The Finite-Sample Bias of the Autoregressive Parameter Under Rank-based Estimation

Monte Carlo simulation is employed to examine the finite-sample properties of the estimated autoregressive parameter associated with the Dickey-Fuller and rank-based Dickey-Fuller unit root tests. While the downward bias associated with estimator under the Dickey-Fuller test has long been noted in the literature, the corresponding properties for the rank-based test have not been considered previously. The results presented show that in comparison to the DF test, increased bias is present for the rank-based test.