RDP 2005-02: The Impact of Monetary Policy on the Exchange Rate: A Study Using Intraday Data 1. Introduction

Recent studies have had some success in identifying the response of the exchange rate to macroeconomic variables by using high-frequency data.[1] This paper makes two important contributions to the literature on the response of the exchange rate to monetary policy. First, we use intraday data which allows us to more precisely control for endogeneity and external factors that may influence both exchange rates and interest rates (such as macroeconomic data releases). With intraday data we can also examine the temporal response of the exchange rate. Our second contribution is to consider how changes in the expected path of future monetary policy that result from a monetary surprise influence the response of the exchange rate. Some interest rate changes may surprise with respect to the timing of the change; for example, the rate rise expected next month occurred this month. Others may surprise with respect to the expected path of monetary policy; for example, a surprise rate rise might be taken to indicate that a tightening phase is going to reach a higher maximum than previously anticipated. Because these surprises will have different effects on the expected future path of monetary policy they are unlikely to have equivalent effects on the exchange rate.

A greater understanding of the impact of interest rates on exchange rates is of interest for several reasons. The theory of uncovered interest parity (UIP), which connects expected changes in the exchange rate to interest differentials, is central to almost all international macroeconomic models. Yet empirically, UIP is a resounding failure (Engel 1996). In addition, the response of the exchange rate to monetary policy is also an important monetary transmission channel in small, open economies (see, for example, Grenville 1995; Thiessen 1995).

Our study includes four countries (Australia, Canada, New Zealand and the United Kingdom) that are relatively small, and so changes in their interest rates are unlikely to affect global interest rates. This is important for isolating the impact of the change in one country's interest rate on the exchange rate. If the country studied was large, such as the United States, then markets might build the likely impact of changes in domestic monetary policy on foreign interest rates into the exchange rate's response. This would contaminate the measured response of the exchange rate. Further, these four countries have highly liquid financial markets, freely floating exchange rates and similar monetary policy regimes.

We use an event study methodology as has become common in the literature on asset prices. An event study is particularly useful because it can abstract from the joint determination of interest rates and exchange rates. The event is a monetary policy decision (either a surprise change in the policy interest rate or no change when a policy announcement was anticipated). We can be confident that we have isolated events in which causality runs in only one direction, from interest rates to exchange rates, for two reasons. Firstly, we use a narrow event window, only examining a short period around the policy change. Secondly, for the countries we study, the institutional structure of monetary policy decision-making means that the decision is made well before the event window we use.

Several papers have recently used high-frequency data to examine the response of asset prices to macroeconomic shocks, including interest rates. This paper most closely follows Zettelmeyer (2004) who examines the response of exchange rates to interest rates using daily data (but not intraday). Unlike Zettelmeyer we restrict our sample to a period in which the central banks we study did not explicitly respond to the exchange rate. We also use a more accurate measure of the monetary surprise (based on 1-month rather than 3-month interest rates) and a larger sample, in part because we include decisions in which monetary policy does not change, that is, ‘no-change’ surprises. We can include these observations because, under the monetary policy regimes we examine, the timing of the announcement of these no-changes was predetermined. Faust, Rogers, Wang and Wright (2003) use intraday data to examine the response of exchange rates to macroeconomic announcements, including interest rates changes. But they only study surprises in US interest rates, and so the exchange rate responses are potentially clouded by anticipated changes in foreign interest rates. Andersen et al (2003) also examine the intraday response of the exchange rate to macroeconomic announcements, but do not consider interest rate shocks. Bernanke and Kuttner (forthcoming), studying the response of equity markets to interest rates using daily data, consider how the impact differs depending on the changes to the profile of anticipated future monetary policy, as we do in this study. A related literature has attempted to consider the longer-run impact of interest rates on the exchange rate. In an early study using a vector autoregression (VAR), Eichenbaum and Evans (1995) suggest that there exists a delayed overshooting. But by identifying surprise interest rate shocks using daily data, Faust, Rogers, Swanson and Wright (2003) find that this result is not robust to allowing the foreign interest rate to respond. Faust and Rogers (2003) also fail to find evidence of delayed overshooting in a less restricted VAR.

The remainder of the paper is structured as follows. Section 2.1 briefly outlines the application of the event study methodology to monetary policy decisions. In Section 2.2 we review the monetary policy operations of the four countries in the study and discuss how this influences the set of events that we consider. The data are described in more detail in Section 2.3. In Section 3.1 we present the results of the instantaneous impact and the timing of the response of the exchange rate. In Section 3.2 we demonstrate how the response of the exchange rate depends on the effect of the monetary surprise on expectations. We examine the robustness of the results in Section 3.3. Section 4 concludes.


For example, see Andersen et al (2003), Faust, Rogers, Wang and Wright (2003) and Zettelmeyer (2004) and the references therein. [1]