoff

UI Effect With Duration Dependence

Target wage Target

Δw τ

Time Layoff Benefit Exhaustion UI Effect With Duration Dependence

Δwe = λ × Δwτ + Δλ × (wτ -we)

Target wage Target

Δw τ

Time Layoff Benefit Exhaustion UI wage effect can be negative Relative role of search and selectiveness in finding a job

UI Extension Effect on Post-unemployment Wage

(+) (−) z }| { z }| { ∆w e = λ × ∆w τ + ∆λ × (w τ − w e ) |{z} |{z} |{z} | {z } Wage Selectiveness Duration Duration Effect Effect Dependence

λ : Hazard rate at time zero (λ = F (w τ, s)) w τ : Target wage at time zero w e : Expected wage (w e = E (w τ)) Model Predictions I

Negative force arises from duration dependence Structural duration dependence Changes in UI benefits over time UI wage effect can be positive or negative Explains variation in empirical estimates Model Predictions II

(+) (−) z }| { z }| { ∆w e = λ ∆w τ + ∆λ (w τ − w e ) | {z } | {z } |{z} Wage Effect Selectiveness Duration Eff.

Relative role of search effort and selectiveness in adjustment of search behavior Strong search effort response increases the duration effect thus strengthens negative force Population with similar job finding rate, but heterogeneity in search effort response Model predicts negative correlation of individual duration and wage effects Negative correlation bw UI duration and UI wage effect

Strategies for empirical tests

Correlation of UI wage and UI duration effects across studies: meta-analysis Correlation within out sample: estimate effects across subsamples Wage vs. Non−enmployment Duration Effects Results across studies .02 .01

AW 0 Lalive(a) SWB(b) CCW SWB(a) −.01 UI effect on wage

Lalive(b) −.02 −.03 −10 0 10 20 30 UI effect on non−enmployment duration

How does UI affect ?

Attenuation of wage drops versus wage increases Jobs in better firms or higher wages for the same job? Effect of UI on other job characteristics UI effect on P(=x) −.01 −.005 0 .005 .01 UI BenefitExtensionAttenuatesWageDrops −.8 PDF (rightaxis) UI EffectonPDF −.6 −.4 −.2 0 x .2 .4 .6 .8

0 .1 .2 .3 .4 P(wage growth=x) Effects of Benefit Extension on Firm Sorting

Firm-level Individual Outcomes level

Change in Firm Size Share Male Av. wage log wage Res. wage Wage (coeff*100)

Covariates No 11.04 0.167 0.324 0.482 0.704 0.680 (14.870) (0.228) (0.125) (0.272) (0.316) (0.313)

Yes 11.36 0.115 0.269 0.491 0.751 0.494 (14.650) (0.194) (0.107) (0.237) (0.307) (0.266)

Dep. Mean 382.6 -2.54 -0.387 -6.21 -0.988 -10.5 Observations 454,990 454,547 454,971 454,401 429,504 456,114 Effects of Benefit Extension on Mobility

Probability of changing Firm Occupation Region Zip-code (coeff*100)

Covariates No -0.318 -0.16 -0.0714 0.189 0.0518 (0.230) (0.196) (0.113) (0.149) (0.225)

Yes -0.227 -0.139 -0.0558 0.166 0.126 (0.154) (0.167) (0.108) (0.140) (0.163)

Dep. Mean 41.9 22.9 6.15 10.9 36 Observations 1,589,178 1,589,174 1,589,177 1,521,919 1,566,755 Effects of Benefit Extension on Job Characteristics

Tenure Separation Wage Pos wage Full-time days in one year growth growth job (coeff*100)

Covariates No 1.202 0.0923 0.000103 -0.126 -0.042 (4.034) (0.208) (0.035) (0.240) (0.237)

Yes 0.289 0.0364 -0.0189 -0.203 -0.184 (3.802) (0.189) (0.034) (0.238) (0.200)

Dep. Mean 567.3 72.4 3.62 72.5 84.5 Observations 1,589,178 1,589,178 1,192,343 1,193,243 755,594 ∆Gov. Budget = L τ∆w − (τw ∆n + b ∆n˜) | {z } | {z } Wage Externality Externality

= (1.4 − 3.2) w

where

L : Duration of the new job n : Duration of non-employment n˜ : Part of n covered by UI ( = min (n, B) )

Policy Implications

Two Fiscal Externalities: Traditional negative externality: Non-employment duration effect Positive externality: Wage effect Policy Implications

Two Fiscal Externalities: Traditional negative externality: Non-employment duration effect Positive externality: Wage effect

∆Gov. Budget = L τ∆w − (τw ∆n + b ∆n˜) | {z } | {z } Wage Moral hazard Externality Externality

= (1.4 − 3.2) w

where

L : Duration of the new job n : Duration of non-employment n˜ : Part of n covered by UI ( = min (n, B) ) Conclusion

Estimated a positive effect of UI duration on wages UI extension attenuates wage drops, no effect on wage raise Wage effect last throughout the first job

Theoretically, average effect of UI on average job quality is determined by Relative role of search and selectiveness in finding a job Decline in job opportunity

Normative Positive wage externality comparable to moral hazard externality