A VAR Forecasting Model of a Regional Economy: Its Construction and Comparative Accuracy

Abstract
The usefulness of the vector autoregression (VAR) approach to forecasting regional economies is explored. A VAR model and a Bayesian VAR (BVAR) model of selected New York State economic variables are constructed using monthly data. Their predictions about these variables are compared with ARIMA and transfer function model forecasts. Overall, the accuracy of BVAR matches or exceeds that of the other techniques. Thus, a previous suggestion that BVAR is promising, as a forecasting tool and as a benchmark for regional forecasts, is supported.