Judging regional forecasting accuracy — Sectoral employment in the San Francisco Bay area

Summary

This paper has briefly discussed the problem of judging regional economic forecasting accuracy from the point of view of assessing the quality of industry employment forecasts for the SF Bay Area on an expost basis. Because of the formidable problems associated with regional economic forecasting a sophisticated econometric approach toex ante analysis was rejected in favor of accuracy analysis of employment projection along lines suggested by Henri Theil. This accuracy analysis involved a slightly modified version of Theil's inequality coefficient. The coefficient permits easy comparison of a given model's output with an assumed naive model, the comparison being predicated upon acceptance of a quadratic loss criteria. The results suggest that the SF Model functions substantially better than two versions of “naive models: the no-change and the average historical change models. Forecasting with the model, however, is independent upon ability to obtain sufficiently detailed forecasts of national economic developments. While the development of econometric art is promising, no national models predict all the exogenous variables required. One model, the Wharton School's, is presently producing results which are being modified for inclusion in the exogenous variables of this model. In the absence of completion of this task, estimates prepared for the State of California's quarterly forecasting model have been used to generate forecasts to 1970, Tests of these forecasts over the period 1964–1966 suggest that errors in the exogenous variables seriously affect the outcome of the regional model. Nonetheless, the results are still superior to the naive model against which the regional model has been tested.