TITLE:
Can Choice of Reference Density Improve Power of M-Estimation Based Unit Root Tests?
AUTHORS:
Tapan Kar, Malay Bhattacharyya
KEYWORDS:
Unit Root, M-Estimation, Local-to-Unity, Johnson SU Distribution, Reference Density
JOURNAL NAME:
Journal of Mathematical Finance,
Vol.12 No.2,
May
12,
2022
ABSTRACT: In this paper, we investigate if the choice of the reference density could improve the power of M-estimation-based unit root tests. For this investigation, we consider models where the AR-coefficient is very close to one (local-to-unity) in the true data generating process. Motivated by the stylized facts that empirical return distributions have large skewness and high leptokurtosis, we explore if Johnson SU and Pearson Type IV distributions can be used as the reference densities to improve the power of the M-estimation based unit root tests. Through extensive simulations, we find that the proposed procedure, in finite samples, is as powerful as the Dickey-Fuller test for normal errors and is significantly more powerful than several existing tests for non-normal errors. We apply the proposed test to the Nelson and Plosser data set and to the nominal monthly interest rate of India.