Very Preliminary and Incomplete;
Very preliminary and incomplete;
please do not quote
SOME EMPIRICAL EVIDENCE ON THE EFFECTS OF MONETARY POLICY SHOCKS
ON FOREIGN INTEREST RATES AND CROSS EXCHANGE RATES
Sarantis Kalyvitis
Athens University of Economics and Business
Ifigeneia Skotida[*]
Athens University of Economics and Business, and Bank of Greece
December 31, 2005
Abstract: This paper examines the impact of US monetary policy shocks on the interest rates and cross exchange rates of UK, Japan and Germany using a Structural VAR approach covering the period 1980-1996. We find that after a US contractionary monetary shock all the major currency crosses (Pound-Yen, Yen-Mark, Mark-Pound) exhibit ‘delayed overshooting’. There is also some evidence of excess returns in the corresponding foreign exchange markets following the US monetary shock whereas a substantial portion of exchange rate variability can be attributed to the foreign monetary shock. We attribute these findings <TRADE BALANCE INTERPRETATION HERE>.
Keywords: monetary policy, cross exchange rate, excess returns.
JEL classification number: E52, F31.
1. Introduction
Following the classical Dornbusch (1976) ‘overshooting’ model, a large empirical literature has focused on the effects of monetary policy changes on financial markets with particular emphasis placed on the impact of US monetary on interest rates and exchange rates.[1] In this paper we explore the impact of US monetary policy on asset prices in other economies; this dimension of US monetary policy was originally noted by Frankel (1986), but has been subsequently ignored in relevant empirical studies.
Existing empirical studies quantifying the effects of US monetary policy shocks are largely based on VARs with US monetary policy typically being identified by exogenous shocks in the Federal Funds rate (FFR) or in the ratio of nonborrowed to total reserves (NBRX) or in the FFR target.[2] In their extensive study on the empirical effects of US monetary policy on USDollar exchange rates, Eichenbaum and Evans (1995) have shown that in response to a tighter US monetary policy the USDollar exhibits a ‘delayed overshooting’ behavior of 2 to 3 years vis-à-vis the major currencies (Japanese Yen, Mark, Italian lira, French Franc, and Sterling Pound), a pattern that is confirmed by Clarida and Gali (1994). ‘Delayed overshooting’ is also assessed by Evans (1994), who uses weekly data and finds that the USDollar overshoots with a delay of 2 to 3 years vis-à-vis the Mark and the Yen, and Lewis (1995), who finds that the USDollar response relative to the Mark and the Yen increases for the first 5 months after the monetary shock.[3] Eichenbaum and Evans (1995) also offer empirical evidence that after a contractionary US monetary shock, domestic interest rates rise and the USDollar appreciates. This accumulated evidence creates in turn a ‘conditional forward premium puzzle’ with opportunities for excess returns by borrowing abroad and investing in the US. Despite the fact that the standard “forward premium puzzle” is well documented in empirical studies of the foreign exchange market (see, for instance, Hodrick (1987) and Froot and Thaler (1990)), the puzzle now arises conditional on an exogenous change in US monetary policy.
In an attempt to address these issues, Cushman and Zha (1997) proposed using a SVAR for two countries, assuming block exogeneity for the domestic variables of a small open economy (Canada) on the external (US) variables. Their specification resolves some standard empirical puzzles (like the domestic currency depreciation following a domestic contractionary policy, as well as the ‘price puzzle’), but still short-run deviations from uncovered interest rate parity persist. Similar results are found in Voss and Willard (2003), who examined the case of US and Australia. Kim and Roubini (2000) used an identified/non-recursive VAR that allows monetary policy to respond contemporaneously to exchange rate shocks and resolved partially the forward discount bias puzzle and the ‘delayed overshooting’ effect. Furthermore, the authors provided evidence that, in response to an unexpected increase of the US policy rate, the short-term interest rates of non-US G7 countries increase and their currencies relative to the US dollar depreciate on impact. ‘Delayed overshooting’ is also addressed by Kalyvitis and Michaelides (2001) who use the Bernanke and Mihov (1998) monetary policy indicator as a measure for the US monetary policy stance and adopt a VAR empirical specification with relative output and price in order to capture the relative business cycle position of the US and the domestic country. Their specification solves the ‘delayed overshooting’ effect, but the authors also report deviations from the uncovered interest rate parity hypothesis. Faust and Rogers (2003) assess the robustness of conclusions drawn from SVAR models relative to changes in the identifying assumptions and find that the ‘delayed overshooting’ result is sensitive to the assumptions adopted. In contrast, they provide evidence that deviations from uncovered interest rate parity are robust to various assumptions and, moreover, that the estimates of the share of exchange rate variance due to monetary policy shocks can take quite large values. Recently, Bluedorn and Bowdler use the Romer and Romer (2004) index of US monetary policy and find that the international transmission effects of US monetary shocks is stronger than those encountered when standard measures are used.
Even though monetary authorities are guided primarily by domestic economic considerations, the adoption of floating exchange rates and the gradual removal of capital controls after the 70s have strengthened the transmission mechanism from internal monetary policy developments to the rest of the world. The ‘delayed overshooting’ and the ‘conditional forward premium puzzle’ could be specific examples of how economies might be influenced by monetary policy abroad. These effects may be exacerbated by the presence of asymmetries in the monetary policy functions of foreign countries, like differential responses in their output gaps or inflation rates.
Hence, a challenging empirical issue in the exchange rate example is whether the magnitude of, say, the USDollar appreciation against other major currencies following a US contractionary monetary shock differs. For instance, does the USDollar appreciate more against the Mark or the Yen or Pound? Motivated by these questions, we ask how do cross exchange rates of major currencies (Pound/Yen, Yen/Mark, Mark/Pound) respond to US monetary policy shocks? In a related vein, we ask how do interest rate differentials between these countries (UK-Japan, Japan-Germany, Germany-UK) respond to tighter monetary conditions in the US. And, also, are there excess returns in the face of tighter monetary conditions in the US from borrowing, for instance, in Germany and investing in the UK or Japan?
In an attempt to assess empirically these open questions, the paper explores the impact of monetary policy shocks on foreign countries’ interest rates and the corresponding cross exchange rates. So far, there is only indirect empirical evidence of the effects of monetary policy on these variables based on their responses vis-à-vis the US. The current study attempts to fill this gap by using a SVAR approach to investigate the impact of US monetary policy changes on the interest rate differential and the cross exchange rates of other economies. We proxy US monetary policy changes by the Romer and Romer (2004) narrative index of US monetary policy that identifies the Federal Reserve’s intended, rather than actual, changes in the Federal Funds rate. The Romer and Romer index is better equipped to overcome the endogeneity and anticipatory biases inherent in standard measures of US monetary policy (like the Federal Funds rate of the non-borrowed reserves ratio), as intended interest rate changes are regressed upon the Federal Reserve’s forecasts of macroeconomic variables with the residuals forming a measure of exogenous monetary policy shocks. We then consider the four major currency blocks, namely the US, UK, Japan and Germany, and we proceed in examining the impact of a US monetary policy shock on the interest rates and cross exchange rates in other countries pairs
Our evidence points towards a ‘delayed overshooting’ pattern for all currencies examined when the Federal Reserve tightens. At the same time, the central bank interest rate in Japan rises. Consequently, we find that an unexpected US monetary contraction generates spillover effects on foreign exchange rates and interest rates, and also leaves room for excess returns. The results are robust to various measures of US monetary policy and alternative empirical specifications. Thus, our findings document a monetary policy effect, which has been unnoticed in existing studies on the monetary policy transmission mechanism.
<TEXT HERE>
The rest of the paper is structured as follows. Section 2 outlines the empirical methodology used in the paper. Section 3 assesses the impact of the monetary policy shocks from two SVAR specifications on foreign interest rate differentials and cross exchange rates, and also discusses their effects on exchange rate variability. Section 4 gives some robustness checks and section 5 concludes the paper.
2. Empirical methodology
In this section we present the identification of US monetary policy and two SVAR specifications based on alternative structural assumptions. Our general empirical strategy is to develop a setup where the external economy (US) is assumed to be large and relatively closed. Thus, a US monetary policy shock reflects an exogenous shock for other countries that in turn respond to domestic and external developments.
In our approach we shall use the as a measure of US monetary shocks the indicator proposed by Romer and Romer (2004), who have further developed the narrative approach initially introduced by Friedman and Schwartz (1963) and formalized by Romer and Romer (1989). In Romer and Romer (2004) a series on intended funds rate changes around the meetings of the Federal Open Market Committee is first derived. In order to eliminate endogenous interest rate changes linked to economic conditions, this series is regressed upon the Federal Reserve’s internal forecasts of inflation and real activity. Thus the residuals represent the intended interest rate changes that are orthogonal to information about future economic conditions. The authors find that the policy effects are substantially stronger and quicker using this new measure, than other conventional US monetary policy indicators, like the change in the Federal Funds Rate or the non-borrowed reserves.[4]
2.1. Specification with country-specific economic variables (Model I)
Most empirical studies of the impact of monetary policy on asset prices involve VARs with identification imposed as restrictions mapping the reduced form shocks to structural shocks. Eichenbaum and Evans (1995) estimated VAR models for US and the other G-7 variables including US and foreign output, prices, interest rates, US non-borrowed reserves, the US federal funds rate and the nominal dollar exchange rate vis-à-vis the foreign currencies. The monetary policy shock is identified in a recursive VAR structure in the form of an orthogonal innovation to the US federal funds rate or non-borrowed reserves. In a series of papers, Cushman and Zha (1997), Kim (1999), Kim and Roubini (2000), and other authors, have emphasized that it is important to use a non-recursive approach to identify innovations such as monetary policy shocks, because it allows for contemporaneous responses between economic variables.
Our first SVAR specification reports the effects of a shock in the monetary policy of a large economy on the interest rates of three pairs of countries and their corresponding cross exchange rates. This specification is broadly based on the identification scheme proposed by Kim and Roubini (2001), but is modified to account for the responses of the two countries’ variables to an exogenous shock in the monetary policy of a large economy. This specification allows us to explore in detail the presence of potential asymmetries in the responses of the two countries as a result of a monetary shock abroad.
In particular, we assume that the structural shocks, e(t), are related to the reduced form residuals, u(t), as follows:
e(t)=G0 u(t)
where the contemporaneous coefficient matrix, G0, is given by:
The vector of variables for countries 1 and 2 consists of the log of output (IP1 and IP2), the log of the price index (CPI1 and CPI 2), the log of a monetary aggregate (M0 and M1), the central bank interest rate (CR1 and CR 2), the log of the nominal cross exchange rate expressed in units of the currency of country 1 in terms of the currency of country 2 (XR1/2), the log of the world commodity price index (CI) and finally the (cumulated) Romer and Romer (2004) shocks measure (RR).
The equations of the model are in turn divided into three blocks that mostly follow, among others, Gordon and Leeper (1994), Kim (1999), Kim and Roubini (2000). The first block with four equations represents the sluggish goods markets in countries 1 and 2, under the assumption that real activity does not respond contemporaneously to the price level, the interest rate and the exchange rate of the domestic economy. Thus, although in general unexpected changes in prices or financial variables affect real activity in the long run, it is assumed here that firms cannot adjust contemporaneously their output due to inertia and other adjustment costs. Furthermore, we assume that there are no contemporaneous mutual effects between the real sectors and the price levels of the two countries and also that both the output and price levels do not respond contemporaneously to innovations in the Romer and Romer (2004) indicator, which makes sense given our monthly data frequency.
The second block describes the demand and supply of money. Specifically, the 5th and 6th equations represent the money demand equations for the two countries where the standard specification with the demand for real money balances depending on real income and the opportunity cost of holding money is adopted. In turn, the next two equations (7th and 8th row) describe the monetary policy rule with the discount rate set by the central bank after observing the current domestic monetary aggregate and the exchange rate vis-à-vis the other foreign country (cross exchange rate), which is consistent with central bank policy.[5] Jang and Ogaki (2004) have recently shown that results from VARs on the effects of US monetary policy on exchange rates are sensitive to this assumption. Notice that the central bank does not react to shifts in the current values of the monetary aggregate of the foreign economy, as well as in the output and price levels at home and abroad, due to information delays in data availability. Also, as in Kim and Roubini (2000) the domestic interest rate is restricted from responding contemporaneously to unexpected shifts in the third country (US) monetary policy shocks.