A vector autoregressive model of the British Columbia regional economy

We construct a small vector autoregressive (VAR) model to investigate the usefulness of this approach to forecasting the British Columbia (BC) macro economy. The forecasts are compared with univariate ARIMA forecasts. Overall, our results suggest that the accuracy of the VAR matches or exceeds that of the univariate ARIMA. The accuracy of the predictions from our VAR model indicates that VAR as a forecasting tool, is a promising approach for regional forecasting and warrants further investigation. Formal tests for Granger causality are conducted and we find evidence of bi-directional causality between BC employment and prices; and uni-directional causality from BC GDP to both BC employment and Canadian GDP. Our results also suggest uni-directional causality from Canadian GDP to both BC employment and prices; and from prices to GDP within BC.