Prediction Using Panel Data Regression with Spatial Random Effects

This paper considers some of the issues and difficulties relating to the use of spatial paneldata regression in prediction, illustrated by the effects of mass immigration on wages andincome levels in local authority areas of Great Britain. Motivated by contemporary urbaneconomics theory, and using recent advances in spatial econometrics, the panel regression haswages dependent on employment density and the efficiency of the labour force. There aretwo types of spatial interaction, a spatial lag of wages, and an autoregressive process for errorcomponents. The estimates suggest that increased employment densities will increase wagelevels, but wages may fall if migrants are under-qualified. This uncertainty highlights the factthat ex ante forecasting should be used with great caution as a basis for policy decisions.