Expectations Hypotheses Tests
NBER Working Paper No. 7609
We investigate the Expectations Hypotheses of the term structure of interest rates and of the foreign exchange market using vector autoregressive methods for the U.S. dollar, Deutsche mark, and British pound interest rates and exchange rates. In addition to standard Wald tests, we formulate Lagrange Multiplier and Distance Metric tests which require estimation under the non-linear constraints of the null hypotheses. Estimation under the null is achieved by iterating on approximate solutions that require only matrix inversions. We use a bias-corrected, constrained vector autoregression as a data generating process and construct extensive Monte Carlo simulations of the various test statistics under the null hypotheses. Wald tests suffer from severe size distortions and use of the asymptotic critical values results in gross over-rejection of the null. The Lagrange Multiplier tests slightly under-reject the null, and the Distance Metric tests over-reject. Use of the small sample distributions of the different tests leads to a common interpretation of the validity of the Expectations Hypotheses. The evidence against the Expectations Hypotheses for these interest rates and exchange rates is much less strong than under asymptotic inference.
Published: Bekaert, Geert and Robert J. Hodrick. "Expectations Hypotheses Tests," Journal of Finance, 2001, v56(4,Aug), 1357-1394.