A Note on the Foreign Exchange Market Efficiency Hypothesis

Article excerpt


This paper examines the weak and strong forms of the foreign exchange market efficiency hypothesis (MEH) (as defined in the paper) using the recently available Harris-Inder null of cointegration procedure, which is powerful enough to distinguish between cointegration and near cointegration, and thus provide more robust results than conventional cointegration tests. Our results indicate that both forms of the MEH are rejected for all the major currencies of the European Economic Community (EEC). (JEL F310)


In this note we re-examine the foreign exchange market efficiency hypothesis, which is a hotly debated topic in the area of international finance. It is basically the theory of informationally efficient markets applied to the foreign exchange arena. The present literature is far from conclusive and inconsistencies abound, as is evident from Lewis (1995).1 With the genesis of the concept of nonstationarity and cointegration came a new approach to testing market efficiency. A multitude of procedures are available, but the standard methodology has been to examine the forward market unbiasedness hypothesis, which tests whether forward rates are unbiased and efficient estimators of the future spot rate.

Acceptance of this hypothesis implies that the spot and forward foreign exchange rates have a tendency to move together over time, i.e., they are cointegrated in the Engle-Granger (EG 1987) sense. The estimated model is

Given this inconclusive state of affairs, we re-examine the MEH for the major EEC currencies over the floating exchange rate period3 using the recently available null of cointegration procedure which is able to distinguish between unit and near-unit roots. This avoids the problems with classical tests mentioned above. We first test stationarity of the log-levels of spot and forward rate series and then examine the bivariate cointegrating properties, using the Harris-Inder (HI 1994) methodology. The MEH is soundly rejected.

This study is divided into three sections. In section 2, we sequentially conduct the classical Dickey-Fuller (DF 1981) and the modern Kwiatkowski et al. (KPSS 1992) unit-root test on the series under consideration. The HI cointegration test is undertaken in section 3, followed by our results and brief concluding remarks.

Tests of Nonstationarity

We examine the nonstationarity of the spot and forward rate series using the standard DF procedure. The results are reported in Table 1.

As shown in Table 1, the value of the test statistic for the log-levels of each series in each case is less than the critical value, implying that the null hypothesis of nonstationarity (presence of unit-roots) cannot be rejected at the five-percent significance level.4

The DF results cannot be taken as conclusive evidence of the existence of unit roots. This method is restrictive because of the unrealistic assumption of identically and independently distributed (iid: 02) Gaussian processes. There is now a substantial body of evidence that financial time series including exchange rates exhibit time-dependent heteroskedasticity.

Another criticism of the classical unit-root testing procedure is that it cannot distinguish between unitroot and near-unit root stationary processes. This prompted the use of the Kwiatkowski et al. (KPSS 1992) method, where the null is stationarity and the alternative is the presence of a unit root. This ensures that the alternative will be accepted (null rejected) only when there is strong evidence for (against) it. The test statistics eta^sub r^ and eta^sub (mu) in Table 2 (for the spot and forward series) use the null of stationarity with and without a time trend, respectively.

Since in each case the test statistic is greater than the critical value, we reject the null of stationarity in favor of the alternative of unit roots. This agrees with the literature.

Null of Cointegration Procedure

The classical or the conventional nonstationarity test procedures examine the null hypothesis of a unit root, meaning the null will be accepted (not rejected) unless there is strong evidence against it. …