Description

This paper tests the hypothesis of linearity against a specific form of nonlinearity in the Data Generating Process (DGP) of the unemployment rate and the difference between the inflation rate (CPI and CPIX1) and the inflation target. The test is performed over each variable using time series models. Under the null hypothesis, the DGP has a linear representation (AR model) and under the alternative, a non linear specification (SETAR model). Unlike traditional ARIMA models, these models allow the endogenous variable to have different regimes across time. The main results are: it is not possible to reject linearity in the deviation of inflation from the inflation target. During the last twenty years, inflation has converged smoothly to the target without any regime switching. Finally, strong evidence is found against linearity in the unemployment rate. On the contrary, it fluctuates with high probability between states or regimes through time.