Specification Issues in Models of Population and Employment Growth

Spatial econometric adaptations of population and employment growth models have been used to study the employment impacts of urban rail transit (Bollinger and Ihlanfeldt, 1997), the links between urban and rural development (Henry, Barkley, and Bao, 1997; Schmitt and Henry, 2000), and causality between intra-metropolitan population and employment location (Boarnet, 1994b). Yet the literature has so far given limited attention to two specification issues that are fundamental to the performance of spatial econometric population and employment growth models. First, the weight matrix, which defines how geographic units of observation relate to one another, must be defined a priori, and alternative versions of the weight matrix have rarely been consistently compared. Second, most recent population-employment growth models are lagged adjustment models, yet the estimated lag parameters often imply that the system does not adjust to a long-run equilibrium, violating one of the maintained hypotheses of the lagged adjustment approach. This paper analyzes those three specification issues, and provides insight into both the validity of various econometric practices that have been common in recent literature and the stability of econometric population and employment growth models when typical assumptions and approaches are changed.