The last decade has been witness to the primacy of monetary policy as the main tool used by policymakers in the stabilization of inflation and output.
With increasing globalization, asset markets appear to move more in tandem with each other. For example, Kim et al. (2005) find that linkages among European stock markets inside and outside the Euro area have strengthened following currency unification. Further, empirical evidence suggests that there are strong correlations between the major bond markets (Ilmanen, 1995) and that these correlations have increased dramatically in recent years (Solnik et al., 1996). There is also evidence of at least partial integration between major international bond markets (Barr and Priestley, 2004) while Driessen et al. (2003) identify common factors in predicting international bond returns. Additionally, Kim et al. (2006) perform a dynamic analysis of integration to try to capture the evolving nature of relationships between markets. They establish strong contemporaneous and dynamic linkages between Germany and other Euroland bond markets but that these links are much weaker and more stable for the U.K. and accession countries.
Not surprisingly, recent research has begun to highlight the likely influence of global, regional and local influences on asset returns. For example, Christiansen (2007, forthcoming) investigates the impact of global and regional spillovers in bond and equity markets and uncovers significant spillovers from both global (U.S.) and regional (E.U.) markets into domestic markets and that the introduction of the Euro has typically strengthened regional effects. While there has also been an increasing number of studies that examine the influence of both domestic and foreign news on domestic and foreign assets, e.g., Andersen et al. (2003), Becker et al. (1995), Ehrmann et al. (2005), and Faust et al. (2007). It is within this context that we seek to investigate the international transmission of monetary policy in terms of its impact on international bond markets.
The price of a bond is a function of the discounted stream of future coupon payments and the redemption value of the bond. Shiller and Beltratti (1992) and Campbell and Ammer (1993) advance an approach to decompose news regarding current excess bond returns into revisions in expectations of future excess returns, inflation and real rates.1 In this study, we focus on the German, U.K. and U.S. long-term bond markets and conduct a decomposition of each respective country’s bond returns based on Campbell and Ammer (1993) decomposition while permitting returns in each country to affect one another.
Given the pivotal role of monetary policy in determining bond returns we next seek to characterize the impact of unanticipated domestic and foreign monetary policy changes on each country’s bond returns and its components. A natural question is how important are domestic monetary surprises in determining domestic bond returns and are there spillovers from foreign monetary policy to domestic returns? It is frequently argued that U.S. monetary policy drives world bond returns - our study seeks to shed light on this view. Related evidence suggests that German bond returns respond more to U.S. macro news than domestic or other Euro area news, see for example Goldberg and Leonard (2003) and Andersson et al. (2006). The reasons cited for such findings include greater financial market integration, the importance of the U.S. to global growth and the earlier release (relative to the Euro area) of U.S. macro announcements.
While it is natural to assume that a surprise tightening in monetary policy would lead to higher long-term rates, Ellingsen and Söderström (2001) argue that the response of long rates to a surprise change in the policy rather is ambiguous. In particular, they suggest that long rates will rise when the change in monetary policy reveals information regarding the economy but if the monetary action reveals changes in the central bank’s preferences then short rates and long rates may move in opposite directions. Thornton (1998) also argues that a tightening of monetary policy may lower inflation expectations.
Bearing this in mind, a critical feature of our paper, in contrast to previous research that simply examines how long-term interest rates respond to monetary policy surprises (e.g., Kuttner, 2001; Rigobon and Sack, 2004; Ehrmann and Fratzscher, 2005), is that we seek to delve further into what lies behind the response (if any) in bond returns, i.e. is the change in excess bond returns due to changes in expectations regarding future excess returns, real rates or inflation? Bernanke and Kuttner (2005) conduct a similar exercise in decomposing the impact of monetary policy surprises on U.S. stock returns and we follow their methodology here.
Using futures market data to derive our measure of the monetary policy surprise, an important feature of our analysis is the decomposition of monetary policy changes into expected and unexpected changes.2 Failure to decompose monetary policy changes into its expected and unexpected components are likely to lead to biased results due to an errors in variables problem. Our key results show that excess bond returns react to domestic monetary policy surprises in both Germany and the U.K. but fail to have a significant impact in the U.S. over the period 1994-2004. Interestingly, we find a strong divergence between the effects of domestic policy on excess bond returns in Germany relative to the U.K. with a surprise monetary tightening in former (latter) leading to a rise (fall) in the excess holding period return.
The rationale behind such contrasting responses becomes clearer when one breaks down news regarding excess bonds returns into its components, i.e. revisions in news regarding future excess returns, future real interest rates and future inflation, and assess how these components are affected by unanticipated monetary policy. In particular, a surprise tightening of monetary policy in Germany (U.K.) leads to a statistically significant revision in inflation expectations downwards (upwards) and this appears primarily responsible for the differing response of bond returns in each respective country. Such contrasting responses of inflation expectations to a tightening of monetary policy could be potentially rationalized by differences in the credibility of the monetary policy authority in respective countries (area). In particular, the Bundesbank has traditionally been viewed as a strong fighter of inflation while the Bank of England less so. Finally, we find little role for monetary policy spillovers, i.e. surprise changes in monetary policy in one country (area) does not appear to affect news regarding excess bond returns in other countries.
The outline of the rest of the paper is as follows. In section 2, we discuss briefly issues regarding the identification of monetary policy, while in section 3 we outline the Campbell-Ammer methodology associated with the variance decomposition of excess bond returns, how we measure monetary policy surprises and how we seek to assess their impact on news regarding current excess bond returns and their respective components. In section 4, we discuss the data used and present the results from and analysis of the variance decomposition as well as the impact of monetary policy surprises. Finally section 5 provides some concluding remarks.
The analysis of the influence of monetary policy actions on asset prices potentially raises two key methodological issues highlighted in the extant literature. First, there is the issue of interest rate endogeneity and potential omitted variables. Second, there is the issue of the correct measurement of unanticipated or surprise component of monetary policy rate changes.
The issue of the appropriate identification of policy changes is most readily witnessed in early studies assessing the impact of changes in the money supply on asset prices. Here, changes in this measure could equally reflect changes in money demand or money supply, e.g., is the announcement of a change in M1 truly exogenous?3 The issue of identification becomes somewhat more subtle when one focuses on short-term rates as the central bank’s main policy variable. In particular, a researcher wishing to isolate the influence of a change in the policy rate on asset prices needs also to be aware that causation may run in the opposite direction, with changes in asset prices leading the monetary authority to change policy rates. This may be particularly acute when considering the impact of monetary policy changes on other interest rates. Further, any adopted framework is unlikely to include all possible channels of influence, i.e. there are potentially omitted variables. Rigobon and Sack (2004) attempt to control for potential endogeneity bias and omitted variables problems by developing a robust estimator. However, they show the impact of failure to account for any such endogeneity appears quite small in practice. Valente (2009), analyzing an international sample of interest rate differentials, further demonstrates that the impact of any endogeneity or omitted variables bias is minimal. Moreover, many central bank practitioners argue that central banks have little role in responding to asset prices per se see, for example, Vickers (2000). Hence we assume that the endogeneity issue will not have a significant bearing on our results and inference.
Theories based on the assumption of efficient markets suggest only unanticipated changes in policy should influence asset prices. Following a change in the policy rate, asset prices should respond only to the surprise element of that change. The anticipated element should already be priced into the asset’s value prior to the announcement. Consequently, analysis that fails to decompose monetary policy changes into expected and unexpected components may lead to biased results due to an errors in variables problem. The most common method used to distinguish between surprises and anticipated changes in monetary policy is to use futures market data see Gurkaynak et al. (2007). Its popularity stems from the fact that futures markets have dramatically increased both their liquidity and the range of instruments on offer. Consistent with the majority of the extant literature we derive measures of monetary policy surprises based on changes in appropriate interest rate futures contracts.
Campbell and Ammer (1993) decompose surprise changes in excess zero-coupon bond returns into revisions in expectations (news) with regard to i) future inflation, ii) current and future real rates and iii) future excess returns. Shiller and Beltratti (1992) derive a similar expression for long-term coupon bonds. Engsted and Tanggaard (2001, 2007) show that the innovation in the excess return on a coupon paying bond is written as;where
Equation (1) is a dynamic accounting identity which can be derived using log-linearization. It states that unexpected excess bond returns
A forecasting VAR is adopted to proxy the components in the above decomposition. However, only three of the four variables in Equation (2) require to be estimated, with the remaining variable being equivalent to the residual. Consistent with previous studies we adopt a linear VAR that includes at a minimum the excess holding period return and the real interest rate. Further variables that are useful in forecasting the two variables of interest can also be included. We elect to include the spread between the long-term bond yield and short-term interest rate since there is wide spread empirical evidence in support of the inclusion of this variable (e.g., Campbell and Shiller, 1991; Fama, 1990).
We represent the forecasting vector autoregression (VAR) as;where
We calculate the relative magnitudes and contributions of each of the news components by normalizing each of the variances and covariances on the right hand side of Equation (5) by TEXT OMITTED FROM SOURCE . As noted by Engsted and Tanggaard (2007), the existence of the covariance terms renders the variance decomposition difficult to interpret. Hence we also measure the relative magnitudes of each of the news components by orthogonalizing them using a Cholesky decomposition and obtaining the
Having derived series for the news regarding the excess holding period return and its constituent components we next seek to take account of possible effects of international monetary policy shocks on each of these variables.
If bond prices reflect the discounted stream of future cash flows, then a surprise change in the policy rate can affect current excess returns by either changing expectations regarding future inflation, real rates or excess bond returns. Using a market derived surprise to domestic and international monetary policy, we seek to identify the impact that the respective surprise has on each of the factors. Bernanke and Kuttner (2005) analyze the impact of monetary surprises on revisions in expected excess stock returns by including the surprise element in monetary policy as an exogenous variable in the forecasting VAR6;where the coefficients represented by the vector
It is now possible to calculate the impact of the monetary surprise on each of the discounted sums of expected future excess returns, real rates and inflation. Moreover, since these sums are related to current excess returns through Equation (2) we can determine the reaction of the bond market to monetary policy shocks. As shown in Equation (4), news regarding future excess holding period returns for each country can be calculated from the VAR coefficients aswhich, when we incorporate the surprise element of policy rate changes as an exogenous variable as in Equation (6), implies
Hence, the response of the present value of expected future excess returns, future real interest rates and future inflation to policy surprises,
The response to unanticipated changes in monetary policy therefore depends not only on
We use a bootstrap simulation (10,000 runs) to compute the small sample standard errors of the components of the variance decomposition - variances, covariances and orthogonalized
Given that we find no evidence of either serial correlation or higher order dependence in our VAR residuals, we adopt a sieve bootstrap, sampling 10,000 observations with replacement from the original empirical sample distribution. However, as a robustness check we also undertake a block bootstrap sampling overlapping blocks of various lengths, the results are quantitatively similar.9
The data is monthly and the sample period for the underlying VAR runs from January 1975 to December 2004. The data is taken from both the International Financial Statistics (IFS) at the IMF and Datastream. Long-term bond yields, treasury yields and consumer price indices (CPI) are all sourced from the IFS for Germany, the U.K. and the U.S. The U.S. Federal Funds futures rate, German 3-month Euromark futures rate (pre euro sample, 1994-1998), 3-month Euroibor futures rate (post euro sample, 1999-2004) and the 3-month Sterling futures contract which are used to derive the respective monetary policy surprises are all sourced from Datastream.
The variables for the three countries included in the VAR are excess returns on bonds, the real short-term interest rate and the spread between the long-term bond yield and the short rate. Using long-term government bond yields for the U.S., the U.K. and Germany, the respective excess holding period return,
Our definition of the monetary surprise for the U.S. follows that of Bernanke and Kuttner (2005). More specifically, in the U.S., the Federal Reserve targets the federal funds rate (an interbank wholesale rate) while the federal funds futures contract is a contract that pays out based on the average level of the federal funds rate and hence can be used to gauge market expectations regarding the expected level of the U.S. policy rate.
Based on this, Bernanke and Kuttner (2005) propose the following measure of the unanticipated element of monetary policy (monetary policy surprise) for month
For both the U.K. and German (Euro area), there are no equivalent futures market instruments that tracks the U.K. or the German (Euro area) policy rate. However, there are interest rate futures contracts that can act as close substitutes since they are likely to be strongly influenced by current expectations of future policy rates. Our proxy for the unanticipated change in the German policy rate between 1994 and 1998 is the 1-month change in the 3-month Euromark futures rate. With the introduction of the euro in January 1999, we proxy surprise changes in the ECB policy rate by the 1-month change in the 3-month Euribor futures rate. Bernoth and Von Hagen (2004) find that the 3-month Euribor futures rate is an unbiased predictor of Euro area policy rate changes.12 For the U.K., the policy rate is the two-week repo rate. Our proxy for the unexpected change in the policy rate is the 1-month change in the 3-month sterling futures contract. This is one of the instruments used by the Bank of England to infer market expectations about the likely course of monetary policy, see Brook et al. (2000).
One concern with using futures rates of a longer maturity than the policy rate, i.e. for the U.K. we use the 3-month sterling futures contract when the policy instrument is the two-week repo rate, is that changes in the former may reflect changes that the market anticipates in the future and not in the immediate horizon. However, Rigobon and Sack (2004) argue that a longer maturity forward contract is more likely to catch a genuine surprise element in the policy rate change rather than a change in timing, i.e. markets are more likely to react to a surprise change in the policy rate relative to when markets had factored in a policy rate change but simply got the timing wrong.13
Although the forecasting VAR is estimated for a sample period running from 1975 to 2004, our measures for the monetary policy surprise are restricted to a shorter sample, February 1994 to December 2004. In the case of the U.S., we restrict the sample to a post 1993 period, given that it was only since February 1994 that the Federal Open Market Committee (FOMC) adopted the present procedure of announcing its policy rate decision after each meeting. Prior to this, markets had to infer what decision had been made by the FOMC after each meeting based on the actions of the Open Market Desk in New York. A second reason for focusing on such a sample is to allow for comparison across countries and so avoid changes in monetary policy regimes when measuring the surprise. The latter is the main consideration in restricting our analysis to post 1994 in both the U.K. and Germany. In the case of the U.K. we concentrate on a post exchange rate mechanism (ERM) currency crisis period, while for Germany the analysis is carried out for a post unification sample.14 Table 1 reports descriptive statistics for each of the three monetary policy surprise measures.
The VAR estimates are reported for the complete sample (1975-2004) in Table 2. The VAR includes the excess return on bonds, the real short-term interest rate and the spread between the long-term bond yield and the short rate for the U.S., Germany and the U.K. Our results are consistent with the anecdotal evidence and the limited empirical evidence that unidirectional spillovers are evident from the U.S. to both the U.K. and the German bond market. As well as the domestic market, U.S. excess returns has a large influence on German returns in particular. The influence of the U.S. is considerably greater for the German bond market, than is the case for the U.K.15 This result is consistent with recent work by Engsted and Tanggaard (2007), although the authors focus solely on the U.S. and German bond markets. As can be seen a somewhat unexpected result is that German excess returns has a small yet statistically significant (negative) effect on U.S. returns. However, consistent with the previous work, U.S. excess returns are dominated by domestic influences. Finally, U.S. variables represent consistently good predictors for both German and U.K. yield spreads, while the German spread has a small negative influence internationally.16
In Table 3, we report the result for the variance decomposition using the news components from Equation (2). The variances and the covariances components from Equation (5) are reported for news about real interest rates, inflation and expected future bond excess returns. Both the total contribution and the respective share as a percentage of current bond excess returns are reported. Consistent with the finding of both Campbell and Ammer (1993) and Engsted and Tanggaard (2007), the news component associated with future inflation, TEXT OMITTED FROM SOURCE , plays the largest role. The dominant influence is news about inflation with the variance of this news term being considerably larger than the other news variance terms although there is some evidence of variation in news about future excess bond returns, this is imprecisely estimated.
As can be seen from Table 3, the dominant influence of the inflation factor is also driving its respective covariance terms, although none are statistically significant. The results in relation to the sign of the covariance between news about future excess returns and news about inflation are theoretically appealing. The negative covariance implies that if market participants receive news that inflation will be higher than expected, they also tend to learn that excess bond returns will be lower than anticipated. Thus bond price variability is reduced since the capital loss from higher expected inflation is partially offset by a capital gain from lower expected excess bond returns. Further, not only is it the case that higher expected inflation is bad news for future excess bond returns (given the negative covariance between TEXT OMITTED FROM SOURCE and TEXT OMITTED FROM SOURCE ), it is bad news for current excess bond returns. The covariance between TEXT OMITTED FROM SOURCE and TEXT OMITTED FROM SOURCE is negative for each of the U.S., Germany and the U.K.17
In the lower panel of Table 3, we assess the magnitude of each of the news components by orthogonalizing each of the news components using the Cholesky decomposition. The table reports the
Table 4 reports all pairwise correlations between the monetary policy surprises and the news components. Consistent with Table 3, there are are large negative correlations ( 0.831 to 0.927) between news about future inflation with both news about future excess returns and future real rates, while the correlation between real rates and future excess returns is positive and much smaller in magnitude (0.039-0.400).
Overall, our results suggest that revisions in expectations regarding current domestic excess returns on long-term bonds is influenced by surprise changes in domestic monetary policy with the exception of the U.S. We find contrasting results for the impact of surprise changes in domestic monetary policy on news of excess returns for German and U.K. bonds with excess returns responding positively in Germany and declining in the U.K. This difference appears to be driven by the contrasting effects of tightening in monetary policy in the two countries. In particular, while a surprise tightening in the U.K. suggests rising inflation expectations, the converse appears to be true in Germany with a contractionary policy leading to declining inflation expectations. One possible explanation for these diverging results is the credibility of the two monetary authorities. It is generally perceived that the Bundesbank had strong inflation fighting credentials. Thus, with enhanced credibility, a surprise tightening of policy would lower inflation expectations. On the other hand, the U.K. has had a mixed history in terms of fighting inflation and may have yet to gain such credibility. Thus, a surprise tightening could suggest to market participants that higher inflation is expected in the future.19
Naturally, during the period under examination, the U.K. monetary policy regime changed with respect to the granting of independence to the Bank of England by the Chancellor of the Exchequer in May 1997. Hence, one potential criticism of our results and subsequent interpretation is that it was only under the new regime that the Bank of England gained a reputation as an inflation fighter and could yet develop the aura of credibility similar to that of the Bundesbank. We examine a shorter sample period 1997:5-2004:12 for a U.K. monetary surprise and find little difference in our results, see Table 6. One potential reason for this is that inflation fighting credibility has to be earned and is not simply gained by granting central bank independence. Thus, it is only over time that markets may change perceptions regarding the effectiveness of a monetary authority in fighting inflation.
Next we seek to assess the impact of surprise changes in monetary policy on revisions in expectations regarding excess bond returns and their constituent components for each country based on Equations (7)-(9) and these results are reported in Table 5. As previously mentioned, we restrict our attention to the sample period 1994:2-2004:12. In the first panel of Table 5 we report the effect of unanticipated US monetary policy on news regarding domestic and foreign current excess bond returns and their respective constituent components, future excess bond returns, future real interest rates and finally future inflation. For example the first row indicates the contemporaneous impact that a U.S. monetary policy surprise has on excess bond returns in the U.S.(0.036%), Germany (0.000%) and the U.K. ( 0.014%). As can be seen the coefficients are very small and not statistically different from zero in each case. In fact U.S. monetary policy does not have a statistically significant impact on any of the three countries’ excess bond returns or their individual components. This result is potentially surprising given the belief that U.S. monetary policy has a strong influence on global bond returns.
In the second panel of Table 5 the impact of surprise German (Euro area) monetary policy changes on news regarding both German and foreign excess bond returns as well as their respective components are reported. There is a significant contemporaneous rise in German excess bond returns in response to a surprise domestic monetary tightening. Specifically, a 1% surprise increase in the policy rate, will lead to a close to 0.05% contemporaneous increase in unexpected domestic bond returns. This result may appear counter intuitive and not economically large at first glance, however as can be seen, the rise in current excess returns is driven by the downward revision in future inflation expectations. This could potentially point to the credibility of German (Euro area) monetary policy with a surprise tightening in the policy rate leading to lower inflation expectations. Surprise changes in German (Euro area) monetary policy do not significantly impact on news regarding excess bond returns in either the U.S. or the U.K.
Finally, in third panel of Table 5 we report the impact of U.K. monetary policy on news regarding domestic and foreign excess bond returns. Here we observe that a surprise tightening by the Bank of England has a significant negative effect on news regarding current excess bond returns. When we focus on the driving force behind this, we see that an unanticipated monetary tightening appears to lead to higher inflation expectations and hence declining excess returns. While there is no evidence that U.K. monetary policy has any impact on U.S. excess returns, there is a significant negative impact on German excess returns. However, the results reported for the impact of a U.K. surprise on the news components are consistent with those reported earlier for both the U.S. and the German (Euro area) surprise.
This paper examines the impact and possible spillovers effects of unanticipated monetary policy on international bond returns. Further, we decompose the response of bond returns to monetary policy surprises to delve deeper into the reasons behind such responses. We seek to assess the impact of surprise changes in monetary policy on revisions in expectations regarding excess bond returns and their constituent components for each country, enabling us to identify whether the response is due to revisions in expectations regarding future excess returns, future real interest rates or future inflation.
The VAR model uncovers evidence of U.S. spillovers in both Germany and the U.K., with U.S. excess returns having a significant impact on German excess returns, German real short rates and both the U.K. and German spreads. However, somewhat surprisingly, when we investigate the effect of a U.S. monetary policy surprise there is no statistically significant impact on any of the three countries’ excess bond returns or their individual components. This result contrasts with the widely-held belief that U.S. monetary policy has a strong influence on global bond returns. This suggests any impact is confined to high-frequency intra-day adjustments which cannot be uncovered at the daily frequency.
Importantly, we demonstrate that there is a significant rise in German excess bond returns in response to a surprise domestic monetary tightening as future inflation expectations are revised significantly downwards. This potentially suggests a credible German (Euro area) monetary policy with a surprise tightening in the policy rate leading to lower inflation expectations. Conversely, we observe that a surprise tightening by the Bank of England has a significant negative effect on news regarding current excess bond returns in the U.K. It appears that an unanticipated monetary tightening in the U.K. leads to higher inflation expectations and hence declining excess returns. Our results for the U.K. are consistent for samples that focus purely on post Bank of England independence. Our results infer that the Bank of England has not yet been successful at establishing credibility in relation to fighting inflation.
The authors would like to thank two anonymous referees, Michael Ehrmann, Laurence Booth, participants at the FMA Europe meeting in Barcelona (May 2007) and seminar participants at the University of Strathclyde for comments and suggestions, and Tom Engsted for sharing his data. The views expressed here are our own and do not necessarily reflect the views of the ESCB or the staff of the Central Bank of Ireland. Any remaining errors are our own.
1
Engsted and Tanggaard (2007) recently apply this decomposition to U.S. and German bond markets while Barr and Pesaran (1997) decompose U.K. bond returns. However, neither study examines the impact of monetary policy shocks.
2
Gurkaynak et al. (2007) find that futures market measures most accurately capture expectations of monetary policy.
3
Failure to properly identify monetary supply changes leads some researchers to find counter intuitive results. See Sellin (2001) for an overview of such problems.
4
The short-term real interest rate is the nominal short rate less the monthly inflation rate while the spread is defined as (1
5
From equation (3), the best prediction of
6
The following formation of the model draws on Bernanke and Kuttner (2005).
7
Bernanke and Kuttner (2005) also claim that since the monetary policy surprise represents the prediction error from a rational forecast,
8
Alternatively, we could include the shock in the monetary policy rate in the forecasting VAR.
9
Results available from the authors on request.
10
The values of
11
We tested the lag length in the VAR using the standard information criteria, Akaike information (AIC) and Schwartz Bayesian (SBC), and found they suggested a lag length of one.
12
Euribor stands for Euro-Interbank Offer Rate.
13
Rigobon and Sack (2004) use the 3-month euro dollar rather than the 1-month Fed funds futures contract in their study of the U.S.
14
A number of sensitivity tests have been carried out in relation to the selected sample period and are discussed later.
15
U.S. real rates have particularly high influence on U.K. real rates, even taking into account the impact of lagged U.K. real rates.
16
The German yield spread is significant at 5% for the case of the U.K. and 10% for the U.S.
17
The covariances are: U.S. 0.004, Germany 0.003 and U.K. 0.003. All are significant at a 5% level.
18
We also conducted a bivariate analysis on each pair of bond markets. The results were consistent with the trivariate analysis and not qualitatively different. Results available from the authors.
19
Indeed, analysis of changes in U.K. monetary policy and actual future inflation reveal that a tightening in monetary policy is associated with higher actual inflation 6 months, 12 months and 24 months in the future. Results are available from the authors on request.
