Forecasting local business activity from aggregate indicators

A regional forecasting technique is developed which combines national and local indicators to provide forecasts of local business activity. A regional business index is first developed to provide a time series from which the historical pattern of local business activity may be evaluated. The relationship between this time series and a similarly constructed national indicator series are first examined with cross-spectral analysis. The information from this analysis is used to guide the formulation of a transfer function time series model. Out-of-sample forecasts are then generated with the transfer model and a univariate ARMIA model. A comparison of these two forecasts to the actual regional index reveals that the transfer model provides a superior forecast.