Adjusting to Trade Liberalization: Reallocation and Labor Market Policies

Transcription

1 Adjusting to Trade Liberalization: Reallocation and Labor Market Policies A. Kerem Cosar Department of Economics The Pennsylvania State University New Faces in International Economics The Pennsylvania State University May 10, 2010

2 Brazilian Trade Liberalization Trade Policy and Openness In Brazil Average Tariff Rate (left axis) Imports/Output (right axis) Export/Output (right axis)

3 Brazilian Trade Liberalization Industry Level Tariff Rates Before and After Trade Reforms Tariff Level in Tariff Level in 1986

4 Brazilian Trade Liberalization Manufacturing Workforce Composition across 3-digit Industries Employment Share in Employment Share in Employment Share in Employment Share in Employment Share in Employment Share in Employment Share in Employment Share in 1990

5 Motivation Evidence: 1 Labor market effect of trade liberalization in the short-run: little inter-sectoral labor reallocation, 2 Sectoral switches are associated with longer than average unemployment spells and earning losses, 3 Higher adjustment burden for older workers. Policy relevance: long-run gains vs. short-run costs Need for dynamic structural models to analyze the transition

6 This paper Model Two-sector small open economy Random search and matching Sector-specific learning-by-doing Overlapping generations Calibration

7 This paper Transition Determinants of sluggish adjustment: search frictions and sector-specificity of human capital Policy experiments: unemployment insurance and employment subsidies

8 Related Literature Sector-specificity of physical capital Mayer (1974), Neary (1978), Rebelo and Albuquerque (2000). Adjustment costs for labor reallocation Leamer (1980), Mussa (1986), Artuc, Chaudhuri and McLaren (2007, 2008). Worker displacements and policies Ljungqvist and Sargent (1998), Rogerson and Schindler (2002). Sector-specificity of human capital Matsuyama (1992), Rogerson (2005). Search models Davidson and Matusz (2004, 2006), Helpman and Itskhoki (2008), Kambourov (2009)

9 THE MODEL

10 Agents and Preferences A measure one of workers and firms Linear momentary utility in consumption with discount factor β (0, 1) Workers own a balanced portfolio of firms

11 Demographics Period t Period t + 1 δ a Old Young 1 δ a Young δ m (newborn) Young Old 1 δ m Old Life-cycle Shocks

12 Production Two intermediate goods with world prices (p 1, p 2 ) = (1, 1) Sector 2 is protected with ad-valorem import tariff τ, p 2d = (1 + τ), p 1d = 1 Final good production Y = Q 1 Q2 Price of the final good p Y = 2 p 1d p2d Intermediate goods produced by worker-firm pairs (jobs)

13 Production Period t Period t + 1 match in sector 1 φ w1t Draw z Accept, produce q1(z, ht) = A1zh1t Reject Sector 1 with (z, h t+1) Unemployed h t+1 Unemployed h t match in sector 2 φ w2t Draw z Accept, produce q2(z, ht) = zh2t Reject Sector 2 with (z, h t+1) Unemployed h t+1 no match Unemployed h t+1 1 φw1t φw2t Sector i with (z, h t) exogenous separation δ jd endogenous separation Unemployed h t+1 no separation Sector i with (z, h t+1) Timing of Matching and Production

14 Human Capital Newborn workers start with h = (1, 1) The law of motion for h (learning-by-doing): { h α it h it+1 = H1 α if employed in sector i, max{1, (1 δ h )h it } otherwise, where α [0, 1] and δ h [0, 1).

15 Human Capital Formation Human Capital, hi H 1 Periods Employed in Sector i

16 Human Capital Newborn workers start with h = (1, 1) The law of motion for h (learning-by-doing): { h α it h it+1 = H1 α if employed in sector i, max{1, (1 δ h )h it } otherwise, where α [0, 1] and δ h [0, 1). State space for h t : h t H = [1, H] [1, H]

17 Firms Beginning of Period End of Period Draw (c 1, c 2 ) Pay c 1 and post vacancy in sector 1 Pay c 2 and post vacancy in sector 2 Match, draw z No match Accept Reject Post no vacancy Timing of Events for Idle Firms

18 Labor Markets Undirected search with aggregate matching function m(u, V ) = UV (U λ + V λ ) 1/λ Composition of vacancies such that μ i = V i V V = V 1 + V 2

19 Labor Markets matching probability for an unemployed worker: φ wi = μ i m(u, V ) U matching probability for an idle firm: φ f = m(u, V ) V Period-by-period Nash bargaining over rents with worker s share σ (0, 1)

20 State Space Workers state: generation: young or old, {y, o} human capital: (h 1t, h 2t ) labor market status: [l 1 (z), l 2 (z), l u ]

21 Value of a Job Sector-i job with an old worker: Π it (z, h t, o) = p id q i (z, h t ) + β(1 δjd)(1 o δ m )Π it+1 (z, h t+1, o) with a young worker: Π it (z, h t, y) = p id q i (z, h t ) + β(1 δ y JD [δ ) a Π it+1 (z, h t+1, o) ] + (1 δ a )Π it+1 (z, h t+1, y)

22 Job Acceptance Policy Outside options: W ut (h, y) for a young worker and J t for the firm Surplus of a sector-i job of productivity z Δ it (z, h t, y) = Π it (z, h t, y) [W ut (h t, y) + J t ] Worker s problem W it [z, h t, y] = max accept,reject { } σδ it (z, h t, y) + W ut (h t, y), W ut (h t, y). jointly accept the job if z z it (h t, y)

23 Firms: Vacancy Posting Policy Non-specialization through heterogeneous vacancy posting cost draws Taking the expected value conditional on matching as given, enter sector 1 if φ ft β [ EJ 1t+1 J t+1 ] py t c 1, φ ft βej 1t+1 p Y t c 1 φ f βej 2t+1 p Y t c 2. which defines a cutoff cost level c 1t = φ ftβ(ej 1t+1 J t+1 ) p Y t

24 Firms: Vacancy Posting Policy c 2 SECTOR 1 NO ENTRY c2 SECTOR 2 c1 c 1 Figure: Sectoral Entry Decision The cutoff rules define (μ 1, μ 2 ), the fraction of vacancies in each sector.

25 Equilibrium Given world prices and trade policy τ, an equilibrium is a collection of paths such that agents optimize, aggregates are consistent with individual behavior, markets clear, trade balance holds, the distribution of workers evolves consistently with the decision rules and idiosyncratic shocks.

26 Steady State Properties of the Model Excess job turnover with simultaneous job creation and destruction Davis, Haltiwanger and Schuh (1998) Wages are increasing in job tenure and labor market experience Murphy and Welch (1990), Topel (1991) Inter-sectoral mobility declines over the life-cycle Kambourov and Manovskii (2008)

27 CALIBRATION

28 Calibration Strategy Model period: quarter Calibrate the steady state to Brazilian pre-liberalization data Parameters Set Without Solving the Model Parameters Obtained by Solving the Model

29 Parameters Set Without Solving the Model Parameter Definition Value Source/Target δ m death probability 1/80 20 years of youth δ a aging probability 1/80 20 years of old-age F z(z) productivity uniform [0, 1] - σ worker s bargaining share A 2 sector 2 productivity 1 normalization τ import tariff 0.63 Pavcnik et al.(2004) β discounting rate 0.97 real interest rate, IPEA δ y JD job destruction for young Bosch and Maloney(2007) δjd o job destruction for old Bosch and Maloney(2007) δ h depreciation of HC 0 Browning et al.(1999)

30 Parameters Obtained by Solving the Model vacancy cost function F c (c): assume log-c is normally distributed with mean 0, and standard deviation C sd Parameter Value Target Source α H 2.6 ave. earnings at 5 years of experience ave. beginning of the career earnings = 1.41 Menezes Filho et al.(2008) ave. earnings at 40 years of experience ave. beginning of the career earnings = 2.43 Menezes Filho et al.(2008) A Export / (Value Added) = Pavcnik et al. (2004),OECD λ 2.16 elasticity of hiring to unemployment = 0.25 Hoek(2007) C sd 1.49 transition probability from U to E = 0.37 Domeland et al.(2006)

31 THE TRANSITION

32 The Transition Trade Liberalization: Unexpected, permanent decrease in τ. Transitional dynamics Algorithm Quantitative exploration of sluggish adjustment Search frictions vs. sector-specificity of human capital Labor Market Policies: Actual policy: unemployment insurance Counter-factual policy: targeted employment subsidy

33 Sources of Sluggish Adjustment 100 % of Reallocation Completed search only search + human capital Years After Trade Liberalization Figure: Employment Share Reallocation During the Transition

34 Brazilian Labor Market Policies 1988 Labor Market Reform in Brazil. Public expenditure on labor market policies 1% of GDP, 70% on unemployment insurance, 15% on wage supplement program, remaining on active labor market policies. Actual policy experiment: introduce unemployment insurance jointly with trade liberalization, financed by a 1% tax on all match revenues

35 Actual Policy: Short Run 1.8 Unemployment Relative Initial Value model data Figure: Unemployment During the Transition

36 Actual Policy: Long Run Net Output unemployment insurance Years Figure: Net Output under Actual Policy

37 Actual Policy: Long Run Net Output no labor market policy unemployment insurance Years Figure: Comparison of Net Output Paths

38 Counter-factual Policy: Employment Subsidy 2002 Alternative Trade Adjustment Assistance, and 2009 Reemployment Trade Adjustment Assistance in the US Insure 50% of earnings losses for trade-displaced workers older than 50 up to $10, 000 over 2 years Counter-factual policy experiment: a targeted limited-duration employment subsidy paid to old workers initially employed in sector 2 upon employment in sector 1. Revenue neutral policy.

39 Counter-factual Policy: Employment Subsidy Net Output targeted employment subsidy no labor market policy unemployment insurance Years Figure: Net Output under Labor Market Policies

40 Role for Policy: Externalities Search externality between firms Human capital externality between workers and firms: sub-optimally low investment in skill acquisition in frictional labor markets

41 Conclusion policy implications for the design of trade adjustment programs: 2002 US Alternative TAA Future work: - Analysis of welfare effects on different groups workers - General versus sector-specific human capital

42 THE END

43

44 Fact 1: Inter-sectoral Reallocation Micro-evidence from Menezes-Filho and Muendler (2007):...tariff cuts and additional imports trigger worker displacements, but neither comparative-advantage sectors nor exporters absorb trade-displaced workers for years. What happens to workers displaced from import-competing industries at a four-year horizon? 30% rehired into the same sectors, 30% employed at formal non-traded sectors, 26% not employed at any formal job, only 14% is hired into export-oriented industries.

45 Fact 1: Inter-sectoral Reallocation Rayner and Lattimore (1990) for New Zealand Ros (1994) for Mexico Currie and Harrison (1997) for Morocco Levinsohn (1999) for Chile IDB (2009) for a group of Latin American countries Wacziarg and Wallack (2004) for a panel of liberalizers

46 Fact 2: Costly Mobility for Displaced Workers Menezes-Filho (2004), Hoek (2007) for Brazil Krebs, Krishna and Maloney (2008) for Mexico US data Murphy and Topel (1987) Kletzer (2001) Artuc, Chaudhuri and McLaren (2008) Krishna and Senses (2009)

47 Fact 3: Life-cycle Effects of Adjustment Burden Older workers face a higher risk of not finding reemployment after being displaced from import competing industries. Evidence from Brazil by Menezes-Filho and Muendler (2008): Failed Reallocation Within a Year of Displacement in Brazil Average Young Workers( 10 years of experience)

48 More Evidence Little inter-sectoral reallocation of labor: Levinsohn (1999) for Chile Currie and Harrison (1997) for Morocco Menezes-Filho and Muendler(2007) for Brazil Wacziarg and Wallack(2004) for a panel of liberalizers Effect on displaced workers: Kletzer (2001) and Artuc, Chaudhuri and McLaren (2007) for US Krishna and Senses (2009) for Mexico Effect on old workers: Boeri and Terrell (2002) for Eastern Block countries Maloney (2003) and IDB (2009) for Brazil

49 Worker s Value Function For l w = l wi, the value of a job to a worker of state (z, h, g) equals the sum of her share of the surplus Σ it(z, h, g), and the value of her outside option of unemployment: W t(l wi, z, h t, g) = σσ it(z, h t, g) + W t(l wu, z, h t, g) The value of unemployment (l w = l wu ) for an old worker is: W t(l wu, h t, o) = p ft b [ 2 + β(1 δ m) i=1 φ wi t z ] + (1 φ w1 t φ w2 t)w t+1(l wu, h t+1, o) 0 W t+1(l wi, z, h t+1, o)i zi (z, h t+1, o)f z(z)dz

50 Worker s Value Function For a young worker: W t(l wu, h t, y) = p ft b [ 2 + β(1 δ a) i=1 + βδ a [ 2 i=1 φ wi t φ wi t z 0 z 0 W t+1(l wi, z, h t+1, y)i zi (z, h t+1, y)f z(z)dz ] + (1 φ w1 t φ w2 t)w t+1(l wu, h t+1, y) W t+1(l wi, z, h t+1, o)i zi (z, h t+1, o)f z(z)dz ] + (1 φ w1 t φ w2 t)w t+1(l wu, h t+1, o)

51 Entrepreneur s Value Function J t [ 2 = β i=1 The surplus is: μ it[φ ft EJ it+1 p ftˆc it] + ( μ 1t + μ 2t)(1 φ ft )J t+1 + (1 μ 1t μ 2t)J t+1 ] Σ it(z, h t, g t) = Π it(z, h t, g t) [ W t(l wu, h t, g t) + J t ].

52 Sector 1 Job Acceptance Rule, z 1t (h t, o) Figure: Cutoff Productivity in Sector 1 BACK

53 Computation of Steady State Equilibrium Step 1. Step 2. Step 3. Step 4. Step 5. Start iteration j with a pair of values for entrants expected values of matching (EJ j 1, EJj 2 ) in the two sectors. Calculate (J(m u), φ f, φ w, μ 1, μ 2 ) by simulating a large number of cost draws for firms from the distribution F c(c), using relevant expressions, and the fact that market tightness θ is equal to μ 1 + μ 2. Solve for the job acceptance cutoffs z it (h, g), and the value functions using the following subroutine: i. Start with old workers. Assume initial set of values for unemployment W (l u,, o) and matches Π i (z,, o) for both sectors. Find the job acceptance cutoffs, and update Π i (z,, o). ii. To update W (l u,, o), use the job acceptance cutoffs. Iterate until convergence. iii. Repeat the same steps for young workers. Simulate the economy with a large number of workers drawing demographic shocks, labor market shocks (matching and separating), and match-specific productivity terms. Aggregate the cross-sections of workers to find the distribution of workers Ψ. Use the distributions to update (EJ j+1 1, EJ j+1 2 ), iterate until the distances EJ j+1 1 EJ j 1, and EJ j+1 2 EJ j 2 are sufficiently small. BACK

54 Steady State Results Moment Model Data Source Share of youth unemployment IDB, ILO Annual excess job reallocation Haltiwanger et al.(2004) Earning losses of old switchers Hoek(2007) 90th/10th wage percentile Sotomayor(2004)

55 Computation of the Transition Solve for the transition using an algorithm similar to Costantini and Melitz (2009) 1 start with {EJ 1t, EJ 2t } t=t t=0 2 solve value functions backwards 3 simulate the model forward to generate distributions {Ψ t } t=t t=0 4 update {EJ 1t, EJ 2t } t=t t=0, iterate until convergence BACK