Detecting breaks on the long memory: a case about the brazilian unemployment

Resumen

This study analyzed the dynamic behavior of the Brazilian unemployment rate, focusing on the persistence level of the series. For this purpose, it was adopted first models of fractionary integration, besides persistence change tests for the series. The first results indicated a non-stationary behavior of the series. However, knowing that neglecting a structural break can cause a bias on the parameter estimative, new estimations were made, which results indicated that the unemployment rate presents two different levels of persistence. On the first level, the series is no-stationary, whereas in the second, it is also non-stationary, but it presents a mean reversion feature