This paper examines the nature of coefficient change in integrated econometric (EC) and input-output (IO) models that employ an embedding strategy. Existing approaches towards coefficient change in integrated models are first reviewed and several substantive processes that give rise to dynamic coefficients at the regional level are identified. The main sources of coefficient change arise from dynamic adjustment processes that affect the regional purchase coefficients, regional technological coefficients and regional integration parameters. Each of these sources of change requires a different treatment when adopting an EC perspective. A general framework is suggested that nests each treatment as a specific case. The different specifications are then implemented in an embedded EC and IO employment demand model for the San Diego region, and evaluated through a series of in-sample and out-of-sample forecasting experiments. Simulation performance is found to improve through the use of a time-varying integration parameter. The results also indicate that model performance is more sensitive to the use of dynamic regional purchase coefficients than to the use of dynamic technical coefficients.
ASJC Scopus subject areas
- Economics and Econometrics