IO and spatial information as Bayesian priors in an employment forecasting model

Abstract

Interindustry input-output (IO) relationships were incorporated into a local labor market forecasting model for the Toledo, OH MSA by Magura (1990); he found that the use of the IO information as a Bayesian prior reduced forecast errors. LeSage and Magura (1991) found similar results using national labor market data. This paper likewise uses IO information as a Bayesian prior in forecasting employment in four industries in five states but also adds spatial information. The purpose of adding the spatial information is to determine if it further reduces forecast errors. Using a mixture of a spatial weight matrix similar to that proposed by LeSage (1993) in addition to the IO information, it is found that forecast errors are reduced beyond that achieved with only the IO information.