Inflation Regimes and Monetary Policy Surprises in the EU

Inflation Regimes and Monetary Policy Surprises in the EU Tatjana Dahlhaus Danilo Leiva-Leon November 7, VERY PRELIMINARY AND INCOMPLETE Abstract This paper assesses the effect of monetary policy during periods of excessively low inflation using aggregate as well as country-specific data on inflation in the euro area. The monetary policy shock is measured by changes in Euribor future contracts around the ECB announcement dates. To obtain regime-dependent responses of inflation we employ a local projection model with Markov-switching. After documenting the price puzzle in a linear framework, we show that a positive monetary policy surprise (comparable to a monetary tightening) decreases prices in the excessively low inflation regime but increases prices in the normal inflation regime. Differences in inflation responses across regimes may be related to a smaller information gap about inflation between the central bank and the public in the low inflation regime. JEL classification: Keywords: Inflation regimes; monetary policy surprises; price puzzle; Markovswitching; Local Projection model. The views expressed in this paper are those of the authors and do not represent the views of the Bank of Canada. Corresponding author. Bank of Canada, International Economic Analysis Department, Laurier Avenue West, Ottawa, ON KA G9, Canada. tel: + (6) 78 76; email: dahl@bankofcanada.ca Bank of Canada, International Economic Analysis Department, Laurier Avenue West, Ottawa, ON KA G9, Canada. tel: + (6) 78 77; email: leiv@bankofcanada.ca

Introduction Over the past years, the Euro Area (EA) has seen a pronounced decline in inflation rates. Inflation stood at three percent in late, and has dropped since then to less than half percent in mid-. The risk of a prolonged period of excessively low inflation imposes challenges on the European Central Bank (ECB) and raises an important question: How does monetary policy affect inflation during times of excessively low inflation? This paper sheds light on this question and assesses the effect of monetary policy during periods of excessively low inflation using aggregate as well as country-specific data on inflation in the EA. In particular, we use changes in Euribor future contracts around the ECB announcement dates to measure a monetary policy surprise or shock. We then study the regimedependent response of inflation to ECB monetary policy surprises in periods of normal inflation and periods of excessively low inflation. The economic literature has studied the monetary policy transmission mechanism and its effect on inflation intensively. In a linear world, i.e., the inflation response is not considered to depend on the level of inflation, economic theory predicts that a monetary contraction reduces nominal aggregate demand, lowering output and prices. However, empirical studies analyzing the monetary policy transmission mechanism often find the opposite result, a positive response of prices after a monetary tightening, the so-called price puzzle. The price puzzle has first been documented by Sims (99) for the United States and subsequently found in the literature for several countries including EA countries (see, for example, Favero and Marcellino () and Boivon, Giannoni, Mojon (9)). From an empirical perspective, a priori it is not clear what response of inflation one would expect either in a linear or a regime-dependent environment. To study the effect of monetary policy surprises on inflation, we use a Local Projection Model (LPM). We model inflation as a Markov-switching

process in the LPM in order to obtain regime-dependent inflation responses. After documenting the price puzzle for aggregate EA data in a linear framework, this paper shows that a positive monetary policy surprise (comparable to a monetary tightening) decreases prices in the excessively low inflation regime but prices respond positively in the normal inflation regime. The price puzzle, thus, disappears once we look only at episodes of prolonged low inflation. Further, we obtain a similar result if we consider EA countries separately. There are several potential explanations of our results related to the source of the price puzzle. One source that has been identified by the literature is asymmetric information between the central bank, i.e., ECB, and the public (see Romer and Romer (), Tas ()). In the presence of asymmetric information, if the ECB tightens policy, market participants may infer that the central bank has unfavourable information about the likely behaviour of inflation, and therefore they revise their expectations of inflation upwards. Consequently, one could infer that if the information gap between the central bank and the public vanishes, the price puzzle disappears. Looking at Google searches it seems that during episodes of excessively low inflation, the public s information set on inflation and particularly on the risks of deflation is bigger. Thus, asymmetric information might be reduced yielding the price puzzle to disappear. Other potential explanations (that need to be studied further) include the inflation predictability of output as well as the configuration with fiscal policy during episodes of low and normal inflation. The remainder of this paper is structured as follows: First, we define and describe monetary policy surprises in the EA. Second, we present the linear empirical framework and document the price puzzle. Third, we introduce the Note that the information gap between the ECB and the public might decrease for two reasons. The public itself might be gathering more information about inflation and/or the central bank s communication is clearer.

regime-dependent LPM and report the effects of monetary policy surprises during episodes of excessively low and normal inflation. Then, we discuss potential sources of the differences in inflation responses across regimes. Section 6 concludes. Monetary Policy Surprises in the EU We measure monetary policy by the impact of ECB policy announcements. We use daily data of Euribor future contracts around ECB interest rate announcements to obtain their surprise component. Following Kuttner (), we estimate the monetary policy surprise by comparing the end-of-day price on the day before the meeting with that on the day of the meeting. Looking only at the surprise component provides an exogenous measure of monetary policy as required by the LPM since news about other macroeconomic indicators are usually not released at the same day as the policy announcement. I contrast to Federal funds future contracts which have monthly maturities, Euribor future contracts exists for a range of quarterly maturities. That is, Euribor future contracts mature months ( quarter), 6 months ( quarters), 9 months ( quarters) ahead, etc. Hence, the simplest monetary policy surprise one could construct is based on the futures rate for the current quarter. However, we focus on a range of maturities for the following reason. As the policy rate in the EA is close to the zero lower bound, surprises based on short horizons may have little movements in the past years. Nowadays, during policy announcements the ECB can influence expectations about the policy rate at longer horizon using for example forward guidance. Therefore, we follow Barakchian and Crowe () and extract the first principal component of the Euribor future contracts with -month to -year maturities. To check for robustness of our results we also use Our analysis considers scheduled ECB announcement meetings. Note that Barakchian and Crowe () focus on contracts for the current month and

surprise for different maturities separately rather than their first common factor. Linear Effects. Local Projection Model In this section we focus on assessing, under a linear context, the effects of monetary policy surprises on EA inflation. Following Jordà () and Owyang and Zubairy (), we use a local projection model to calculate impulse responses. Specifically, we estimate a set of the following parsimonious models for each horizon h as follows: π t+h = φ h + φ h π t + φ h y t + φ h MP t + ε t, ε t N(, σ ), () where π and y refer to annualized inflation and monthly industrial production growth, respectively. The monetary policy surprise shock is denoted by MP, which is obtained by extracting the first principal component of Euribor future contracts with horizons from until quarters ahead. We employ monthly data starting in January until April. The model is estimated by Bayesian methods, i.e., a Gibbs sampler using a noninformative Normal-Gamma prior.. Empirical Results The effect of monetary policy surprises on EA inflation is shown in Figure. Note that positive surprises are comparable to a monetary tightening and negative surprises to expansionary monetary policy. Accordingly, the results indicate that positive monetary policy surprises have a positive and up to months ahead. Specifically, we use a zero vector and an identity matrix as hyper-parameters for the vector of means and the variance-covariance matrix, respectively.

significant effect on aggregate EA inflation. This result is consistent with the so-called price puzzle, since a surprise increase is expected to lower price inflation from standard economic theory. To assess if that is a consistent pattern across EA countries or whether there is heterogeneity in the responses to surprises, we estimate Equation () for 6 euro area economies separately. We use the corresponding inflation rate and industrial production growth for each country, but we use the measure of ECB monetary policy surprises across all models. The results are reported in Figure, showing a robust pattern of positive and significant responsiveness of inflation to a positive monetary policy surprise across various EA countries. Inflation responds strongest to monetary policy surprises in countries such as Germany, France, Italy and Spain. Further, notice that the countries showing the lowest responsiveness to monetary policy surprises are those that have faced severe debt problems in the recent years, i.e. Greece, Ireland and Portugal. Overall, these results represent strong evidence of the price puzzle in the Euro Area. Regime-dependent Effects. Regime-Switching Local Projection Model Since the implementation of the Euro, inflation in the euro area has experimented periods of stability, defined as inflation close to the target of per cent. However, inflation has also experimented periods of significant downturns, reaching even periods of deflation. Therefore, it is natural to ask whether the effect of monetary policy surprises may vary across regimes of excessively low and normal inflation rates. To assess the regime-dependent effects of monetary policy surprises on inflation we estimate the model in Equation () with two modifications. First, we allow the coefficients of the regression to be regime-dependent, i.e. 6

to be driven by a latent variables S t that follows a first order Markov chain and that takes the value of when the economy is under a normal inflation regime, and the value of when the economy faces a low inflation regime. Second, as documented in Camacho and Perez-Quiros (7), the positive autocorrelation in relevant macroeconomic variables can be better captured by shifts between regimes rather than by the standard view of autoregressive coefficients. Therefore, we replace the autocorrelation term in Equation () by a Markov-switching mean, yielding the following Regime-Switching Local Projection model (RSLPM) : π t+h = φ h (S t ) + φ h (S t )y t + φ h (S t )r t + ε t, ε t N(, σ ), () p(s t = j S t = i, S t = h,...) = p(s t = j S t = i). () We first estimate the RSLPM at horizon h = to obtain the different inflation regimes that are plotted in Figure. The results indicate two clearly defined episodes of excessively low inflation in the EA. The first episode of low inflation occurs between December 8 and March, and the second one starts in April and continues until the end of the sample (April ). Similar to the linear case, to obtain the regime-dependent responses we re-estimate the RSLPM for each horizon h. However, inferences on the latent variable S t are performed only at horizon h =. Such inferences are then used to construct an indicator variable for the regimes of inflation. For horizons h, instead of producing new inferences on S t, we directly use the indicator variable. In other words, inflation regimes are estimated at horizon h =, and that information is carried out through horizons h. We estimate the RSLPM using Bayesian methods implementing a Gibbs sampler with a Normal-Gamma prior. For horizon h = we use a zero vector for the hyper-parameters of the vector of coefficients (uninformative 7

prior). However, for h, we use the posterior estimates obtained from the previous estimation as new hyper-parameters. We use these priors in order to smooth responses as the LPM can produce quite volatile responses.. Empirical Results The regime-dependent effects of monetary policy surprises on euro area inflation are plotted in Figure. The results show that during episodes of normal inflation, monetary policy surprises have a positive effect on euro area inflation, consistent with the price puzzle. This is not surprising as most of the time the model is in the normal inflation regime. However, during times of excessively low inflation, monetary policy surprises have a negative and significant effect on aggregate EA inflation, a behaviour that is consistent with standard economic theory. In the next section, we discuss potential economic reasons for the differences in inflation responses across regimes. To assess the robustness of our result, we perform the same analysis across euro area countries. First, we estimate the RSLPM at horizon h = for each country to estimate the regimes of inflation. The results are reported in Figure, showing two main features. First, there is a substantial heterogeneity of inflation regimes across countries. Second, despite such heterogeneity, the inflation regimes of EA countries overlap with ones obtained from aggregate data. The only exception is Iceland during 9, however, this country has experimented inflation dynamics significantly different from the other euro area countries, reaching inflation rates higher than 7 per cent in 9. The effect of monetary policy surprises on inflation in each country and across inflation regimes is reported in Figure 6. The responses indicate a similar pattern as in the case of using aggregate euro area data, indicating robustness of our results. During normal inflation regimes, the price puzzle 8

holds for all the countries, with the exception of Iceland and Greece. In these countries the ECB monetary policy surprise seem to have no significant effect on inflation. However, during episodes of excessively low inflation, positive monetary surprises decrease inflation in nearly all EA countries. Why are inflation responses different across regimes? There are several potential explanations for the difference in the inflation response across regimes that are related to the source of the price puzzle. One source that has been identified by the literature is asymmetric information between the central bank, i.e., ECB, and the public (see Romer and Romer (), Tas ()). In the presence of asymmetric information, if the ECB tightens policy, market participants may infer that the central bank has unfavourable information about the likely behaviour of inflation, and therefore they revise their expectations of inflation upwards. Consequently, one could infer that if the information gap between the central bank and the public vanishes, the price puzzle disappears. To provide a first look at this hypothesis, we assess the relation between the information gathered by the public and the two inflation regimes. Figures 7 and 8 plot the excessively low inflation regime with the number of google searches of the word deflation. As one would expect Google searches of the word deflation peak during the episodes of low inflation. Therefore, during episodes of excessively low inflation, the public gathers more information about inflation or in that case deflation. This may suggest that the information gap between the ECB and the public vanishes in the low inflation regime and thus the price puzzle disappears. Other potential explanations (that need to be studied further) include the inflation predictability of output as well as the configuration with fiscal policy during episodes of low and normal inflation. 9

6 Conclusion

References Barakchian, S. M. and Crowe, C. (). Monetary policy matters: Evidence from new shocks data, Journal of Monetary Economics 6(8): 9 966. Camacho, M. and Perez-Quiros, G. (7). Jump-and-Rest Effect of U.S. Business Cycles, Studies in Nonlinear Dynamics & Econometrics (): 9. Jordà, O. (). Estimation and Inference of Impulse Responses by Local Projections, American Economic Review 9(): 6 8. Kuttner, K. N. (). Monetary policy surprises and interest rates: Evidence from the Fed funds futures market, Journal of Monetary Economics 7():. Owyang, Michael T., V. A. R. and Zubairy, S. (). Are Government Spending Multipliers Greater during Periods of Slack? Evidence from Twentieth-Century Historical Data, American Economic Review: Papers and Proceedings (): 9. Romer, D. H. and Romer, C. D. (). Federal Reserve Information and the Behavior of Interest Rates, American Economic Review 9(): 9 7. Sims, C. A. (99). Interpreting the macroeconomic time series facts : The effects of monetary policy, European Economic Review 6(): 97. Tas, B. K. O. (). An explanation for the price puzzle: Asymmetric information and expectation dynamics, Journal of Macroeconomics (): 9 7.

Figure : Effect of Monetary Policy Surprises on Inflation in the Euro Area.... Note: The solid black line represents the median responses in euro area inflation to market s surprises in the ECB s interest rate obtained with a linear model. The dashed blue lines correspond to the.7 and quantiles. The horizontal axis denotes the horizon in months.

Figure : Effect of Monetary Policy Surprises on Inflation in Euro Area Countries.6... 8 6.6. 8....6... 8 6 AUSTRIA BELGIUM ESTONIA FINLAND.6 8.6. 6. 8 8.6. 8 - - - FRANCE GERMANY GREECE ICELAND - -.6. 8 7 6 - - - IRELAND ITALY LUXEMBOURG NETHERLANDS. 8 6... 8 6 -... 8 6 -... PORTUGAL SLOVAKIA SLOVENIA SPAIN Note: The solid black line represents the median responses in the corresponding euro area country inflation to market s surprises in the ECB s interest rate obtained with a linear model. The dashed blue lines correspond to the.7 and quantiles. The horizontal axis denotes the horizon in months.

Figure : Low Inflation Regimes in the Euro Area -..8.6 6 8 Note: The blue dashed line is the euro are inflation and the solid red line is the probability of low inflation regime. Figure : Regime Dependent Effect of Monetary Policy Surprises on Inflation in the Euro Area.. -. - -. - LOW NORMAL: EURO AREA Note: The solid blue (red) line represents the median responses in euro area inflation to market s surprises in the ECB s interest rate under normal (low) inflation regime. The associated dashed lines correspond to the.7 and quantiles. The horizontal axis denotes the horizon in months.

Figure : Low Inflation Regimes in Euro Area Countries - 6 8..8.6 6 - - 6 8..8.6 8 6 - - 6 8..8.6 - - 6 8..8.6 AUSTRIA BELGIUM ESTONIA FINLAND - 6 8..8.6 - - 6 8..8.6 - - - 6 6 8..8.6 6 8 6 8..8.6 FRANCE GERMANY GREECE ICELAND 8 6 - - -6-8 6 8..8.6 6 8..8.6-6 8..8.6 6 8..8.6 IRELAND ITALY LUXEMBOURG NETHERLANDS 6 - - 6 8..8.6 6 8-6 8..8.6 8 6-6 8..8.6 6 - - 6 8..8.6 PORTUGAL SLOVAKIA SLOVENIA SPAIN Note: For each chart, the blue dashed lines represent the inflation in the corresponding euro area country, the solid red lines are the associated probability of low inflation regime in that country. The shaded bars represent low inflation regimes in the euro area (based on Figure ).

Figure 6: Regime Dependent Effect of Monetary Policy Surprises on Inflation in Euro Area Countries.. -. - -. -.6 - - -.6.8.6 - - -.6.. -. - -. - LOW NORMAL: AUSTRIA LOW NORMAL: BELGIUM LOW NORMAL: ESTONIA LOW NORMAL: FINLAND..8....6 - -. - -. - - - -.6 -. - -.6 -.8 LOW NORMAL: FRANCE LOW NORMAL: GERMANY LOW NORMAL: GREECE LOW NORMAL: ICELAND.... -. -. -.. -. - -.... -. - -... -. - -. - LOW NORMAL: IRELAND LOW NORMAL: ITALY LOW NORMAL: LUXEMBOURG LOW NORMAL: NETHERLANDS.8.6 - - -.6...8.6 - -..8.6 - -.6 - - -.6 LOW NORMAL: PORTUGAL LOW NORMAL: SLOVAKIA LOW NORMAL: SLOVENIA LOW NORMAL: SPAIN Note: The solid blue (red) line represents the median responses in the corresponding euro area country inflation to market s surprises in the ECB s interest rate under normal (low) inflation regime. The associated dashed lines correspond to the.7 and quantiles. The horizontal axis denotes the horizon in months. 6

Figure 7: Google Searches for Deflation in Euro Area Countries AUSTRIA BELGIUM FINLAND 8 8 8 6 6 6 6 8 6 8 6 8 FRANCE GERMANY ITALY 6 8 6 8 6 8 NETHERLANDS PORTUGAL SPAIN 8 8 8 6 6 6 6 8 6 8 6 8 Note: The charts plot the number of Google searches for the word Deflation translated to the language of the corresponding country. Shaded bars represents low inflation regimes in the euro area. 7

Figure 8: Google Searches for Deflation in the Euro Area EURO AREA 8 6 6 7 8 9 Note: The charts plot the number of Google searches for the word Deflation in the euro area. This measures is constructed based on the information in plotted Figure 7. Shaded bars represents low inflation regimes in the euro area. 8