Neural Network Modeling as a Tool for Forecasting Regional Employment Patterns

This article analyzes artificial neural networks (ANNs) as a method to compute employment forecasts at a regional level. The empirical application is based on employment data collected for 327 West German regions over a period of fourteen years. First, the authors compare ANNs to models commonly used in panel data analysis. Second, they verify, in the case of panel data, whether the common practice of combining forecasts of the computed models is able to produce more reliable forecasts. The technique currently employed by the German authorities to compute such regional employment forecasts is comparable to a simple naïve no-change model. For this reason, ANNs are also compared to this undemanding technique. © 2005 Sage Publications.