Inflation Dynamics During the Financial Crisis

Transcription

1 Inflation Dynamics During the Financial Crisis S. Gilchrist 1 1 Boston University and NBER MFM Summer Camp June 12, 2016 DISCLAIMER: The views expressed are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of anyone else associated with the Federal Reserve System.

2 Cyclical Dynamics of Producer Prices and Inflation In particular, the absence of more substantial deflationary pressures during the "Great Recession" is difficult to square with the Phillips curve common to most macroeconomic models. Economic forces that dampen the response of inflation to adverse demand or financial shocks reflect Cyclical Dynamics of Producer Prices and Industrial Production Core producer prices* Industrial production* Percentage points Percentage points Peak: Jan1980 Peak: Jul1981 Peak: Jul1990 Peak: Mar2001 Peak: Dec Peak: Jan1980 Peak: Jul1981 Peak: Jul1990 Peak: Mar2001 Peak: Dec Months to and from business cycle peaks * Deviations from a linear trend estimated over the 24 months preceding the specified recession Months to and from business cycle peaks * Deviations from a linear trend estimated over the 24 months preceding the specified recession. -30 Our answer

3 MOTIVATION What accounts for the resilience of inflation in the face of significant and long-lasting economic slack? In particular, the absence of more substantial deflationary pressures during the Great Recession is difficult to square with the Phillips curve common to most macroeconomic models (including standard financial accelerator models). In a customer-markets model with financial frictions, firms have the incentive to raise prices to increase cash flow at the cost of future market share (Gottfries [1991]; Chevalier and Scharfstein [1996]).

4 DATA SOURCES GILCHRIST, SCHOENLE, SIM AND ZAKRAJSEK (2015) Monthly good-level price data underlying the PPI. (Nakamura & Steinsson [2008]; Goldberg & Hellerstein [2009]; Bhattarai & Schoenle [2010]) Match 584 PPI respondents to their income and balance sheet data from Compustat. Sample period: Jan2005 Dec2012

5 RELATIVE INFLATION Financially unconstrained vs constrained firms 3-month moving average Percent Low liquidity firms High liquidity firms NOTE: Weighted average monthly inflation relative to industry (2-digit NAICS) inflation.

6 ACCUMULATED INDUSTRY-ADJUSTED PPI PPI INFLATION By By selected financial characteristics as inof 2006 Monthly Low liquidity firms High liquidity firms Index (Dec = 100) NOTE: NOTE: Relative Weighted to industry average (2-digit NAICS) monthly inflation. inflation relative to industry (2-digit NAICS) inflation.

7 PRICE ADJUSTMENT AND FIRM CHARACTERISTICS Multinomial logit specification: + Pr(p i,j,t+3 p i,j,t ) = 0 (base) Price change regression: = Λ(X jt ; β t ) log(p i,j,t+3 ) log(p i,j,t ) = βx j,t + ɛ i,j,t+3 X jt = liquidity ratio and other controls. Includes fixed time effects and 3-digit inflation. Estimated using four-quarter rolling window.

8 DIRECTIONAL PRICE CHANGE REGRESSIONS Explanatory Variables + LIQ j,t 1[CRISIS t = 1] (0.107) (72) LIQ j,t 1[CRISIS t = 0] (68) (50) log(s j,t /S j,t 12 ) (25) (25) log(c j,t /C j,t 12 ) (13) (11) [N/S] j,t (21) (24) π IND(3m) t (0.333) (0.170)

9 DIRECTIONAL PRICE CHANGE REGRESSIONS Coefficient on liquidity ratio (4-quarter rolling window estimates) Negative price changes Positive price changes Coefficient 0.4 Coefficient 0.4 Estimate +/- 2 S.E Quantitative implication: a two std. dev. reduction in liquidity implies a 33% higher probability of a price increase. -0.8

10 INFLATION REGRESSIONS Explanatory Variables LIQ j,t 1[CRISIS t = 1] LIQ j,t 1[CRISIS t = 0] -29 (09) -12 (04) log(s j,t /S j,t 12 ) 04 (03) log(c j,t /C j,t 12 ) -02 (02) [N/S] j,t 01 (01) π IND(3m) t (55)

11 INFLATION Marginal effect with respect to liquidity ratio Estimate +/- 2 S.E. Coefficient Quantitative implication: A two std. dev. reduction in liquidity implies a 5% increase in annualized inflation.

12 Industry-level evidence ( Examine sensitivity of 6-digit industry-level PPI inflation to changes in aggregate financial conditions. Regress industry-specific year-ahead inflation on Current and lagged inflation and industrial production. Current financial conditions excess bond premium (EBP) Coefficient on EBP varies across 4-digit industry groups. Is variation in industry-specific EBP coefficieent related to financial constraints across industries? Measure severity of financial constraints using size-age index.

13 Coefficients on EBP and commodity price inflation vary across 4-digit indu Inflation Response to EBP - Is variation in industry-specific EBP coefficients related to the likelihood across industries? - Use industry-specific size-age index to identify the likelihood of financia 12-month PPI inflation and commodity prices 4 12-month PPI inflation and financial conditions By industry-specific indicator of financial constraints Coefficient on EBP (4-digit NAICS) Median Size-Age Index (4-digit NAICS) p <.10 p >=.10 β ^ = 1.11 t = 4.88 R-sq = 0.29 Note: Smaller values of the size-age index indicate a smaller likelihood of financial constraints.

14 Output Response to the EBP Figure 7: Sensitivity of Industry-Level Output to Financial Conditions, (By Industry-Specific Indicator of Financial Constraints) Coefficient on EBP Median Size-Age Index p <.10 p >=.10 β ^ = t = R-sq = 0.22 Note: No. of (4-digit NAICS) industries = 52. The figure shows the relationship between the median SA-index of financing constraints at the 4-digit NAICS level during the period and the corresponding industry-specific estimates of the coefficient on the EBP; the dependent

15 GE MODEL Customer markets imply that firms trade off current profits for future market share. Financial market frictions imply that firms discount the future more when demand is low and therefore charge high markups. Embed this intuition into a GE model with nominal price rigidities.

16 Deep habits Ravn, Schmitt-Grohe and Uribe [2006] Demand for monopolistically competitive good: where c it = ( pit p t ) η s θ(1 η) i,t 1 c t s it = ρs i,t 1 + (1 ρ)c it Firms are forward looking set low price today to build future stock of customer base.

17 FINANCIAL FRICTIONS Firms make production decisions prior to realization of marginal cost. [ ] α hit y it = φ i ; 0 < α 1 a it If realized operating income is negative, firms must raise costly equity finance: ϕ (0, 1) = constant per-unit dilution costs of new equity Expected shadow value of internal funds: E a t [ξ it ] > 1

18 Optimal Pricing without Deep Habits Assume flexible prices and no customer markets. When α = 1, optimal pricing p i,t = η η 1 }{{} accounting markup Ea t [ξ it a it ] E a t [ξ it] } {{ } economic markup [ ] wt /p t A t }{{} real marginal cost Financial frictions E a t [ξ it a it ] E a t [ξ it] = 1 + Cov[ξ it a it ] 1

19 Optimal Pricing with Deep Habits Bring back customer markets (still flexible prices!) Growth-adjusted, compounded discount rate: s s+1 /s s ρ β t,s m s,s+1 1 ρ s t Optimal pricing p i,t = j=1 η E a t [ξ it a it ] η 1 E a t [ξ it] [ χ η 1 E t [ ρ + χ s t+j/s t+j 1 ρ 1 ρ [ ] wt /p t s=t+1 A t β t,s E a s[ξ i,s ] E a t [ξ i,t] ] m t+j 1,t+j ( p h,s w ) ] s/p h,s A s

20 Optimal Pricing with Deep Habits Bring back customer markets (still flexible prices!) Growth-adjusted, compounded discount rate: s s+1 /s s ρ β t,s m s,s+1 1 ρ s t Optimal pricing p i,t = j=1 η E a t [ξ it a it ] η 1 E a t [ξ it] [ χ η 1 E t [ ρ + χ s t+j/s t+j 1 ρ 1 ρ [ ] wt /p t s=t+1 A t β t,s E a s[ξ i,s ] E a t [ξ i,t] ] m t+j 1,t+j ( p h,s w ) ] s/p h,s A s

21 LOG-LINEARIZED PHILLIPS CURVE New Keynesian model with cost channel [ ω(η 1) ] ˆπ t = ˆµ t + E t χ δ s t+1 ˆµ s+1 + βe t [ˆπ t+1 ] γ p s=t + 1 [ ] [η ω(η 1)] E t χ δ s t+1 (ˆξ t γ ˆξ s+1 ) ˆβ t,s+1 p s=t ˆµ t = (financially-adjusted) mark-up ˆβ t,s+1 = capitalized growth of customer base ˆξ t = shadow value of internal funds

22 LOG-LINEARIZED PHILLIPS CURVE The role of deep habits [ ω(η 1) ] ˆπ t = ˆµ t + E t χ δ s t+1 ˆµ s+1 + βe t [ˆπ t+1 ] γ p s=t + 1 [ ] [η ω(η 1)] E t χ δ s t+1 (ˆξ t γ ˆξ s+1 ) ˆβ t,s+1 p s=t ˆµ t = (financially-adjusted) mark-up ˆβ t,s+1 = capitalized growth of customer base ˆξ t = shadow value of internal funds

23 LOG-LINEARIZED PHILLIPS CURVE The role of financial frictions [ ω(η 1) ] ˆπ t = ˆµ t + E t χ δ s t+1 ˆµ s+1 + βe t [ˆπ t+1 ] γ p s=t + 1 [ ] [η ω(η 1)] E t χ δ s t+1 (ˆξ t γ ˆξ s+1 ) ˆβ t,s+1 p s=t ˆµ t = (financially adjusted) mark-up ˆβ t,s+1 = capitalized growth of customer base ˆξ t = shadow value of internal funds

24 CALIBRATION Benchmark model: homogeneous firms Parameter Value Preferences and Technology Relative risk aversion: γ x 1.00 Deep habit: θ Persistence of deep habit: ρ 0.95 Elasticity of labor supply: 1/γ h 5.00 Elasticity of substitution: η 2.00 Fixed operating costs: φ 0.21 Idiosyncratic volatility (a.r.): σ 0.20 Financial Frictions Equity dilution costs: ϕ Persistence of financial shock: ρ ϕ 0.90

25 DEMAND SHOCK: FINANCIAL CRISIS(ϕ = 0.5) (a) Output w/o FF w/ FF % (b) Hours worked % (c) Inflation pps (d) Real wage % (e) Markup pps. 0.4 (f) Value of internal funds pps. 2.0 (g) Value of marginal sales % 1.5 (h) Interest rate pps

26 FINANCIAL SHOCK:(ϕ = 0.3) WITH VS WITHOUT CUSTOMER MARKETS (a) Output w/ deep habit w/o deep habit % (b) Hours worked % (c) Inflation pps (d) Real wage % (e) Markup pps. 1.0 (f) Value of internal funds pps. 5 (g) Value of marginal sales % 5 (h) Interest rate pps

27 Alternative monetary policy rules DEMAND SHOCK DURING A FINANCIAL CRISIS (a) Markup pps. τ y = 0 τ y = τ y = (b) Inflation pps (c) Output % (d) Real interest rate pps Policy Implications: Divine Coincidence breaks down!

28 HETEROGENEOUS FIRMS Sectors differ by operating efficiency: 0 φ 1 < φ 2 Fixed measures of firms: Ξ 1 = Ξ 2 = 1 2 Equilibrium dispersion of relative prices: π t = [ 2 k=1 Ξ k p 1 η k,t 1 π1 η kt ] 1 1 η p kt P kt P t = sector-specific relative price π kt = sector-specific inflation rate P kt P k,t 1 Benchmark case: φ 1 = 0 financially strong firms φ 1 = 0.3 financially weak firms

29 PRICE WAR IN RESPONSE TO FINANCIAL SHOCKS Heterogeneous firms (a) Relative prices (b) Output % % Financially strong Financially weak Financially weak -1.0 Financially strong Aggregate Aggregate Case I: φ 1 = 0.8 φ, φ 2 = φ and ω 1 = ω 2 = 0.5 Case II: φ 1 = 0, φ 2 = φ and ω 1 = ω 2 = 0.5

30 PARADOX OF FINANCIAL STRENGTH Heterogeneous firms (a) Relative prices % (b) Output % 0.6 Financially strong 0.5 Financially weak Financially weak Financially strong Aggregate Aggregate Case I: φ 1 = 0.8 φ, φ 2 = φ and ω 1 = ω 2 = 0.5 Case II: φ 1 = 0, φ 2 = φ and ω 1 = ω 2 = 0.5

31 PARADOX OF FINANCIAL STRENGTH Heterogeneous firms (a) Relative prices % (b) Output % 0.6 Financially strong 0.5 Financially weak Financially weak Financially strong Aggregate Aggregate Aggregate Aggregate Case I: φ 1 = 0.8 φ, φ 2 = φ and ω 1 = ω 2 = 0.5 Case II: φ 1 = 0, φ 2 = φ and ω 1 = ω 2 = 0.5

32 CONCLUSION Empirical results imply that financially healthy firms decreased prices, while financially weak firms increased prices during the financial crisis. Industry-level and Eurozone evidence suggest this is a regular feature of business cycles. DSGE model: financial theory of countercylical markups implies attenuation of inflation dynamics in response to demand shocks and severe contraction in response to temporary financial shocks. Implications for monetary policy: Tradeoff between inflation and output in response to demand and financial shocks.