You are on page 1of 31

German Economic Review 10(1): 1–31

Which News Moves the Euro Area Bond Market?
Magnus Andersson
European Central Bank

Lars Jul Overby
University of Copenhagen/ Danmarks Nationalbank

´n Szabolcs Sebestye
Catholic University of Portugal

Abstract. This paper explores a long dataset (1999–2005) of intraday prices on German long-term bond futures and examines market responses to major macroeconomic announcements and ECB monetary policy releases. German bond markets tend to react more strongly to the surprise component in US macro releases compared with aggregated and national euro area and UK releases, and the strength of those reactions to US releases has increased over the period considered. We also document that the numbers of German unemployed workers consistently have been known to investors before official releases. JEL classification: E43, E44, E58. Keywords: Monetary policy; high-frequency data; macroeconomic announcements.

1. INTRODUCTION
What causes financial market prices to undergo the sometimes strong swings observed during a trading day? The answer to this basic question is of utmost interest to anyone monitoring financial markets – from central banks using asset prices to gauge investors’ macroeconomic expectations, to fund managers and traders exploring buying and selling opportunities from the prices fluctuating on their computer screens. As financial assets are inherently forward-looking, only new news should cause revisions to what is currently built into asset prices, thereby immediately affecting prices. The availability of prices at very high frequencies allows for an in-depth analysis of the price discovery process. This in turn enables a ‘cleaner’
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009, 9600 Garsington Road, Journal Compilation r Verein fu Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

M. Andersson et al. analysis of market reactions surrounding major market-moving events compared with commonly used daily data, where other news during a trading day may blur instantaneous market reactions to events. In addition, the use of high-frequency data makes it possible to depict the dynamic market reactions to the constant flow of information. The purpose of this paper is twofold. First, to examine the effects of macroeconomic data releases and of the ECB’s monetary policy statements on five-minute intraday German government bond returns from the beginning of 1999 until the end of December 2005. Second, allowing for the fact that long-term bond yields in theory should be affected by the expected monetary policy path, the paper examines whether the German bond price sensitivity differs depending on the monetary policy stance taken by the ECB. Macro releases from a wide range of economies are tested. As a result, we account for both euro area-specific news as well as international spillover effects. We use aggregated euro area releases as well as German, Italian, French, Spanish and Belgian announcements to gauge the effects of euro arearelated news. To account for the international linkages of the euro area economy and its markets, we include data from the two largest trading partners – the United States and the United Kingdom. The main findings are that German bond markets tend to react more strongly to the surprise component in US macro releases compared with aggregated and national euro area releases and UK releases, and the strength of those reactions to US releases has increased over the period considered. The level of bond returns appears to adjust quickly to new information, whereas macro announcements have a stronger and more long-lasting impact on volatility. In addition, by splitting our sample into three different monetary policy regimes in the euro area, we find that the sensitivity of bond returns to US economic activity and employment news differs depending on the regime. The international spillover to the German bond market mainly stems from US announcements, whereas UK releases do not seem to have any significant impact. This is perhaps slightly surprising, given that the United Kingdom is one of the largest trading partners for the euro area. Thus, the international financial linkages observed in the German bond market do not appear to be based on real economic linkages. The paper contributes to the existing literature in a number of ways. First, the responses of the benchmark German long-term bond returns on macroeconomic and monetary policy announcements are examined, a topic that has received little attention in the empirical literature. Second, the studies within this strand of the literature have usually narrowed the examination to US and/or aggregated euro area announcements. Therefore, the combined use of euro area, German, French, Italian, Spanish, Belgian and UK macroeconomic releases, in addition to the ‘traditional’ US macroeconomic announcements, is novel. Third, we identify problems with macroeconomic releases that have not been noticed in previous announcement studies. In particular, evidence is provided that the outcome of the German employment reports is known to

2

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? investors ahead of the prescheduled release. This finding explains why our and previous high-frequency studies find no significant impact on financial market prices around the time of the official release. To enable an accurate examination of the impact from the German employment report, we identify prescheduled release times using news wires, which reduces the measurement errors. The results from the corrected specification, however, still suggest that the employment report does not significantly impact long-term bond yields in Germany. Finally, we provide evidence that the sensitivity of bond returns to news is not constant but varies across monetary policy regimes. The paper is structured as follows: Section 2 briefly reviews the literature within these fields, while Section 3 elaborates on the data used in the study. The econometric model and the results are reported in Section 4. Section 5 examines whether the price discovery process for the German bond markets differs depending on the monetary policy stance. Finally, Section 6 makes some concluding remarks.

2. RELATED LITERATURE
Overall, the literature about the macro announcements’ impact on asset prices is large and spans across asset classes. Numerous studies have analysed financial markets in the United States (see, for instance, Andersen et al., 2007; Balduzzi et al., 2001; Dwyer and Hafer, 1989), while less attention has been paid to euro area, UK and Japanese markets. As regards announcement studies applied on the euro area bond markets, the focus in these papers has been on the impact stemming from US macro announcements. In general, the findings support the notion that US data releases indeed not only affect US markets but also exert a significant effect on the European bond markets (see Andersen et al., 2003, 2007; Faust et al., 2007). The procedure used by Ehrmann and Fratzscher (2003, 2005) is slightly different as it focuses on the impacts of monetary policy and macroeconomic announcements both in the euro area and in the United States on the money market rates in the two economies. They show a high and increasing interdependence between the euro area and the United States, with euro area money market rates reacting more strongly to US data releases in comparison with US interest rate reaction to euro area announcements. This increasing interdependence may be linked to higher financial integration between the two economies, and US data may be considered as a leading indicator for euro area markets by market participants. The impact of monetary policy announcements on financial markets has received considerable attention, although the focus has primarily been on the impact on the stock and foreign exchange markets.1 The effect on bond
1. See Bentzen et al. (2004), Bomfim (2000) or Wongswan (2005) for the impact on stock markets, and Andersen and Bollerslev (1997, 1998) or Faust et al. (2002) for the impact on foreign exchange markets.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

3

M. Andersson et al. markets has received less attention. Similar to the macro announcement literature, the bulk of the studies have been applied to the United States (see Fleming and Piazzesi, 2005; Fleming and Remolona, 1997). For the euro area, both Bernoth and von Hagen (2004) and Ehrmann and Fratzscher (2003, 2005) study the volatility reaction on money market rates following ECB Governing Council announcements. They both find that volatility generally tends to be higher on these days.

3. DATA ISSUES
The bond market data consist of five-minute intraday prices of long-term German government bond futures contracts from the beginning of 1999 until the end of December 2005. The dataset was purchased from TickData Inc. The eligible delivery bonds are usually a basket of both non-benchmark and benchmark German government bonds with a remaining term of between 8.5 and 10.5 years. From the price data, bond returns are calculated as 100 times the logarithmic difference between consecutive prices. Table 1 shows all macro announcements and also highlights the distribution of the release days of the 44 macroeconomic announcements used in this study. As can be seen in Table 1, most euro area macro data are released later than the US equivalents. The delayed release of the aggregate euro area statistics is linked to the time needed to compile the statistics from all EMU member states. The delayed release of euro area macroeconomic statistics also implies that they potentially contain less new news as the national releases are already known to the investors at the time of publication. To account for this, the most important national German, French, Italian, Spanish and Belgian macro releases are also included in this study.2 German GDP is not included, as it is typically published before the German bond markets open, but overall, the dataset covers most of the macroeconomic information that is typically considered important for a fundamental analysis. In addition to the 44 macro announcements, the paper also examines the German bond market responses following actual monetary policy decisions by the ECB and the accompanying releases of the Introductory Statements. Given the strong real economic linkages between the United Kingdom and the euro area, a comparable set of UK announcements are also included.
2. We kept a broad set of US indicators and used a reduced and comparable set of CPI, business confidence, employment and industrial production indicators for other countries. It was only possible to compile this set of indicators for German, French, Italian and Spanish releases among the euro area countries, as survey expectations for the remaining euro area countries were either missing or too irregular. Finally, Belgium business confidence was included as this indicator frequently is quoted among financial analysts due to its close connection to the IFO indicator.
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

4

Which News Moves the Euro Area Bond Market? Table 1 Release days of macroeconomic announcements used in this study

Notes: The table shows the timing of macroeconomic news releases for month X. The earliest available release date for the macroeconomic releases has been chosen, which implies that some releases are only preliminary and subject to further revisions, such as University of Michigan and German CPI. Furthermore, the release times are only indicative, as holidays and other events may move publication dates and represent the current typical publication times. US GDP Final is not included – this release is published in month X þ 3, just as initial jobless claims (United States) is not included, as this data release comes weekly on Thursdays with data from Friday the week before.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

5

M. Andersson et al. When measuring financial market impact from news, it is common to use the standardized surprise component of the news and also test for unbiased market expectations (see, for instance, Balduzzi et al., 1998, 2001). The surprise component is calculated as Si;t ¼ Ai;t À Ei;t si ð1Þ

where Ai,t and Ei,t are the actual and the expected outcomes of data release i at time t, respectively, and si is the standard deviation of the forecast error of data release i. The expected outcomes of the macroeconomic data releases are collected from Bloomberg and consist of median expectations of the survey panellists. The expectations regarding the outcome of the ECB decisions consist of the mean of analysts’ survey-based expectations, collected one week before the Governing Council meetings and published by Reuters. The use of mean expectations thus differs with the use of median expectations for macroeconomic announcements. However, the use of median expectations for the ECB monetary policy decisions would give rise to very few non-zero surprises compared with the mean expectations. The potential impact of the Fed’s monetary policy decisions is not examined in this paper as they are released after the close of the trading day. In order to test for unbiasedness in the expectations data, standard techniques used in the literature are used (see Balduzzi et al., 1998, 2001). For the majority of the data releases, the null hypothesis of unbiased expectations cannot be rejected at the 10 per cent level, which suggests that the survey expectations are of good quality.3 One issue that has not been systematically addressed in earlier announcement studies is potential problems arising from leaks and early releases. Clearly, if the outcome of an announcement becomes available to investors ahead of the official release time, it will have an impact on our analysis as the market reaction will then take place earlier. If the release of macroeconomic statistics before official release times only takes place infrequently, the overall effect on our results will typically be limited. However, if the macroeconomic numbers are released regularly before official release times, owing to early releases or systematic leaks to the media ahead of official announcement times, our analysis will clearly be biased. For the macroeconomic announcements used in this study, there is no compelling evidence of the statistics being released early or of alleged leakages, with one notable exception: the German unemployment figures. By collecting news reports from Reuters and other market news agencies we document clear evidence that the numbers of German unemployed workers consistently have been known to investors before official releases; see
3. The complete results of the bias test can be found in the working paper version of this study (see Andersson et al., 2006).
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

6

Which News Moves the Euro Area Bond Market? Appendix A. These leaks have taken place before all releases in our sample, with the exception of a few releases at the beginning of 1999.4 It is therefore reasonable to assume that bond markets should have incorporated the latest news into the German unemployment report already before the official release time, with little or no reaction taking place at the scheduled time. Indeed, the econometric results shown in Section 4 suggest that the German unemployment statistics do not induce a significant impact on German bond markets at the time of the official release. However, given the attention that financial markets tend to give the equivalent US employment reports, the above analysis does not reveal whether the German unemployment figures have a market-moving impact or not. To examine this, ‘unofficial’ release times from Reuters were collected, corresponding to the point in time when the figures actually became known to the public. As a next step, the intraday bond returns surrounding these unofficial release times were regressed on the surprise component embedded in the German unemployment releases.5 The result of the exercise rejects any significant impact on bond returns following the releases of the German unemployment figures. This non-significant response is in line with the previous literature. For example, Goldberg and Leonard (2003) find that the releases of German employment statistics do not significantly affect German bond yields. Previous studies have, however, drawn that conclusion based on the no-significant impact surrounding official release times. The approach adopted in this paper can more accurately back up this finding. There is, by contrast, no evidence that other German macroeconomic statistics are systematically leaked to the media: leaks only appear to affect ¨ r Arbeit. This is the German unemployment figures from the Bundesanstalt fu to some extent also supported in our later analysis, where other German data releases are found to have a market impact at official release times.

4. ECONOMETRIC ANALYSIS AND RESULTS
In this section we investigate the influence that macro announcements have on intraday returns in the German bond market by utilizing a general econometric model, which simultaneously estimates both the level and the volatility of intraday returns on German bonds. In order to capture the
4. 5. This topic has also received some media attention; see ‘German Jobless Leaks Annoy Analysts, Investors – And Officials’, by Andreas Cremer, Bloomberg News, 9 April 2002. Results from this simple OLS regression are Rt ¼ À0:0059 þ 0:0034 GEU ; ð0:0040Þ ð0:0041Þ where Rt denotes five-minute returns surrounding the unscheduled release time (collected from news wires) of the German unemployment report. GEU represents the standardized surprise of the German unemployment report.
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

7

M. Andersson et al. time-varying feature of intraday return variability, a semiparametric model is used. The intraday statistical properties of the data suggest that three important features of the data should be taken into account in the econometric model. First, bond returns react sharply to macroeconomic announcements (announcement effect). This effect may be present in both the conditional mean and in the conditional volatility of the series. Second, the intraday pattern with higher observed volatility in opening and closing sections of the trading days should, together with the interday and day-of-the-week effects, also be properly captured in the volatility equation (calendar effect). Third, the conditional heteroscedasticity of daily returns – commonly known to be present in financial time series at lower frequencies – should also be taken into account (ARCH effect). The conditional mean of the five-minute German bond futures returns is specified as Rt þ 1 ¼ a 0 þ
P X i¼ 1

a i Rt À i þ 1 þ

K X R X k¼1 j¼0

k a MA kj MAt Àj þ

Q X j¼0

a MS j MSt Àj þ et

ð2Þ

where the five-minute bond return Rtþ1 is modelled as a linear function of: (1) P 5 2 values of lagged bond returns, (2) contemporaneous and R 5 2 lagged values of the standardized surprise of the K 5 44 announcements and (3) contemporaneous and Q 5 3 lagged values of the ECB’s monetary surprises. Note that this model is able to separate the effects of concurrent announcements. The lag lengths were suggested by the Akaike and Schwarz information criteria. The total number of observations amount to 233,269. Regarding the conditional variance, we model the disturbance term in equation (2) to be heteroscedastic and approximate its volatility by the following model: ^dðt ÞÀ1 s jet j ¼ b0 þ b1 pffiffiffiffi N ( )     X Z  D n n2 X 2pzt 2pzt i þ m1 þ m2 þ f sin gi WDwdðt Þ þ jz cos þ N1 N N N2 z¼1 z i¼ 1 þ
K X X k¼1 j¼0;3;6 MA lMA kj D t Àj þ

X
j¼0;3;6

MP lMP j D t Àj þ

X
j¼0;3;6

IS lIS j D t À j þ ut

ð3Þ

where N denotes the number of five-minute intervals on Pa trading day, n is the nth five-minute interval on a trading day, and N ¼ 1 i¼1; N ; i ¼ N ðN þ 1Þ=2, P and N2 ¼ i¼1; N ; i2 ¼ N ðN þ 1Þð2N þ 1Þ=6, are normalizing constants. This general set-up for the conditional mean and the volatility equations follow the procedure proposed by Andersen and Bollerslev (1997, 1998)

8

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? and Andersen et al. (2003, 2007). For two of the terms in the equations, however, we depart from their procedure. First, the second term in the ^dðt ÞÀ1 represents the estimated daily conditional standard volatility equation s deviation. This term is usually approximated by a GARCH-type model. ´ n (2006) argues that realized volatility captures better the However, Sebestye intraday return movements and consequently provides a better fit for the ^dðt Þ is calculated model than GARCH. Consequently, the s P a parametric 1=2 ^dðt Þ ¼ ½ m¼1; M R2 as s , where Rm is the mth 30-minute return.6 It is notable mŠ that a one-day lag, d(t) À 1, is used in the regression as all the observations needed to calculate realized volatility of day d(t) are not available in an intraday estimation framework. Evidently, the previous day’s realized volatility may not reflect correctly the return variability on the current day, but, as the realized daily volatility is highly autocorrelated, the volatility on day d(t) À 1 is likely to be a good proxy for day d(t)’s volatility. Second, in the conditional variance equation three dummy variables are used to capture decay in volatility: the first at the time of the announcement, the second from five to 15 minutes after the announcement and the third between 20 and 30 minutes after the announcement. This differs from Andersen and Bollerslev (1998), who propose a polynomial decay structure of the volatility response pattern, and they estimate the degree to which an announcement loads into this pattern. In addition, they allow for an ´ n (2006), whose adjustment of one hour. This paper instead follows Sebestye findings suggest that generally 30 minutes are sufficient for the complete volatility adjustment.7 The terms in brackets in equation (3) serve to capture intraday, interday and interweekly patterns of the data. The second-order polynomial (i.e. n/N1 and n2 /N1) approximates the intraday U-shaped pattern of the volatility. The interday pattern is by standard trigonometric terms, whereas the dummies WDi wdðt Þ account for interweekly impacts. Dummies accounting for the monetary policy communications are also included. DMP and DIS represent dummies for the monetary policy announcements and the Introductory Statement, respectively. Initial estimations suggested that the Introductory Statement dummy was insignificant in the conditional mean equation and, hence, it has been omitted from equation (2). The model is estimated by two-step weighted least squares (WLS). In the first step, equation (2) is estimated by ordinary least squares (OLS). Thereafter, equation (3) is estimated, and the fitted residuals j^ et j are used to perform a WLS estimate of equation (2).
6. Calculating daily-realized volatility using very high frequencies such as five-minute returns results in endogeneity in the conditional variance equation as the returns appear in the right-hand side of the equation. Returns at the 30-minute frequency do not exhibit a serial correlation; hence, the endogeneity problem no longer exists. We also allowed for longer adjustment (up to one hour), but the results showed that 30 minutes were sufficient to capture the entire response pattern.

7.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

9

M. Andersson et al. The econometric specification outlined above contains many variables and lags and therefore only the most interesting features are reported; see Table 2. With regard to the conditional mean equation, Table 2 MS presents estimates of aMA k;j¼0 and aj¼0 . These correspond to the contemporaneous point estimates of the surprises in the most relevant US/euro area/ national macro announcements and the ECB’s monetary policy decision, respectively, on the German bond market. In the same vein, Table 2 also reports on the lMA , lMP and lIS coefficients, which correspond to immediate and lagged volatility response from the macroeconomic surprises, the ECB monetary policy decisions and the volatility induced by the Introductory Statement read by the president at the press conference following the decisions. Overall, the regression results reveal several interesting features. First, it turns out that many announcements (27 out of 44) exert a significant impact on the level of German bond yields. In general, a higher than expected release should result in a negative sign of the surprise component coefficient apart from the US initial jobless claims and the unemployment data releases where a higher than expected number indicates that more people than anticipated are unemployed. As can be seen in the table, all the significant estimates result in an expected sign. In this context, it is also worth mentioning that the estimates of the lagged point estimates of the surprises in the US/euro area/national/UK macro announcements and the ECB’s monetary policy decision, respectively, on the euro area bond markets are, with a few exceptions, not significant. This in turn suggests an immediate jump in the returns at the time of the announcements and little reaction thereafter.8 Second, for the volatility impact, also here the bulk of the estimates at the time of the releases turned out to be significant. In stark contrast to the mean estimates, though, the volatility remains elevated longer, and for some of the announcements volatility remains high up until 30 minutes after the actual release. Interestingly, both the monetary policy decision and the Introductory Statements induce higher than normal volatility up to half an hour after the announcements. This prolonged heightened volatility may arise as a consequence of differences in opinions among investors. Third, actual and forward-looking measures of real economic activity and unemployment releases have a larger impact compared with price announcements. Fourth, US announcements influence the German bond returns more than euro area, national and UK macro releases. Finally, we find no impact from UK releases on euro area bond markets, indicating that the international spillover effects mainly stem from the United States. There are several reasons for the strong influence that US data exerts on the German bond markets. First, the United States can be perceived as the engine of global growth, which therefore explains its importance for the global financial markets, including Germany. Second, it may also be argued that
8. See, for instance, Charts 9 and 10 in Andersson et al. (2006).
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

10

Which News Moves the Euro Area Bond Market? Table 2 Estimated contemporaneous news response, the standard deviation for the surprise components (second and third columns) and volatility response coefficients
Contemporaneous news and volatility response coefficients Contemporaneous Volatility responses minutes after news response the announcement (l coefficients) Std. dev. surprise

a coefficients

0

15

30

US activity and employment US GDP advance À 0.0925 *** 0.84 0.0600 *** 0.0110 *** 0.0063 *** US GDP preliminary À 0.0147 0.23 0.0170 *** 0.0032 0.0003 US GDP final À 0.0182 0.24 0.0087 À 0.0007 0.0019 0.29 0.0169 *** 0.0060 *** 0.0047 *** US industrial production À 0.0239 *** US non-farm payroll À 0.0943 *** 100.61 0.1189 *** 0.0289 *** 0.0113 *** US non-farm payroll – À 0.0490 * 0.00 NA NA NA revisions US initial jobless claims 0.0166 *** 19.01 0.0145 *** 0.0071 *** 0.0039 *** US retail sales À 0.0494 *** 0.69 0.0371 *** 0.0078 *** 0.0043 *** US factory orders À 0.0158 *** 0.58 0.0072 * 0.0027 0.0029 * US durable goods orders À 0.0472 *** 2.98 0.0216 *** 0.0071 *** 0.0035 *** US business inventories 0.0041 0.23 0.0181 *** 0.0060 *** 0.0008 US forward-looking 4.00 0.0136 *** 0.0083 *** 0.0020 US University of Michigan À 0.0192 *** consumer sentiment index 2.17 0.0580 *** 0.0120 *** 0.0056 *** US ISM manufacturing À 0.0395 *** confidence US Chicago PMI À 0.0315 *** 4.36 0.0241 *** 0.0073 *** 0.0028 US consumer confidence À 0.0500 *** 5.01 0.0200 *** 0.0069 *** 0.0041 ** US Philadelphia Fed index À 0.0328 *** 8.88 0.0316 *** 0.0118 *** 0.0084 *** *** US ISM non-manufactur- À 0.0489 3.42 0.0151 *** 0.0062 *** 0.0032 * ing confidence US prices US consumer price index À 0.0209 * 0.13 0.0399 *** 0.0094 *** 0.0029 * * US producer price index À 0.0158 0.43 0.0197 ** 0.0081 *** 0.0035 ** UK activity and unemployment UK industrial production À 0.0008 0.65 0.0013 À 0.0031 0.0003 UK employment À 0.0068 * 8.80 0.0017 À 0.0002 À 0.0003 UK prices UK RPI À 0.0018 0.13 À 0.0007 0.0014 À 0.0003 Euro area activity and employment EA unemployment 0.0000 0.07 0.0007 À 0.0014 À 0.0001 EA industrial production À 0.0079 ** 0.50 0.0044 * À 0.0003 0.0000

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

11

M. Andersson et al. Table 2 Continued
Contemporaneous news and volatility response coefficients Contemporaneous Volatility responses minutes after news response the announcement (l coefficients) Std. dev. surprise

a coefficients

0

15

30

Euro area forward-looking EA business climate 0.0004 0.17 0.0018 0.0046 0.0018 EA PMI À 0.0052 0.70 0.0032 0.0000 0.0009 Euro area prices EA flash HICP À 0.0072 ** 0.09 À 0.0004 À 0.0010 À 0.0019 EA HICP À 0.0015 0.07 À 0.0012 À 0.0012 À 0.0005 National activity and employment GE unemployment À 0.0035 28.04 À 0.0007 0.0012 0.0029 * FR unemployment 0.0050 20.89 0.0022 0.0009 À 0.0013 ES unemployment 0.0005 0.39 À 0.0024 À 0.0035 À 0.0006 GE industrial production À 0.0058 1.42 0.0076 *** 0.0008 0.0017 *** FR industrial production À 0.0116 0.70 0.0010 À 0.0007 0.0004 IT industrial production À 0.0015 1.11 À 0.0024 À 0.0022 À 0.0020 ES industrial production À 0.0035 2.29 À 0.0033 * À 0.0015 À 0.0002 National forward-looking ZEW À 0.0392 *** 8.63 0.0149 *** 0.0079 *** 0.0037 ** IFO À 0.0255 *** 1.13 0.0234 *** 0.0093 *** 0.0050 *** FR business confidence À 0.0136 *** 2.08 À 0.0003 À 0.0013 0.0011 IT business confidence À 0.0076 ** 2.60 0.0008 0.0004 À 0.0011 BE business confidence À 0.0146 *** 3.06 0.0003 À 0.0003 À 0.0035 National prices GE consumer price index À 0.0034 0.14 0.0010 À 0.0001 À 0.0005 FR consumer price index À 0.0162 *** 0.14 0.0056 * 0.0025 À 0.0001 IT consumer price index 0.0042 0.33 À 0.0006 0.0006 0.0007 ES consumer price index 0.0004 0.18 0.0017 0.0012 À 0.0014 ECB ECB monetary policy 0.0128 ** NA 0.0190 *** 0.0125 *** 0.0058 *** decision ECB Introductory NA NA 0.0186 *** 0.0084 *** 0.0082 *** Statements
Notes: The news response coefficient represents the average market return in the five-minute interval following a macroeconomic release for a standardized surprise of one [see equation (1)]. Columns four, five and six represent the response in volatility immediately, 15 and 30 minutes after release. One, two and three asterisks denote significance at the 10%, 5% and 1% levels, respectively. Sample period; January 1999–December 2005.

12

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? business cycles have become more integrated and globalization therefore has led to a higher degree of interdependence between economies. Third, US macro data are typically released earlier than equivalent euro area and national data. Thus, market participants may therefore draw inferences about the euro area economy from the US data releases. In this respect, only euro area and national releases that cause investors to revise these inferences should lead to market reactions.

5. DO MONETARY REGIMES MATTER?
The constant estimates from the previous sections may not be completely representative as the impact of macro announcements and monetary policy decisions can change over time. There is no consensus in the literature on how to accurately gauge time-varying features of macro and monetary policy announcements. Ehrmann and Fratzscher (2003, 2005) use regression analysis in a rolling window, whereas Andersen et al. (2007) measure the impacts of macroeconomic variables in different business cycles. This paper adopts a different approach and considers various monetary policy regimes that are of particular interest for the German long-term bond markets given the introduction of the euro in January 1999. Changes in news sensitivity may occur for several reasons, of which the following three are of most interest. First, policy-makers can sometimes signal a preference for one or more macroeconomic indicators as input to their policy decisions for a given period, and thus may lead to increased responses in financial returns to those announcements. Second, a macroeconomic release may behave in an unusual manner at a certain point in the business cycle. Market participants may then perceive, at least temporarily, this variable as being particularly important. For example, employment data for the United States in late 2003 and early 2004 probably fell into both these categories, given the growing concerns about a so-called jobless recovery. This in turn led to heightened attention being paid to the monthly non-farm payroll release and unemployment data releases. Third, it is reasonable to assume different market reactions depending on the state of the business cycle. For instance, if a turning point of economic activity is expected, but the magnitude of the subsequent up- or downturn is unknown, some forward-looking variables may be monitored more closely by market participants. In order to gauge whether the reaction on the German bond markets to macroeconomic announcements and monetary policy decisions differs across monetary regimes, the sample is divided into three different subsamples. As in previous studies, our regimes are to some extent ad hoc and the estimations can thus be sensitive to the definition of the periods. Therefore, the results should be interpreted with some caution. However, the monetary cycle regime approach is in our view better suited for bond market data than
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

13

M. Andersson et al. the business cycle approach, adopted by Andersen et al. (2007), as one intuitively would expect investors to pay more attention to the short-term interest rate outlook when pricing bonds, rather than the current state of the business cycle, although the two concepts are obviously very closely intertwined. As markets are forward-looking and hence likely to anticipate future interest changes, we defined the tightening and accommodative monetary regimes as ending at the last interest rate change. The ending of the neutral regime is set according to market commentaries. Consequently, the first, a tightening regime, is assumed to start in May 1999 until October 2000, corresponding to the end of the month of the ECB’s last decision to increase its key interest rate. The second, an accommodative regime, is defined from November 2000 to June 2003, corresponding to the end of the month of the ECB’s interest rate reduction in June 2003. Finally, the third, a neutral regime in which short rates have remained unchanged, start in July 2003, and lasting until September 2005, when the ECB, according to market participants, signalled a less accommodative monetary policy stance using the wording ‘strong vigilance’ in the October 2005 Introductory Statement (see Barclays, 2005a, 2005b). This was widely considered by market participants to be signalling increases in the main refinancing rate at the December meeting. To gauge the price sensitivity across the three monetary policy regimes, the following econometric specification is used: Rt ¼ ai þ b1i D1t Si;t þ b2i D2t Si;t þ b3i D3t Si;t þ et ð4Þ

where D1t, D2t and D3t represent time dummies controlling for the three monetary regimes, respectively, i.e. they take on the value one in the corresponding monetary regime and zero otherwise. Si,t represents the ith surprise variable. The shortcoming of this approach is obviously that the length of the three regimes is relatively short (the first 18 months, the second 30 months and the third 27 months) and hence our estimates will suffer from a small sample bias. Therefore, quarterly releases are dropped and only those for which there are observations available for almost each month of the corresponding regime are included. It is also noteworthy that expectations for most euro area announcements started in early 2001, which clearly also leads to some problems when comparing results across regimes. Table 3 summarizes the contemporaneous news response coefficients in the first, second and third monetary policy regimes. Several interesting features can be observed. First, US activity and employment announcements seem to increase in importance over time, in line with Bernanke et al. (2004). One may argue that this is due to the smaller sample size as it requires a larger t-value to reject the null hypothesis of zero response. However, the t-values for the US announcements during the ECB’s tightening cycle are generally

14

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table 3 Contemporaneous news response coefficients under different monetary policy regimes
First regime US activity and employment US industrial production US non-farm payroll US initial jobless claims US retail sales US factory orders US durable goods orders US forward-looking US University of Michigan consumer sentiment index US ISM manufacturing confidence US Chicago PMI US consumer confidence US Philadelphia Fed index US ISM non-manufacturing confidence National activity and employment FR industrial production National forward-looking ZEW IFO FR business confidence IT business confidence BE business confidence National prices FR consumer price index À 0.0515 *** À 0.0622 0.0112 À 0.1217 * À 0.0258 * À 0.0438 *** À 0.0385 *** À 0.1498 *** À 0.0581 *** À 0.0250 *** À 0.0248 * À 0.0033 À 0.0058 NA À 0.0368 NA NA NA 0.0079 Second regime À 0.0311 *** À 0.0406 0.0115 * À 0.0464 *** À 0.0157 * À 0.0439 * À 0.0044 À 0.0220 À 0.0427 *** À 0.0492 *** À 0.0388 *** À 0.0660 *** À 0.0269 ** À 0.0361 *** À 0.0224 *** À 0.0154 *** À 0.0128 *** À 0.0206 *** À 0.0210 *** Third regime À 0.0189 ** À 0.2073 *** 0.0230 *** À 0.0530 ** À 0.0316 *** À 0.0599 * À 0.0260 ** À 0.0265 À 0.0148 À 0.0582 ** À 0.0277 * À 0.0450 *** À 0.0107 *** À 0.0375 *** À 0.0081 ** À 0.0068 À 0.0055 À 0.0091 À 0.0070

Notes: The news response coefficients represent respectively b1i, b2i and b3i in equation (4). Only the coefficients for the announcements which are significant at the 1% level in the full sample using the regression setup in equations (2) and (3) are shown in the table (GDP Advance is left out due to small sub-sample sizes). One, two and three asterisks denote significance at the 10%, 5% and 1% levels, respectively. Some announcements are not available for the entire first regime.

much smaller than in the other periods. Regarding the magnitude of the estimated significant coefficients, the most interesting characteristics concern the US employment data where the size (in absolute value) of both the non-farm payroll and the initial jobless claims estimates has increased over time. This higher asset return sensitivity to unemployment data in the United States may be related to the fact that the US economic recovery since the 2001 recession has been accompanied by a relatively large degree of slack in the labour market, raising concerns about a ‘jobless recovery’. Again, note that, in contrast to other variables, a positive sign is expected a priori for initial jobless claims and other unemployment variables.
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

15

M. Andersson et al. Second, national announcements seem to have a larger impact on the German bond returns during the ECB’s accommodative policy regime than in the neutral period observed between mid-2003 and mid-2005. This extra sensitivity can probably be linked to the macro announcements signalling an increased likelihood of changes in the monetary and fiscal stance compared with the neutral period. As a cross-check we also tested for structural breaks, using the standard methodology described by Bai and Perron (1998, 2003).9 The structural break tests are rather restrictive, but nonetheless indicated one or two breaks in 11 out of 44 announcements. For instance, the test indicates a structural break in important US non-farm payroll announcements around mid-2003, which coincides with our defined break point. However, for other announcements, it is difficult to provide economic interpretations for the results obtained. The tests therefore, on the one hand, appear to warrant the use of modelling breaks in the sample but, on the other, the inconclusiveness of the break tests also indicates that the results should be interpreted with some care.

6. CONCLUDING REMARKS
This paper finds that United States and to some extent euro area and national macro releases exert a significant impact on the returns of long-term German government bonds. Overall, the announcements have a longer-lasting impact on volatility than on the level of bond returns. US announcements seem to influence German bond returns more than UK, aggregate euro area and national euro area macro announcements. There are three probable explanations for these findings. First, the United States can be perceived as the engine of global growth, which therefore explains its importance for the global financial markets, including the euro area. Second, it may also be argued that business cycles have become more integrated and globalization therefore has led to a higher degree of interdependence between economies. Third, US macro data are typically released earlier than equivalent euro area data. Thus, market participants may therefore draw inferences about the euro area economy from the US data releases. In this respect, only euro area releases that cause investors to revise these inferences should lead to market reactions. By splitting our sample period into three subsamples, reflecting three different monetary policy regimes (tightening, accommodative and neutral), we show that the impact of public information about US activity and employment on German bond markets has increased over time. A possible explanation may be that in late 2003 and early 2004, US employment data were closely monitored by policy-makers owing to growing concerns about the so-called ‘jobless recovery’.
9. The structural break test results are available upon request.
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

16

Which News Moves the Euro Area Bond Market? With regard to the ECB’s monetary policy decisions and statements, the financial market tends to have predicted the outcomes of monetary policy decisions with a high degree of precision so far, possibly due to transparency around the intentions of the ECB. Nonetheless, heightened volatility is observed following both monetary policy decisions and the Introductory Statement read by the President at the Press Conference following the decisions.

APPENDIX A
Table A.1 Selected media reports concerning German unemployment releases, 1999–2005
Announcement date 8 January 1999 09:55 9 February 1999 09:53 9 March 1999 10:20 8 April 1999 09:55 7 May 1999 09:50 Actual release 34,000 À 59,000 À 6,000 À 3,000 10,000 Early release/ leak No leak detected No leak detected No leak detected No leak detected 7 May 1999 09:27 Quote

10 June 1999 10:00

11,000

9 June 1999 16:21

6 July 1999 09:52

15,000

6 July 1999 09:02

Only unadjusted numbers leaked. German unemployment fell in April by 275,383 from a year earlier to 4.145 million, equivalent to 10.7 per cent of the workforce, figures obtained by Reuters on Friday ahead of their official release showed. Only unadjusted numbers leaked. German unemployment fell in May to 3.989 million from 4.145 million the month before on an unadjusted basis, a source familiar with data due to be officially released on Thursday said. The source told Reuters the unadjusted number of jobless stood at 3.938 million. The seasonally adjusted jobless number rose by 15,000.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

17

M. Andersson et al. Table A.1 Continued
Actual release 1,000 Early release/ leak Quote

Announcement date 5 August 1999 09:51

7 September 1999 09:52

4,000

5 October 1999 09:55

9,000

9 November 1999 10:28

À 11,000

7 December 1999 10:08

À 29,000

5 January 2000 09:50

À 68,000

8 February 2000 10:12

À 31,000

8 March 2000 10:21

À 34,000

1 August 1999 Only unadjusted numbers leaked. In 18:52 June, a total of 3.94 million people were jobless in Germany in unadjusted terms, compared with 3.99 million in May and 4.15 million in April. 7 September German unemployment rose by a 1999 09:06 seasonally adjusted 4,000 in August, a source told Reuters on Tuesday ahead of the official release of unemployment data by the Federal Labour Office. 5 October The number of people out of work in 1999 09:11 Germany in September rose by a seasonally adjusted 9,000 from August, a political source said on Tuesday. 9 November Pan-German seasonally adjusted 1999 09:20 unemployment fell by 11,000 in October, opposition party sources told Market News International on Tuesday. 7 December German unemployment fell by 1999 09:35 29,000 from the previous month in November on a seasonally adjusted basis, according to sources. 5 January German unemployment fell last 2000 09:13 month by 68,000 from the previous month on a seasonally adjusted basis, sources told Reuters late on Tuesday. 8 February Inaccurate number. German 2000 09:54 adjusted unemployment fell in January by 33,000 from the previous month, sources said on Tuesday. 8 March 2000 German seasonally adjusted 09:18 unemployment fell in February by 34,000 from the previous month, sources said on Wednesday.

18

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table A.1 Continued
Actual release 8,000 Early release/ leak 5 April 2000 09:10 Quote

Announcement date 5 April 2000 10:24

9 May 2000 09:55

À 8,000

8 June 2000 09:55

À 32,000

6 July 2000 09:55

À 14,000

8 August 2000 09:55

À 16,000

6 September 2000 09:55

À 19,000

5 October 2000 09:55

À 18,000

German seasonally adjusted unemployment rose in March by 8,000 from the previous month, sources said on Wednesday. 9 May 2000 German unemployment, adjusted for 09:21 seasonal factors, fell by 8,000 in April from March, sources with access to data due to be released later by the Federal Labour Office said on Tuesday. 8 June 2000 Inaccurate number. The sources also 09:01 said the number of unemployed on a seasonally adjusted basis fell by 27,000 in May from April. 6 July 2000 Only unadjusted numbers leaked. 14:32 German unemployment fell by at least 60,000 in June from 3.788 million in May without adjustments for seasonal factors, a trade union official who sits on the Federal Labour Office’s board said on Wednesday. 7 August 2000 Only unadjusted numbers leaked. 17:50 Bild, whose past leaks of jobless data have been accurate, gave no firm adjusted figure for unemployment in July, but said the Federal Labour Office – due to release the figures on Tuesday – would report a fall after June’s surprise rise. 6 September Inaccurate number. German seasonally 2000 09:05 adjusted unemployment fell 18,000 from 3.891 million in July, a source told Reuters on Wednesday. 5 October Pan-German seasonally adjusted 2000 09:32 unemployment fell by 18,000 in September compared to August, informed sources told Market News International on Thursday.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

19

M. Andersson et al. Table A.1 Continued
Actual release À 25,000 Early release/ leak Quote

Announcement date 7 November 2000 09:50

5 December 2000 09:26

À 15,000

9 January 2001 09:22

À 27,000

6 February 2001 09:27

7,000

6 March 2001 09:50

3,000

4 April 2001 09:50

12,000

8 May 2001 09:24

6,000

7 November Following are economists’ comments 2000 09:23 after German seasonally adjusted unemployment fell in October by 25,000 from the previous month, according to data provided by an official source to Reuters. GERMANY MEDIAN FORECAST – NOV SA UNEMPLOYMENT À 20,000 – LEAK À15K. 8 January Only unadjusted number leaked. 2001 17:33 German unemployment rose by 163,700 in December from the previous month to around 3.8 million, without adjustment for seasonal factors, the Bild daily said on Monday. The newspaper, which has a good track record in publishing Germany’s notoriously leaky jobs figures, gave no figures for seasonally adjusted unemployment. Following are economists’ reactions to leaked German January jobless data showing a seasonally adjusted rise of 7,000 from December. German seasonally adjusted unemployment rose by 3,000 in February, sources said on Tuesday, confounding analysts’ expectations of a resumption of the downward trend. 4 April 2001 German seasonally adjusted 08:47 unemployment rose in March by 12,000, a source told Reuters on Wednesday, the third month in a row the adjusted figure has risen. Following are economists’ reactions to a report that Germany’s unemployment rose 6,000 in April on a seasonally adjusted basis.

20

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table A.1 Continued
Actual release 18,000 Early release/ leak 7 June 2001 09:16 Quote German unemployment rose in May by 18,000 from the previous month, a source said on Thursday, the fifth month in a row the seasonally adjusted figure has risen. Germany’s adjusted number of jobless rose 22,000 in June, a source told Reuters on Wednesday. Germany’s Federal Labour Office said on Tuesday seasonally adjusted unemployment rose by 11,000 in July, confirming earlier leaks. German unemployment fell by a seasonally adjusted 2,000 in August, with a rise by 5,000 in western Germany offset by a 7,000 fall in the east, a source said on Wednesday. German unemployment rose by a seasonally adjusted 20,000 in September, with a rise of 13,000 in western Germany and a 7,000 increase in the east, largely in line with expectations, a source said on Monday. German adjusted unemployment rose by 27,000 in October, a source said on Tuesday, while the unadjusted figure fell 17,550 in October from 3.725 million to give an unchanged rate of 9 per cent. German seasonally adjusted unemployment rose in November by 17,000 from the previous month, a source told Reuters on Tuesday. German seasonally adjusted unemployment rose by 6,000 in December and the November figure was adjusted to show an increase of 19,000, a government source in Berlin told Reuters on Wednesday.

Announcement date 7 June 2001 09:25

5 July 2001 09:24

22,000

4 July 2001 18:30

7 August 2001 09:28

11,000

5 September 2001 09:33

À 2,000

5 September 2001 08:18

9 October 2001 09:03

20,000

8 October 2001 18:57

6 November 2001 09:26

27,000

5 December 2001 09:33

17,000

5 December 2001 09:28

9 January 2002 09:21

6,000

9 January 2002 09:08

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

21

M. Andersson et al. Table A.1 Continued
Actual release 31,000 Early release/ leak Quote

Announcement date 6 February 2002 09:28

6 March 2002 09:21

1,000

9 April 2002 09:10

À 8,000

7 May 2002 09:05

6,000

7 June 2002 09:34

60,000

9 July 2002 09:40

39,000

7 August 2002 09:02

8,000

German unemployment rose 31,000 in January on a seasonally adjusted basis with the unadjusted unemployment rate up at 10.4 per cent from 9.6 per cent in December, a source said on Wednesday. 6 March 2002 German unemployment rose 1,000 08:50 in February on a seasonally adjusted basis, a deputy head of the Labour Ministry said on Wednesday. German unemployment, adjusted for seasonal factors, fell in March for the first time since December 2000, sources said on Tuesday. 7 May 2002 Germany’s adjusted unemployment 08:32 rate rose by 6,000 in April, a source told Reuters on Tuesday, above a consensus forecast for a monthly rise of 600. 7 June 2002 Following are economists’ reactions 08:32 to a report that German seasonally adjusted unemployment rose by a higher than expected 60,000 in May. 8 July 2002 Only unadjusted number leaked. 18:18 Germany’s headline jobless total rose by an unadjusted 8,000 in June to 3.954 million, mass-circulation Bild newspaper reported on Monday, bad news for Chancellor Gerhard Schroeder ahead of September elections. 6 August 2002 German adjusted unemployment 19:30 rose by 8,000 in July from the previous month, a source told Reuters on Wednesday, well below most analysts’ expectations, but not much consolation for Chancellor Gerhard Schroeder.

22

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table A.1 Continued
Actual release 2,000 Early release/ leak Quote

Announcement date 5 September 2002 09:19

8 October 2002 09:19

À 1,000

7 November 2002 09:20

22,000

4 December 2002 09:17

35,000

9 January 2003 09:17

28,000

5 February 2003 09:17

62,000

6 March 2003 09:45

67,000

4 September Only unadjusted number leaked. 2002 17:44 Germany’s unemployment rate fell to 4.018 million in August, implying a drop of around 29,000 in the month, a source told Reuters on Wednesday. 7 October Only unadjusted number leaked. The 2002 18:24 source said that the unadjusted figure was down by nearly 70,000 on the 4.018 million registered in August, bringing the headline figure below four million. West German seasonally adjusted unemployment rose to 2.716 million from 2.689 million, while eastern German unemployment fell to 1.403 million from 1.408 million, the source said. 3 December German seasonally adjusted un2002 19:25 employment rose in November by 35,000 from the previous month, a source told Reuters on Tuesday. Confirming market expectations of a strong rise, the number of unemployed in Germany rose by a seasonally adjusted 28,000 in December from the November level, a source with access to the official German Labour Office data said Thursday. 4 February The unemployment total adjusted 2003 20:44 for seasonal factors rose by 62,000 in January compared with December, the source said. 6 March 2003 The number of unemployed in 09:37 Germany rose by a seasonally adjusted 67,000 in February from the January level, a source with access to the official German Labour Office data said Thursday.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

23

M. Andersson et al. Table A.1 Continued
Actual release 52,000 Early release/ leak 3 April 2003 09:23 Quote

Announcement date 3 April 2003 09:25

7 May 2003 09:32

44,000

5 June 2003 09:16

À 4,000

8 July 2003 09:24

À 33,000

6 August 2003 09:27

7,000

4 September 2003 09:25

0

9 October 2003 09:14

À 14,000

German unemployment adjusted for seasonal factors rose by 52,000 in March, a labour office source told Reuters on Thursday. 6 May 2003 Only unadjusted number leaked. 14:30 German headline unemployment, not adjusted for seasonal factors, fell just below 4.5 million in April from 4.608 million in March, a government source told Reuters on Tuesday. A surprise 4,000 month-on-month decline in German unemployment was caused by government labour market policies rather than any cyclical upturn, a Labour Office source said on Thursday. 8 July 2003 German unemployment adjusted for 08:17 seasonal factors fell by 33,000 in June, said a source with knowledge of data to be released by the Federal Labour Office on Tuesday. 6 August 2003 German unemployment adjusted 08:21 for seasonal factors rose by 7,000 in July to 4.408 million, said a source with knowledge of data to be released by the Federal Labour Office on Wednesday. 4 September German unemployment adjusted for 2003 08:35 seasonal factors was unchanged in August from July’s 4.408 million, said a source with knowledge of data to be released by the Federal Labour Office on Thursday. 9 October German unemployment adjusted for 2003 00:35 seasonal factors fell 14,000 in September, a source with knowledge of the data said on Thursday, attributing the surprising drop to recent labour market reforms.

24

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table A.1 Continued
Actual release À 12,000 Early release/ leak Quote

Announcement date 6 November 2003 09:14

4 December 2003 08:59

À 18,000

8 January 2004 08:53

À 21,000

5 February 2004 09:29

À 81,000

4 March 2004 09:17

26,000

6 April 2004 09:32

44,000

6 November German unemployment fell 2003 00:34 by 12,000 in seasonally adjusted terms in October as a result of labour market reforms, an informed source told Reuters on Thursday, a bigger drop than analysts had expected. 4 December German unemployment fell by a 2003 00:41 seasonally adjusted 18,000 in November as measures to get jobless people into training schemes or part-time jobs took effect, a source familiar with data to be released by the Federal Labour Office later on Thursday told Reuters. 8 January German unemployment fell by a 2004 08:16 greater than expected 21,000 in seasonally adjusted terms in December, a source who has seen data to be released later on Thursday told Reuters. 5 February German unemployment fell by 2004 08:23 81,000 in seasonally adjusted terms in January, under a new method of calculation that excludes some people on training programmes, but was up more than 20,000 under the old method, a source who has seen the data told Reuters on Thursday. 4 March 2004 German unemployment rose by 08:21 26,000 in seasonally adjusted terms in February, a source who has seen the data told Reuters on Thursday. 6 April 2004 German seasonally adjusted 08:15 unemployment rose by a higherthan-expected 23,000 in April from the previous month, according to a source who has seen data due to be released on Wednesday.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

25

M. Andersson et al. Table A.1 Continued
Actual release 23,000 Early release/ leak 5 May 2004 07:06 Quote German seasonally adjusted unemployment rose by a higherthan-expected 23,000 in April from the previous month, according to a source who has seen data due to be released on Wednesday. German unemployment rose for a fourth consecutive month in May by a seasonally adjusted 9,000 to 4.374 million, according to a source with knowledge of data due to be released on Tuesday. Germany’s seasonally adjusted unemployment total fell by 1,000 in June from May, the first such drop since January, according to a source familiar with the data to be released on Tuesday. German unemployment adjusted for seasonal factors rose in July by 11,000 month-on-month to 4.386 million, a source with knowledge of official data due to be released on Wednesday said. German unemployment adjusted for seasonal factors rose in August by 24,000 to 4.414 million, a source with knowledge of official data due to be released on Thursday told Reuters. German unemployment adjusted for seasonal factors rose by a bigger-than-expected 27,000 in September, a source with knowledge of official data due to be released on Tuesday told Reuters on Monday. German seasonally adjusted unemployment rose in October by 12,000 month-on-month, a source with knowledge of the data told Reuters on Tuesday.

Announcement date 5 May 2004 09:30

8 June 2004 08:17

9,000

8 June 2004 07:54

6 July 2004 09:25

À 1,000

6 July 2004 06:01

4 August 2004 09:29

11,000

4 August 2004 08:11

2 September 2004 09:10

24,000

2 September 2004 08:06

5 October 2004 09:27

27,000

4 October 2004 17:38

3 November 2004 09:21

12,000

2 November 2004 15:04

26

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table A.1 Continued
Actual release 7,000 Early release/ leak Quote

Announcement date 2 December 2004 09:22

4 January 2005 09:14

17,000

2 February 2005 09:07

227,000

1 March 2005 09:55

161,000

31 March 2005 08:55

92,000

28 April 2005 09:30

À 79,000

2 December The number of unemployed people 2004 09:18 in Germany rose 7,000, in seasonally adjusted terms, in November from October, a source with access to the official German Labour Agency data told Dow Jones Newswires on Thursday. 4 January The number of Germans out of work 2005 08:33 rose by a seasonally adjusted 17,000 in December from November, Reuters said, citing unidentified people with knowledge of the figures from the Federal Labour Agency. 1 February Germany’s adjusted jobless total 2005 18:06 increased by 227,000 in January from the previous month, a Federal Labour Office source told Reuters on Tuesday. 28 February The source said that unemployment 2005 20:00 rose by 161,000 in February versus the prior month on an adjusted basis, almost double the amount forecast by economists. 30 March Germany’s seasonally adjusted 2005 13:29 jobless total rose by a bigger-thanexpected 92,000 in March, a Federal Labour Office source told Reuters on Wednesday. 27 April 2005 The number of Germans out of work 16:15 unexpectedly fell on a seasonally adjusted basis in April, the first monthly decline since January 2004, a Labour Office source told Reuters on Wednesday. The adjusted total dropped by 80,000 in April from March, the source said. Economists had been forecasting an adjusted rise of 17,500, according to a Reuters poll.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

27

M. Andersson et al. Table A.1 Continued
Actual release 0 Early release/ leak 30 May 2005 15:04 Quote German unemployment adjusted for seasonal factors was unchanged in May, or slightly better than economists’ expectations for a drop of around 10,000, a source familiar with the data told Reuters on Monday. The source said that the unadjusted total, a politically sensitive number in Germany, fell by 161,000 to 4.807 million. German unemployment adjusted for seasonal factors fell a larger-thanexpected 23,000 month-onmonth in June, a source familiar with official Labour Office data said on Wednesday. Unemployment in Germany adjusted for seasonal swings fell by a much-bigger-than-expected 40,000 in July from the previous month, a source with knowledge of the official data told Reuters on Wednesday. Adjusted for seasonal swings, German unemployment fell by a smaller-than-expected 12,000 month-on-month in August, a source with knowledge of the official data due for release on Wednesday told Reuters. German unemployment adjusted for seasonal factors rose around 40,000 in September, bucking expectations for a decline, following a statistical change, a source with knowledge of official data said on Wednesday. German unemployment adjusted for seasonal factors fell in October by 36,000, more than double the drop expected, after a statistical blip pushed the total higher in September, a source familiar with the data said on Tuesday.

Announcement date 31 May 2005 09:30

30 June 2005 09:25

À 23,000

29 June 2005 20:36

28 July 2005 09:55

À 42,000

27 July 2005 16:33

31 August 2005 09:55

À 12,000

30 August 2005 16:51

29 September 2005 09:55

39,000

28 September 2005 20:26

2 November 2005 09:55

À 36,000

1 November 2005 19:08

28

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market? Table A.1 Continued
Actual release À 53,000 Early release/ leak Quote

Announcement date 1 December 2005 09:55

3 January 2006 09:55

À 110,000

1 December The number of Germans out of work 2005 08:05 fell by 53,000 month-on-month in November on a seasonally adjusted basis, a source with knowledge of official data told Reuters on Thursday. 3 January The number of Germans out of work 2006 08:16 fell by 110,000 month-on-month in December on a seasonally adjusted basis, a far greater drop than expected, a source with knowledge of official data told Reuters on Tuesday.

Note: It has not been possible to identify the exact release times of all media reports. The presumed leaks may have occurred earlier than the cited release times, as it may have been reported earlier in other news flashes or by other media. Source: Factiva and Reuters.

ACKNOWLEDGEMENTS
We are grateful for useful discussions with and suggestions from Sven Astheimer, Francesco Drudi, Michael Ehrmann, Michael Fleming, Marcel Fratzscher, Manfred Kremer, Garry Schinasi, and two anonymous referees. We would also like to thank seminar participants from presentations at the International Conference on High Frequency Finance in Konstanz, at the ECB, Danmarks Nationalbank and Sveriges Riksbank for useful comments. The views expressed in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the ECB or Danmarks Nationalbank. Any remaining errors are the authors’ responsibility. Address for correspondence: Magnus Andersson, European Central Bank, Postfach 160319, 60066 Frankfurt am Main, Germany. Tel.: þ 49 69 1344 7410; fax: þ 49 69 1344 6514; e-mail: magnus.andersson@ecb.int

REFERENCES
Andersen, T. G. and T. Bollerslev (1997), ‘Intraday Periodicity and Volatility Persistence in Financial Markets’, Journal of Empirical Finance 4, 115–158. Andersen, T. G. and T. Bollerslev (1998), ‘Deutsche Mark–Dollar Volatility: Intraday Activity Patterns, Macroeconomic Announcements, and Longer-Run Dependencies’, Journal of Finance 53, 219–265.
r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

29

M. Andersson et al.
Andersen, T. G., T. Bollerslev, F. X. Diebold and C. Vega (2003), ‘Micro Effects of Macro Announcements: Real-Time Price Discovery in Foreign Exchange’, American Economic Review 93, 38–62. Andersen, T. G., T. Bollerslev, F. X. Diebold and C. Vega (2007), ‘Real-Time Price Discovery in Stock, Bond and Foreign Exchange Markets’, Journal of International Economics 73, 251–277. ´ n (2006), ‘Which News Moves the Euro Andersson, M., L. J. Overby and S. Sebestye Area Bond Market?’, ECB Working Paper No. 631. Bai, J. and P. Perron (1998), ‘Estimating and Testing Linear Models with Multiple Structural Changes’, Econometrica 66, 47–78. Bai, J. and P. Perron (2003), ‘Computation and Analysis of Multiple Structural Change Models’, Journal of Applied Econometrics 18, 1–22. Balduzzi, P., E. J. Elton and T. C. Green (1998), ‘Economic News and the Yield Curve: Evidence from the US Treasury Market’, Salomon Center Working Paper No. S-98-5, Stern School of Business. Balduzzi, P., E. J. Elton and T. C. Green (2001), ‘Economic News and Bond Prices: Evidence from the US Treasury Market’, Journal of Financial and Quantitative Analysis 36, 523–543. Barclays (2005a), ‘ECB Press Conference: Words not Deeds’, market commentary by Julian Callow, Barclays Capital, 2 September 2005. Barclays (2005b), ‘ECB Watching: Just Do It!’, market commentary by Julian Callow, Barclays Capital, 6 October 2005. Bentzen, E., P. Hansen, A. Lunde and A. Zebedee (2004), ‘The Greenspan Effect on Equity Markets: An Intraday Examination of US Monetary Policy Announcements’, unpublished manuscript. Bernanke, B. S., V. R. Reinhart and B. P. Sack (2004), ‘Monetary Policy Alternatives at the Zero Bound: An Empirical Assessment’, Finance and Economics Discussion Series of the Federal Reserve Board, September 2004. Bernoth, K. and J. von Hagen (2004), ‘Euribor Futures Market: Efficiency and the Impact of ECB Policy Announcements’, International Finance 7, 1–24. Bomfim, A. (2000), ‘Pre-Announcement Effects, News Effects, and Volatility: Monetary Policy and the Stock Market’, Journal of Banking and Finance 27, 133–151. Dwyer, B. and R. W. Hafer (1989), ‘Interest Rates and Economic Announcements’, Federal Reserve Bank of St. Louis Review, March/April, 34–46. Ehrmann, M. and M. Fratzscher (2003), ‘Monetary Policy Announcements and Money Markets: A Transatlantic Perspective’, International Finance 6, 309–328. Ehrmann, M. and M. Fratzscher (2005), ‘Equal Size, Equal Role? Interest Rate Interdependence between the Euro Area and the United States’, Economic Journal 115, 928–948. Faust, J., J. H. Rogers, E. Swanson and J. H. Wright (2002), ‘Identifying the Effect of Monetary Policy Shocks on Exchange Rates using High-Frequency Data’, Journal of European Economic Association 1, 1031–1057. Faust, J., J. H. Rogers, S. B. Wang and J. H. Wright (2007), ‘The High-Frequency Response of Exchange Rates and Interest Rates to Macroeconomic Announcements’, Journal of Monetary Economics 54, 1051–1068. Fleming, M. J. and M. Piazzesi (2005), ‘Monetary Policy Tick-by-Tick’, unpublished working paper.

30

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

Which News Moves the Euro Area Bond Market?
Fleming, M. J. and E. M. Remolona (1997), ‘What Moves the Bond Market?’, Federal Reserve Bank of New York Economic Policy Review, December, 31–50. Goldberg, L. S. and D. Leonard (2003), ‘What Moves Sovereign Bond Markets? The Effects of Economic News on U.S and German Yields’, Current Issues in Economics and Finance, Federal Reserve Bank of New York, 9, No. 9. ´ n, S. (2006), ‘What Drives Money Market Rates?’, unpublished working paper. Sebestye Wongswan, J. (2005), ‘The Response of Global Equity Indexes to U.S. Monetary Policy Announcements’, International Finance Discussion Paper No. 844, Board of Governors of the Federal Reserve System.

r 2009 The Authors ¨ r Socialpolitik and Blackwell Publishing Ltd. 2009 Journal Compilation r Verein fu

31