On the Small Sample Properties of Dickey-Fuller and Maximum Likelihood Unit Root Tests on Discrete-Sampled Short-Term Interest Rates

Abstract

Testing for unit roots in short-term interest rates plays a key role in the empirical modelling of these series. It is widely assumed that the volatility of interest rates follows some time-varying function which is dependent of the level of the series. This may cause distortions in the performance of conventional tests for unit root nonstationarity since these are typically derived under the assumption of homoskedasticity. Given the relative unfamiliarity on the issue, we conducted an extensive Monte Carlo investigation in order to assess the performance of the DF unit root tests, and examined the effects on the limiting distributions of test procedures (t- and likelihood ratio tests) based on maximum likelihood estimation of models for short-term rates with a linear drift.