53rd Congress of the European Regional Science Association: "Regional Integration: Europe, the Mediterranean and the World Economy", 27-31 August 2013, Palermo, Italy
Forecasting regional variables provides very important information for political, institutional and economic agents. However, in the present context characterized by important decline of economies, heterogeneous data and regional interdependencies, it is even more difficult to carry out accurate forecasts for any economic variable. In this paper, we assess the predictive performance of alternative models such as certain spatial panel models, the space-time autoregressive (ST-AR) model or the spatial Global VAR model. In all the cases, models also take into account the significant structural breaks in time. The capacity of forecasting is measured through traditional measures such as Theil?s U statistics, mean absolute error or root mean square error. Furthermore, comparison with some non-spatial models is also carried out. The empirical application refers to the explanation of employment in European Regions at NUTS II administrative level in terms of Eurostat. Results show that spatial models outperform non-spatial ones by a great extent.
Keywords: Forecasting regional data heterogeneity interdependencies structural breaks European employment