International Shocks, Variable Markups and Domestic Prices

Transcription

1 International Shocks, Variable Markups and Domestic Prices Mary Amiti Oleg Itskhoki Jozef Konings NY FRB Princeton Leuven and BNB Stanford IO Lunch May 2018 Previously circulated as International Shocks and Domestic Prices: How Large are Strategic Complementarities? 1 / 23

2 Motivation How strong are strategic complementarities across firms in price setting? do firms mostly respond to own cost shocks or put a large weight on the prices of their competitors? a fundamental mechanism in both micro- and macroeconomics empirical challenge: separate marginal costs from markups 1 / 23

3 Motivation How strong are strategic complementarities across firms in price setting? do firms mostly respond to own cost shocks or put a large weight on the prices of their competitors? a fundamental mechanism in both micro- and macroeconomics empirical challenge: separate marginal costs from markups A particular application: How do international shocks affect domestic markups? International transmission and low exchange rate pass-through Other applications: slope of the Phillips Curve in Monetary 1 / 23

4 Our Approach 1 Directly estimate extent of strategic complementarities across a broad set of manufacturing industries (a) Develop a general theoretical framework decompose firm price changes into response to own cost shocks and to changes in competitor prices (b) Develop an identification strategy major challenges: (i) measurement error in marginal costs, (ii) simultaneity of price setting, (iii) correlated demand shocks (c) Construct a new micro-level dataset for Belgium manufacturing to carry out this estimation intensive data requirements on firm prices, marginal costs, and firm s competitor prices within sectors 2 Develop a quantitative framework to evaluate counterfactual scenarios a flexible model tightly calibrated to the Belgian microdata explore response of firm markups and prices to various shocks 2 / 23

5 Main Findings 1 Strong evidence of strategic complementarities: (i) response to own marginal cost 0.6 (or own/idiosyncratic cost pass-through elasticity) (ii) response to competitor prices 0.4 (or strategic complementarities elasticity) (iii) cannot reject that the two sum to one 2 These are average (sales-weighted) responses. Yet, a lot of heterogeneity in responses across firms: small firms: no strategic complementarities and constant markups (as in standard MC-CES models) large firms: strong strategic complementarities and variable markups 3 Implications for aggregate pass-through into domestic prices conditions for aggregate markup adjustment and low ERPT decrease in aggregate home markup in response to devaluation when less foreign competition and more foreign inputs 3 / 23

6 Related Literature 1 IO-style studies: Industry studies: Feenstra et al. (1996, cars), Nakamura and Zerom (2010, coffee), Goldberg and Hellerstein (2013, beer) Alternative: De Loecker and Warzynski (2012), De Loecker and Goldberg (2014) 2 International prices and exchange rates: Gopinath and Itskhoki (2011) Berman, Martin and Mayer (2012), Amiti, Itskhoki and Konings (2014) 3 Domestic prices and exchange rates: industry-level: Goldberg and Campa (2010) product level: Auer and Schoenle (2013), Cao, Dong and Tomlin (2012), Pennings (2012) 4 Domestic prices and trade shocks: De Loecker, Goldberg, Khandelwal and Pavcnik (2015) Edmond, Midrigan and Xu (2015) 4 / 23

7 THEORETICAL FRAMEWORK 5 / 23

8 Estimating Equation p it = }{{} α mc it + γ }{{} own cost pass-through strategic elasticity complementarity elasticity index of competitor price changes {}}{ p it + ε it 5 / 23

9 Log price identity: p it = mc it + µ it Price setting 6 / 23

10 Price setting Log price identity: p it = mc it + µ it Proposition 1 For any given invertible demand system q t = {q i (p t ; ξ t )} i competition structure (monopolistic or oligopolistic, price or quantity) there exists a markup function µ it = M i (p t ; ξ t ) such that the firm s static profit-maximizing price p it solves: p it = mc it + M i ( p it, p it ; ξ t ) (1) 6 / 23

11 Price setting Log price identity: p it = mc it + µ it Proposition 1 For any given invertible demand system q t = {q i (p t ; ξ t )} i competition structure (monopolistic or oligopolistic, price or quantity) there exists a markup function µ it = M i (p t ; ξ t ) such that the firm s static profit-maximizing price p it solves: p it = mc it + M i ( p it, p it ; ξ t ) (1) Markup function: µ it = log σ it σ it 1 = M i(p t ; ξ t ) Fixed point (1) implicitly defines best response schedule p it does not depend on mc it conditional on p it Industry equilibrium further requires p t = mc t + M(p t ; ξ t ) 6 / 23

12 Price change decomposition Totally differentiate best response (1) around some (p t ; ξ t ) and rearranging to decompose the firm s price change: dp it = Γ it dmc it + Γ it 1 + Γ it dp it + ε it (2) Own markup elasticity: Γ it M i(p t ; ξ t ) p it Competitor markup elasticity: Γ it j i Index of competitor price changes: M i (p t ; ξ t ) p jt dp it ω M i (p t ; ξ ijtdp jt with ω ijt t )/ p jt j i k i M i(p t ; ξ t )/ p kt Residual demand shock: ε it 1 1+Γ it N j=1 M i (p t;ξ t ) ξ jt dξ jt 7 / 23

13 Price change decomposition Totally differentiate best response (1) around some (p t ; ξ t ) and rearranging to decompose the firm s price change: dp it = Γ it dmc it + Γ it 1 + Γ it dp it + ε it (2) Proposition 2 (i) If q it = q i (p it, z t ; ξ t ), where z t is log expenditure function, then ω ijt = S jt /(1 S it ) and dp it = j i S jt 1 S it dp jt. (ii) If, further, σ it = σ i (p it z t ; ξ t ), then Γ it = Γ it and α it + γ it = 1. holds for a broad class of models. offers a testable implication intuition: Shephard s lemma 7 / 23

14 Examples Demand and competition structure Monopolistic competition: 1 CES with σ it σ and hence Γ it = Γ it 0 2 Non-CES with σ it q i(p it, p i,t ; ξ t ) p it and Γ it, Γ it 0 demand vs strategic complementarities Γ it Γ it if q it = q i (p it p it ; ξ t ) linear demand (e.g., Melitz and Ottaviano 2008), translog demand (e.g., Feenstra and Weinstein 2010), Kimball demand (e.g, Gopinath and Itskhoki 2010), nested logit demand (e.g., Goldberg 1995), etc. 8 / 23

15 Examples Demand and competition structure Monopolistic competition: 1 CES with σ it σ and hence Γ it = Γ it 0 2 Non-CES with σ it q i(p it, p i,t ; ξ t ) p it and Γ it, Γ it 0 demand vs strategic complementarities Γ it Γ it if q it = q i (p it p it ; ξ t ) linear, translog, Kimball, nested logit demand, etc. Oligopolistic competition: [ qi (p t ; ξ σ it t ) + j i p it q i (p t ; ξ t ) p jt ] p jt p it 1 Price (Bertrand) competition: p jt / p it = 0 2 Quantity (Cournot): p jt p it = q j (p t;ξ t )/ p it q j (p t;ξ t )/ p jt dq j (p t, ξ t )=0 8 / 23

16 A Model of Variable Markups Nested CES and Oligopolistic Competition (Atkeson and Burstein 2008) Firm-product demand: Q it = ξ it P ρ it Pt ρ η D t, ρ > η 1 Sectoral price index: P t [ N ] i=1 ξ itp 1 ρ 1 1 ρ it = [ ξ it P 1 ρ it ] + (1 ξ it )P 1 ρ 1 1 ρ i,t Market share: S it ( ) P it Q 1 ρ it Pit N j=1 P = ξ it [0, 1] jtq jt P t 9 / 23

17 A Model of Variable Markups Cournot (quantity) competition: Why Cournot? P it = σ [ it 1 σ it 1 MC it, where σ it η S it + 1 ] 1 ρ (1 S it) Markup elasticity: Γ it = Γ it = (ρ η)(ρ 1)σ its it (1 S it ) ρη(σ it 1) Markup elasticity increases in market share: Γ it = Γ(S it ) (over relevant range) Log price change: dp it = Γ(S it ) dmc it + Γ(S it) 1 + Γ(S it ) dp it + κ it dξ it 10 / 23

18 Marginal cost: MC it = W 1 φ it it Ω it Cost Structure V φ it it Y ζ i it Average Variable Cost: AVC it = TVC it Y it = ζ i MC it Marginal cost in log changes: mc it = avc it mc it = φ it v it + (1 φ it ) w it + (v it w it ) φ it + ζ i y it ω it mc it = φ it v it 11 / 23

19 DATA 12 / 23

20 We merge 3 micro-level datasets: Dataset 1 PRODCOM: Belgium firm-product level data on values and quantities PC 8-digit (2,500 products) All manufacturing firms with minimum of 10 employees Notaion: i corresponds to firm-product at this level 2 Customs: Import and export data on values and quantities at firm-product-country level CN 8-digit (over 10,000 products) Notation: m corresponds to firm-product-country for inputs 3 Census: firm-level data on firm characteristics includes material costs, wagebill and employment Baseline industry s definition: NACE 4-digit level 12 / 23

21 Variables Domestic Prices: p it = log Domestic Value it Domestic Quantity it Average Variable Cost as proxy for Marginal Cost: Instrument: mc it = log TVC it Y it mc it = φ it m ωc imt v imt e it = φ it m ωc imt e mt where v imt is the log change in firm input prices 13 / 23

22 Competitor price index: Competitor prices p it = j i S jt 1 S it p jt. Instruments (for home, non-ez and EZ competitors): mc it = j D i e it = j D i e X it = j X i p E it = j E i S jt 1 S it mc jt S jt 1 S it e jt S jt 1 S it e k(j)t S jt 1 S it p m k(j)s(i)t 14 / 23

23 Identification We estimate best response in changes over time: p it = α }{{} it mc it + γ it p }{{} it + ɛ it own cost strategic pass-through complementarity elasticity elasticity Allow α it and γ it to vary with firm size and test the null α it + γ it = 1 15 / 23

24 Identification We estimate best response in changes over time: p it = α }{{} it mc it + γ it p }{{} it + ɛ it own cost strategic pass-through complementarity elasticity elasticity Allow α it and γ it to vary with firm size and test the null α it + γ it = 1 Identification challenges: 1 measurement error in mc it 2 measurement error and endogeneity of p it 3 heterogeneity in Γ it and Γ it across observations 4 multi-product firms Show reduced form 15 / 23

25 EMPIRICAL FINDINGS 16 / 23

26 Strategic Complementarities Baseline p it = α mc it + γ p it + ε it OLS IV Dep. var.: p it (1) (2) (3) (4) (5) mc it (0.040) (0.041) (0.094) (0.112) (0.103) p it (0.079) (0.095) (0.097) (0.118) # obs. 64,823 64,823 64,823 64,823 64,823 Year F.E. yes yes yes yes yes Industry F.E. no yes no yes yes H 0 : ψ + γ = yes [p-value] [0.00] [0.00] [0.05] [0.16] Overid J-test χ [p-value] [0.30] [0.69] [0.70] Weak IV F -test IV regressions: 1 pass over-id and weak instrument tests First Stage 2 cannot reject the equality to 1 of the sum of the coef. 16 / 23

27 Strategic Complementarities Small versus Large Firms Large i definition: Employment 100 S it >2% Sample: Small Large All All All All Dep. var.: p it (1) (2) (3) (4) (5) (6) mc it (0.160) (0.211) (0.128) (0.195) (0.123) mc it Large i (0.203) (0.344) (0.178) (0.432) (0.201) p it (0.194) (0.237) (0.229) (0.102) p it Large i (0.175) (0.320) (0.403) (0.187) # obs. 49,469 15,354 64,823 64,823 64,822 64,823 Ind.&Year F.E. yes yes yes yes Ind. Year F.E. no no no 4-digit 2-digit no Overid. J-test χ [p-value] [0.32] [0.78] [0.23] [0.29] [0.29] Weak IV F -test Continuous splits 17 / 23

28 Robustness 1 Alternative instrument sets show 2 Quality and productivity upgrading show 3 Alternative samples and selection show multi-product firms selection into exporting, importing, and FDI levels of aggregatation price stickiness 4 Alternative measures of competitor prices show placebo regressions 18 / 23

29 AGGREGATE MARKUPS and ERPT 19 / 23

30 From Micro to Macro Consider an exchange rate devaluation e t > 0: ψ it { dpit { }} } { E de t = ϕ it { { }} } { 1 dmcit E 1 + Γ it de t Ψ it { dp it {}} + Γ }{ it E 1 + Γ it de t 19 / 23

31 From Micro to Macro Consider an exchange rate devaluation e t > 0: ψ it { dpit { }} } { E de t = ϕ it { { }} } { 1 dmcit E 1 + Γ it de t Ψ it { dp it {}} + Γ }{ it E 1 + Γ it de t We are interested in aggregate ERPT: { } dpt Ψ t E = { } N de S dmct itψ it vs ϕ t E = t i=1 de t N S it ϕ it i=1 19 / 23

32 From Micro to Macro Consider an exchange rate devaluation e t > 0: ψ it { dpit { }} } { E de t = ϕ it { { }} } { 1 dmcit E 1 + Γ it de t Ψ it { dp it {}} + Γ }{ it E 1 + Γ it de t We are interested in aggregate ERPT: { } dpt Ψ t E = { } N de S dmct itψ it vs ϕ t E = t i=1 de t Aggregate markup adjustment Ψ t ϕ t = N S it(ψ it ϕ it ) i=1 Individual markup adjustment: N S it ϕ it i=1 ψ it ϕ it = κ it (ϕ it Ψ t ), where κ it 1 S it + Γ it Γ it 19 / 23

33 Proposition 3 Aggregate ERPT: Ψ t = 1 1 κ t where κ it ERPT and Aggregate Markups N S it (1 κ it )ϕ it = ϕ t cov( ) κ it, ϕ it, 1 κ t i=1 Γ it 1 S it +Γ it and cov ( κ it, ϕ it ) = N i=1 S it(κ it κ t )ϕ it 20 / 23

34 Proposition 3 Aggregate ERPT: Ψ t = 1 1 κ t where κ it ERPT and Aggregate Markups N S it (1 κ it )ϕ it = ϕ t cov( ) κ it, ϕ it, 1 κ t i=1 Γ it 1 S it +Γ it Γ it and cov ( κ it, ϕ it ) = N i=1 S it(κ it κ t )ϕ it Corollary 1 If 1 S it = const for all i, then Ψ t = ϕ t, and aggregate markup is constant, even if all Γ it > 0 markup adjustment at the micro level washes out in the agg. 20 / 23

35 Proposition 3 Aggregate ERPT: Ψ t = 1 1 κ t where κ it ERPT and Aggregate Markups N S it (1 κ it )ϕ it = ϕ t cov( ) κ it, ϕ it, 1 κ t i=1 Γ it 1 S it +Γ it Γ it and cov ( κ it, ϕ it ) = N i=1 S it(κ it κ t )ϕ it Corollary 1 If 1 S it = const for all i, then Ψ t = ϕ t, and aggregate markup is constant, even if all Γ it > 0 markup adjustment at the micro level washes out in the agg. Corollary 2 If Γ it and ϕ it increase with firm size S it, then Ψ t < ϕ t, and aggregate markup declines with depreciation our evidence suggests this is the empirically-relevant case 20 / 23

36 Stylized Example Three types of firms Type Cum. Import Markup of firm share intensity elasticity Small Home λ S ϕ S = 0 Γ S = 0 Large Home λ L ϕ L = ϕ > 0 Γ L = Γ > 0 Large Foreign λ F ϕ F = ϕ > ϕ Γ F = Γ Γ Aggregate ERPT and markup adjustment: ϕ Ψ = where ϕ = λ L ϕ + λ F ϕ 1 + λ S Γ 21 / 23

37 Stylized Example Three types of firms Type Cum. Import Markup of firm share intensity elasticity Small Home λ S ϕ S = 0 Γ S = 0 Large Home λ L ϕ L = ϕ > 0 Γ L = Γ > 0 Large Foreign λ F ϕ F = ϕ > ϕ Γ F = Γ Γ Aggregate ERPT and markup adjustment: ϕ Ψ = where ϕ = λ L ϕ + λ F ϕ 1 + λ S Γ Average markup adjustment by domestic firms: [ ] λ L Γ ϕ Ψ D ϕ D = λ S + λ L 1 + Γ 1 + λ S Γ ϕ conventional logic that Ψ D > ϕ D does not apply in general 21 / 23

38 Quantitative Model Table: Strategic complementarities Dep. var.: p it All Small Large Interaction mc it mc it Large it p it p it Large it Table: ERPT and Markup adjustment Sets of firms ERPT into: All Home Large Small Foreign Costs ϕ J Prices Ψ J Markups Ψ J ϕ J / 23

39 Quantitative Model Figure: ERPT and Markup adjustment (by size bins of home firms) 0.3 ERPT Markup Bins of home firms by market share 22 / 23

40 Conclusions We provide direct evidence on the strength of strategic complementarities in firm price setting on average, responsiveness to own cost is 0.6 and to competitor prices is 0.4 Uncover substantial heterogeneity in the extent of strategic complementarities across firms: small firms do not respond to competitor prices and have complete pass-through of own cost shocks large firms exhibit substantial strategic complementarities and variable markups The interplay of these forces with heterogeneous exposure to foreign inputs shapes aggregate markups and ERPT 23 / 23

41 APPENDIX 24 / 23

42 Reduced form p it = a it mc it + b it mc it + ε it, where (assuming Γ it = Γ i,t = Γ): a it = b it = Γ 1 S it Γ 1 S it + Γ, 1 mc it = 1 Γ j 1 Γ j i 1 Γ j S jt 1 S jt +Γ j i S jt 1 S jt +Γ S jt 1 S jt +Γ, S jt 1 + Γ 1 S jt mc jt. Back to slides 25 / 23

43 Price Setting Fixed Point In each industry, given a vector of firm marginal costs {MC it }, we find the equilibrium vector of prices {P it } This is a fixed point problem: P it = M it MC it, M it = σ it /(σ it 1), σ it = [ η 1 S it + ρ 1 (1 S it ) ] 1, ( ) 1 ρ, S it = ξ it Pit /P t ] 1/(1 ρ) P t = [ N i=1 ξ itp 1 ρ it All prices are determined simultaneously The solution to this problem can be found numerically by iteration Market Structure 26 / 23

44 Sensitivity Demand and Market Structure 1.6 Markup, M it = σ it σ it 1 Pass-through, Ψ it = 1 1+Γ it ρ = Bertrand 0.8 Markup, Mit Pass-through, Ψit η = 2 ρ = η = 2 Bertrand Market share, sit [ Market share, sit 1 Note: σ it = S η it + 1 (1 S ρ it)] under Cournot and σ it = [ηs it + ρ(1 S it )] under Bertrand. Market Structure 27 / 23

45 Strategic Complementarities First Stage Regressions For column 3 For column 4 Dep. var.: mc it p it mc it p it mc it (0.117) (0.034) (0.120) (0.033) mc it (0.363) (0.217) (0.372) (0.238) e X it (0.363) (0.217) (0.372) (0.238) p E it (0.226) (0.149) (0.281) (0.113) # obs. 64,823 64,823 64,823 64,823 Industry F.E. no no yes yes First stage F -test [p-value] [0.00] [0.00] [0.00] [0.00] Back to slides 28 / 23

46 Strategic Complementarities Bins of Firms Varying employment cutoff from 100 to 8,500 workers 1 Fraction below cutoff, % Observations Sales Employment cutoff Cost pass-through, ^, Strategic complementarity, ^ Share of sales below employment cuto, Back to slides 29 / 23

47 Alternative Instrument Sets Robustness Robustness to: p E it e X it mc it mc it and mc it Dep. var.: p it (1) (2) (3) (4) (5) (6) (7) (8) mc it (0.099) (0.154) (0.150) (0.123) (0.525) (0.315) (0.293) (0.240) p it (0.114) (0.174) (0.147) (0.239) (0.401) (0.283) (0.314) (0.267) Notes: All regressions are counterpart to column 4 of Table 1, with baseline instrument set ( mc it, mc it, e X it, p E it). Each column drops one or two of these instruments in turn, sometimes replacing them with alternative more conservative instruments. Column 1 replaces p E it with one that only uses export prices to non-eurozone destination, and column 2 drops p it E altogether. Column 3 drops e it, X and hence excludes exchange rate variation from the instrument set. Column 4 drops mc it. Columns 5 8 drop both mc it and mc it. Column 5 adds instead exchange-rate-based alternatives e it and e it described in the text. Column 6 (7) additionally adds two new instruments analogous to mc it and mc it, which replace firm import prices with proxies based on source-country export prices to countries other than Belgium (to outside the eurozone). Column 8 is like column 7, but with time-invariant firm-level weights used to construct the instruments. In all cases, the regressions pass the weak instrument F -test and the overidentification J-test, and the null that the coefficients sum to one cannot be rejected; the number of observations is 64,823, as in the baseline regression. Back to slides 30 / 23

48 Quality and Productivity Upgrading Robustness Rauch index Firm R&D Large firm R&D TFP VA/worker Dep. var.: p it (1) (2) (3) (4) (5) mc it (0.175) (0.151) (0.258) (0.116) (0.118) mc it R i (0.215) (0.213) (0.283) p it (0.191) (0.207) (0.346) (0.122) (0.124) p it R i (0.270) (0.247) (0.360) log TFP it log(va it /L it ) (0.018) (0.018) # obs. 64,823 64,823 15,354 64,247 64,405 Notes: All regressions are counterpart to column 4 of Table 1. In column 1, R i is a dummy for whether firm-product i is in a differentiated sector according to the Rauch classification. In columns 2 and 3, R i is a dummy for whether firm i records any positive R&D expenditure during the sample; column 3 limits the sample to the large firms only. Columns 4 and 5 add controls for firm-level log changes in measured TFP and value added per worker, respectively. Back to slides 31 / 23

49 Alternative Samples and Selection Robustness Alternative input Finer industry Two-period Main product definition 5-digit 6-digit diff Dep. var.: p it (1) (2) (3) (4) (5) (6) (7) mc it (0.162) (0.128) (0.145) (0.126) (0.145) (0.145) (0.161) p it (0.135) (0.131) (0.192) (0.177) (0.174) (0.143) (0.210) # obs. 64,823 64,823 27,031 48,284 64,350 62,713 51,322 Back to slides 32 / 23

50 Competitor Prices and Placebo Robustness Placebo with random Largest Placebo industry assignment competitor(s) with mc it Dep. var.: p it (1) (2) (3) (4) (5) mc it (0.101) (0.139) (0.114) (0.100) (0.155) p it (0.159) (0.408) S it L pl it (0.238) (0.223) (1 S it L ) p L it (0.161) (0.245) p it (0.114) (0.094) mc it (0.424) # obs. 64,823 64,823 64,823 64,823 64,780 Back to slides 33 / 23