TIME AND SPACE MODELLING: ASYMPTOTICALLY EFFICIENT INSTRUMENTAL VARIABLES ESTIMATORS OF EXPORT BASE MULTIPLIERS

The study shifts the focus of the export-base model of regional economic growth from the spatial independence of the individual region to the spatial interdependence of all fifty states and Washington, D.C. The model is also expanded to include aspects of the non-export sector, as well as other time varying and time invariant shift variables which have been identified from the regional economics literature. These changes to the model imply a different interpretation of the error term in the regression equation, the errors between regions are assumed to be contemporaneously correlated, and the errors within regions are assumed to be correlated with some of the explanatory variables.;Using the U.S. Department of Commerce's state personal income by place of work data, income multipliers are estimated using an asymptotically efficient instrumental variables technique. The technique uses time varying variables in two ways--to estimate their own coefficients and to serve as the instruments for the endogenous time-invariant models. The modified theory fits the facts very well.;Alternative estimating techniques are considered. The seemingly unrelated regressions technique is selected because it is the only method which can handle the complex structure of the error term. The extremely large data base required to estimate the coefficients for such a model exceeds the University's software capacity, and a compromise is made. The assumption of the contemporaneous correlation between regions is dropped. A consequence of not using the seemingly unrelated regressions technique is that the estimates are national weighted averages rather than the region specific multipliers that were anticipated.