Spatial and sectoral linkages in regional models: A Bayesian vector autoregression forecast evaluation

A Bayesian vector autoregression (BVAR) approach is used to assess whether prior information on spatial and economic base-sectoral linkages improves forecast accuracy of employment for the metropolitan areas of the state of Oklahoma and their proximate metropolitan areas. Compared to autoregressive and vector autoregressive alternatives, a BVAR with a gravity-based prior is found to improve forecast accuracy of aggregate metropolitan employment. In bifurcated employment models, a prior of proportionality between basic and nonbasic sectors outperformed a prior of an inverse relationship between the two sectors implied by a fixed-factor-full-employment general equilibrium model. Copyright (c) 2008 the author(s). Journal compilation (c) 2008 RSAI.