Resurrecting the Role of the Product Market Wedge in Recessions

Transcription

1 Resurrecting the Role of the Product Market Wedge in Recessions Mark Bils, University of Rochester and NBER Pete Klenow, Stanford University and NBER Ben Malin, Federal Reserve Bank of Minneapolis 1 Federal Reserve Bank of San Francisco Macroeconomics and Monetary Policy Conference March 4, Views expressed here are those of the authors and do not necessarily reflect the views of the Federal Reserve System.

2 Decomposing the Labor Wedge Hours worked appear to be inefficiently low in recessions. Labor Wedge is high: µ mpn mrs

3 Decomposing the Labor Wedge Hours worked appear to be inefficiently low in recessions. Labor Wedge is high: µ mpn mrs Labor Wedge is the product of: 1 Labor Market Wedge: µ w w/p mrs 2 Product Market Wedge: µ p mpn w/p p mc

4 The Standard Decomposition Approach Uses (aggregate) wage data E.g., Gali, Gertler, Lopez-Salido (2007), Karabarbounis (2014) Measure of Price of Labor: w/p = average wage Key Assumption: all workers employed in spot markets. Conclusion: µ w accounts for nearly all cyclicality of µ.

5 The Standard Decomposition Approach Uses (aggregate) wage data E.g., Gali, Gertler, Lopez-Salido (2007), Karabarbounis (2014) Measure of Price of Labor: w/p = average wage Key Assumption: all workers employed in spot markets. Conclusion: µ w accounts for nearly all cyclicality of µ. BUT, conclusion depends critically on wage measure used. Alternative theories emphasize durable nature of employment and wage smoothing. w/p can be much more procyclical using other wage measures.

6 This Paper Decomposes Labor Wedge µ without using wage data. Recall: µ p p mc

7 This Paper Decomposes Labor Wedge µ without using wage data. Recall: µ p p mc Consider 2 alternative inputs:

8 This Paper Decomposes Labor Wedge µ without using wage data. Recall: µ p p mc Consider 2 alternative inputs: 1 Self-Employed p mc = p p mrs/mpn = mpn mrs,

9 This Paper Decomposes Labor Wedge µ without using wage data. Recall: µ p p mc Consider 2 alternative inputs: 1 Self-Employed p mc = p p mrs/mpn = mpn mrs, or µp = µ

10 This Paper Decomposes Labor Wedge µ without using wage data. Recall: µ p p mc Consider 2 alternative inputs: 1 Self-Employed p mc = p p mrs/mpn = mpn mrs, or µp = µ 2 Intermediate Inputs p mc = p p m/mpm

11 Preview of Findings Our point estimates: µ p accounts for the cyclical variation in µ Self-Employed µ is just as cyclical as all-worker µ Intermediate Inputs µ p is just as cyclical as µ Thus, countercyclical price markups deserve a central place in business cycle research, alongside labor market frictions.

12 Outline for Remainder of Talk Measuring the Labor Wedge Focus on Intensive Margin Decompose using Wage Data

13 Outline for Remainder of Talk Measuring the Labor Wedge Focus on Intensive Margin Decompose using Wage Data Our 2 Alternative Decompositions 1 Self-Employed 2 Intermediate Inputs

14 Intensive-Margin Wedge ln(µ t ) ln(mpn t ) ln(mrs t ) ( ) [ yt 1 = ln n t σ ln(c t) + 1 ] η ln(h t) h t = hours per worker η = 0.5 σ = 0.5

15 Cyclicality of Intensive-Margin Labor Wedge ln(µ t ) = α + β ln(cyc t ) + ɛ t Elasticity wrt GDP Labor Wedge (0.13) Labor Productivity (0.08) Cons per capita 0.61 (0.03) Hours per worker 0.30 (0.07) Quarterly data, with σ = 0.5, η = 0.5

16 Decomposing the Wedge Decomposition: [ ln(µ t ) = ln ( yt n t ) ln = ln(µ p t ) + ln(µw t ) ( wt p t )] [ + ln ( wt p t ) 1 σ ln(c t) 1 ] η ln(h t) Cyclicality: ln(µ t ) = α + β ln(cyc t ) + ɛ t ln(µ p t ) = αp + β p ln(cyc t ) + ɛ t ln(µ w t ) = α w + β w ln(cyc t ) + ɛ t Note: β = β p + β w.

17 Wedge Decomposition: Standard Approach Elasticity wrt GDP µ (0.13) µ p ( w p = AHE ) (0.13)

18 Wedge Decomposition: Alternative Wage Measures Elasticity wrt GDP µ (0.13) µ p ( w p = AHE ) (0.13) µ p ( w p = NH ) µ p ( w p = UC ) (0.16) (0.21)

19 Outline Measuring the Labor Wedge Fous on Intensive Margin Decompose using Wage Data Our 2 Alternative Decompositions 1 Self-Employed 2 Intermediate Inputs

20 Approach 1: Self-Employed Idea: Compare the wedge for the self-employed (µ se ) to the wedge for all workers (µ). Assuming µ se = µ p se = µ p, comparison yields µ p vs. µ. Focus on intensive (hours) margin Extensive movements could reflect costs of starting business

21 Data on Self-Employed Hours and Earnings: March CPS Self-employed Primary job is (nonag) self-employment. 95% of earnings from primary job Trim sample to deal with top and bottom coding Hours: usual weekly hours (also total annual hours) Earnings from primary job Examine year-to-year changes for matched workers Consumption: Consumer Expenditure Survey Construct relative consumption of self-employed

22 Cyclicality of the Labor Wedge: All vs. Self-Employed Labor Wedge Elasticity wrt (1) (2) (3) (4) Real GDP (0.10) Hours MPN Consumption All Agg. y/n NIPA PCE

23 Cyclicality of the Labor Wedge: All vs. Self-Employed Labor Wedge Elasticity wrt (1) (2) (3) (4) Real GDP (0.10) (0.17) Hours All SE MPN Agg. y/n Agg. y/n Consumption NIPA PCE NIPA PCE

24 Cyclicality of the Labor Wedge: All vs. Self-Employed Labor Wedge Elasticity wrt (1) (2) (3) (4) Real GDP (0.10) (0.17) (0.25) Hours All SE SE MPN Agg. y/n Agg. y/n SE earn/hr Consumption NIPA PCE NIPA PCE NIPA PCE

25 Cyclicality of the Labor Wedge: All vs. Self-Employed Labor Wedge Elasticity wrt (1) (2) (3) (4) Real GDP (0.10) (0.17) (0.25) (1.00) Hours All SE SE SE MPN Agg. y/n Agg. y/n SE earn/hr SE earn/hr Consumption NIPA PCE NIPA PCE NIPA PCE NIPA PCE + CE adj.

26 Labor Wedge for Self-Employed vs. All Workers.06 Labor Wedge for Self-employed vs. All Workers All-worker Labor Wedge Self-employed Labor Wedge

27 Self-Employed Conclusions (Baseline) self-employed wedge is at least as countercyclical as all-worker wedge. Robustness: 1 Use only unincorporated self-employed 2 Weight CPS observations by industry 3 Weight CPS observations by share of self-employed in industry-occupation that have employees Conclusion: µ p accounts for the bulk of cyclical variation in µ.

28 Outline Measuring the Labor Wedge Focus on Intensive Margin Decompose using Wage Data Our 2 Alternative Decompositions 1 Self-Employed 2 Intermediate Inputs

29 Approach 2: Intermediate Inputs Production function: y = [ θm ε 1 ε + (1 θ) [ [z v αk ω 1 ω ] + (1 α)(zn n ω 1 ω ω 1 ω ) ] ε 1 ] ε ε 1 ε Marginal Product wrt Intermediates: Product Market Wedge: ( ) 1 yt ε mpm t = θ m t µ p t = p t mc t = p t p mt /mpm t

30 Constructing µ p i Product Market Wedge µ p it = p ( ) 1 it y it yit ε 1 p m,it m it m it BLS Multifactor Productivity Database Annual data, industries (18 manufacturing) Output and KLEMS inputs, nominal and real Baseline: ε = 1 Robustness: ε < 1

31 Cyclicality of Intermediate Share

32 Cyclicality of Intermediates-based µ p ln ( µ p it) = αi + β p ln(cyc t ) + ɛ it Elasticity wrt GDP All Industries (0.24) Manufacturing (0.32) Non-Manufacturing (0.24) Baseline estimates with ε = 1.

33 Cyclicality of Industry-level Labor Wedge (µ i ) ( pi v i ln (µ i ) = ln pn i ) + ln ( yi v i ) [ 1 σ ln (c) + 1 ] η ln (h i) Elasticity wrt GDP All Industries (0.26) Manufacturing (0.39) Non-Manufacturing (0.24) Baseline estimates with ε = 1.

34 Intermediates-based µ p vs. Total Labor Wedge µ

35 Role of µ p in µ, with ε < 1 ε < 1 µ p i ln ( µ p it more countercyclical ( ) ( ) ( ) ) pit y it 1 = ln + p m,it m it ε 1 yit ln m it ε < 1 µ i less countercyclical ( ) ( ) pit y it 1 ln (µ it ) = ln + p t n it ε 1 ln ( yit v it ) ( ) ln mrsit h ε < 1 µ p accounts for > 100% of cyclicality of µ.

36 Conclusion Our point estimates: µ p accounts for the cyclical variation in µ Self-Employed µ is just as cyclical as all-worker µ Intermediate Inputs µ p is just as cyclical as µ Countercyclical price markups deserve a central place in business cycle research, alongside labor market frictions. Sticky prices Customer base and/or learning-by-doing + financial shocks Countercyclical risk or risk-aversion

37

38 Representative-Agent Labor Wedge Preferences: Production: Labor Wedge: { } E 0 β t c 1 1/σ t 1 1/σ ν n1+1/η t 1 + 1/η t=0 y t = z t k α t n 1 α t ln(µ t ) ln(mpn t ) ln(mrs t ) ( ) [ yt 1 = ln σ ln(c t) + 1 ] η ln(n t) n t

39 Extensive and Intensive Margin Labor Wedges Consider extensive and intensive margins of labor supply Why? Can base Frisch elasticity on micro estimates using hours margin Self-employed wedge will be on intensive margin only Product market distortions should impact wedge on both margins If wedge is only important on one margin, product market distortions must have little cyclical importance.

40 Theory with Both Extensive and Intensive Margins Preferences: { ( ) } E 0 β t c 1 1/σ t 1 1/σ ν h 1+1/η t 1 + 1/η + ψ e t Production: t=0 Search Frictions: y t = z t k α t (e t h t ) 1 α Matching Technology: m t = v φ t f (u t ) Vacancy-posting cost: κ Separation rate: δ

41 Extensive Margin Wedge Consider spending today to generate one more matched worker, then reduce spending next period to cut matches by 1 δ workers: EMW ln(y/n) 1/σ ln(c) dynamic cost of vacancy matching So: EMW IMW = 1 ln(h) dynamic cost of vacancy matching η

42 EMW vs. IMW In Logs EMW (based on VAR) IMW

43 Alternative Wage Measures Semi-elasticities wrt the Unemployment Rate (s.e. s): Average Hourly Earnings -1.8 (0.7) New-hire Wage -3.0 (0.8) User Cost of Labor -5.2 (0.8) Source: Kudlyak (2015) using the NLSY

44 Wedge 0.1 MPN Tax adjusted MRS

45 Wedge Decomposition: Avg Wage 0.1 MPN 0.05 Wage (Avg) Tax adjusted MRS

46 Wedge Decomposition: User Cost of Labor 0.1 MPN Wage (UC) 0.05 Tax adjusted MRS

47 Alternative Wage Measures Wage (Avg) 0 Wage (UC)

48 Weekly Hours: Wage-Earn vs. Self-Emp (Matched) Cyclicality (wrt GDP): 0.17 (0.03), 0.28 (0.07)

49 Productivity: All Workers vs. Self-Emp BLS Labor Productivity Self-employed Earnings per Hour Cyclicality (wrt GDP): (0.07), (0.19)

50 Consumption: All Workers vs. Self-Emp.12 Figure 6: Alternative Consumption Measures Aggregate Nondurable & Services Aggregate plus Relative Estimate for Self-Employed from CE Survey Cyclicality (wrt GDP): 0.64 (0.04), 1.27 (0.56)

51 Cyclicality of Labor Wedge: Robustness All Workers Labor Wedge Self-employed Wedge, Earnings per hour excluding incorporated Self-employed Wedge, Mimicing all-worker industry mix

52 Industry Composition of the Self-Employed Construction 17.2 Personal Services 6.3 Retail Trade 15.9 Repair 5.0 Business 12.7 Manufacturing 6.0 Medical & Legal 8.6 Other 4.7 FIRE 8.5 Wholesale Trade 4.3 Other Professional 7.8 Recreation 3.0 Entries are percent of all self-employed. Other = Transportation, Communications, Utilities and Mining.

53 Outline Measuring the Labor Wedge Examine both Extensive and Intensive Margins Decompose using Wage Data Our 2 Alternative Decompositions 1 Self-Employed 2 Intermediate Inputs Discuss Other Non-Wage Decompositions

54 Other ways to get price markups without wage data Capital expenditures (Galeotti and Schiantarelli, 1998) Advertising (Hall, 2014) Inventories Finished goods inventories Bils and Kahn, 2000 Kryvtsov and Midrigan, 2012 Work-in-process inventories (appendix)

55 Summary of other ways to get price markups Capital expenditures countercyclical markups Advertising acyclical markups (maybe) Inventories countercyclical markups All involve dynamics, requiring one to measure any adjustment costs and the stochastic discount factor. Self-Employed and Intermediates require only static measurements.

56 Advertising Approach Hall (2014) Implication of simple theory: [ ] max p,a {(p c) Zp ɛ A α κa} κa pq 1 1 p/c Thus, acyclical κa pq acyclical p c. But this implication is not robust to reasonable alterations: 1 Advertising { could affect the reservation price max p,a (p c) Z ( p ) } ɛ A κa α κa pq independent of desired markup movements. 2 Advertising could affect future demand Bagwell (2007)

57 Campello, Graham and Harvey (2010) % Change in Policy Variable Panel B - Two Constraint Categories Policies of Constrained vs. Unconstrained Firms Constrained Tech R&D Expenditures Marketing Expenditures Cash Holdings Unconstrained Capital Expenditures Number of Employees Dividend Payments Figure 2: This figure displays U.S. firms planned changes (% per year) in technology expenditures, capital expenditures, marketing expenditures, total number of domestic employees, cash holdings, and dividend payments as of the fourth quarter of 2008 (crisis peak period). Responses are averaged within sample partitions based on the survey measure of financial constraint. See text for additional details.

58

59 Constructing Extensive-Margin Wedge Optimal vacancy creation: φm t v t Can re-arrange to get [ u (c t ) y t n t h t Ω t h t { +β(1 δ)e t u (c t+1 )κ y t+1 h n t+1 EMW t = ln ( yt n t where ( ) h Ω t = 1+1/η t 1+1/η + ψ /h t ] u (c t )κ y t n t h } m t /v t = 0. m t+1 /v t+1 ) [ ] 1 σ ln(c t) + ln (Ω t ) S t, S t = f (m t, v t, h t, E t g(r t+1, y t+1 /n t+1, v t+1, m t+1 ))

60 Cyclicality of EMW and IMW Elasticity wrt GDP IMW (0.13) EMW (0.28) Quarterly data, σ = 0.5, η = 0.5 δ = 0.105, φ = 0.5, γ = 0.16 r ss = 0.004, ( ) κv m ss = 0.4 Expectational terms in EMW constructed using VAR approach

61 Cyclicality of EMW and IMW Elasticity wrt GDP Total Hours EMW (0.28) (0.15) IMW (0.13) (0.05) Quarterly data, σ = 0.5, η = 0.5 δ = 0.105, r = 0.004, φ = 0.5, κv m = 0.4, γ = 0.16 Expectational terms in EMW constructed using VAR approach

62 EMW and IMW Decomposition EMW = [ ( ) ] [ ( ) y/n ln w/p S w + ln + p S S 1σ ] ln(c) ln(ω), where S = S, but with φ = 1. IMW = [ ( )] [ ( ) y/n w ln + ln 1σ w/p p ln(c) 1η ] ln(h) Elasticity wrt GDP EMW IMW µ (0.28) (0.13) µ p ( w p = AHE ) (0.13) (0.13) µ p ( w p = NH ) (0.16) (0.16) µ p ( w p = UC ) (0.21) (0.21)

63 Hours: Self-Emp vs. Wage-Earn (Repeated CPS) Weekly Hours cyclicality (wrt GDP): 0.37 (0.14), 0.20 (0.02) Annual Hours cyclicality (wrt GDP): 0.57 (0.18), 0.39 (0.04)

64 Annual Hours: Self-Emp vs. Wage Earn (Matched) Cyclicality (wrt GDP): 0.54 (0.13), 0.57 (0.07)

65 Self-Employed Consumption in the PSID 3% 10% 2% Real GDP Growth (Right Axis) 8% 1% 6% 0% 4% 1% 2% 2% 0% 3% 4% Growth of Self Employed/Employed Consumption (Left Axis) % 4%

66 Industry-level Labor Wedge (µ i ) Preferences: { E 0 β t t=0 c 1 1/σ t 1 1/σ ν i Marginal Product wrt Labor (for ε = ω = 1): Labor Wedge (intensive-margin): ( ) pit mpn it ln (µ it ) = ln = ln p t mrs it mpn it = y it n it [( ) ]} 1+1/η h it 1 + 1/η + ψ e it ( pit p t y it n it ) [ 1 σ ln(c t) + 1 ] η ln(h it)

67 Intuition for Intermediates Results If w and p m reflect true shadow prices, then (for ε = 1) w n p m m = const. But empirically, intermediate expenditures more procyclical than labor expenditures intermediates-based µ p is more countercyclical. ( ) p y ( p y ) ( ) w n ln (µ p ) = ln = ln + ln p m m w n p m m Possible reconciliation: w doesn t reflect true shadow price.

68 Finished Goods Inventories (Simplified) Bils-Kahn and Kryvtsov-Midrigan first order condition: [ mc p u c = E φ s ( a u c + β 1 φ s ) ] mc a p u c a = finished inventories, s = sales Gives: [ ( E φ s ) ( p/mc a Γ + 1 p /mc ) uc u c ] = 1 β where Γ = ( p mc / p mc / ( p p u c p βu c u c βu c ) )

69

70

71 Issues with Finished Inventories 1 Could be scale effects for holding or ordering finished inventories. 2 Impact of inventories on sales could vary over cycle i.e., expect lower elasticity in recessions if demand less elastic.

72 Work-in-Process Inventories Follow Christiano (1988) in making work-in-process inventories a factor of production. Gives: E [(ψ y ) ( p/mc q + 1 p /mc ) uc u c ] = 1 β q = work-in-process inventories y = production

73

74 Inventory-based µ p vs. Total Labor Wedge µ Note: For Manufacturing Industries

75 Work-in-Process Inventories Production Technology: Marginal Product wrt Inventories: y it = g(z it, k it, n it )q ϕ it it q i,t+1 = (1 δ q )q it + y it s it mpq it = ϕ it y it q it Euler equation for shifting from WIP to sales (and back next period): [ mr it βu ( ) ] (c t+1 ) y i,t+1 mri,t+1 = E t p t u 1 δ q + ϕ i,t+1 (c t ) q i,t+1 p t+1

76 Constructing Inventory-based µ p Iterate forward and take logs to get ln ( µ p it ) = 1 σ ln(c t) + ln ( pit p t ) E t s=1 ϕ i,t+s 1 δ q y i,t+s q i,t+s NIPA Underlying Detail Tables Quarterly data, Manufacturing industries (aggregated to 14) q it : Work-in-process inventories y it : Sales plus change in (total) inventories p it : Sales price deflator

77 Return to Inventories vs. MUC

78 Cyclicality of Inventory-based µp

79 Cyclicality of Inventory-based µ p ln ( µ p it ) = 1 σ ln(c t) + ln ( pit p t ) E t s=1 ϕ i,t+s 1 δ q y i,t+s q i,t+s Elasticity wrt GDP µ p (0.12) MUC (0.06) Relative Price 0.67 (0.11) Output/Inventory Path 0.25 (0.03)

80 Role of µ p in µ, based on Inventories ln ( µ p ) / it ln (µit ) ln (cyc t ) ln (cyc t ) µ p vs. µ Manufacturing 109%