Spatial aggregation and regional economic forecasting

Abstract

This research examines the effect of spatial aggregation on the accuracy of regional economic forecasts. Literature in econometrics, business forecasting, regional economic forecasting, spatial econometrics, and spatial time series analysis suggests that the issue is far from settled at either the theoretical or empirical levels. Univariate time series analysis is used to generate both state-level and sub-state regional forecasts of monthly total employment in the State of Idaho and its economic sub-regions. All spatial aggregation methods provide accurate forecasts, but direct state-level forecasts, and disaggregation of state-level forecasts to the sub-state level are slightly more accurate than more directly disaggregated approaches. As state-level data are often subject to less measurement error, and are also more complete and timely, this result should be encouraging to regional forecasting practitioners.