Unbiased Estimation as a Solution to Testing for Random Walks

Abstract:     
A new formula for the mean squared error of a unit root estimator is derived. Its minimization gives rise to an estimator which is almost unbiased, unlike the maximum likelihood estimator. For this reason, it is suggested that the problem of testing for unit roots may be solved more satisfactorily in the future by using estimation techniques that lead to smaller biases.

working papers series