Disentangling Qualitative and Quantitative Central Bank ... - OFCE

Disentangling Qualitative and Quantitative Central Bank Communication

Paul Hubert* OFCE – SciencesPo

March 2012 Preliminary version

Abstract The effects of ECB inflation projections and Governing Council members’ speeches on private inflation forecasts are identified through an Instrumental-Variables estimation using a Principal Component Analysis to estimate valid instruments. We find that ECB projections has an effect on private current year forecasts, while ECB speeches impact private next year forecasts along with the ECB rate. Moreover, ECB projections are found to have asymmetric effects when they are interacted with the ECB rate or ECB qualitative communication. Both qualitative and quantitative central bank communication appear to be a crucial element of the conduct of monetary policy. JEL classification: E52, E58 Keywords: European Central Bank, Monetary Policy, Central Bank Communication, ECB Projections, Private Inflation Forecasts.

* I would like to thank particularly Michael Ehrmann for sharing his dataset and for numerous helpful discussions. I am also grateful to Andre Romahn, Alexander Jung, Patrick Hürtgen, Harun Mirza and Jirka Slacalek as well as participants at seminar at the ECB for helpful comments, and Michael Lamla for sharing the KOF index. This research was conducted while the author was visiting the Monetary Policy Strategy Division at the European Central Bank. Email: [email protected]. Address: OFCE, 69 quai d’Orsay, 75340 Paris cedex 07, France.

1

1. Introduction Expectations matters in determining current and future macroeconomic outcomes. Hence, the management of private expectations has become a central feature of monetary theory, as emphasized by Woodford (2005). This raises to the forefront the communication of central banks to signal what they intend to do, but does not diminish the importance of actions by the monetary authorities because “actions speak louder than words”: central bank should thus take the actions consistent with how they want private expectations to be influenced. The question whether central bank communication (for a comprehensive overview of the literature, see Blinder et al., 2008) has been successful to affect financial markets or to help explain or predict interest rate decisions has given rise to an abundant empirical literature. The part of this literature focusing on the ECB communication has been surveyed by De Haan (2008). Many studies have already coded ECB communications and report evidence that they influence financial markets in the intended direction and improve the predictability of ECB interest rate decisions. Two ways have been followed: several authors including Connolly and Kohler (2004), Andersson et al. (2006), Berger et al. (2006), Brand et al. (2006), Musard-Gies (2006), Andersson (2007), Ehrmann and Fratzscher (2009), Gerlach (2007), Heinemann and Ullrich (2007), Rosa and Verga (2007) and Sturm and De Haan (2011) have constructed or used indicators based on coding words in the ECB Introductory Statement or the Monthly Bulletin, while some others, for instance Ehrmann and Fratzscher (2007) and Jansen and De Haan (2009) have coded indicators of the tone in all speeches and statements made by policymakers during and between Governing Council meetings. In general, most of the literature has focused on the central bank qualitative communication: either formal statements and reports or more informal speeches and interviews. Moreover, most of the literature assesses the impact of central bank communication on financial markets or on the predictability of interest rate decisions. Only few authors, Fujiwara (2005) and Ehrmann, Eijffinger and Fratzscher (2009), have up to now focused on the effects of a) quantitative communication, i.e. the publication of central bank macroeconomic forecasts, on the dispersion of b) private inflation expectations; but without taking into account qualitative communication. Finally, the effects of the interaction of qualitative and quantitative central bank communication to influence private inflation forecasts have not been evaluated so far. This paper extends the analysis in the existing literature in two directions. First, it assesses whether ECB projections and ECB Governing council members’ speeches considered together impact the level of private inflation expectations. To our knowledge, no studies compare the influence of ECB qualitative and quantitative communication, while one might argue that the effects of one are a proxy for the effects of the other, or at the opposite that their effects are differentiated and respond to two different purposes. Second, it evaluates the interaction of these communication devices together and with the ECB rate. Indeed, one might expect that ECB projections have more impact on private inflation expectations if they are explained through speeches or consistent with the ECB rate decisions. One could expect that ECB qualitative communication (speeches about the future policy stance) has a negative effect on private inflation forecasts because it signals future decisions, while one could also expect that ECB qualitative communication has a positive effect on private inflation forecasts because it signals inflationary pressures. The same line of reasoning applies for ECB projections and the ECB rate. Extending the concept of monetary policy tools to quantitative and qualitative signals in addition to interest rate decisions, this paper aims at highlighting the interpretation made by private agents to ECB tools; in other words, it intends to identify the potentially differentiated and interacted effects of ECB tools on private inflation expectations. The closest paper to this study is Andersson et al. (2006) who assess how 2

interest rate changes, inflation reports and speeches of the Swedish central bank affect the term structure of interest rates in Sweden. The contribution of this paper is therefore to provide original empirical evidence on the individual and interacted effects of ECB projections and ECB qualitative communication on private inflation forecasts. An original monthly index encompassing all speeches of the members of the ECB Governing Council is constructed by coding the stance of each communication between June 2004 and June 2011 following the methodology of Ehrmann and Fratzscher (2007). The effects of ECB tools on private inflation forecasts are identified through Instrumental Variables (IV). In addition to lags of endogenous variables, Principal Components Analysis (PCA) is used to generate satisfying valid instruments according to Bai and Ng (2010) and Kapetanios and Marcellino (2010), and thus overcome the weak identification issue. Macroeconomic controls including core inflation, the output gap, the credit growth rate, the oil price growth rate, M3, wages growth rate and a measure of the macroeconomic uncertainty, are also added to the regression model. We specifically focus on private inflation expectations as the dependent variable of this analysis since price stability corresponds to the main objective and mandate of the ECB. In addition, it is worth specifying that both ECB projections and Consensus Forecasts (CF) used this paper are fixed-event forecasts for current and next years. One inconvenient of this type of forecasts compared to fixed-horizon forecasts is their decreasing forecasting horizon. However, one advantage of using these fixed-event forecasts, in addition to consider data as they are really published and to avoid data transformation to convert them in fixed-horizon, is that private current year forecasts are not directly affected by policy decisions, as interest rate changes take time to affect the real economy. It is particularly interesting to focus on private current year forecasts when the ECB rate has no control over inflation as it enables to assess the effect of communication as a tool to affect private expectations and to shorten the transmission lags of monetary policy. The main findings of the paper are the following. The main determinants of private inflation forecasts for current year are lagged private forecasts, ECB inflation projections, the output gap and oil prices. ECB current year inflation projections positively influence private inflation forecasts, as well as increasing ECB projections and ECB projections higher than private forecasts. Neither the ECB qualitative communication nor the ECB rate appears significant. Concerning next year forecasts, the main determinants of private inflation forecasts are lagged private forecasts, ECB qualitative communication, the ECB rate, the output gap, credit growth, oil prices and macroeconomic uncertainty. It appears that speeches increase inflation expectations. While ECB qualitative communication captures the future policy decisions on the ECB rate, its level appears to signal the inflation risk to private agents. Hence, ECB projections rather seem to be a short-term tool to affect private inflation forecasts while the ECB qualitative communication which captures the future stance of monetary policy seems to act as a longer-term tool for private inflation forecasts. Interacting central bank communication types and action, it appears that ECB projections have more influence on current year private forecasts when ECB qualitative communication is more hawkish or the ECB rate is higher. However, changes in ECB projections have a negative effect on private forecasts when interacted with ECB qualitative communication, and this negative effect is more pronounced when the stance of communication is more hawkish. In the same vein, changes in ECB projections and ECB projections higher than CF forecasts have a negative effect on private forecasts when the ECB rate is high, whereas the effect is positive when the ECB rate is low. The pattern is quasi-similar for next year 3

forecasts: ECB projections, changes in the stance of qualitative communication and ECB projections higher than CF forecasts have a negative effect on private inflation forecasts when the ECB rate is high whereas no effect when the ECB rate is low. These outcomes are consistent with Andersson et al. (2006) who find that Riksbank inflation forecasts affect interest rates with a maturity of one year or less while speeches are found to impact the longer end of the term structure. They also find that that the effects of speeches on the Swedish term structure are higher for interest rate increases than decreases. They are also in line with the main argument of Gürkaynak, Sack and Swanson (2005) that both policy actions and statements have different effects on asset prices. The key argument of the present paper is that both communication types have differentiated effects and that the optimal design of one communication device should depend on ECB policy decisions and on the other communication device. Indeed, this paper suggests that both qualitative and quantitative central bank communication are a crucial part of the conduct of monetary policy as stressed by Guthrie and Wright (2000). ECB projections should be used to explain the underlying reasons for ECB rate decisions (a given decision has been taken because of a given projection) and not to explain the expected effects of ECB rate decisions (a given decision will produce a given projection). Indeed, publishing increasing ECB inflation projections or ECB projections above CF forecasts when the ECB rate is high (so referring to the case: a given decision has been taken because of a given projection) has a negative effect on private inflation forecasts. One mechanism that may explain this effect is that the central bank appears credible when it justifies and explains its decisions. At the opposite, publishing increasing ECB inflation projections when the ECB rate is low would have a positive effect on private inflation forecasts, possibly because private agents would not understand then why the central bank is not increasing its interest rate. Another interesting policy implication refers to the stance of qualitative communication. Since increases in the stance of communication have a negative effect when the ECB rate is high and no effect when the ECB rate is low, policymakers should not hesitate to make hawkish announcements. The rest of the paper is organized as follows. Section 2 presents the data. In sections 3 and 4 respectively, the individual and interacted effects of ECB projections and ECB qualitative communication on private inflation forecasts are assessed. Section 5 concludes.

2. Data This section describes the variables used to estimate the effects of the ECB qualitative communication and the publication of ECB projections on private forecasts along with controls that typically drive expectations formation. 2.1. The ECB Qualitative Communication Index Speeches, interviews and testimonies related to monetary policy made by the individual committee members are measured by a monthly index in the vein of the one of Ehrmann and Fratzscher (2007). This index covers the 6 Executive Board members of the ECB and the governors of the national central banks of the Eurosystem. It starts in June 2004 and ends in June 2011 to match the publication of ECB inflation projections. For this time period, Reuters News, a standard newswire service is used to gather all reports about forward-looking policy

4

statements. The focus is specifically on the future monetary policy inclination and explanations or clarifications of past decisions are not taken into account. Following Kohn and Sack (2004) and Ehrmann and Fratzscher (2007), the classification is kept as straightforward and general as possible. The search commands used are governing council, member, president, or vice president along with interest rate or monetary policy. Moreover, only the first report in Reuters News which directly follows the statement and is rather descriptive is considered, the updates or analyses are not included. Each statement is then classified into three categories: those that give an inclination of tighter monetary policy, no change, or lower interest rates:

MPt =

+1 0 -1

tighter inclination no change easing inclination

For each month between June 2004 and June 2011, four variables are then available. MP_NBt is the number of statements during the given month. MP_STt provides the average inclination of the statements, in other words the policy stance of the Governing council members communication and is comprised between [-1; 1]. MP_INTt displays the intensity of the communication which is the number of statements times the stance. Finally, MP_DISPt measures the dispersion of statements and is computed as the standard deviation of the stance of all statements during a given month. Figure 1 plots the main ECB communication variables along with the ECB interest rate, while Table 1 provides some descriptive statistics. Some interesting facts appear from the preceding figure and table. There are much more statements that have a tightening inclination than neutral or easing ones. 56.1% of the statements made between June 2004 and June 2011 have a hawkish inclination and this makes sense as interest rates were increasing or high for half of the sample period. It is nevertheless interesting to note that the ECB signals much more interest rate hikes than decreases. This is in line with Jansen and De Haan (2009) for the ECB or Hayo and Neuenkirch (2010) for the Federal Reserve who find these central banks seems cautious about mentioning too much rate cuts. It can also be noted from Figure 1 that ECB projections, the ECB rate and ECB qualitative communication, either MP_ST or KOF, are consistent with each other. This classification methodology is usually referred to as “content analysis” because of the systematic analysis of the content of a message (Holsti, 1969) and it is worth noting that this work is by nature judgmental and subjective. In particular, the choice has been made to focus on forward-looking conventional and unconventional monetary policy announcements with a possible effect on price stability and inflation, and not on policies providing liquidity to money markets and banks. We believe that it is important to differentiate policies aiming at price stability and financial stability following the usual segmentation of monetary policy mandates. Moreover, most of liquidity interventions were realized in the meantime than announced and no forward-looking communication were made for them. Another possible caveat is that Reuters News may have not reported or misinterpreted some statements. The present index is therefore compared to the KOF Monetary Policy Communicator for the Euro Area which provides a quantitative measure of the ECB communication with a special focus on forward-looking statements concerning price stability (see Conrad and Lamla (2010) or the KOF website1 for more details) and is available on the same time span than the present 1

http://www.kof.ethz.ch/en/indicators/monetary-policy-communicator/

5

index. It enables to assess the robustness and relevance of the latter. However, the KOF index translates the ECB president’s statement concerning risks to price stability as made during the monthly press conference (and only this specific Governing council day) as well as ECB projections into a unique common index. In contrast, the index constructed in this paper encompasses all qualitative communication of each month and focuses specifically on speeches and statements in contrast with ECB projections. Table 1 also shows the correlation matrix between the KOF index and all indices of this study. 2.2 The ECB Projections The ECB/Eurosystem staff macroeconomic projections2 for the euro area are produced biannually since December 2000, and quarterly since June 2004 with a special emphasis on their disclosure to the public. They are usually published during the first week of March, June, September and December and are presented as ranges for both HICP (the Harmonized Index for Consumer Prices) and real GDP. These projections are published as average annual percentage changes and are for current and next years. These fixed-event projections might have seasonal effects as the forecasting horizon decreases from quarter to quarter: one might for instance suppose that the effects of ECB inflation projections on private ones are stronger in the beginning of each year and smaller at the end when much more information is known on actual variables. In a companion paper, Hubert (2012) constructs one-year-ahead fixedhorizon3 forecasts as a weighted average of fixed-event forecasts based on the number of quarters forecasted in both the current and next years following Dovern, Fritsche and Slacalek (2011), and shows that this does not alter the estimated results. The ranges are based on twice the mean absolute projection error of historical projection errors to reflect uncertainty. As common for the FOMC (Federal Open Market Committee at the Federal Reserve) forecasts, the midpoint of the range is used to represent the ECB projections. The underlying scenarios for interest rates and commodity prices were until 2006Q1 that these variables remain constant over the projection horizon; since 2006Q2 they are based on market expectations derived from future rates4. Finally, the sample considered here starts in June 2004 when ECB projections became quarterly so as to combine the need for highfrequency data to measure qualitative communication and the relative low-frequency of publication of ECB projections. In addition, we interpolated quarterly ECB projections to monthly frequency by filling the gaps of the two months following their disclosure to the public with the value of the last projection published. This assumption introduces a bias against ECB projections which remain constant during two months whatever the macroeconomic or policy developments. 2.3 Private Forecasts See ECB (2001, 2009) for more details. An advantage of these fixed-horizon forecasts to check the robustness of the fixed-event estimates is that there is a break in the forecasts series as the current year Q1 forecast estimate the underlying variable for the subsequent year compared to the preceding Q4 forecast. One argument to overcome the effect of this break is that we are interested in the signalling content of the projections which is not calendar-year based, and not in their actual accuracy. In other words, if the ECB decides to disclose a policy signal, it should move both current and next year projections in the same direction. Since we use monthly frequency to assess qualitative communication effects, and therefore interpolate quarterly ECB projections to a monthly frequency, constructing fixed-horizon projections would either make ECB projections artificially change every month along with the number of months in current and next years, or mismatch the horizons when comparing with private forecasts if filling the following 2-month gap after the publication of ECB projections with the constructed one-year-ahead forecast. 4 We have checked that these technical assumptions for constructing ECB forecasts have no impact on the results. Although it should matter whether one assume constant interest rates or market-expected interest rates, estimates on the whole sample and on the post-2006Q2 provide similar effects. Results are available upon request. 2 3

6

The private inflation forecasts come from Consensus Economics Inc. The Consensus Forecasts (CF) is a monthly survey of quantitative predictions of private sector forecasters, with an average of 30 institutional respondents, for about fifteen macroeconomic variables including HICP for the euro area, measured as average annual percentage change for current and next years. Surveys are collected at the very early of each month and published in average around the 10th of each month. The overall sample starts in June 2004, ends in June 2011 and this gives 85 monthly observations. 2.4 Other variables To ensure that the effects identified are those of the qualitative and quantitative ECB communication, various controls for the macroeconomic environment which may impact the private expectations formation are added. The ECB interest rate (the Main Refinancing Operations interest rate) enables to check whether ECB communication may be a proxy for ECB decisions or whether ECB communication really adds some value to private inflation expectations’ formation. Indeed, it has to be noted that the ECB qualitative communication variable may measure the “procyclical” effect of the speech (e.g. when central bankers say they fear high future inflation, then private inflation expectations should increase), and in the meantime may capture the “countercyclical” effect of the same speech (e.g. when central bankers say they will increase interest rates, private inflation expectations should decrease). The presence of ECB rate should thus control for the countercyclical effect of ECB qualitative communication while the presence of ECB projections should thus control for the procyclical effect of ECB qualitative communication. Core HICP (the Harmonised Index of Consumer Prices considering all-items excluding energy and unprocessed food), the output gap (the HP-filtered monthly interpolated real GDP), the credit growth rate (to euro area residents other than governments), the oil price growth rate (based on the Brent crude oil spot price) and a measure of the macroeconomic uncertainty (the standard deviation of individual point estimate responses to Consensus Forecasts’ surveys corrected for the decreasing forecasting horizon during each year) are included in the analysis as control variables. M3 and wages (the indicator of negotiated wage rates published by the ECB) growth rates have also been tested without improving the regression output.

3. Do ECB Speeches and Projections influence Private Agents? This section is articulated in two parts. First, the empirical model and the estimation method are described along with the construction of instruments needed to identify the causal effects of ECB communication on private expectations. Second, the independent effects of each communication type are assessed. 3.1 Empirical Model The following empirical model has for objective to estimate the effects of both central bank qualitative communication and projections on private inflation forecasts. Because private forecasts are constructed and published at the beginning of each month, the assumption is made that they are produced based on the information set from the previous month. In order to disentangle the effects of qualitative and quantitative communication on private forecasts along with the effects of the control variables and beyond the information contained in the previous private forecast, private forecasts at date t are regressed on all variables of interest at date t-1. In other words, one may suppose that the basic expectations’ formation process

7

follows an AR(1) process and the analysis consists of adding variables to the AR(1) model in order to assess their effects beyond lagged private inflation forecasts which explain approximately 85% of the variance of private inflation forecasts. One would want to include a macroeconomic news component in the equation, but this seems impractical because of the format of private forecasts. Usually the news variable encompassing the information set released between t-1 and t is computed as the difference between the forecast of a given variable (inflation) in t-1 and the actual value of the given variable in t, and this is not possible with CF forecasts as the monthly forecasts are not for the next month horizon. One may nevertheless argue that the news component is small as the ECB qualitative communication variable encompasses all speeches during a given month and all macroeconomic variables are generated at the end of this given month, while private forecasts are formed at the early beginning of the following month. In addition, since ECB projections are published at the beginning of the month together with ECB rate decisions, and one might consider that potential news released until private forecasts are formed next month are orthogonal to these ECB variables, it might be argued that their estimated effect is not biased by this omitted variable. We therefore implicitly assume that price and monetary policy news which affect expected future inflation (Beechey and Wright, 2009) are comprised in t-1 variables and that no news is published between the end of a month and the early beginning of the following month when private agents form their forecasts. The following equation where yt,h is the private forecast for a given event h is therefore estimated: yt , h = α + β y ⋅ yt − 1, h + β x ⋅ xt − 1, h + β z ⋅ zt − 1 + β Π ⋅ Π t − 1 + β Ω ⋅ Ω t − 1 + ε

where xt is the ECB projection for a given event h, zt is the qualitative communication variable, either MP_ST or MP_INT, Πt are the exogenous controls: the oil price and the macroeconomic uncertainty, and Ωt are the endogenous controls: namely the ECB rate, core HICP, the output gap and credit growth. Both ECB quantitative and qualitative communication variables are also endogenous to the dependent variable, and this equation is therefore estimated with instrumental variables (IV) using two-stages-least-squares5 (2SLS). Since random shocks that affect private forecasts are likely to also affect previous private forecasts, ECB projections, the ECB communication and most of the controls considered, these regressors are considered as endogenous (said differently, their correlation with the error term ε is not equal to zero) and require additional variables that are correlated with these endogenous regressors but not with the error term ε to estimate causal effects. Another issue arises. The IV estimator may be biased in the same direction as the ordinary least squares estimator, and Stock, Wright and Yogo (2002) call this problem ‘weak identification’ as instruments are only weakly correlated with the included endogenous variables. To overcome this issue, lags of the endogenous regressors are considered since they are highly correlated to endogenous regressors. In addition, based on a principal component analysis6 of the private forecasts, the ECB projections, the ECB communication stance and the ECB rate, orthogonal components, constructed as linear combinations of the five preceding variables and maximizing the common variance explained between these four variables, are estimated. The first component captures most of the common variance and the The IV estimation is also performed with the Limited-Information Maximum Likelihood (LIML) estimator which is supposed to yield less bias than the 2SLS estimator according to Stock, Wright and Yogo (2002). 6 Bai and Ng (2010) and Kapetanios and Marcellino (2010) proposes to use factor analysis (or equivalently principal component analysis) to overcome weak identification in IV estimation since they show that the estimated factors (or components) can be more efficient instrumental variables than observed variables. 5

8

following orthogonal components contain less and less information than the preceding components. Two criteria are used to determine the number of components kept: the components whose eigenvalue (the variance of the component) is superior to one are kept (meaning that the component captures more variance than its nominal share of the total variance of variables) as well as those which allow for retained components to capture 80% of the cumulative variance of the series. In all specifications, it corresponds to the first 2 components and these ones are used as additional instruments. The Kaiser-Meyer-Olkin measure of sampling adequacy provides a simple way to assess the relevance of the principal component analysis on the selected variables by comparing the partial correlations and correlations between variables and is provided in Table 2 which summarizes the estimation and characteristics of these components. In order to assess the validity of the set of instruments, the instrumentals variables must be uncorrelated with the error term. Two tests (the Sargan and Basmann χ2 tests) are provided to confirm that the instruments are not correlated with the error term and whether one or more of the instruments should be included in the structural equation. The Sargan test is a statistical test used to check for over-identifying restrictions in a statistical model and is based on the observation that the residuals should be uncorrelated with the set of exogenous variables if the instruments are truly exogenous. The Basmann over-identification test of instrument validity consists of regressing the estimated error term from the second stage regression on the set of instruments and all other exogenous variables, and multiplying the resulting R-squared value by the number of observations. If the instruments are indeed uncorrelated with the dependent variable from the original regression, they should be uncorrelated with the error term from the second stage regression. In addition to these tests, the R2 of the regression of the structural IV equation residuals on the instruments is also provided. The set of instruments therefore includes lags of both the endogenous variables and the two components, which are highly correlated with instrumented variables but not correlated to the structural IV equation residuals, minimize the R2 of the regression of the structural IV equation residuals on the instruments and maximize p-values of the Sargan and Basmann tests. 3.2 Results Table 3 displays estimates of the model for comparing separately the contribution of both the qualitative and the quantitative ECB communication variables to current year private forecasts. The first column exhibits the most complete model tested: lagged private inflation forecasts are positively significant together with ECB inflation projections, the output gap, and the oil price, while macroeconomic uncertainty has a slightly significant negative effect on private inflation forecasts. The ECB qualitative communication, measured by MP_ST the stance of overall communications, is not significant as well as the ECB rate. The second column removes all non-significant control variables and exhibit similar results. The third column removes ECB projections from the regression and displays a slightly significant and positive coefficient for the ECB qualitative communication variable. This suggests a potential complementarity between both types of ECB communication. The fourth and fifth column use respectively the LIML estimator and the MP_INT qualitative communication variable for robustness purposes and confirm the previous results: ECB projections are significant whereas ECB qualitative communication is not. After having assessed whether ECB communication variables affect private forecasts, it is interesting to evaluate how they are affected. The sixth column then shows the effect of the difference between t-2 and t-1 in both qualitative and quantitative communication variables. The effect of ECB projections is still positive but only slightly significant, while qualitative communication is still not. The 9

seventh column assesses whether the (lagged) difference between ECB projections and private forecasts impacts private forecasts. It appears that a positive difference (when ECB projections are above CF forecasts) has positive and significant effect on private forecasts. Last, the eighth column does not focus anymore on the content of ECB qualitative communication but on its dispersion. ECB projections still have a positive effect whereas the dispersion in qualitative communication has not. To summarize, the main determinants of private inflation forecasts for current year are lagged private forecasts, ECB inflation projections, the output gap and oil prices. ECB current year inflation projections positively influence private forecasts, as well as increasing ECB projections and ECB projections higher than private forecasts. Neither the ECB qualitative communication nor the ECB rate appears significant. One may argue for the latter that the forecasting horizon being inferior to the delays of transmission of monetary policy, it is consistent that the ECB rate has no effect on inflation forecasts for current year. Concerning the effect of qualitative communication, the same argument may apply to this variable which captures the future orientation of monetary policy by focusing on all forward-looking statements referring to policymaking. Table 4 then displays estimates of the model for comparing separately the effects of both the qualitative and the quantitative ECB communication variables on next year private forecasts following the same order of specifications and robustness tests. Columns one to five evidence that lagged private inflation forecasts, the output gap and oil prices are still significant determinants of private inflation forecasts. Some changes also appear. ECB projections are not anymore significant, whereas the ECB qualitative communication variables now have a significantly positive effect. Besides, the ECB rate has a significantly negative effect and credit growth a significantly positive effect, both in line with what economic theory would suggest. Another novelty is that macroeconomic uncertainty has a strongly significant positive effect on next year inflation forecasts. The sixth column confirms that the previous result for the ECB rate (a negative effect) and the change in ECB projections (no effect) whereas the change in ECB qualitative communication has a negative effect. The seventh and eighth column also confirms that the ECB rate a negative effect on private next year forecasts, while showing that neither the difference between ECB projections and private forecasts nor the dispersion of ECB qualitative communication has an effect on private next year forecasts. To summarize, the main determinants of private inflation forecasts for next year are lagged private forecasts, ECB qualitative communication, the ECB rate, the output gap, credit growth, oil prices and macroeconomic uncertainty. It appears that speeches increase inflation expectations, but more hawkish speeches reduce inflation expectations. This might be due to the policy content of changes in ECB qualitative communication which signals the future policy decision of an increase in the ECB rate, while the level of ECB qualitative communication signals the inflation risk. A first reason for which ECB projections should be less important in determining private next year forecasts is that the ECB rate is supposed to impact next year inflation and so, private agents are more prone to focus on the policy instrument when forming their expectations, as confirmed by the significant effect of the ECB rate. A second reason might be that ECB projections are based on market-expected interest rates for which the forecast error increases with the forecasting horizon. These explanations coupled with the negative bias of monthly data for quarterly ECB projections might explain why they are not anymore a significant determinant of private inflation forecasts.

10

All in all, it seems from those estimates that ECB projections are rather a short-term tool to affect private inflation forecasts while the ECB qualitative communication which captures the future stance of monetary policy seems to act as a longer-term tool for private inflation forecasts. This interpretation supposes that the signalling content of both types of communication may be different and the next section focuses on the interaction of both communication tools and their interaction with policy decisions.

4. Interacting Communication Types After having estimated the separate effects of both ECB communication types on private inflation forecasts, the objective of this section is to assess the interaction effects between ECB projections and speeches, and with the ECB rate. One may for instance expect that ECB qualitative communication enables central bankers to explain ECB projections and so that some hawkish speeches reinforce the positive effect of ECB projections on private current year inflation forecasts, as estimated in the previous section. Starting from the baseline model for both current and next year forecasts, an interaction term is added. Several potentially relevant interacting combinations between ECB projections, the change in ECB projections or the difference between ECB projections and private forecasts, and the stance of communication, the change in the stance of communication, the dispersion of the communication, the number of communications during each month and the ECB rate are tested. The estimated model is the following: yt , h = α + β y ⋅ yt − 1, h + β x ⋅ xt − 1, h + β z ⋅ zt − 1 + β I ⋅ I t − 1 + β Π ⋅ Π t − 1 + β Ω ⋅ Ω t − 1 + ε

where I is the interaction term. The latter is considered endogenous to the dependent variable. Since all variables interacted are continuous, predictor and moderator variables are defined to interpret the results: the effect of the former on private forecasts is presented for low and high values of the latter. Low and high values of the moderator variables are defined as mean - 1 S.D. and mean + 1 S.D. of the relevant variable (Table 1 provides these descriptive statistics for all variables). Naturally, all interaction terms can be interpreted in the other direction. Table 5 provides a review of the significance of the different interaction terms tested for both current and next year forecasts. It shows that interaction of ECB projections and ECB qualitative communication, and the interaction of ECB projections and the ECB rate have a significant effect on private current year forecasts. It also shows that ECB projections and the ECB rate, and the interaction of the change in ECB qualitative communication and the ECB rate also have an effect on private next year forecasts. Table 6 details the significant interacting effects7. The first column shows the interaction between ECB current year projections and the ECB qualitative communication variable. The average effect of ECB projections can be calculated as the coefficient associated to ECB projections (0.220) plus the coefficient associated to the interaction term (0.374) times the mean of MP_ST (0.44) and is 0.384. This is in line with the linear estimates of the previous section. The effect of ECB projections are however non-linear with high and low values of MP_ST: with a hawkish stance of communication, the effect of ECB projections on private inflation forecasts are high and very significant compare to a neutral stance of communication. This effect is confirmed in column 2 with MP_INT. The third column exhibits the interaction effects between changes 7

All regressions with interaction terms have also been run with the LIML estimator, which confirms 2SLS results.

11

in ECB projections and MP_ST. It is very interesting to note that when the stance of ECB communication is hawkish, increasing ECB projections has a negative effect on private forecasts whereas this effect is four times smaller when the stance of communication is neutral. A similar pattern emerges from the interaction of ECB projections and the ECB rate. The effect of ECB projections on private forecasts is positive and stronger with a high interest rate than a small one (fourth column), whereas increasing ECB projections (fifth column) and ECB projections higher than CF forecasts (sixth column) have a negative effect on private forecasts when the ECB rate is high and a positive effect when the ECB rate is low. For next year forecasts, ECB projections (seventh column), ECB projections higher than CF forecasts (eighth column), and changes in MP_ST (ninth column) all have a negative effect on private inflation forecasts when the ECB rate is high whereas they have a null or positive effect when the ECB rate is low. For all nine estimations, lagged private forecasts, the output gap, oil prices, credit growth and macroeconomic uncertainty have the same effects than in the previous section. To summarize the effects of central bank communication and action on current year private forecasts, it appears that ECB projections are more influential when ECB qualitative communication is more hawkish or the ECB rate is higher. However, changes in ECB projections have a negative effect on private forecasts when interacted with ECB qualitative communication, and this negative effect is more pronounced when the stance of communication is more hawkish. In the same vein, changes in ECB projections and ECB projections higher than CF forecasts have a negative effect on private forecasts when the ECB rate is high, whereas the effect is positive when the ECB rate is low. The policy implications are the following. First, policymakers should then pay more attention to the effect of their quantitative communication when the ECB rate is high or the stance of communication is hawkish. Second, ECB projections should be used to explain the underlying reasons for ECB rate decisions (a given decision has been taken because of a given projection) and not to explain the expected effects of ECB rate decisions (a given decision will produce a given projection). Indeed, publishing increasing ECB inflation projections or ECB projections above CF forecasts when the ECB rate is high (so referring to the case: a given decision has been taken because of a given projection) has a negative effect on private inflation forecasts. One mechanism that may explain this effect is that the central bank appears credible when it justifies and explains its decisions. At the opposite, publishing increasing ECB inflation projections or ECB projections above CF forecasts when the ECB rate is low would have a positive effect on private inflation forecasts, possibly because private agents would not understand then why the central bank is not increasing its interest rate. The effects of central bank communication and action on next year private forecasts can be summarized as follows. ECB projections and ECB projections higher than CF forecasts have a negative effect on private forecasts when the ECB rate is high whereas a positive effect when the ECB rate is low. Similarly, changes in the stance of qualitative communication have a negative effect on private inflation forecasts when the ECB rate is high. The policy implications are similar to those for current year forecasts: ECB projections and the ECB rate should be consistently set following this pattern: a given decision has been taken because of a given projection. This would produce a negative effect on private forecasts. Another interesting policy implication refers to the stance of qualitative communication. Since increases in the stance of communication have a negative effect when the ECB rate is high and no effect when the ECB rate is low, policymakers should not hesitate to make hawkish announcements.

12

The result that ECB projections or speeches have a positive effect on private forecasts when the ECB rate is low and a negative one when the ECB rate is high might be interpreted as the fact that the ECB rate is a proxy for the business cycles and signals turning points. However, first, the output gap is included as a control and should capture business cycles. Second, the potential signal disclosed by the ECB projections or speeches on turning points might be exact for low rates (ECB projections have a positive effect on private forecasts) but not for high rates (ECB projections still have a positive effect on private forecasts). One further argument might then be that the interaction between the ECB rate and ECB projections captures the mean reverting process of forecasts’ dynamics. However, as the ECB qualitative communication variable also captures business cycle dynamics (its correlation with the output gap is 0.61) and is correlated to the ECB rate (0.38), the question remains why the three different interaction terms between ECB qualitative communication and ECB projections (column 1 to 3) does not display the same properties. This suggests that the identified effect is from the ECB rate not from business cycles.

5. Conclusion This paper extends the analysis in the existing literature in two directions. First, it assesses whether ECB projections and ECB Governing council members’ speeches considered together impact the level of private inflation expectations. Second, it evaluates the interaction of these communication types together and with the ECB rate. The main findings are the following. ECB current year inflation projections positively influence private current year inflation forecasts. Neither the ECB qualitative communication nor the ECB rate appears significant, whereas concerning next year forecasts, hawkish speeches increase and the ECB rate decreases inflation expectations, and ECB projections do not impact anymore private forecasts. Interacting central bank communication types and action, it appears that ECB projections have more influence on current year private forecasts when ECB qualitative communication is more hawkish or the ECB rate is higher. However, changes in ECB projections have a negative effect on private forecasts when interacted with ECB qualitative communication, and this negative effect is more pronounced when the stance of communication is more hawkish. In the same vein, changes in ECB projections and ECB projections higher than CF forecasts have a negative effect on private forecasts when the ECB rate is high, whereas the effect is positive when the ECB rate is low. The pattern is quasi-similar for next year forecasts: ECB projections, changes in the stance of qualitative communication and ECB projections higher than CF forecasts have a negative effect on private inflation forecasts when the ECB rate is high whereas no effect when the ECB rate is low. The key argument of the present paper is that both communication types have differentiated effects and that the optimal design of one communication device should depend on ECB policy decisions and on the other communication device. Indeed, this paper suggests that both qualitative and quantitative central bank communication are a crucial part of the conduct of monetary policy as stressed by Guthrie and Wright (2000).

13

References Andersson, M., H. Dillén, and P. Sellin (2006), “Monetary policy signaling and movements in the term structure of interest rates“, Journal of Monetary Economics, 53, 1815–1855 Andersson, M. (2007), “Using intraday data to gauge financial market responses to Fed and ECB monetary policy decisions“, ECB working paper 726. Andersson, M., L. Hansen and S. Sebestyén (2006), “Which news moves the Euro area bond market? “, ECB working paper 631. Bai, J. and S. Ng (2010), “Instrumental Variable Estimation in a Data Rich Environment”, Econometric Theory, 26, 1577–1606. Beechey, M., J.H. Wright (2009), “The high-frequency impact of news on long-term yields and forward rates: Is it real?”, Journal of Monetary Economics, 56, 535–544. Berger, H., J. De Haan and J.-E. Sturm (2006), “Does money matter in the ECB strategy? New evidence based on ECB communication“, CESifo working paper 1652. Blinder, A., M. Ehrmann, M. Fratzscher, J. De Haan and D.-J. Jansen (2008), "Central Bank Communication and Monetary Policy: A Survey of Theory and Evidence", Journal of Economic Literature, 46(4), 910-45. Brand, C., D. Buncic and J. Turunen (2006), “The impact of ECB monetary policy decisions and communication on the yield curve“, ECB working paper 657. Connolly, E. and M. Kohler (2004). “News and interest rate expectations: A study of six central banks“, In C. Kent, & S. Guttman (Eds.), The future of inflation targeting, 108– 134). Sydney: Reserve Bank of Australia. Conrad, C. and M. Lamla (2010), "The High-Frequency Response of the EUR-US Dollar Exchange Rate to ECB Monetary Policy Announcements", Journal of Money, Credit and Banking, 42 (7), 1391–1417. De Haan, J. (2008), “The effect of ECB communication on interest rates: An assessment“, The Review of International Organizations, 3(4), 375–398. Dovern, J., U. Fritsche and J. Slacalek (2011), “Disagreement among Forecasters in G7 Countries”, The Review of Economics and Statistics, forthcoming. ECB (2001), “A guide to Eurosystem staff macroeconomic projection exercises”, Month Bulletin, European Central Bank, June. ECB (2009), “New procedure for constructing ECB staff projection ranges” , Note, ECB website, December. Ehrmann, M., S. Eijffinger and M. Fratzscher (2009), “The Role of Central Bank Transparency for Guiding Private Sector Forecasts”, CEPR Discussion Paper 7585. Ehrmann, M. and M. Fratzscher (2007), “Communication by central bank committee members: Different strategies, same effectiveness“, Journal of Money, Credit, and Banking, 39(2–3), 509–541. Ehrmann, M. and M. Fratzscher (2009), “Explaining Monetary Policy in Press Conferences”, International Journal of Central Banking, 5, 41–84. Fujiwara, I. (2005), “Is the central bank’s publication of economic forecasts influential?”, Economics Letters, 89, 255-261. Gerlach, S. (2007), “Interest rate setting by the ECB, 1999–2006: Words and deeds“, International Journal of Central Banking, 3(3), 1–46. Gürkaynak, R., B. Sack, and E. Swanson (2005), “Do Actions Speak Louder Than Words? The Response of Asset Prices to Monetary Policy Actions and Statements”, International Journal of Central Banking, 1(1), 55–93. Guthrie, G. and J. Wright (2000), “Open mouth operations“, Journal of Monetary Economics, 46, 489–516.

14

Hayo, B. and M. Neuenkirch (2010), “Do Federal Reserve communications help predict federal funds target rate decisions? “, Journal of Macroeconomics, 32, 1014–1024. Heinemann, F., and K. Ullrich (2007), “Does it pay to watch central bankers lips? The information content of ECB wording“, Swiss Journal of Economics and Statistics, 143(2), 155–185. Holsti, O. (1969), Content Analysis for Social Sciences and Humanities. Reading, PA: AddisonWesley. Hubert, P. (2012), “ECB Projections as a Tool for Understanding Policy Decisions”, mimeo. Jansen, D.-J. and J. De Haan (2009), “Has ECB communication been helpful in predicting interest rate decisions? An evaluation of the early years of the Economic and Monetary Union“, Applied Economics, 41(16), 1995–2003. Kapetanios, G. and M. Marcellino (2010), “Factor-GMM estimation with large sets of possibly weak instruments“, Computational Statistics and Data Analysis, 54, 2655-2675. Kohn, D. and B. Sack (2004), “Central Bank Talk: Does it Matter and Why?”, In Macroeconomics, Monetary Policy, and Financial Stability, edited by Bank of Canada, 175–206. Ottawa: Bank of Canada. Musard-Gies, M. (2006), “Do ECB’s statements steer short-term and long-term interest rates in the eurozone“, The Manchester School, 74, 116–139 (supplement). Rosa, C. and G. Verga. (2007), “On the consistency and effectiveness of central bank communication: Evidence from the ECB“, European Journal of Political Economy, 23(1), 146–175. Sturm, J.-E., and J. de Haan (2011): "Does central bank communication really lead to better forecasts of policy decisions? New evidence based on a Taylor rule model for the ECB", Review of World Economics, 147(1), 41-58. Stock, J. H., J. H. Wright and M. Yogo (2002), “A survey of weak instruments and weak identification in generalized method of moments”, Journal of Business and Economic Statistics, 20(4), 518–529. Woodford, M. (2005), “Central-bank communication and policy effectiveness“, In The Greenspan era: Lessons for the future, 399–474. Kansas City: Federal Reserve Bank of Kansas City.

15

Figure 1 – Main ECB variables and the KOF index 4.5

2

4 3.5

1

3 2.5 0 2 1.5 -1

1 0.5 0 2004:6

-2 2005:6 ECB CY Forecasts

2006:6

2007:6

2008:6

ECB rate

2009:6

MP_ST (right scale)

16

2010:6 KOF (right scale)

2011:6

Table 1 - Introductory Statistics Communication on Monetary Policy Inclination Tightening

Neutral

Easing

Total

331

184

75

590

56.1%

31.2%

12.7%

100.0%

Descriptive Statistics Variable

Obs

Mean

Std. Dev.

Min

Max

CF_CY

85

1.90

0.80

0.27

3.61

CF_NY

85

1.80

0.32

1.13

2.53

ECB_CY

85

1.95

0.84

0.30

3.50

ECB_NY

85

1.83

0.42

1.00

2.60

ΔECB_CY

84

0.01

0.37

-2.90

0.90

ΔECB_NY

84

0

0.19

-1.20

0.50

(ECB-CF)_CY

85

0.06

0.42

-0.92

2.52 0.79

(ECB-CF)_NY

85

0.03

0.19

-0.46

MP_ST

85

0.44

0.53

-1

1

ΔMP_ST

84

0.00

0.35

-1.27

1

MP_INT

85

3.01

5.41

-13.01

15.00

MP_NB

85

6.93

3.52

1

15

ECB rate

85

2.33

1.15

1.00

4.25

Core HICP

85

1.65

0.48

0.70

2.70

Output Gap

85

-0.02

1.17

-2.49

2.26

Credit

85

7.32

4.45

0.10

13.20

Oil price

85

24.43

37.05

-54.63

86.56

Uncertainty_CY

85

0.00

0.04

-0.07

0.11

Uncertainty_NY

85

0.00

0.07

-0.13

0.17

Correlation between ECB Communication, Projections, rate and the KOF index ECB_CY

ECB_NY

MP_ST

MP_INT

MP_DISP

KOF

ECB_CY

1

ECB_NY

0.74

1

MP_ST

0.19

0.47

MP_INT

0.15

0.42

0.91

1

MP_DISP

0.08

-0.09

-0.14

-0.14

1

KOF

0.56

0.71

0.60

0.56

-0.07

1

ECB rate

0.63

0.77

0.38

0.32

-0.06

0.66

ECB rate

1

17

1

Table 2 - Factors as Instruments Current Year Forecasts Principal components/correlation

Next Year Forecasts Obs = 85 Principal components/correlation

Rotation: (unrotated=principal)

Obs = 85

Rho = 0.87 Rotation: (unrotated=principal)

Eigenvalue Difference Proportion Cumulative

Rho = 0.92

Eigenvalue Difference Proportion Cumulative

Comp1

2.621

1.749

0.66

0.66

Comp1

2.975

2.287

0.74

0.74

Comp2

0.872

0.478

0.22

0.87

Comp2

0.688

0.443

0.17

0.92

Comp3

0.394

0.281

0.10

0.97

Comp3

0.246

0.155

0.06

0.98

Comp4

0.113

.

0.03

1

Comp4

0.091

.

0.02

1

Principal components (eigenvectors) Comp2 Unexplained

Principal components (eigenvectors)

Variable

Comp1

Variable

Comp1

CF_CY

0.574

-0.179

0.11

CF_NY

0.554

-0.123

0.08

ECB_CY

0.549

-0.374

0.09

ECB_NY

0.541

-0.190

0.10

MP_ST

0.312

0.909

0.02

MP_ST

0.376

0.913

0.01

ECB rate

0.521

0.047

0.29

ECB rate

0.508

-0.340

0.15

Kaiser-Meyer-Olkin measure of sampling adequacy Overall

Comp2 Unexplained

Kaiser-Meyer-Olkin measure of sampling adequacy

0.662

Overall

18

0.773

Table 3 - IV 2SLS estimation of the effects of ECB forecasts and ECB Qualitative Communication on Private Current Year Inflation Forecasts Dependant variable: CF inflation forecasts at date t for current year All regressors are considered at date t-1 ECB Qual. Com: CF forecasts ECB projections

Complete

Baseline

w/o Quant.

LIML

Robustness

ΔECBF

(ECB-CF)

Baseline

MP_ST

MP_ST

MP_ST

MP_ST

MP_INT

ΔMP_ST

MP_ST

MP_DISP

0.385**

0.393**

0.761***

0.370**

0.382**

0.754***

0.732***

0.522**

(0.18)

(0.18)

(0.05)

(0.18)

(0.18)

(0.06)

(0.05)

(0.22)

0.288*

0.304**

-

0.322**

0.314**

0.269º

0.190*

0.225º

(0.15)

(0.15)

(0.25)

(0.10)

(0.16)

0.056

0.006

0.071

0.109º

-0.546

(0.16)

(0.14)

ECB Qual. Com.

0.068

0.053

0.088º

(0.14)

(0.09)

(0.08)

(0.09)

(0.01)

(0.22)

(0.09)

(0.61)

ECB rate

-0.064

-0.003

0.003

-0.003

-0.001

0.029

-0.004

-0.020

(0.10)

(0.04)

(0.04)

(0.04)

(0.04)

(0.05)

(0.04)

(0.04)

0.118

-

-

-

-

-

-

-

0.147**

0.143***

0.092**

0.145***

0.141***

0.079*

0.113***

0.138***

(0.07)

(0.05)

(0.04)

(0.05)

(0.05)

(0.05)

(0.04)

(0.05)

0.007

-

-

-

-

-

-

-

0.005***

0.003***

0.005***

0.005***

0.003***

0.004***

0.005***

Core HICP

(0.17) Output Gap Credit

(0.03) Oil price

0.005*** (0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Uncertainty

-1.060º

-0.929º

-0.364

-0.941º

-0.983º

-0.710

-0.599

-0.505

(0.92)

(0.84)

(0.77)

(0.85)

(0.82)

(0.78)

(0.78)

(1.03)

0.372º

0.436***

0.331***

0.441***

0.438***

0.324***

0.370***

0.537***

(0.24)

(0.12)

(0.11)

(0.12)

(0.12)

(0.12)

(0.11)

(0.15)

Constant Nb. of obs

82

82

82

82

82

82

82

82

Wald chi2 stat

818.4

818.4

937.2

797.5

807.4

796.2

900.2

834.7

R2

0.91

0.91

0.92

0.91

0.91

0.91

0.92

0.91

0.90

0.90

0.92

0.90

0.90

0.90

0.91

0.90

Adj.

R2

Test of overidentifying restrictions Sargan p-value

0.38

0.51

0.51

0.51

0.74

0.85

0.75

0.80

Basmann p-value

0.43

0.55

0.55

0.51

0.77

0.86

0.78

0.83

R2

0.02

0.00

0.01

0.01

Regression of IV residuals on Instruments 0.02

0.02

0.02

Adj. R2

0.02

-0.13 -0.11 -0.09 -0.11 -0.12 -0.11 -0.12 -0.13 º,*,**,*** means coefficients are significant at 30%, 10%, 5% and 1% respectively. Standard Errors in parentheses. Instruments are lags of endogenous variables and components. Oil price and the measure of macroeconomic uncertainty are considered exogenous. The dependent variable is private inflation forecasts at date t, while all regressors are from date t-1 except uncertainty. For the LIML estimator, the Sargan test for overidentifying restrictions is replaced by the Anderson-Rubin test.

19

Table 4 - IV 2SLS estimation of the effects of ECB forecasts and ECB Qualitative Communication on Private Next Year Inflation Forecasts Dependant variable: CF inflation forecasts at date t for next year All regressors are considered at date t-1 Complete Baseline w/o Quant. ECB Qual. Com: CF forecasts

LIML

Robustness

ΔECBF

(ECB-CF)

Baseline

MP_ST

MP_ST

MP_ST

MP_ST

MP_INT

ΔMP_ST

MP_ST

MP_DISP

0.512***

0.507***

0.507***

0.499***

0.499***

0.432***

0.507***

0.499**

(0.16)

(0.14)

(0.08)

(0.15)

(0.15)

(0.13)

(0.08)

(0.24)

ECB projections

0.001

-0.003

-

0.001

0.035

-0.189

-0.027

-0.030

(0.10)

(0.10)

(0.10)

(0.10)

(0.23)

(0.09)

(0.17)

ECB Qual. Com.

0.120*

0.106**

0.106**

0.106**

0.010***

-0.173º

0.054º

-0.548

(0.07)

(0.04)

(0.04)

(0.05)

(0.00)

(0.14)

(0.04)

(0.56)

-0.117***

-0.117***

-0.116***

-0.117***

-0.102***

-0.187***

-0.134***

-0.169**

(0.04)

(0.04)

(0.04)

(0.04)

(0.04)

(0.05)

(0.04)

(0.07)

0.029

-

-

-

-

-

-

-

0.090***

0.089***

0.091***

0.076***

0.120***

0.100***

0.137**

ECB rate Core HICP

(0.07) Output Gap

0.081**** (0.03)

(0.02)

(0.02)

(0.02)

(0.02)

(0.03)

(0.02)

(0.05)

Credit

0.042***

0.044***

0.044***

0.044***

0.040***

0.062***

0.048***

0.056***

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

(0.02)

(0.01)

(0.02)

0.001*

0.001*

0.001*

0.001*

0.001*

0.001**

0.001**

0.001**

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

0.795***

0.819***

0.815***

0.825***

0.738***

0.508º

0.658**

0.870º

(0.26)

(0.27)

(0.27)

(0.27)

(0.24)

(0.31)

(0.26)

(0.56)

0.730***

0.781***

0.775***

0.786***

0.732***

0.978***

0.805***

1.047***

(0.13)

(0.13)

(0.13)

(0.13)

(0.13)

(0.21)

(0.13)

(0.35)

82

83

83

83

83

82

82

83

Wald chi2 stat

992.4

995.9

994.0

986.8

986.6

563.3

977.8

336.7

R2

0.92

0.92

0.92

0.92

0.92

0.87

0.93

0.78

0.92

0.92

0.92

0.92

0.92

0.86

0.92

0.75

Oil price Uncertainty Constant Nb. of obs

Adj.

R2

Test of overidentifying restrictions Sargan p-value

0.67

0.40

0.42

0.40

0.80

0.59

0.82

0.29

Basmann p-value

0.72

0.43

0.44

0.40

0.82

0.61

0.84

0.32

Regression of the IV residuals on Instruments R2

0.02

0.01

0.01

0.01

0.01

0.00

0.00

0.01

Adj. R2

-0.15

-0.11

-0.10

-0.11

-0.13

-0.12

-0.14

-0.11

º,*,**,*** means coefficients are significant at 30%, 10%, 5% and 1% respectively. Standard Errors in parentheses. Instruments are lags of endogenous variables and components. Oil price and the measure of macroeconomic uncertainty are considered exogenous. The dependent variable is private inflation forecasts at date t, while all regressors are from date t-1 except uncertainty. For the LIML estimator, the Sargan test for overidentifying restrictions is replaced by the Anderson-Rubin test.

20

Table 5 - Interaction effects among ECB Communication Tools and Instrument Current Year ECBF ΔECBF

(ECB-CF)

MP_ST

ΔMP_ST

MP_DISP

MP_NB ECB rate

ECBF

-

ΔECBF

-

-

(ECB-CF)

-

-

MP_ST

*** No

º

No

-

ΔMP_ST

No

No

-

MP_DISP

No

No

No

No

No

-

MP_NB

No

No

No

No

No

-

-

ECB rate

***

º

º

No

No

ECBF ΔECBF ECBF

-

ΔECBF

-

-

(ECB-CF)

-

No No Next Year MP_ST

ΔMP_ST

MP_DISP

MP_NB ECB rate

-

(ECB-CF)

-

-

-

MP_ST

No

No

No

-

ΔMP_ST

No

No

No

-

MP_DISP

No

No

No

No

No

-

MP_NB

No **

No No

No ***

No No

No *

No

ECB rate

-

No

º,*,**,*** means coefficients are significant at 30%, 10%, 5% and 1% respectively.

21

-

Table 6 - Interacting ECB Action and Communication Dependant variable: CF inflation forecasts made at date t All regressors are considered at date t-1 Current Year (1) ECBF

Predictor Moderator

ECBF

ECB Qual. Com.

ECB Qual. Com: MP_ST CF forecasts

ΔECBF

Next Year (2) ΔECBF

(ECB-CF)

ECBF

ECB rate

MP_INT MP_ST

(3) (ECB-CF) ΔMP_ST ECB rate

MP_ST

MP_ST

MP_ST

MP_ST

MP_ST

ΔMP_ST

0.553***

0.533***

0.713***

0.248º

0.632***

0.757***

0.587***

0.492***

0.470*

(0.18)

(0.19)

(0.11)

(0.23)

(0.13)

(0.07)

(0.12)

(0.07)

(0.25)

0.220º

0.260*

-0.527º

0.115

1.975º

1.770º

0.130º

0.604***

-0.141

(0.14)

(0.15)

(0.47)

(0.13)

(1.44)

(1.14)

(0.12)

(0.23)

(0.17)

ECB Qual. Com.

-0.788***

-0.078**

0.009

0.258*

-0.190

0.166º

0.129***

0.031

0.285º

(0.25)

(0.03)

(0.18)

(0.13)

(0.21)

(0.12)

(0.04)

(0.04)

(0.24)

Interaction

0.374***

0.036***

-1.124º

0.198***

-1.025º

-0.831º

-0.084**

-0.256***

-0.222*

(0.11)

(0.01)

(1.05)

(0.07)

(0.76)

(0.60)

(0.04)

(0.08)

(0.13)

ECB rate

-0.061º

-0.044

0.002

-0.481***

0.140º

-0.039

0.089

-0.128***

-0.239***

(0.05)

(0.05)

(0.07)

(0.17)

(0.11)

(0.05)

(0.10)

(0.03)

(0.08)

0.157***

0.148***

0.144**

0.140***

0.129*

0.073º

0.078***

0.105***

0.151***

(0.05)

(0.05)

(0.07)

(0.05)

(0.07)

(0.06)

(0.02)

(0.02)

(0.05)

-

-

-

-

-

-

0.034***

0.052***

0.078***

(0.01)

(0.01)

(0.02)

0.001*

0.001**

0.001**

ECB projections

Output Gap Credit Oil price

0.003***

Uncertainty

(0.91)

(0.95)

(2.08)

(1.27)

(1.70)

(1.00)

(0.27)

(0.23)

(0.39)

Constant

0.483***

0.413***

0.503**

1.100***

0.354*

0.414***

0.372*

0.818***

1.153***

(0.12) 0.579***

(0.13) 0.559***

(0.23) -1.606º

(0.30) 0.805***

(0.20) -1.606º

(0.14) -1.133º

(0.19) -0.164*

(0.11) -0.292***

(0.31) -0.491*

Predictor Coef. with High Moderator Predictor Coef. with Low Moderator

0.003***

0.004***

0.004***

0.005***

(0.00)

(0.00)

-1.765*

-1.921**

0.004***

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

1.299

-3.287***

0.529

-0.864

1.024***

0.824***

0.618º

(0.19)

(0.19)

(1.43)

(0.29)

(1.27)

(0.96)

(0.09)

(0.10)

(0.29)

0.186º

0.171º

-0.424º

0.349**

0.759º

0.785*

0.030

0.300**

0.021

(0.14)

(0.16)

(0.40)

(0.16)

(0.58)

(0.44)

(0.10)

(0.14)

(0.14)

0.393***

0.388***

-1.181º

0.456***

-2.366º

-1.918º

-0.194**

-0.592***

-0.512*

(0.12)

(0.14)

(1.11)

(0.16)

(1.76)

(1.38)

(0.09)

(0.19)

(0.30)

82

82

81

82

81

82

82

82

81

Wald chi2 stat

767.6

716.3

364.0

822.5

296.1

574.8

1206.8

1430.4

342.8

R2

0.90

0.90

0.80

0.91

0.76

0.87

0.94

0.95

0.79

Adj. R2

0.89

0.88

0.78

0.90

0.73

0.86

0.93

0.94

0.76

Sargan p-value

0.48

0.79

0.71

0.83

0.81

0.96

0.41

0.67

0.80

Basmann p-value

0.52

0.83

0.73

0.85

0.84

0.96

0.46

0.71

0.82

0.02

0.01

0.01

Difference Nb. of obs

Test of overidentifying restrictions

Regression of the IV residuals on Instruments R2

0.02

0.01

0.00

0.00

Adj. R2

0.01

0.00

-0.12 -0.14 -0.12 -0.14 -0.15 -0.14 -0.13 -0.15 -0.15 We compute the predictor coefficient while holding the value of the moderator variable constant at either a high value (mean + 1SD) or a low value (mean - 1SD). º,*,**,*** means coefficients are significant at 30%, 10%, 5% and 1% respectively. Standard Errors in parentheses. Instruments are lags of endogenous variables and components. Oil price and the measure of macroeconomic uncertainty are considered exogenous. The dependent variable is private inflation forecasts at date t, while all regressors are from date t-1 except uncertainty. For the LIML estimator, the Sargan test for overidentifying restrictions is replaced by the Anderson-Rubin test.

22