Journal of International Money Journal of International Money and Finance anc| Finance 24 (2005) 343-361 ^^^^=^^^^ www.elsevier.com/locate/econbase Talking heads: the effects of ECB statements on the euro–dollar exchange rate David-Jan Jansena,*, Jakob De Haanb,c aDe Nederlandsche Bank, Research Division, P.O. Box 98, 1000 AB Amsterdam, The Netherlands bDepartment of Economics, University of Groningen, P.O. Box 800, 9700 AV Groningen, The Netherlands cCESifo, Munich, Germany Abstract This paper studies the reaction of the conditional mean and volatility of the euro-dollar exchange rate to statements by European Central Bank and national central bank officials. We focus on comments on monetary policy and the external value of the euro. We find that the Bundesbank has dominated the news coverage. We conclude that ECB statements have mainly influenced conditional volatility. In some cases there are effects of statements on the conditional mean of the euro-dollar exchange rate. Efforts to talk up the euro have generally not been successful. There is also evidence of asymmetric reactions to news. © 2004 Elsevier Ltd. All rights reserved. JEL classification: E50; F31 Keywords: ECB; Exchange rates; Euro; News approach; EGARCH * Corresponding author. Tel.: +31 20 5243170; fax: +31 20 5242529. E-mail address: d.jansen@dnb.nl (D. Jansen). 0261-5606/$ - see front matter © 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jimonfin.2004.12.009 344 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 1. Introduction Currency markets are continuously flooded with information. In spite of intensive research, the adjustment of exchange rates to this stream of information is still not fully understood. Previous research on the reaction of exchange rates to news has mainly studied the adjustment to announcements of new macroeconomic information. There is, however, hardly any research explicitly analyzing the impact of statements by central bank officials, the study by Fatum and Hutchison (2002) being one exception. Still, it is likely that monetary policy has an influence on exchange rates, especially in the short run. The central bank can change interest rates and intervene in foreign exchange markets. Central bankers often also try to influence exchange rates directly by their comments. This paper studies the effects of statements made by European Central Bank (ECB) and national central bank officials on the conditional mean and volatility of the euro– dollar exchange rate during the first years of the European Economic and Monetary Union (EMU). During this turbulent period, the depreciation of the euro caused great problems for the ECB. Before its introduction, the euro was widely expected to be a strong currency. Some even hoped (or feared) that the euro would compete with the dollar in its role of most important international currency (see Eijffinger and De Haan (2000) for a discussion).1 As the new currency was an important symbol for the EMU, its decline led some observers to question the overall success of the EMU. Moreover, the strong depreciation of the euro was a serious threat to price stability in the euro area. In the end, the ECB reacted by intervening in the foreign exchange market. The success of these interventions is under debate.2 Apart from interventions, the ECB has been very active to support the euro verbally (a strategy known as talking up the currency). ECB officials have often expressed the view that the euro was undervalued during the period under consideration. This paper investigates whether ECB statements have affected the euro–dollar exchange rate. Firstly, we test whether efforts to talk up the euro have been successful. Secondly, we test whether statements on monetary policy have fed through into the euro– dollar exchange rate. For example, have comments indicating higher interest rates led to changes in the exchange rate? Thirdly, we study whether statements have had an impact on the conditional volatility of the exchange rate. We follow the so-called news approach which rests on two assumptions (see e.g. De Grauwe et al. (1993)): the exchange rate is modeled as an asset price and expectations 1 The exchange rate regime may also affect other financial markets. See, for instance, Cheung and Westermann (2001) who report that even though the introduction of the euro did not affect the relationship between German and US equity markets, the volatility of the German stock index has fallen significantly since the introduction of the euro. 2 Whereas many observers question the effectiveness of interventions, Fatum and Hutchison (2002) claim that the first of four ECB interventions had some success. See also Frenkel et al. (2001). See Sarno and Taylor (2001) for a review of the literature on foreign exchange interventions. A recent study finding support for the effectiveness of interventions is Fatum and Hutchison (2003). According to the survey by Cheung and Chinn (2001), traders in currency markets think that interventions increase volatility. They do not consider central bank interventions very successful. D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 345 are formed rationally. These assumptions imply that only unexpected information (or news) influences the exchange rate as all currently available information is already reflected in its current price. We use an EGARCH specification that allows us to test for possible effects on the conditional mean as well as on volatility. Even though news is found to be rapidly incorporated into exchange rates (see, for example, Cheung and Chinn (2001) who find that currency traders believe that the bulk of the adjustment takes place even within just 1 minute after the release of new information), we use daily observations like Fatum and Hutchison (2002) and Galati and Ho (2003). This is justified, as we are mainly interested in the effectiveness of statements of ECB officials. If the effects of talking up the currency can only be observed using higher frequency data, ECB statements cannot be considered a very powerful instrument to influence developments in the foreign exchange market. We use statements of European central bankers - including statements of officials of national central banks in the euro area - as reported by the Bloomberg News Service. We find that statements of Bundesbank officials outnumber those of other central bankers. Our results indicate that ECB statements have had a larger impact on the conditional volatility than on the conditional mean of the exchange rate. Efforts of the ECB to talk up the currency have generally not been successful during the period under investigation. Our results also indicate that there has been an asymmetric reaction to news. For example, comments indicating interest rate rises cause the exchange rate to move, whereas comments pointing towards lower interest rates do not. The remainder of the paper is structured as follows. Section 2 surveys the literature on the relationship between news and the exchange rate. Section 3 discusses our methodology and the data used, while Section 4 describes the statements in detail. Section 5 presents the results. Section 6 offers our conclusions. 2. News and the exchange rate: a selective survey 2.1. The news approach If currency markets are efficient, only unexpected information (news) has an effect on the exchange rate. The application of the news approach to exchange rates has a history going at least as far back as Frenkel (1981). The conceptual idea behind the news approach is very straightforward and may be represented by Eq. (1): Rt = a + b1 {Xnt — Xf) +b2Zt + et (1) where Rt denotes the exchange rate return, defined as the first difference of the log of the exchange rate. Xn is a matrix with new realizations of macroeconomic variables and Xet is a matrix with expectations. The difference between realizations and expectations is then defined as news. Zt is a matrix with control variables and a is a constant. Table 1 presents a selection of the numerous papers that have studied 3 Rigobon and Sack (2002) and Evans and Lyons (2003) advocate the use of a new approach, based on state dependent heteroscedasticity. Table 1 The news approach to exchange rates: selected papers News items Hakkio and Pearce (1985) Ito and Roley (1987) Goodhart et al. (1993) Ederington and Lee (1993) DeGenarro and Shrieves (1997) Almeida et al. (1998) Andersen and Bollerslev (1998) Fatum and Hutchison (2002) Tivegna (2002) Andersen et al. (2003) Galati and Ho (2003) Macroeconomic announcements Macroeconomic announcements Two specific events Scheduled macroeconomic announcements Macroeconomic news, economic policy news, interest rate reports Macroeconomic announcements Scheduled macroeconomic announcements Statements and rumors on ECB intervention Macroeconomic announcements, events, opinions, statements Macroeconomic announcements Macroeconomic announcements Measure Xtn-Xte Xtn-Xte Dummies Dummies Number of news items Xnt-Xte Dummies Dummies Xnt — Xet and dummies Xnt — Xte and dummies Xtn-Xte Estimation OLS with 3 observations per day OLS with 4 observations per day OLS and GARCH-M using continuous data OLS using 5 min returns ARMA using 10 min returns OLS using 5 min returns Volatility modeling using 5 min returns OLS using daily data Multivariate GARCH using 2 observations per day ARMA using 5 min returns ARMA using daily observations D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 347 the relationship between news and the exchange rates. The table highlights some of the choices researchers have made with respect to the news that was studied, how news was measured and how the estimation was done. 2.2. Which news is studied? Most papers in this field study reactions to macroeconomic announcements. Hakkio and Pearce (1985), for example, study short-run responses of spot exchange rates to three categories of economic announcements: money growth, inflation and real activity. They find that the exchange rate only reacts to unexpected changes in the money stock. 4 Ito and Roley (1987) study reactions of the yen–dollar exchange rate to news from Japan and the US. These authors find that the yen–dollar exchange rate was mainly driven by news from the US. Improved data availability enabled researchers to use high frequency data in order to separate the effects of news from the effects of other factors influencing the exchange rate. For example, Ederington and Lee (1993) examine the impact of 19 types of scheduled macroeconomic news announcements on the volatility of the dollar–German mark exchange rate. They find that the volatility of the exchange rate is influenced by announcements on merchandise trade, employment, retail sales, the producer price index and GNP. Goodhart et al. (1993) focus on two particular events: a release of US trade figures and an interest rise in the UK. These authors conclude that these events significantly change the time-series behavior of the exchange rate. DeGenarro and Shrieves (1997) and Tivegna (2002) study macroeconomic announcements and other factors. DeGenarro and Shrieves (1997) distinguish three categories of news: macroeconomic news, economic policy news and interest rate reports. The last two categories contain unscheduled news. They use quote arrival as a proxy for market activity and the number of news headlines in the Reuters news service as a proxy for news. These authors conclude that both private information and news are important determinants of exchange rate volatility. Tivegna (2002) studies both scheduled (mostly quantitative) and non-scheduled (mostly qualitative) news items. He concludes that statements by policy makers are an important market mover for exchange rates. The seminal paper by Andersen and Bollerslev (1998) has been very important in bringing together three factors that influence the volatility of exchange rates. The first category is calendar effects, which can be split into intraday effects, weekly effects, holiday effects and time change effects. The second category is macroeco-nomic announcements. The final category is interdaily volatility dependencies or ARCH effects (also see Hsieh (1988) and Hsieh (1989)). One important conclusion is that the announcement effects are of secondary importance in explaining overall volatility. Announcements have significant effects shortly after they were made, but they explain less of the volatility than the other two factors. However, Andersen and For a similar analysis, see Doukas and Rahman (1986) and Doukas and Melhem (1986). 348 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 Bollerslev (1998) conclude the effect of the announcements is large. Almeida et al. (1998) support this conclusion. They study the reaction of the German mark– dollar exchange rate to macroeconomic announcements using high frequency data. They conclude that after 3 hours most of the effect from the announcement has disappeared. The effects are dwarfed by other events at lower frequencies. The study by Andersen et al. (2003) is similar to Andersen and Bollerslev (1998) in scope and modeling approach. Andersen et al. (2003) characterize the conditional mean of the spot rate of the US dollar against five other currencies. They find that, in general, news has a statistically significant effect on the exchange rate. The adjustment of the conditional mean in reaction to news is very quick, whereas the adjustment in volatility is more gradual. There is also evidence that the sign of the news matters for the response. Negative surprises have greater impact than positive surprises. 2.3. Studies on the euro–dollar exchange rate Understandably, research on the euro–dollar exchange rate is still underrepre-sented in the literature. Two recent papers are Galati and Ho (2003) and Fatum and Hutchison (2002). Galati and Ho (2003) investigate the relationship between scheduled macroeconomic news and the daily change in the euro–dollar rate, focusing on the impact of US and European news announcements during the first two years of EMU. They conclude that the unexpected part of macroeconomic news announcements can explain up till 10% of the variation in the euro–dollar rate. There is also evidence that the reaction to news is asymmetric: good news on the euro area has been ignored, while bad news has had an influence. Fatum and Hutchison (2002) study the effects of ECB intervention and interventionrelated news on the euro–dollar exchange rate. These authors use a regression approach to determine which news variables have had effect on the euro and use an event study methodology to assess whether the ECB interventions have been successful. They use four categories of news: rumors of intervention by the ECB, statements by officials in support of the euro, statements by officials not supportive of the euro and reports of actual intervention. Rumors on intervention have a positive effect on the exchange rate and statements not supportive of the euro have a negative effect on the euro–dollar exchange rate. Interestingly, Fatum and Hutchison (2002) find that only the negative statements have persistent effects. 3. Methodology 3.1. Modeling strategy We test whether statements by central bank officials contain news affecting the conditional mean and the conditional volatility of the euro–dollar exchange rate. Following the above mentioned literature, we model the first difference of the log of the exchange rate series, using AR-terms and control variables. We also take D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 349 clustered volatility into account. Given the qualitative nature of our data, we measure news by using dummies.5 We use the EGARCH model proposed by Nelson (1991), both because EGARCH allows us to take asymmetric effects on volatility into account and because nonnegativity constraints in the variance equation were violated using GARCH. The model is specified as follows: n j Rt = b0+ \~)biRt-i + 2_\bsiXst-i"*"bwDw + et, e t ~ N(0,ht) i=1 2) i=0 j d3 e- + J2 dsiXst-i + dw Dw log ht = d0 + d1 log ht-1 + d2 -— + i=0 h t-1 where Rt represents the daily change in the natural logarithm of the exchange rate and Xs contains the dummies that represent the ECB statements. The variables in Xs enter both in the mean and variance equation. The mean Eq. (2) has n autoregressive terms in the dependent variable and has dummies for weekdays, denoted as Dw, as control variables.6 The error term is assumed to have a conditional normal distribution with a zero mean and a conditional variance ht which follows an EGARCH (1,1) process.7 The weekday dummies also enter the variance equation (as in Hsieh (1988) and Hsieh (1989)). To choose between alternative specifications, we use both the Akaike information criterion and the Schwartz criterion, following a general to specific approach. 3.2. Constructing dummies to measure news As noted, we use dummies to measure news. Firstly, we use binary dummies to test for effects whenever a central banker makes a statement or when a certain topic is commented on. However, as Tivegna (2002) argues, this could incorrectly lead to nonsignificant results as the effects of statements with contrasting contents may cancel out. Therefore, he uses ternary variables that record whether a statement has an expected positive, neutral or negative effect on the exchange rate. Galati and Ho (2003) follow a similar approach. We partly follow this suggestion and also use dummies that measure the content of a statement. However, we do not completely follow the method suggested by Tivegna (2002). Firstly, when using ternary dummies, the interpretation of the coefficients is not straightforward. Therefore, we prefer to construct three dummies per category of statements. An additional benefit 5 Following the news approach, we would have to contrast the ECB statements with market expectations. As the statements are mainly in qualitative terms, there is no obvious proxy we can use. 6 We also estimated this model using interest differentials, stock returns and dummies for ECB rate decisions and ECB interventions as additional control variables. This did not influence our basic results. 7 In general, our results are robust to using higher order of ARCH/GARCH. Since exchange rate data is often characterized by fat tails, we have checked our results using a conditional t-distribution and a conditional generalized error distribution instead of the conditional normal distribution. In general, our results are robust. 350 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 is that we can study whether there have been asymmetric reactions to news. Secondly, we find it too subjective to assign, a priori, positive, neutral or negative values to statements based on their expected effect on the exchange rate. Therefore, we start with a more neutral approach and record whether an ECB official states that a macroeconomic variable, for example inflation, will go up, down or remain the same. We record the following basic characteristics of each statement: the day on which the statement was made, the person who made the statement, the topic of the statement and the content of the statement. Firstly, we construct (0,1) dummies that measure whether individual central bankers or different groups of ECB officials made a statement. Then, we study topics and content.8 We identify eight categories of statements that we expect to be relevant for the exchange rate (see Table 2). There are two main categories: monetary policy and the external value of the euro. There are five subcategories in the monetary policy category. First of all, there are statements on interest rates in the euro area. The remaining sub-categories relate to the two pillar monetary policy strategy of the ECB. Firstly, there are statements on money growth in the euro area (the former first pillar). Secondly, there are statements on economic growth, inflation and the effects of the euro on inflation (the former second pillar). For statements in the monetary policy category, we record whether according to the ECB official a variable, say interest rates, is expected to go up, or down, or remain stable. Why would these statements affect the exchange rate? With respect to the conditional mean, economic theory suggests a few possible outcomes. A statement suggesting higher interest rates may, for instance, affect the exchange rate via three possible channels (see also Harris and Zabka (1995)): an investor channel, an inflation channel and a growth channel. According to the first channel higher interest rates make buying European securities more attractive which will induce demand for euros. The second channel works through prices. If a central bank raises the interest rate, this will put downward pressure on inflation. Assuming that purchasing power parity holds, this will lead to an appreciating euro. The last channel predicts a negative relationship between statements on higher interest rate and the exchange rate: higher interest rates hamper investments and thus hamper growth, leading to a weaker currency. The second main category that we distinguish are statements on the external value of the euro. The first sub-category under this heading contains statements expressing beliefs on the future direction of the exchange rate. In Section 4 we show that during the period under investigation ECB officials repeatedly expressed the view that the euro had potential to appreciate. Section 5 tests whether this strategy of talking up the euro has been successful. In addition, we test whether the conditional variance of the euro–dollar series has been affected by these statements. 8 We tried to construct variables that combine the results from dummies for central bankers with those from the content of the statements. However, combining these two perspectives results in variables with very low variation and few data-points (i.e. a lot of zeros out of a possible 880). Although this route is certainly very interesting, our data is not rich enough to explore it further. D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 351 Table 2 Categories of ECB Statements Category Contains statements on (1) Monetary policy Interest rates M3 Economic growth Inflation Pass-through ECB interest rates Growth of money supply Real GDP growth in the euro area Inflation in the euro area How the euro exchange rate will affect inflation (2) The euro Future value of euro No target versus inflation Intervention Expectations on the external value of the euro The euro in the monetary policy strategy The possibility of intervention The second sub-category contains statements trying to explain the position of the euro in the ECB monetary policy strategy. There have been many of these statements in the sample period as the ECB has intensively communicated about its monetary policy strategy in recent years. Time and again, ECB officials stressed that the external value of the euro was not an objective for monetary policy. Nevertheless, the extensive depreciation of the euro, which affected inflation in the euro area, forced them to take the euro–dollar rate into account when making monetary policy decisions. We characterize statements that stress that the exchange rate is not a target as negative, i.e. leading to a depreciating euro. We consider statements that stress the possible effect on inflation as positive. This no target/inflation problem may have led to confusion in currency markets, thus potentially increasing volatility. Finally, there are statements on the possibility of an intervention in the currency market. During the period under consideration, there has been a lot of speculation on possible intervention. In a similar vein as Fatum and Hutchison (2003), we test how markets have reacted to this news. Statements that suggest the possibility of intervention are expected to lead to an appreciating euro, whereas statements that deny possible intervention may lead to (further) depreciations. In addition, this type of statement is also likely to have an effect on volatility. Finally, note that there is a difference in the number of observations between the statements series and the exchange rate series. There is only one data point per day for the exchange rate series. However, there can be more than one statement per day. In addition, statements can be made during the weekend when trading activity is generally low. To synchronize the series, we add the statements made during the weekend to the scores for Monday. We also added up the different scores if we found more than one observations per trading day. 3.3. Data sources The sample period ranges from 4 January 1999 until 17 May 2002. This amounts to a total of 881 trading days. The data set with ECB statements was collected from the Bloomberg News Service. This was done by scanning the headlines for keywords like ECB or names of ECB officials. We selected some 930 news reports in which 352 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 ECB statements are reported. The criterion for including a statement is that it has to refer to the entire euro area. Statements on developments in individual countries are not included. We focus on unexpected statements, because we expect to find the greatest effects in these cases. For this reason, the monthly press conferences are not included in our data set. The exchange rate series is taken from Datastream. The exchange rates are New York noon spot rates (dollars per euro). This means that the cut-off point for a trading day is 18.00 CET.9 4. Description of the statement dummies 4.1. Central bankers Table 3 displays the statement dummies for three groups of central bankers and some selected individuals. National central bank presidents made statements on 336 trading days compared to 256 statements for members of the Executive Board of the ECB. We find 82 statements of Bundesbank officials (excluding the president of the German central bank). For no other European central bank do we find such a large amount of statements by its officials.10 Table 3 also shows that the Bundesbank president has made more comments than any other ECB official. We find statements by either Tietmeyer or Welteke for 179 trading days, compared to 92 for Duisenberg and 95 for Trichet. Together with the 82 statements of other Bundesbank officials, the total number of statements by the German central bankers is even higher than that of the Executive Board of the ECB. There are two possible explanations for this. First, financial markets may still attach greater importance to the Bundesbank than to the ECB. The ECB is a new institution that has not yet established a firm reputation. In contrast, the German central bank is a venerable institution that has proven its capability in monetary policy-making over the years. Financial markets do not yet know how to judge the new situation and may still attach great importance to Bundesbank statements for this reason. However, this explanation assumes that Bloomberg selects news in such a way, that the statements of Bundesbank officials receive most attention. Since it is Bloomberg’s business to bring as much news as possible, this is not very likely. Alternatively, it may be argued that the Bundesbank has not yet learnt to cope with its new role within the European System of Central Banks. For years, the Bundesbank has dominated monetary policy in Europe. All this has changed, however, and the Bundesbank had to step back in favor of the ECB. The fact that Bundesbank officials communicated so intensively to financial markets may reflect their difficulty in accepting this new situation. 9 Except during two weeks per year when US and European D.S.T. are not synchronized. The cut-off point is then 19.00 C.E.T. 10 There is quite some variation in Bundesbank statements. Bundesbank officials commented on most of the topics mentioned in Table 2. Moreover, the message in their comments was not uniform, as we discuss in Jansen and De Haan (2004). D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 353 Table 3 Statement dummies for central bankers Total number of statements Groups Executive Board National central bank presidents Bundesbank officials (excluding president) 256 336 82 Total 674 Individuals Duisenberg Bundesbank president Tietmeyera Weltekeb Trichet 92 179 24 155 95 Note: the numbers in this table show the number of trading days on which statements were made. a 4 January 1999 until 31 August 1999. b 1 September 1999 until 17 May 2002. 4.2. Content of statements Table 4 presents the number of ECB comments and their content regarding the topics as described in Table 2. Table 4 shows the number of trading days with statements of a particular category. For example, there were neutral statements on interest rates on 173 out of the 881 trading days, while there were indications for higher interest rates on 36 trading days and indications for lower rates on 12 trading days. It follows from Table 4 that the number of statements on money growth is relatively small compared to those on the second pillar of the monetary policy strategy. The statements on economic growth generally indicate that the outlook for Table 4 The content of the ECB statements Monetary policy Interest rates M3 Economic growth Inflation Pass-through The euro Future value of euro No target versus inflation Intervention Up Neutral Down Total 36 27 194 79 51 173 27 25 111 39 12 29 35 90 10 221 83 254 280 100 Positive Neutral Negative Total 172 112 17 6 24 22 2 42 18 180 178 57 Note: The numbers in this table show the number of trading days on which particular statements were made. Totals for Tables 3 and 4 do not add up, because one news item may report comments on more than one category. 354 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 growth is good, whereas the statements on inflation are more balanced. One reason could be that statements on inflation may sooner lead to speculation on interest rate changes. This may cause ECB officials to be more careful when talking about inflation. Finally, there are statements on the pass-through of the depreciation of the euro into higher inflation on 100 trading days. Most of these statements (51) indicate higher risks for inflation. There are many statements referring to the second main category. Of special interest is the sub-category with statements on the future value of the euro. ECB officials made positive statements on the euro on 172 trading days. Only two statements were negative and six were neutral, reflecting the efforts of the ECB to talk up the euro. The subcategory on the no target/inflation trade-off contains 178 statements. This large number of statements indicates that the ECB has made great efforts to explain the role of the euro in its monetary policy strategy. The majority of statements stress the possible effects of a depreciation of the euro on inflation in the Euro-zone. Finally, we recorded statements on (possible) intervention for 57 trading days. Of these statements 17 indicated that interventions could take place, 22 were neutral and 18 lowered the probability of intervention by the ECB. 5. Results: did ECB statements contain news? 5.1. ECB statements: groups and individuals As a benchmark, we estimate Eqs. (2) and (3) initially without the dummies for the ECB statements. Using the Akaike and Schwartz criteria, we find that one autoregressive term in the mean equation is sufficient. The weekday dummies in the mean equation are always insignificant, so they are excluded in the remainder of the analysis. In the variance equation, we find a lower conditional variance on the Monday and Friday.11 Starting from our benchmark model, we proceed to add statements dummies to our model. Firstly, we study whether there are effects from statements by different groups of ECB officials. The second column of Table 5 presents the regression results for the case in which all ECB officials are included. The number of lags for the statements dummies have been chosen on the basis of the Akaike information criterion and the Schwartz criterion, starting with a maximum of 10 lags. We do not find any significant effect of statements of ECB officials on the mean of the exchange rate. However, there is a significant increase in volatility after the day a statement was made. Volatility (measured by the conditional standard deviation) is approximately 12% higher in these cases. Also, statements by Executive Board members increase volatility (as reported in column 3 of Table 5). The coefficient of the lagged value of the statement dummy is 0.20, indicating, once again, an increase 11 Given the exponential specification a coefficient in the variance equation that is smaller (larger) than zero indicates lower (higher) variance. D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 355 Table 5 Results for groups of ECB officials (1) Benchmark (no dummies) (2) All (3) Executive Board (4) Bundesbank officials (excl. President) Mean equation Constant AR(1) Statements Statement (1st lag) -0.00 0.03 – – -0.00 0.03 – – -0.00* 0.03 – – -0.00* 0.04 0.0017*** —0.0012* Variance equation Constant Statements Statements (1st lag) d1 d2 d3 Monday Friday Akaike IC -10.2*** – – -0.05 -0.18* 0.01 -0.50*** -0.27* -7.295 —8.47*** -0.04 0.22* -0.02 -0.16* -0.0 -0.51*** -0.27* -7.296 —9 75*** 0.08 0.20 0.01 -0.17* 0.01 -0.57*** -0.26 -7.296 -10.15*** – – -0.04 -0.19* 0.00 -0.53*** -0.27* -7.301 Note: This table shows the estimation results for Eqs. (2) and (3). In column (1) no statement dummies are included. In column (2) statements by all ECB officials are taken up. In column (3) only statements by members of the Executive Board are included, while in column (4) statements by officials of the German central bank (excl. the president of the Bundesbank) are shown. Bollerslev–Woolridge robust standard errors and covariance are used. Reported results assume a conditional normal error distribution. */**/*** denotes significance at the 10/5/1% level, respectively. in volatility of 11%. However, the coefficient of the statement dummy is only marginally significant in this case ( p Z 0.11). Interestingly, statements by Bundesbank officials (excluding the president) have no effect on volatility, but there are temporary and off-setting effects on the conditional mean. Column (4) of Table 5 shows that there is an appreciation of 0.17% on the trading day itself and a subsequent depreciation of 0.12% on the day thereafter. According to a Wald test, the total effect is not significantly different from zero. Our next step is to focus on individual central bankers. Therefore, we estimate Eqs. (2) and (3) using dummies for all 18 members of the ECB Governing Council. As before, we select our preferred specification by evaluating the Akaike and Schwartz criteria. Statements by Duisenberg, the Bundesbank president, and Trichet turn out to be of importance. Table 6 reports the results. We find that statements affect the conditional mean, but not the conditional variance. Statements by Duisenberg and the Bundesbank president coincide with a depreciating euro, whereas statements by Trichet coincide with an appreciating euro.12 There is no evident explanation for this. The data, for example, does not show that these three central bankers made different sorts of comments (see Jansen and De Haan (2004)). 12 However, the dummy for statements by Duisenberg is only marginally significant under the assumption of a generalized error distribution ( p Z 0.16). 356 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343-361 Table 6 Results for Duisenberg, Bundesbank president and Trichet (1) Mean equation Constant AR(1) Duisenberg Bundesbank president Trichet Variance equation Constant <51 <52 <53 Monday Friday Akaike IC —0.00 0.03 —0.0013** —0.0012** 0.0014** —10.32*** -0.06 -0.17 0.00 —0.58*** —0.25 —7.304 Note: This table shows results for Eqs. (2) and (3) assuming a conditional normal distribution with BollerslevWoolridge robust standard errors and covariance. */**/*** denotes significance at the 10/5/1% level, respectively. 5.2. Looking at topics We now turn to the topics of the statements. First, we use occurrence dummies to test for effects of statements, regardless of the direction in which a variable was said to develop. We analyze two cases. In the first regression we include dummies for statements on interest rates and the euro category (see bottom half of Table 2). In the second regression we have replaced the dummy for statements on interest rates by dummies for statements on the first and second pillar of the monetary policy strategy of the ECB. We present two regressions because of a possible simultaneity problem: statements on inflation and growth may influence the exchange rate as well as trigger statements on interest rates. By estimating two equations, we try to separate these effects. The results are reported in Table 7. In the first regression, none of the statement dummies is significant in the mean equation (column 1 of Table 7). It follows that the ECB has not been successful in talking up the euro. However, statements on the future value of the euro have a significant effect in the variance equation. Volatility is about 15% lower on the day that statements on the euro are made. However, the day after the statement is made volatility increases 25%. This may be explained by the fact that heterogeneous agents start delivering comments after the statement, thus enlarging market uncertainty. The results for our second regression are comparable to those of the first. For intervention, however, we now find effects for the lagged value of the intervention dummy. This also means that over the two day period, the effect is significantly We thank the referee for this suggestion. D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343-361 357 Table 7 Results using dummies for content of statements (1) Mean equation Constant AR(1) Pass-through Intervention Intervention(1st lag) Variance equation Constant Pass-through Future euro Future euro (1st lag) No target versus inflation 51 0.42** <52 <53 Monday Friday Akaike IC (2) —0.00 0.02 - —0.00 0.03 —0.0018** —0.0005 0.0020** —5.66*** —0.28* 0.47*** - —6.81*** 0.53*** —0.34** 0.48*** —0.26* -0.12 0.03 0.29* -0.15* 0.01 —0.65*** —7.306 —0.57*** —0.23 —7.323 Note: This table shows results for Eqs. (2) and (3). Column (1) reports the results for the specification with dummies for statements on interest rates and dummies for statements on the external value of the euro, intervention and the no target/inflation trade-off. Column (2) reports the results if the dummy for statements on interest rates is replaced by dummies for statements on the two-pillar strategy of the ECB. Results are shown for errors that have a conditional normal distribution. Bollerslev-Woolridge robust standard errors and covariance are used. */**/*** denotes significance at the 10/5/1% level, respectively. different from zero. This result is in line with the findings of Fatum and Hutchison (2002). In addition, we find a small effect of statements on pass-through of changes in the euro-dollar exchange rate into inflation (depreciation of 0.20%). Once again, we do not find evidence that talking up the euro has been successful. Again, the coefficient of the dummy for statements on the future potential of the euro is significant in the variance equation. There is also evidence that statements on the trade-off between the euro as a target and inflation have led to lower volatility. 5.3. Looking at content Do the results change if we use dummies that measure the message contained in ECB statements? As before, we use two regressions. Table 8 shows the results. However, these authors have a dummy reflecting the presence of rumors on interventions, while our dummy refers to statements of ECB officials on interventions, which may not be the same. 15 Our analysis assumes that the causality runs from news to changes in the exchange rate. There may be an issue of reverse causality if central bankers react to currency market conditions. There is no evidence of this in probit regressions of all ECB statements on the current and lagged (average) change in the euro-dollar exchange rate. Finding causes for central bank communication may be a fruitful avenue for future research, nonetheless. 358 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 Table 8 Taking the message in the statements into account (1) (2) Mean equation Constant AR(1) Higher rates (1st lag) Higher M3 growth (1st lag) Pass-through neutral -0.00 0.03 -0.0027*** – – -0.00 0.03 – -0.0028*** -0.0017** Variance equation Constant Higher M3 growth Higher inflation Higher degree of pass-through Stronger euro Stronger euro(1st lag) Higher chance of intervention Higher chance of intervention(1st lag) d1 d2 d3 Monday Akaike IC -5.84*** – – – -0.36** 0.49*** 0.62* -0.68 0.41** -0.13* 0.03 -0.61*** -7.313 -5.50*** -0.96*** 0.41*** 0.44** -0.40** 0.58*** – – 0.44*** -0.17** 0.02 —0 70*** -7.333 Note: This table shows results for Eqs. (2) and (3). Column (1) reports the results for the specification with dummies for statements on interest rates and dummies for statements on the external value of the euro, intervention and the no target/inflation trade-off. Column (2) reports the results if the dummy for statements on interest rates is replaced by dummies for statements on the two-pillar strategy of the ECB. Results are reported for error terms that follow a conditional normal distribution. Bollerslev–Woolridge robust standard errors and covariance are used. */**/*** denotes significance at the 10/5/1% level, respectively. Columns (1) and (2) show that, once again, there is no evidence that talking up the euro has been successful. Furthermore, column (1) shows that statements indicating an interest rate rise are followed by a depreciation of approximately 0.3% the next day. This negative reaction may be explained by the negative effect of an interest rate rise on economic growth. In the conditional variance equation we find that volatility was lower on days during which positive statements on the euro were made. However, the next day volatility was almost 30% higher. Statements raising the possibility of intervention lead to higher volatility during the day, as may be expected, while the next day volatility is lower.16 In the second regression we find significant coefficients in the mean equation for money growth and pass-through. Comments on higher money growth lead to 16 The coefficient is marginally significant when using normally distributed errors: p Z 0.13. However, the significance is higher when assuming a t-distribution ( p Z 0.08) or a generalized error distribution ( p Z 0.11). D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 359 a depreciation of about 0.3%. Neutral statements on pass-through coincide with a depreciation of 0.17%. In addition, statements on money supply, inflation and passthrough enter the conditional variance equation. Statements on higher money growth lead to lower volatility, whereas statements on higher inflation and a higher degree of passthrough of the exchange rate into inflation lead to higher volatility. The results for statements belonging to the euro category are similar to those of the first regression. One difference is that positive statements on intervention no longer lead to changes in volatility. Overall, we find evidence for asymmetric reactions to news as markets react differently to positive and negative statements on similar topics. For example, comments indicating interest rate rises cause the exchange rate to move, whereas comments pointing towards lower interest rates do not. This result is in line with the findings of Galati and Ho (2003), Fatum and Hutchison (2002), and Andersen et al. (2003). 6. Conclusion To what extent did ECB statements influence the euro–dollar exchange rate? We show that effects of ECB statements on the conditional mean of the exchange rate, if any, are comparatively small. Moreover, the effects are not persistent: statements rarely have significant effects over the two day period after the statement. In contrast, ECB statements have had considerable impact on the conditional volatility. Firstly, there is evidence that for the ECB officials as a group volatility increases after policy statements are made. Secondly, more statement dummies enter the variance equation than the mean equation. Thirdly, effects on volatility may be quite large. We conclude that central bankers should be careful with their comments, as long as they perceive higher volatility as welfare decreasing. The implication may be that, in order to secure this, communication should be more centralized. One way to achieve this would be for national central banks to step back and let the ECB president handle the communication to financial markets more efficiently. We show that efforts of the ECB to talk up the euro have been without result. Statements that were intended to bolster the external value of the euro did not lead to an appreciation of the euro. Evidently, there was no reason for the market to react to these statements. Importantly, these statements have led to higher volatility. From this perspective, the advice to the ECB would be to not use this particular strategy again in the future, as volatility may increase uncertainty. However, an interesting avenue for future research is to examine the more recent experience with ECB verbal intervention intended to talk the euro down, as there is tentative evidence that this strategy has been more successful. Some types of statements on monetary policy have had influence on the conditional mean of the exchange rate. In particular, the results show a negative relationship between statements on higher interest rates and the strength of the euro against the dollar. Finally, there is evidence of asymmetric reactions to news. Markets respond differently to positive or negative news from the same category. 360 D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 Acknowledgements The authors thank Elmer Sterken, Lex Hoogduin, Peter van Els and Peter Vlaar for useful discussions. This paper benefited greatly from comments by Frank Heinemann, Yin-Wong Cheung, Frank Westermann and other participants of the CESifo Area Conference on Macro, Money and International Finance (Munich, 21– 22 February 2003). Comments from participants of the 6th European Workshop on EMU (University of Crete, 25–31 August 2003) and the 8th International Conference on Macroeconomic Analysis and International Finance (University of Crete, 27–29 May 2004) are also gratefully acknowledged. Finally, the authors thank Ansgar Belke and an anonymous referee for useful comments. We also thank Het Financieele Dagblad for giving us the opportunity to use Bloomberg. All remaining errors are our own. The views in this article do not necessarily coincide with those of De Nederlandsche Bank. References Almeida, A., Goodhart, C.A.E., Payne, R., 1998. The effects of macroeconomic news on high frequency exchange rate behaviour. Journal of Financial and Quantitative Analysis 33 (3), 383–408. Andersen, T.G., Bollerslev, T., 1998. Deutsche mark–dollar volatility: intraday activity patterns, macroeconomic announcements, and longer run dependencies. Journal of Finance 53 (1), 219–265. Andersen, T.G., Bollerslev, T., Diebold, F.X., Vega, C., 2003. Micro effects of macro announcements: real-time price discovery in foreign exchange. American Economic Review 93 (1), 38–62. Cheung, Y.W., Westermann, F., 2001. Equity dynamics before and after the introduction of the euro: a note. Multinational Finance Journal 5 (2), 113–128. Cheung, Y.W., Chinn, M.D., 2001. Currency traders and exchange rate dynamics: a survey of the US market. Journal of International Money and Finance 20 (4), 439–471. De Grauwe, P., Dewachter, H., Embrechts, M., 1993. Exchange Rate Theory: Chaotic Models of Foreign Exchange Markets. Blackwell, Oxford. DeGenarro, R.P., Shrieves, R.E., 1997. Public information releases, private information arrival and volatility in the foreign exchange market. Journal of Empirical Finance 4 (4), 295–315. Doukas, J.A., Melhem, M., 1986. The reaction of spot and forward rates to new information. European Economic Review 30 (2), 305–324. Doukas, J.A., Rahman, A., 1986. Foreign currency futures and monetary policy announcements. Journal of Futures Markets 6 (3), 343–373. Ederington, L.H., Lee, J.H., 1993. How markets process information: news releases and volatility. Journal of Finance 48 (4), 1161–1191. Eijffinger, S.C.W., De Haan, J., 2000. European Monetary and Fiscal Policy. Oxford University Press, Oxford. Evans, M.D.D., Lyons, R.K., 2003. How is macro news transmitted to exchange rates? National Bureau of Economic Research, NBER Working Paper 9433. Fatum, R., Hutchison, M.M., 2002. ECB foreign exchange intervention and the euro: institutional framework, news, and intervention. Open Economies Review 13 (4), 413–425. Fatum, R., Hutchison, M.M., 2003. Is sterilized foreign exchange intervention effective after all? An event study approach. Economic Journal 113 (487), 390–411. Frenkel, J.A., 1981. Flexible exchange rates, prices and the role of ‘news’: lessons from the 1970s. Journal of Political Economy 89 (4), 665–705. Frenkel, M., Pierdzioch, C., Stadtmann, G., 2001. The foreign exchange market interventions of the European Central Bank. Banca Nazionale del Lavoro Quarterly Review 54 (218), 249–287. D. Jansen, J. De Haan / Journal of International Money and Finance 24 (2005) 343–361 361 Galati, G., Ho, C., 2003. Macroeconomic news and the euro/dollar exchange rate. Economic Notes 32 (3), 371–398. Goodhart, C.A.E., Hall, S.G., Henry, S.G.B., Pesaran, B., 1993. News effects in a high-frequency model of the sterling–dollar exchange rate. Journal of Applied Econometrics 8 (1), 1–13. Hakkio, C.S., Pearce, D.K., 1985. The reaction of exchange rates to economic news. Economic Inquiry 23 (4), 621–636. Harris, E.S., Zabka, N.M., 1995. The employment report and the dollar. Current issues in Economics and Finance, Federal Reserve Bank of New York. Hsieh, D.A., 1988. The statistical properties of daily foreign exchange rates: 1974–1983. Journal of International Economics 24 (1/2), 129–145. Hsieh, D.A., 1989. Modeling heteroscedasticity in daily foreign exchange rates. Journal of Business and Economic Statistics 7 (3), 307–317. Ito, T., Roley, V.Vance, 1987. News from the US and Japan: which moves the yen/dollar exchange rate? Journal of Monetary Economics 19 (2), 255–277. Jansen, D., De Haan, J., 2004. Look who’s talking: ECB communication during the first years of EMU, CESifo Working Paper 1263. Nelson, D.B., 1991. Conditional heteroscedasticity in asset returns: a new approach. Econometrica 59 (2), 347–370. Rigobon, R., Sack, B., 2002. The impact of monetary policy on asset prices. National Bureau of Economic Research, NBER Working Paper 8794. Sarno, L., Taylor, M.P., 2001. Official intervention in the foreign exchange market: is it effective and, if so, how does it work. Journal of Economic Literature 39 (3), 839–868. Tivegna, M., 2002. News and dollar exchange rate dynamics. Rivista Italiana degli Economisti 7 (1), 3–48.