JOURNAL OF APPLIED ECONOMETRICS, VOL. 6, 125-142 (1991) THE TIME-SERIES PROPERTIES OF THE RISK PREMIUM IN THE YEN/DOLLAR EXCHANGE MARKET FABIO CANOVA Department of Economics, Box B, Brown University, Providence, RI 02912 and Department of Economics University of Rochester, Rochester, NY 14627, USA AND TAKATOSHI ITO Department of Economics, Hitotsubashi University, Kunitachi-shi, Tokyo, 186 Japan; Department of Economics University of Minnesota, Minneapolis, MN 55455; and National Bureau of Economic Research, Cambridge, MA, 02138 SUMMARY In this paper a VAR model is employed to construct a measure of the conditional expectations of the future yen/dollar spot rate. This measure allows us to examinethe dynamics of an ex-antetime-series for the risk premium in the market. The VAR model produces 'better' forecasts thanthe survey responses for turbulent periods such as 1981-1982and 1984-1985. The VAR-generated expectations are then used to construct a risk premium time-series. Thisrisk premium ieerie seriesseemsto be morereliable than the ones obtained using either survey data on expectations of the future spot exchange rateor the ex-post realized spot exchange rate. Tests on the risk premium series suggest that a risk premium was present, but that it was virtually constant throughout the sample. The conditional variance of the risk premium changed over time, but its unconditional distribution seemed stableacross subsamples. Despitethese features, the volatility of the series was substantial and variedconsiderably throughout the sample. 1. INTRODUCTION The size and the variability of the risk premium in foreign exchange markets are of crucial importance in analysing issues concerning the efficiency of thy ese markets. Several authors (for example, Frankel, 1986; Frankel and Meese, 1987) have argued that, theoretically, the risk premium should be small and approximately constant. The experience of the 1980s in several foreign exchange markets indicates that the ex-post bias of the forward rate is very large and volatile. Since the ex-post forward bias is the sum of a forecast error and of a risk premium, many researchers find it hard to attribute the empirical features of the forward bias to a risk premium, and conclude that there are persistent patterns in the forecast errors made by agents. Some have interpreterd this evidence as an indication that a 'peso problem' may have occurred in the 1980s (see e.g. Krugman, 1987). Others have used this result to question the rationality of investors' expectations (see e.g. Frankel and Froot, 1987). Since the risk premium is unobservable, a crucial step in investigating the efficient market hypothesis is the construction of (a proxy for) the risk premium. The empirical evidence on the properties of risk premium proxies is somewhat contradictory. Hansen and Hodrick (1980, 1983), Hodrick and Srivastava (1984, 1986), Cosset (1984), Domowitz and Hakkio (1985), and Cumby (1988) suggest that the risk premium is important in explaining the ex-post bias of the forward rate and the existence of complex pattern of heteroskedasticity and nonlinearities in 0883-7252/91/020125-18$09.00 Received May 1988 © 1991 by John Wiley & Sons, Ltd. Revised October 1990