Excerpt
Table of Content
List of figures and tables
1. Introduction
2. How do markets react to macro news?
2.1. What is news and how does it affect the mean and volatility of exchange rates?
2.2. Velocity of news absorption
2.3. Context of news: What impact does it have?
2.4. Asymmetric reaction to news
3. Central Banks
3.1. Why are Central Banks and especially the ECB of special academic interest?
3.2. How can we classify the existing research in the ECB-field?
3.3. What can central banks achieve by communication?
4. Empirical Evidence for exchange rate communication efficiency
4.1. Which existing papers have similar empirical focus?
4.2. How do existing papers measure communication?
4.3. What econometric methodology has been used?
4.4. What effects could the existing research discover?
4.5. What do those results mean for the actions of the ECB?
5. Own empirical findings
5.1. Data sources
5.2. Which is the most appropriate regression?
5.3. Why could lag specification be helpful?
5.4. What is the economic reasoning behind the sign and the size of the coefficients?
5.5 Which results does further testing yield?
5.6 Interpretation of Results / Implications for ECB
6. Conclusion
7. References
8. Tables
9. Graphs
List of figures and tables
Table 5.0.: ECB President announcements about future monetary policy moves
Table 5.1.: Communication variable as explanatory factor for federal funds rate changes: High t-stat indicates that communication variable dummy has high explanatory power.
Table 5.2: Various specifications of the Y variable to find the best specification:
a) Actual value of exchange rate
b) Log of exchange rate.
c) First difference of actual value
d) First difference of log
Table 5.3.: Initial result of OLS equation that intends to predict exchange rate changes through a communication dummy variable and changes of the federal funds rates in the Euro- Area and the US
Table 5.4.: First test if lagged x-values could be relevant by using geometric lag:
Table 5.5: Regression in which the change in the Euro/Dollar exchange rate is partly explained by lagged value of the change in the federal funds rate of the US:
Table 5.6: Dickey-Fuller test to see if the time series are stationary: They all are!
a) Dependent variable:
b) Communication dummy
c) ECB interest rate changes
d) US Fed interest rate changes
Table 5.7: Correlation matrix shows that multicollinearity is no issue (low correlation between dependent variables):
Table 5.8: Application of the White test to see if Heteroskedasticity is present, which is not the case:
Table 5.9: Application of the Brusch-Godfrey serial correlation test to check if error terms in our sample are related, which is not the case:
Table 5.10: Results of the Ramsey-Reset test show that the chosen functional form of the estimated regression equation is obviously correct:
Table 5.11.: Chow breakpoint test to see if market reaction to news might have changed throughout the sample period. This does not seem to be the case
Table 5.12.: Explanation of the actual level of the exchange rate by interest rate levels in the us and Europe
Graph 1.1.: Hierarchy of News:
Graph 5.1.: Euro/Dollar spot exchange rate:
Graph 5.2.: Federal Funds Rates USA and Euro-Area 2000-2006:
Graph 5.3.: First difference of the log of the Euro/Dollar exchange rate:
Graph 5.4: The cumulative sum of squared errors stays within the error band throughout most of the observed period:
Graph 5.5: Explanation of exchange rate level by interest rate levels
1. Introduction
This paper intends to analyze the effect of ECB communication on exchange rates. Starting with an overview of relevant theoretical aspects and previous research in the field we finally conduct our own study to see if communication does effectively influence the Euro/Dollar exchange rate. Hence this paper follows the subsequent structure:
In the first part (chapter 2) we present how currency markets in general react to news. This means we analyze studies that show how news are incorporated in prices for foreign exchange. These news have a wide origin and are not necessarily related to central bank communication in general or ECB communication in particular. The exceedingly complex process of how markets process these information has not been entirely understood despite high research efforts in this field as different studies come to diverging results, a fact that will be exemplified further on in this paper.
In the second part (chapter 3) we turn towards central banks and especially the ECB and explain why it has lately drawn so much research attention towards its actions. In our third part (chapter 4) we focus in more detail on those studies that specifically analyze the effects of ECB communication and thereby open the field for our own study that will be presented in the fourth part (chapter 5). In alignment with part of the analyzed existing research our findings support the fact that ECB communication does not significantly affect exchange rate returns. Nevertheless we are well aware of shortcomings with regard to the frequency of our data and the deficiency of the used econometric methods. Part five (chapter 6) concludes.
2. How do markets react to macro news?
In order to start our analysis of how markets respond to news we first have to understand what we want to express when we talk about “news” and its impact on exchange rates in general. Furthermore we need a deeper comprehension of how quick news are absorbed and if there are special contextual elements that are capable of influencing the overall effect. Moreover special environmental conditions can cause asymmetric market response to news, an issue that is discussed in the last sub-section of this chapter.
2.1. What is news and how does it affect the mean and volatility of exchange rates?
Before we can analyze the effect that the occurrence of macroeconomic news has on markets it is necessary to define the term “news” since its general meaning differs from the scientific terminology. Markets for currencies are confronted with a perennial stream of diverse information. Before we analyze different aspects of news it is necessary to distinguish between news and information. Then we have to classify different categories of news. News are generally regarded as unexpected information. If certain information are already expected markets can and will anticipate their occurrence and adjust accordingly. These anticipated information are not new anymore when they are officially announced, therefore they cannot be “news”. Old news do not exist. In line with the efficient market hypothesis it can thus be claimed that, as Frenkel (1981) puts it: “If the foreign exchange market is efficient, we should expect current prices to reflect all currently available information.” Within the field of news it is possible to classify news according to their character: There are news relating to certain economic agents (micro-news) that cannot have a systematic effect on the market in total because the total market is not affected. Then there are macroeconomic news of which it can be expected that the market will react to.
Macroeconomic news can then be further classified according to their origin: One example are news that are conveyed by research institutions, such as the ifo-Geschäftsklima-Index. Another example on which we will focus more strongly in the second half of this paper are news of which a central bank is the originator, e.g. the ECB. Anything unexpected that is done by a central bank can be regarded as central bank news. This implies that actions as well as communication can fall in this category of news, although they do not have to since they could already be expected.
An example for the fact that news can also be separated in several alternative ways is the study of DeGenarro, Shrieves (1997). In their work news are separated into three categories: Macro announcements, interest rates news and economic policy reports. Graph 1.1 shows our understanding of news-classification which we outlined in this paragraph. After having understood what news is we can now shift our attention to its effect on the mean and the volatility of exchange rates. Extensive research has been performed in this field as is now illustrated. Some studies which focus on the ECB will also be analyzed in later paragraphs of this paper however under a different perspective so we think this not to be redundant. Jansen, de Haan (2005) find out th]at news issued by the ECB mainly influence the conditional volatility whereas concrete impact on the mean was relatively rare in the first years of the European Monetary Union (EMU). The authors follow from that that efforts of talking up the euro have not been successful but increased market uncertainty through volatility increase.
Fratzscher (2005) contradicts by ascertaining that in his study exchange rate communication proved to be an effective tool for central banks with a perennial effect on exchange rates. Another study from the same author in 2004 focuses on the short-term already and suggests “that oral interventions might constitute, on a short-term basis, an effective and largely autonomous policy tool” and that the effectiveness of this tool is independent of the presence of real interventional steps by the central bank. Moreover the paper points out that oral interventions reduce the volatility of exchange rates in contrast to an increasing momentum created by actual interventions. Ehrmann, Fratzscher (2005) come to the result that economic news are a crucial force in influencing daily movements of the us-dollar-euro/deutsche mark exchange rate in a period lasting from 1993 to 2003. Their model explains around 75% of monthly directional changes. However it does not give satisfying results concerning the magnitude of the alterations in the exchange rates.
Another study that analyzes the effect of news on exchange rates is Andersen, Bollerslev, Diebold, Vega (2003). The authors come to the result that generally speaking news has a statistically significant impact on the exchange rate. In this context they ascertain that the adjustment of the mean is very rapid whereas the adjustment in volatitility follows a more gradual pattern.
Ederington, Lee (1993) arrive at the conclusion that the major adjustments in prices take place during the first minute after the announcement whereas volatility remains significantly higher than the normal level for around fifteen minutes and slightly higher for a couple of hours. Galati, Ho (2003) find a statistically significant correlation of macro news and daily movements of the euro-dollar exchange rate. At the same time they grasped that this relationship exhibits a considerable time variation and “may be short term at best”. As can be seen from this brief literature review the findings of previous research are mixed and somewhat inconsistent, which can also relate to the fact that this field of study is relatively new compared to the vast amount of research that has been conducted with regard to interventions of central banks.
2.2. Velocity of news absorption
We now know what news is and what preceding research has found out about its impact on the mean and volatility of exchange rate. To extend this general knowledge we continue by tackling the issue of the speed with which news are integrated into exchange rates and how persistent the observed effects are.
According to the findings of Cheung, Chinn (2001) the biggest part of the adaptation of the market as a result of news occurs within 60 seconds after news are released. Ederington, Lee (1993) come to the same result that the bulk of adjustments takes place in the first postannouncement minute whereas volatility remains significantly higher than the normal level for around fifteen minutes. Also Almeida, Goodhart, Payne (1998) ascertain that the effect is not persistent as most announcements only create a significant impact in the 15 minutes postannouncement phase. If this interval is extended the significance decreases at a great pace. These findings are in line with Jansen, de Haan (2005) who state that the effects of macroeconomic announcements are not persistent.
This contradicts the findings of Evans, Lyons (2005) which conclude in their study that macro news cause a dynamic impact which can need several days until it is entirely integrated into foreign exchange market prices. This is confirmed by Fratzscher (2005) who arrives at the result that communication by the ECB moves the exchange rate in the desired direction in a five-day post-event period in the majority of scrutinized cases. This directional effect even proves a high persistence and similar success rates if the period is extended up to 40 days after the news announcements.
As news absorption appears to be a quick process, the results of research papers could depend on data frequency they use. A comparison of sources reveals the following quintessence: Major adjustments take place very quickly after announcements thus the less precise the timeintervals of data recording are, the less significant are the results. Therefore studies that rely on high frequency data such as Almeida, Goodhart, Payne (1998), Ederington, Lee (1993), Andersen, Bollerslev, Diebold, Vega (2003) and Evans, Lyons (2003) find statistically more significant results in the short time after an announcement was made.
Consequently the use of high-frequency or real-time data i.e. information available to economic agents when they make their investment and trade decisions seems to provide more valuable conclusions as the used data is more precise. Unfortunately our own study will only rely on daily data.
2.3. Context of news: What impact does it have?
After having understood the general impact of news on the exchange rate we know take into consideration the larger context of news which can possibly alter the overall effect. First the timely context is scrutinized (A), second divergences in the place of origin (B).
(A) It is not merely the content of news that matters but also the timely context i.e. the point of time of a macronomic announcement can influence the impact of news on the exchange rate. Andersen, Bollerslev (1998) call this phenomenon the “calendar effect” which can be further divided into time change effects, holiday effects, weekly effects and intraday effects. Fratzscher (2005) adds to this that exchange rate communication tends to be more successful if it goes in the same direction as the trend before the announcement (“leaning with the wind”), if it appears in periods of extensive uncertainty in the market and when the exchange rates are not aligned appropriately. However, Fratzscher (2004) states that oral interventions have a greater influence “if they deviate from the prevalent policy mantra”.
Ederington, Lee (1993) realize that prices in interest rates and foreign exchange markets are a lot more volatile between 8:30 and 8:35 A.M. eastern time (ET) in comparison to any other five-minute trading period including the open and close. The authors explain this by the fact that several major macro news e.g. CPI or employment report are released at 8:30 A.M. (ET). From that they conclude that the observed volatility patterns are mainly due to the timing of major macro announcements. When the influence of these announcements is removed, the volatility, according to the observations of the authors appears to be basically flat across both the trading week and the trading day.
(B) Apart from the point of time, the place of origin of news offers room to elucidate possible influences on the impact on exchange rates. Almeida, Goodhart, Payne (1998) compare the effects of U.S. and German central bank announcements over a three year period from 1992 to 1994. Their study points out important differences in the reaction of currency markets. It appears that the deutsche mark-dollar exchange rate was driven more by US than by German announcements In this context it is important to ascertain that German Bundesbank announcement did not have pre-set and pre-advertised release dates. Hence the authors make up the hypothesis that an adaptation of the exchange rate is completed more rapidly if the announcement is scheduled in advance as the market participants are thus enabled to make higher commitments to thorough pre-planning. This hypothesis is confirmed by the result that the impact of an announcement becomes greater the closer it is to Bundesbank Council Meetings and that German announcements retain significance for a longer period of time in comparison to their US counterparts.
Other studies also focus on market reactions to macro news from different countries. An example for this give Ito, Roley (1987) who focus on the yen-dollar exchange rate and find out that the currency market responds to macro news from the US but do not react to those which are of Japanese origin.
Another reason for the fact that different announcements yield different reactions may be explained by different levels of credibility of the respective central banks as “communication is so crucial for the success of monetary policy precisely because it has an important influence on expectations and credibility” (Fratzscher 2004).
However central banks in general might even have an incentive to give out false information if they have political goals that diverge with the predicted currency market development. Goodfriend (1986) and Stein (1989) analyze this problem which is known as “time inconsistency problem”. Different solutions are proposed to this problem. Vitale (2003) suggests that the central bank should do actual interventions as they increase credibility by a direct commitment whereas Stein (1989) states that it is a solution for the central bank to provide imprecise announcements concerning its views about the currency markets.
2.4. Asymmetric reaction to news
If the context of news which was scrutinized in general in the preceding section plays an crucial role, this can lead to asymmetric market reactions e.g. exchange rates only react to announcements from a certain country or markets may pay more attention to bad news while ignoring good. This phenomenon has been subject to extensive research of which we here want to present the major findings.
Jansen, de Haan (2005) find some evidence for asymmetric foreign exchange market response during the first years of the EMU. This means the market shows different reactions to positive and negative statements on comparable topics e.g. news indicating a rise in interest rates provokes an exchange rate move whereas on the other end news commenting on lower interest rates do not.
Further evidence for asymmetric response is found by Andersen, Bollerslev, Diebold, Vega (2003). In their study which focuses on the spot rate of the US dollar against five other currencies they find that negative surprising news influence the exchange rate more significantly than positive do. Moreover negative surprises have a greater impact if they occur in times that are dominated by a flow of goods news.
Fatum, Hutchinson (2002) scrutinize among other influential factors the effect of statements by officials not supportive of the euro and statements by officials in support of the euro. They come to the result that only the negative statements have persistent impact on the euro-dollar exchange rate.
Galati, Ho (2003) reveal in their study certain indicators for asymmetric market behavior which is present to different extents at different points of time. Their results indicate that the scrutinized markets appear to remain fixated on bad news and ignore good news, especially from the euro area in a certain periods of time during the first two years of EMU. Moreover they discover that the impact of macro news is stronger when a switch from bad news to good news (or the other way round) is chronicled.
Finally, Ehrmann, Fratzscher (2005) find out that macro news from the US have a larger impact than German or euro area news, between 1993 and 2003. They explain this phenomenon “at least partly” by the earlier release time of the US news apart from the fact that the US economy has a relatively greater importance on the world economy. Moreover they find supporting evidence for asymmetric response as “negative and large news have a disproportionately larger effect on exchange rates than positive and smaller announcement surprises”.
3. Central Banks
3.1. Why are Central Banks and especially the ECB of special academic interest?
Markets are interested in central bank activities for two major reasons (a) and (b). A: central bank activity can convey information i.e. knowledge that has been private before such as private information of the monetary authorities about the economy and information about future monetary policy (see Mussa, 1981). B: Besides that a central bank can influence market parameters (interest rate and amount of domestic credit) that open or close profit opportunities for economic agents. For these reasons actions of central banks have been the focus of increasing research attention in recent years. Moreover the general attitude of central banks vis-à-vis the general public has changed considerably: While in the past central banks tried to hide their monetary policy (Goodfriend, 1986), today they prefer to be rather transparent and reliable (Ferguson, 2004) thus allowing for precise and accurate data that is suitable for academic endeavours, some of which is presented in this paper.
One institution that has drawn a specifically high amount of attention to its activities is the European Central Bank (“ECB”). A reason for this fact is the introduction of the Euro and the responsibility shift from the different national central banks towards a newly created institution. Financial markets consequently have to learn, and as Jansen, de Haan (2004) state “to get fully acquainted with the way the ECB would operate.”
Another reason why the analysis of the ECB in particular is of high interest relates to the previously mentioned credibility issue, as credibility is especially crucial for relatively new central banks as the ECB. Jansen, de Haan (2004) find out that the Deutsche Bundesbank still has a great reputation among market participants and therefore could be a source of disturbance for the ECB if they communicate diverging opinions although the study points out that “in some respects ECB communication has improved over the years”. A later study of Jansen, de Haan (2006) confirms that the relative youth of the ECB can cause problems as it “faces the continuous challenge of explaining monetary policy to a diverse audience.” Our paper will thus be another example of this rapidly increasing list of research.
3.2. How can we classify the existing research in the ECB-field?
This part of our paper intends to briefly outline the different branches of academic research in the above-highlighted area and then classifies our contribution within the existing portfolio of research.
The first branch of literature tries to find out the intentions of the ECB. One example of these kind of papers is Berger, de Haan, Sturm (2006) in which the authors try to find out whether monetary policy actually matters or not. (They find out that it only plays a minor role in overall ECB policy.)
The second branch focuses on the ECB and whether it is able to achieve its goals via those means that it makes use of: Interventions and communication. One example for an empirical intervention effectiveness test is Fatum, Hutchison (2003). They find out that sterilised intervention is able to influence the exchange rate, however only in the short run. Since our paper focuses more on communication we abstain from listing further literature with regard to this research issue, however a good overview of existing research and its findings can be found in Sarno, Taylor (2001).
The focus of our study however lies on ECB communication. Here it can be analyzed whether this communication can be looked upon as effective. In the literature the notion of “effectiveness” generally means that its communication had a certain noticeable, for e.g. via regression detectable, statistically significant effect. If the object of study is the exchange rate, the effectiveness hypothesis proposes that the exchange rate should have been altered as a result of the act of communication, or as Fratzscher, 2004 writes: “Effectiveness is defined as the systematic change of the exchange rate level and volatility in the desired direction on intervention days.” Although this last statement refers to intervention, this notion can certainly include communication which then is simply classified as “oral intervention”.
We also intend to analyse the effectiveness of ECB communication by conducting various types of regression analysis to find out whether “news” that origin from the central bank can change the exchange rate.
3.3. What can central banks achieve by communication?
Literature review reveals that there seem to be two dominant ways in which communication can be positive for central banks: According to Jansen, de Haan (2004) communication can influence the market expectations with regard to future actions and long term behaviour. This gives the central bank more flexibility to respond to short term shocks, as these short-term responses will not astonish the market. King (1997) points out that these short term reactions could otherwise be interpreted as an “indication of a lack of commitment”. If investors would, contrary to the intentions of a central bank, interpret their short-term action from a long-term perspective they might be mislead and may take decisions that are ineffective both for them as individual economic agents and for the overall level of welfare.
The second positive effect of central bank communication relates to the fact that its “communication may be used to reduce noise in the market” (Posen 2003): If a central bank is predictable, the potential risk resulting from an unpredictable central bank action is eliminated and consequently the uncertainty in financial markets is lower. This reason is further backed up by Kuttner (2001).
4. Empirical Evidence for exchange rate communication efficiency
4.1. Which existing papers have similar empirical focus?
Before we present the results of our own study we use the next part of our paper to analyze and compare the existing research with regard to the effects of ECB communication from several different angles. We especially focus on the used data and proxies to measure ECB communication, the undertaken econometrical analysis and the results. Finally we agglomerate the derived implications for the ECB. This cross-analysis draws on the evidence found in five different papers: Jansen, de Haan (2005), Jansen, de Haan (2006), Fratzscher (2004), Fratzscher (2006) and Rosa, Verga (2006). The following analysis does not aim at systematically mentioning all aspects and details of existing papers as could be expected from a comprehensive and exhaustive (but also exhausting!) review but rather intends to draw attention to those points that are remarkable and influenced the way we set up our own study.
4.2. How do existing papers measure communication?
Jansen, de Haan (2005) focus their analysis on the statements that are made by central bank officials. These statements are divided into two main categories. (monetary policy (a) and the Euro itself (b)). Category (a) is then further detailed into the relevant factors of monetary policy such as growth, inflation etc. while (b) is split up along aspects that relate to the euro currency itself such as interventions or the future value of the euro. For each data point certain characteristics are recorded such as “day of announcement” (this implies an analysis based on daily data), “origin of statement” (such as ECB president or any other ECB official) and “content”. Since statements are qualitative, empirical analysis however draws heavily on econometric (=quantitative) methods dummy variables are constructed to capture the effect of the ECB statements. These dummies can assume the values “-1”, “0” and “1” to incorporate whether statements concerning economic variables (for category (a) such as interest rates) are supposed to go down (“-1”), remain neutral (“0”) or go up (“1”) while the same dummy values with regard to the euro and its value (thus for category (b)) signify negative, neutral or positive. The analyzed time horizon goes from 4 January 1999 until 17 May 2002 which amounts to 881 trading days. Data source is the Bloomberg News service. Jansen, de Haan (2006) makes basically use of the same data set.
Fratzscher (2004) also looks for statements made by relevant policy makers. His dataset however is considerably longer (1 January 1990 to 30 June 2004) and he uses Reuters News instead of Bloomberg. The only information that is recorded is whether the statements are positive (“strengthening”), negative (“weakening”) or neutral (“ambiguous”). Here the object of consideration is solely the exchange rate. Further detailing as can be observed in Jansen, de Haan (2005) does not take place. Fratzscher (2005) uses the same data set.
Rosa, Verga (2006) do not make use of news-service transmissions but focus exclusively on the introductory statement that is made at the beginning of the monthly press conferences of the President of the ECB. Along the lines of content analysis (see Krippendorf, 2004) the authors systematically extract the content of the given messages and forge it into a five-value scale (-2 to +2). The value zero for them means that the intended statement means that the ECB sees the present level of the Repo rate as suitable, while positive values are assigned to comments that imply as Rosa, Verga (2006) state it “a (strong) likelihood of future monetary policy tightening”. As can be inferred from these details, Rosa, Verga (2006) do specifically focus on the monetary policy moves. Since these data are used to explain movements in the Euribor Futures Market, however no analysis with this data set with regard to exchange rate changes has been carried out, our paper intends to fill this gap. This endeavour is even more reasonable as the other two data sets mentioned in this part of our paper have already been used twice and thus reliably proven their worth while our intention is to do this very task for the Rosa, Verga (2006) data set that has only been used once until now, according to our knowledge.
[...]