Forecasting the sectoral GVA of a small Spanish region

Our main goal in this paper is to evaluate the point forecasting accuracy of several time series econometric models when applied to a small Spanish region. The variable of interest is the sectoral GVA of the Basque Country. The results support the use of causal models, which outperform univariate models, such as ARMA and SETAR, in forecasting accuracy. The use of a causal model, such as a simple Dynamic Regression model, offers a systematic advantage in the case of a small regional economy for which abundant regional statistical information is available.