Resumo

LIZARAZU-ALANEZ, Eddy  e  VILLASENOR-ALVA, José A.. Effects of breaks under the null hypothesis with the Dickey-Fuller test for unit root. Agrociencia [online]. 2007, vol.41, n.2, pp.193-203. ISSN 2521-9766.

The Dickey-Fuller (DF) test has low power if it is based on the least squares estimator for the parameter of an autoregressive process (Y t −μ)=ρ (Y t-1 −μ)+u t , with error u t ∼iidN (0,σ2), where H 0: ρ=1 and H1: |ρ| < 1. When the series that corresponds to a unit root process has a break, the Dickey-Fuller test leads to a spurious rejection of the null hypothesis. This article studies, through Monte Carlo simulation, the effects in the DF test for two types of breaks under the null hypothesis: (1) the coexistence of breaks in the level of the series and in the variance of the error term; (2) two breaks in the level of the series. The simulation results confirm the property of robustness of the recursive mean adjustment Dickey-Fuller (DFR) test regarding the DF test.

Palavras-chave : Recursive mean adjustment; structural break; unit root, autoregressive processes.