Review of Quantitative Finance and Accounting, 6 (1996): 39-46
© 1996 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Testing the Unbiasedness Hypothesis of Foreign
Exchange Rates and the Analysis of Transformations
ALBER!' A. OKUNADE AND H. HARYANTO
Department of Economics, 400BB (FCBE), The University of Memphis, Memphis, TN 38152
DWIGHT B. MEANS, JR.
Department of Finance, 402BB (FCBE), The University of Memphis, Memphis, TN 38152
Abstract. Recent researchers have utilized various functional forms for testing the hypothesis that the forward
rate is an unbiased predictor of future spot rates in foreign exchange markets. We compare a large number of
these functional forms for a similar time period and test their consistency with the data for five major currencies.
Our results imply that certain functional form models may be inappropriate for some currencies. Researchers
must, therefore, be cautious of misspecification due to erroneous functional forms when testing the unbiased
forward rate hypothesis.
Key words: foreign exchange rate, functional forms, transformation, forward rate, spot rate
1. Introduction
Recent researchers of the foreign exchange market have tested the hypothesis that the for-
ward rate is an unbiased predictor of the future spot rate. The hypothesis, based on expec-
tation theory, contends that all information which influence the future spot rates is already
reflected in the forward rate (see Chiang (1986». This conjecture relies on the efficient
market hypothesis, akin to the rational expectation hypothesis in economic theory.
The two popular a priori restrictive functional forms used in previous studies to test the
unbiased forward rate hypothesis (UFRH) are: (1) The double-log form (see Fama (1984),
Chiang (1986), and Barnhart and Szakmary (1991); and (2) the linear functional form (see
Bilson (1981) and Boyer and Adams (1988)). Prior studies produced mixed results on the
tenability of the UFRH. Chiang (1986), for example, argued that the statistical rejection
of UFRH implied that either the market was inefficient or the model was misspecified,
or both. Barnhart and Szakmary (1991) showed that the conflicting results derived from
the different econometric specifications and periods of estimation. Therefore, they looked
at other models which they classified as "level" and "percent change" specifications. In-
terestingly, however, none of the previous researchers conducted a statistical hypothesis
test concerning the appropriateness of using double-log and linear functional forms for
the data in their analysis.
Rather than testing the UFRH, this research note focuses on conducting statistical
hypotheses tests regarding the appropriateness of specific functional forms when specify-
ing the UFRH for econometric estimation. Applying the Box-Cox (1964) transformation
model, we test the linear, double-log, log-linear, linear-log, classical Box-Cox, Tidwell