Changes in regional input-output coefficients and regional forecasting

The issue of constant regional input-output coefficients is studied in the context of regional forecasting. Under the constant coefficient assumption and given known final demand vectors, output and intermediate output prediction errors average five and 20 percent, respectively. Neither price nor product-mix adjustments improve predictions, and measurement error in base-year tables accounts for only a portion of the observed variation in coefficients. On the other hand, the interindustry structure is sensitive to short-run disturbances in the region's propensity to import. Tests further indicate that input-output forecasting is superior to a series of naive methods, but that the problem of predicting regional final demands is a relatively serious one.