Desempenho dos testes de raiz unitária com mudança no nível de dependência cross-section

Abstract

Unit root tests have been widely used to validate or reject economic model’s hypotheses. Because of this, many authors have created different versions of this kind of test in order to generate statistics which are more precise in identifying the presence of a unit root. Some authors have increased the power of these statistics using panel data. However, the use of panel data brings the possibility of dependence between cross-sections, this has been initially handled by the independence between cross-sections hypothesis. Only the second generation tests consider dependence between cross-sections. Nevertheless, in the literature there is no test which allows changes in the dependence between cross-sections over time. Thus, this paper uses Monte Carlo experiments to analyze the small sample properties of some statistics used to identify the presence of a unit root. It is noticed that the size of these statistics has a large distortion when the level of dependence between cross-sections changes.