Predicting the National Unemployment Rate in Romania Using a Spatial Auto-regressive Model that Includes Random Effects

Abstract

The main objective of this research is to predict the national unemployment rate using unemployment rates at regional level that are registered for the Romanian counties. In this study the annual unemployment rates for the 42 counties (including Bucharest) are used over the period 2005-2010 to construct one spatial autoregressive process (SAR) on panel data and to make predictions for 2011-2014. The spatial rho is quite large and it shows that the neighbouring changes in unemployment rate in the previous period will affect the county unemployment in the current period. At 5% level of significance for 2013 the spatial auto-correlation for unemployment rate in Romanian counties is not significant. Most of the counties had the unemployment rate in 2013 in the interval [5.36%; 6.54%]. For 2014 the forecasts that use the SAR process including random effects anticipated that the national unemployment rate will be around 4.7%, but there are high chances to be an underestimated value as in the previous years (from 2011 to 2013) was.