An autoregressive short-run forecasting model for unemployment rates in Romania and the European Union

The paper discusses an autoregressive model that captures through a residual analysis the dependence structure of unemployment rates. The model is designed for the analysis and time-forward prediction of spatio-temporal econometric data. Linearity tests are performed for a number of quarterly and monthly, seasonally adjusted, unemployment series from EU-27 countries, focusing on Romania. For a number of series, we found by testing that unemployment rate can be modeled satisfactorily by use of a first-order linear autoregressive model AR(1), but also by a second-order autoregressive model AR(2). The properties of the estimated models, including persistence of the shocks related to them, are illustrated in various ways and discussed within the paper.