The link between labor costs and price inflation in the euro area

Transcription

1 The link between labor costs and price inflation in the euro area E. Bobeica M. Ciccarelli I. Vansteenkiste European Central Bank* Paper prepared for the XXII Annual Conference, Central Bank of Chile Santiago, 26 October 2018 (*) The views expressed in this paper are of the authors only and do not necessarily reflect those of the European Central Bank or the Eurosystem

2 Motivation and literature A better understanding of the signal that labor cost developments provide for the inflationary process is a key (monetary) policy issue Economic theory consistent with both directions of causality Empirical evidence inconclusive and mostly USbased A systematic analysis for euro area is missing 2

3 Aim and contribution Empirical exploration of the wage-price link in the 4 biggest euro area countries and across the three main sectors Question: Is the ECB rhetoric justified? Does a link between wages and prices exist in the euro area at a cyclical frequency? Does this link depend on the state of the economy (in particular on the level of inflation)? Does the link depend on the shocks hitting the economy? 3

4 Outline Data Unconditional evidence Conditional evidence Summary and conclusion 4

5 Outline Data Unconditional evidence Conditional evidence Summary and conclusion 5

6 Data Four countries: DE, FR, IT, ES. Three sectors and total economy Three variables (y-o-y growth rates): Real value added (RVA) RVA deflator Productivity-adjusted compensation per employees Robustness: Unemployment Monetary policy stance Why disaggregated data Institutional settings Cost structure and factor share Labor market characteristics, wage formation specificities Degree of global exposure and international competition; workers turnover rates; capital intensity Degree of wage and price rigidity (see e.g. WDN report, 2009) 6

7 Data and de-trending High correlation between y-o-y growth rates and common (mainly downward trend) We assume that the gap between productivityadjusted wage growth and inflation reflects deviations away from a common long-run equilibrium (Consensus survey-based long-run inflation expectation) This adjustment might be crucial to understand the link between ULC and price inflation 7

8 Data and de-trending Figure 1: ULC, GDP deflator and common trend, year on year % change 8

9 Data and de-trending Figure 1a: ULC, GDP deflator and common trend, year on year % change 9

10 Data and de-trending See US Figure 2: Adjusted and unadjusted ULC (6 months prior) and price inflation in EA 10

11 Outline Data Unconditional evidence Conditional evidence Summary and conclusion 11

12 Unconditional evidence Consistent and relatively strong link across euro area countries and sectors at a cyclical frequency Direction of causality not obvious but possible to find in- or out-of-sample forecasting power of labor costs for price inflation No obvious country or sector specific pattern 12

13 Cross correlations leads/lags leads/lags Figure 3: Cross-correlations between price inflation and labor cost inflation 13

14 Granger causality Figure 4: Recursive Granger causality test results (p-value) 14

15 Forecast Evaluation Unconditional: labor cost for price inflation Conditional: labor cost for price inflation 199Q1-2018Q1 1999Q1-2007Q4 2008Q1-2018Q1 199Q1-2018Q1 1999Q1-2007Q4 2008Q1-2018Q1 GERMANY steps total total total total total total FRANCE steps total total total total total total Note: For the unconditional forecast the table reports the ratio between the RMSEs of the forecast based on a VAR with three variables and a VAR without ULC. For the conditional forecast the table reports the ratio between the forecast conditional on the observed path of ULC and the one conditional on a constant path for ULC 15

16 Forecast Evaluation Unconditional: labor cost for price inflation Conditional: labor cost for price inflation 199Q1-2018Q1 1999Q1-2007Q4 2008Q1-2018Q1 199Q1-2018Q1 1999Q1-2007Q4 2008Q1-2018Q1 ITALY steps total total total total total total SPAIN steps total total total total total total Note: For the unconditional forecast the table reports the ratio between the RMSEs of the forecast based on a VAR with three variables and a VAR without ULC. For the conditional forecast the table reports the ratio between the forecast conditional on the observed path of ULC and 16 the one conditional on a constant path for ULC

17 Outline Data Unconditional evidence Conditional evidence Summary and conclusion 17

18 VAR Analysis Baseline VAR system contains three growth rate variables: real value added unit labor cost (in deviation from trend) value added deflator (in deviation from trend) Estimation period is: 1985Q1 to 2018Q1 The 16 VARs are estimated with 4 lags and Bayesian techniques Identification with Choleski Pass-through computed as a dynamic multiplier: ratio of cumulative IRF of price inflation and ULC inflation to a shock to ULC 18

19 Results The VAR-based correlations are as high as the unconditional ones A pass-through exists as in micro evidence but there is little evidence that it is higher in sectors with high labor share The pass-through depends on the level of inflation in all sectors and countries Results are robust when controlling for MP 19

20 VAR correlations with Choleski The table reports estimates of conditional correlations between ULC and price inflation computed from the Moving Average representation of a tri-variate VAR (Gali, 1999) 20

21 VAR pass-through with Choleski Figure 5: The chart s report the ratios of cumulative IRF of price inflation and ULC inflation to a shock to ULC 21

22 Labour shares Figure 6: Labor shares by sector 22

23 State-dependent pass-through Figure 7: The chart s report the ratios of cumulative IRF of price inflation and ULC inflation to a shock to ULC computed on samples where sector price inflation was above (red) or below (grey) historical average 23

24 Structural VAR Analysis Sign restrictions to identify demand and supply shocks Two counterfactual experiments: 1. Correlations between counterfactual labor cost and price inflations generated by demand or supply shocks 2. Counterfactual IRF of price inflation when assuming that ULC does not react to demand and to supply shocks (shut down the ULC amplification channel) 24

25 Results counterfactual #1 Demand shocks affect prices and labor costs in a similar manner, with prices leading labor costs Supply shocks affect prices and labor costs differently, with labor costs leading price inflation The correlation between labor cost and price inflation tends to be higher for demand than for supply shocks 25

26 ULC and price inflation generated by demand shocks Figure 8: Counterfactual ULC and price inflation obtained from a historical decomposition when all shocks but demand are shut down 26

27 ULC and price inflation generated by demand shocks Figure 9: Counterfactual ULC and price inflation obtained from a historical decomposition when all shocks but supply are shut down 27

28 Conditional correlations from a SVAR Note: Maximum correlation as computed with the SVAR. A positive lag for the maximum correlation suggests that labor costs are leading. 28

29 Results counterfactual #2 The pass-through of labor costs to prices is bigger and more significant when the economy is hit by a demand than by a supply shock The peaks of this pass-through tend to occur at a higher lag for demand-type shocks than for supply-type shocks and the pass-through is also more persistent Results are robust to a richer specification (only available for total economy) that allows to identify labor market shocks 29

30 Amplification due to ULC channel DE FR IT ES Demand shocks Supply shocks quarters quarters Total mfg constr serv Total mfg constr serv Total mfg constr serv Total mfg constr serv Note: blue shows where there is a statistically significant difference between the impulse responses to price inflation with and without the labor cost channel. The white diamond indicates the quarter 30 for maximum impact of the price inflation response.

31 SVAR: Robustness DE FR IT ES Total Total Total Total periods Demand shocks periods Supply shocks DE FR IT ES Total Total Total Total Labour supply shocks periods Wage mark-up shocks periods 31

32 Outline Data Unconditional evidence Conditional evidence Summary and conclusion 32

33 Summary The cost-push view of inflation has some support in the euro area data High (but imperfect) pass-through from costs to prices, in line with available firm-level evidence (Druant et al. 2009) The link between price and ULC is heterogeneous across countries and sectors, is time-varying and depends on high/low inflation The pass-through depends on the shocks hitting the economy and is higher for demand shocks than for supply shocks 33

34 Rationalizing results 1. Higher pass-through in France Prominent role of unions Informal wage indexation and strong second round effects 2. Medium-moderate pass-through in Italy Formal wage indexation at sectoral level and strong second round effects 3. Moderate pass-through in Germany Decentralization of wage formation and wage moderation Wage setting at sectoral level (distribution margins) 4. Low pass-through in Spain No important role of inflation developments for the determination of negotiated wages Negotiation more at firm than sector level with indexation reduced after

35 Rationalizing results 1. Causal relationship difficult to identify in the data Results depend on sample, state of the economy and combination of shocks hitting the economy over that sample and under that state 2. That pass-through is higher with high inflation is consistent with several stories Intertemporal smoothing of the profit path and monetary policy Uncertainty: the higher degree of uncertainty associated with high inflation regime induces firms to avoid buffering with margins Expectation theory: low inflation (persistence) cause low passthrough (Taylor, 2000) Search intensity of consumers (Head et al. 2007) 3. Why firms are more willing to increase prices with positive demand shocks Demand shocks increase the share of high income consumers with lower demand elasticity. This in turn raises firms ability and power to pass-through cost increases to prices (e.g. Dornbusch, 1987; Bergin and Feenstra, 2001) 35

36 Current implications Results support the current ECB forecast narrative (ECB 2018), which relies on the pick-up in wage growth for the one in underlying inflation Under the current circumstances of predominantly demand shocks (ECB 2018) the findings give us confidence that the expected increase in ULC will support underlying inflation in the euro area However, results also call for patience as, coming from a period of low inflation, this pass-through could be moderate at least until inflation stably reaches a sustained path 36

37 Background slides 37

38 Motivation and literature A better understanding of the signal that labor cost developments provide for the inflationary process is a key policy issue Underlying inflation is projected to rise on the back of the improving cyclical position of the economy and the related increase in wage growth. Sept 2018 ECB staff macroeconomic projections for the euro area The Committee forecasts a gradual pickup in domestic cost growth that would help keep inflation slightly above target two and three years ahead even as currency effects fade M. Saunders (20/4/2018), MPC member, Bank of England The inflation rate is projected to rise in line with wage increases. K. Iwata (31/1/2018), Deputy Governor, Bank of Japan BACK 38

39 Motivation and literature Economic theory consistent with both direction of causality Cost-push/Price markup New Keynesian theory wage increases in excess of productivity put upward pressure on prices, and wages are the exogenous variable determining the future direction of inflation (e.g. Faso and Huq, 1988) This view has mostly determined the past and current central bank monitoring of labor cost pressure Firms charge whatever the market will bear, regardless of their actual costs If the markets acceptance of higher prices is the dominant determinant of inflation, the cost-push model would have less validity (e.g. Banerji 2005) Presence of monetary policy can revert the causality (Mehra, 2000) The link between labor cost inflation and price inflation depends on price and wage rigidities and on the type of shock that hits the economy. Hence we should in fact expect this link to vary over time and across countries and sectors. BACK 39

40 Motivation and literature Empirical evidence inconclusive and mostly US-based No clear in-sample causal relation between wage and price inflation price inflation causes wage inflation ([Hu and Toussaint-Comeau, 2010],[Emery and Chang, 1996], [Sbordone, 2002], [Bidder, 2015]) none or double Granger causality ([Hess and Schweitzer, 2000], [Gordon, 1988] and[darrat, 1994]). Some leading properties of wage inflation found by [Banerji, 2005] and [Rissman, 1995]. Out-of-sample, wages do not add information to forecast inflation [Stock and Watson, 2008], [Knotek and Zaman, 2014] The link between wage and price inflation has weakened over time [Knotek and Zaman, 2014],[Peneva and Rudd, 2017] 40

41 Motivation and literature For Euro area scant empirical evidence that finds something Models including ULC improve the inflation forecasting performance [Dees and Gunter (2014)] In most euro area countries there is a plausible relationship between inflation and ULC growth [Tatierska 2010] At the firm level, wage and price changes feed into each other [WDN 2009] The pass-through of wage to price inflation is higher for demand than for supply shocks [ECB2018] BACK 41

42 Data and de-trending Figure 2a: Adjusted and unadjusted ULC (6 months prior) and price inflation in US BACK 42

43 VAR: Robust when controlling for MP Figure 5a: The chart s report the ratios of cumulative IRF of price inflation and ULC inflation to a shock to ULC in a VAR which controls also for a measure of monetary policy stance BACK 43

44 VAR with sign restrictions Figure 5b: The chart s report the ratios of cumulative IRF of price inflation and ULC inflation to a shock to ULC in a structural VAR identified with (alternative) sign restrictions 44

45 Decline in US pass-through? Peneva-Rudd data Peneva-Rudd detrended Figure 5c: The chart s report the time varying steady state ratios of cumulative IRF of price inflation and ULC inflation to a shock to ULC in a VAR identified with choleski using Peneva and Rudd (2017) data and approach. 45