The Employment and Output Effects of Short-Time Work in Germany Russell Cooper Moritz Meyer 2 Immo Schott 3 Penn State 2 The World Bank 3 Université de Montréal Social Statistics and Population Dynamics Seminar McGill March 8th, 207
Motivation In Germany the 2008 recession led to: Large negative effect on GDP & total hours worked Small effect on unemployment Stark contrast with other OECD economies German Labor Market Miracle Possible explanation: Short-Time Work (STW) Our question: Can STW save jobs? And if yes, at what cost?
GDP Growth (year-to-year) GDP growth 0 5 0 5 995q 2000q 2005q 200q 205q Time DEU OECD ESP USA AUT FRA Micro Data Hours Change
Unemployment Rate 2005q = 00 50 00 50 200 250 995q 2000q 2005q 200q 205q Time DEU OECD ESP USA AUT FRA Micro Data
What is Short-Time Work (STW)? Labor market policy instrument Goal: Mitigating cyclical shocks Change labor demand via intensive margin (hours vs. workers) UI compensates workers for lost income (60-67%) Absent STW, unilateral reductions in hours worked are illegal Use of STW is subject to strict set of legal requirements Details The STW policy : 2009-200 Gov t dramatically reduced eligibility criteria & burden of proof Maximum duration increased from six to 8, and then 24 months June 2009: Around 60 000 establishments and 500 000 workers Graph
Summary of Results Can STW save jobs? Economic press, Government, Unions We find a positive effect on employment What are the costs? Reduced form vs. structural model Reallocation channel STW prevents reallocation of labor negative effect on GDP
Literature Work Sharing: Burdett & Wright (989), Hunt (998, 999), Marimon & Zilibotti (2000), Kudoh & Sasaki (20) German Labor Market: Krause & Uhlig (20), Burda & Hunt (20), Cahuc & Carcillo (20), Balleer et al. (206) Factor allocation: Hsieh & Klenow (2007), Bartelsman et. al (203) Multi-worker firms: Cooper, Haltiwanger, & Willis (2007), Elsby & Michaels (203), Stole & Zwiebel (996)
Data Afid-Panel Indusriebetriebe from German Statistical Office Universe of manufacturing plants, annual panel 995-200 Up to 68 000 observations, of which we use 39 000 Variables: Revenue, Employment, Hours Worked,... Sumstats Advantages June 2009: 80.4% (4%) of workers (firms) using STW were located in manufacturing Heavy concentrating of employment in Mittelstand No sampling bias Disadvantages No direct information on STW
Changes in Total Hours: Extensive and Intensive Margins 996 997 998 999 2000 200 2002 2003 2004 2005 2006 2007 2008 2009 200..05 0.05 Employment Hours per Employee
Distribution of changes in annual hours per worker: 995-2008 Frequency 0 5 0 5 x>.3.2>x>.3.>x>.2.05<x<..025<x<.05.0<x<.025 inactive.025<x<.0.05<x<.025.<x<.05.2<x<.%.3<x<.2 <.3
Distribution of changes in annual hours per worker: 995-2009 Frequency 0 5 0 5 20 x>.3.2>x>.3.>x>.2.05<x<..025<x<.05.0<x<.025 inactive.025<x<.0.05<x<.025.<x<.05.2<x<.%.3<x<.2 <.3 995 2008 2009
Distribution of changes in annual hours per worker: 995-200 Frequency 0 5 0 5 20 x>.3.2>x>.3.>x>.2.05<x<..025<x<.05.0<x<.025 inactive.025<x<.0.05<x<.025.<x<.05.2<x<.%.3<x<.2 <.3 995 2008 2009 200
Model - Overview Basic Model Hours Contraints & STW Aggregate Shocks Quantitative Results: Counterfactuals
Model - Ingredients Workers and multi-worker Firms Firms face idiosyncratic productivity shocks ε Decreasing returns to scale in production Total labor input L = h n Frictional labor market produces rents Nash-Bargaining Matching Function M = m(u, V ), Labor Market Tightness θ = V U Vacancy-filling probability q = M V Distribution of firms over (ε, n)
Model - Timing Firm enters period with n workers and productivity ε Choose n workers and average hours h Negotiate wage with n workers Produce output
Model - Firm s Problem { V (ε, n ) = max εf (h n) ω(h, n, ε) h n h,n } c v q (n n ) + + β V (ε, n)dg(ε ε), ω( ) is a wage schedule c v is a linear vacancy creation cost + is an indicator for when a firm is hiring
Model - Firm s Problem FOC Hours εf L (h n) ω(h, n, ε) ω h (h, n, ε) h = 0
Model - Firm s Problem FOC Hours εf L (h n) ω(h, n, ε) ω h (h, n, ε) h = 0 FOC Employment (if n 0) εhf L (h n) ω(h, n, ε) h ω n (h, n, ε) nh c v q + +βd(ε, n) = 0, where D(ε, n) V n (ε, n)dg(ε ε)
Model - Worker s Problem W e [ (ε, n) = ω(h, ε, n) h ξ(h)+βe ε ε sw u + ( s)w e (ε, n ) ]. W u [ = b + βe (ε,n ) ( φ)w u + φw e (ε, n ) ].
Model - Wages Workers and Firm share surplus of match Decreasing return to scale surplus changes for each worker Nash bargaining over marginal surplus (Stole & Zwiebel (996)) Firm s marginal surplus for matching with a worker: S(ε, n) = εhf L (h n) ω(h, n, ε)h ω n (h, n, ε)hn + βd(ε, n) Surplus is shared according to W e (ε, n) W u = η S(ε, n). η
Model - Wages Wage solves differential equation ω(h, ε, n) h = ( η) [b + ξ(h)] + [ η εhf L (h n) + φ c ] v q ω n(h, n, ε) h n Assume F (L) = L α = n α h α [ εαh α n α ] ω(h, ε, n) h = ( η) [b + ξ(h)] + η η( α) + φc v q Negotiated at t = 0
Model - Optimal Labor Demand Graph Combine wage with FOCs to get H(ε, n) and N (ε, n ). The optimal hours choice: [ H(ε, n) = εαn α ξ (h) ( η( α)) ] α The optimal employment choice: ψv (ε) if ε > ψ v (n ), N (ε, n ) = n if ε [ψ(n ), ψ v (n )], ψ (ε) if ε < ψ(n ),
Hours Constraint and STW Standard hours = h. Firm cannot set h < h Policy parameter for STW: Ξ Ξ [0, h] Constraint changes to h Ξ The optimal hours policy function becomes { [ H(ε, n) = max h Ξ, STW use has to be approved by gov t εαn α ξ (h) ( η( α)) ] α }.
Model - Calibration (Ξ = 0) Parameter Meaning Value Reason Calibrated β Discount factor.9967 Annual r = 4% γ Matching elasticity.6 Petrongolo & Pissarides (200) µ Matching efficiency.622 θ = 0.09 α F (L) = L α.65 Cooper et al. (2007) ε Mean of ε Normalization b Unemployment benefit.024 Average employment = 98.5 ξ 0 Disutility of work (scale).24 Average hours = η Worker bargaining power.43 Labor share 0.76 Table: Model Parameters.
Model - Estimation (Ξ = 0) Moment Data Model L N = δ L φ+δ.09.09 h < 5% (annual).538.542 n < 5% (annual).476.440 cv(n)/cv(h) 5.63 5.66 Distance L(Θ) - 0.00382 Table: Estimated Parameters
Steady state results - no policy Match inactivity regions of Hours and Employment changes Match the relative variability of hours and employment Value of leisure = 3.24% of average wages Firms spend on average.07% of monthly wage bill on recruiting costs Labor costs of posting vacancies are 32.66% of the average monthly worker wage Labor market tightness θ = V U = 0.09 Monthly job-finding rate of 6.22% US 30% (Hall (2006))
Steady state results - Hourly wage 0.7 0.65 Wage 0.6 0.55 0.5 0.45 80 00 Employment 20.5.4.3 Hours.2. Wage is decreasing in n and h Effect via marginal product of labor & disutility More productive firms are large Positive relationship between size and wages
Steady state results - The Hours Constraint h = Empirical CDF of Hours 0.9 0.8 0.7 0.6 F(x) 0.5 0.4 0.3 0.2 0. no STW STW 0 0.7 0.8 0.9..2.3.4.5 Average Hours Constraint can be binding in steady state h prevents hours reductions, firms use extensive margin
Aggregate Shocks Π = A high A low A Ξ A high ρ ρ 0 A low ρ ρ 0 A Ξ ρ ρ π π Average duration of STW is six months: π Solve similarly to Krusell & Smith (998) Firms need to forecast q which depends on the cross-sectional distribution
Effect of STW Simulation of economy Let STW policy become active in period t = 200 no negative productivity shocks
IRF - Effect of STW 2 Aggregate Productivity.0 Output.5 0.99 0.5 0 50 00 50 200 250 300 350 0.98 50 00 50 200 250 300 350.06 Employment.0 Total Hours Worked.04.02 0.99 0.98 0.96 0.98 0.97 50 00 50 200 250 300 350 50 00 50 200 250 300 350 Average Hours Vacancy-filling probability 0.68 0.98 0.66 0.96 0.64 50 00 50 200 250 300 350 0.62 50 00 50 200 250 300 350.05 Hourly Wages 0.5 Fraction of Firms using STW.0 0.4.005 0.3 0.2 0.995 0. 0.99 50 00 50 200 250 300 350 0 50 00 50 200 250 300 350
Effect of STW Simulation of economy Let STW policy become active in period t = 200 no negative productivity shocks Partial Equilibrium: Keep q fixed
IRF - Effect of STW - PE 2 Aggregate Productivity.0 Output.5.005 0.5 0 GE PE 50 00 50 200 250 300 350 0.995 0.99 0.985 0.98 50 00 50 200 250 300 350. Employment.0 Total Hours Worked.05 0.99 0.98 0.97 0.95 50 00 50 200 250 300 350 50 00 50 200 250 300 350 Average Hours 0.69 0.68 Vacancy-filling probability 0.98 0.67 0.66 0.96 0.65 0.64 0.94 0.63 50 00 50 200 250 300 350 50 00 50 200 250 300 350 Hourly Wages 0.6 Fraction of Firms using STW.0 0.5.005 0.4 0.3 0.2 0.995 0. 0.99 50 00 50 200 250 300 350 0 50 00 50 200 250 300 350
Effect of STW STW increases employment but has a negative effect on output. Key: endogeneity of q Positive employment response more than twice as large in PE Output falls by almost % Heterogeneous effect on firms
IRF - Recession without STW.0 Aggregate Productivity Output 0.99 0.99 0.98. 5 0 5 20 25 Unemployment Rate 0.97 5 0 5 20 25 Total Labor Input L.05 0.99 5 0 5 20 25 Average Hours 0.98.03 5 0 5 20 25 q 0.995.02 0.99.0 0.985 0.999 0.998 5 0 5 20 25 Hourly Wages 5 0 5 20 25 Fraction of Firms using STW no STW 0.5 0.997 5 0 5 20 25 0 5 0 5 20 25
IRF - Recession with STW.0 Aggregate Productivity Output 0.99 0.98. 5 0 5 20 25 Unemployment Rate 0.96 5 0 5 20 25 Total Labor Input L.05 0.98 0.96 5 0 5 20 25 Average Hours 0.94.03 5 0 5 20 25 q 0.98.02 0.96.0 0.94.005 5 0 5 20 25 Hourly Wages 5 0 5 20 25 Fraction of Firms using STW 0.6 no STW 0.4 STW 0.2 0.995 5 0 5 20 25 0 5 0 5 20 25
Productivity Effects.04.02 Correlation of Employment and Productivity no STW STW.0.008.006.004.002 0.998 0.996 5 0 5 20 25 30
Employment Effects for firms with ε < 0-7.8 Average Employment Change -8-8.2-8.4-8.6-8.8-9 -9.2-9.4 5 0 5 20 25.02 Average Hours 0.98 0.96 0.94 0.92 0.9 0.88 5 0 5 20 25 no STW STW
Job Creation and Job Destruction 0.8 0.6 Job Destruction no STW STW 0.4 0.2 0-0.2 5 0 5 20 25 0.06 0.05 0.04 0.03 0.02 0.0 0-0.0-0.02 Job Creation 5 0 5 20 25
Robustness Role of parameters (see paper) Role of labor market institutions Flexibility, h < Alternative: Hiring Credits cheaper, but less effective Large initial effect on U via JD
Model Predictions Germany 2009: labor productivity per worker -4.9% labor productivity per hour -2.2% Less job creation in sectors with more STW Graph in line with model prediction
Conclusion Can STW save jobs? Economic press, Government, Unions We find a positive effect on employment What are the costs? Reduced form vs. structural model Reallocation channel STW prevents reallocation of labor negative effect on GDP of around %
Thank you
Employment Policy 8 Employment Policy Function for levels of productivity 7 6 Log Employment Tomorrow 5 4 3 2 0 0 2 3 4 5 6 7 8 Log Employment Today Figure: Firm s Employment Policy N (ε, n ) as a function of productivity. back
Change in Total Hours Worked Total Hours growth, % 6 4 2 0 2 4 995 2000 2005 200 205 Time DEU OECD ESP USA AUT FRA back
Summary Statistics Count Mean SD IQR p0 p50 p90 N 38,839 98.5 42.6 73.8 9.4 48.2 228.0 H 33,67 56,300 20,576,694 3,578 8,366 35,07 H/N 34,303 35.8 35.7 3.6 04.5 34.0 67.9 PY 39,80,53,785 3,06,538,6,285 0,242 474,343 3,766,944 Table: Summary Statistics Note: Summary statistics for Employment N, Hours H, Hours per Employee H/N, and Revenues PY. The table shows average values over all years. Revenues are deflated to 2005 Euros. back
Rules for STW Hours reduction must not be preventable (overtime, holidays) 2 The firm must be unable to compensate the work stoppage with permissible variations in intra-firm working hours 3 At least a third of the firm s workforce must suffer an earnings loss of at least 0%. 4 Reduction in working time must be temporary. The maximum duration of STW is six months. After this time full-time employment should be restored. back Hours worked will be paid as usual Remanence costs for the firm The gov t will compensate workers for 60% (67%) of earnings loss
STW use by Workers and Firms Firms 0 20 40 60 0 500 000 500 Workers 2008m 200m 202m 204m 206m Time Firms Workers back
Hours Change Distribution 0.6 Data Model 0.5 0.4 0.3 0.2 0. 0 <-.20 -.20--.0 -.0--.05 Inactive.05-.0.0-.20 >.20 back
Employment Change Distribution 0.5 0.45 Data Model 0.4 0.35 0.3 0.25 0.2 0.5 0. 0.05 0 <-.20 -.20--.0 -.0--.05 Inactive.05-.0.0-.20 >.20 back
Job Creation Log Job Creation (2004 = ).98.99.0.02.03 2005 200 205 year Manufacturing Services Figure: Job Creation, in logs, normalized to 2004 values. Source: German Employment Agency. back