Simulation of a gravity model

Conclusions Three conclusions have come from this research. First, the elements of the gravity model, its three variables (distance, population and per capita income) and one parameter (), cannot be separated and investigated independently. Secondly, to say more about how the model predicts and the number of grid divisions for a desired accuracy depends on empirical investigation of actual income density gradients.The final conclusion, while tentative and subject to empirical verification is the feeling that perhaps the usefulness of a very fine grid system has been over emphasized. This conclusion in no way obviates concern by location practitioners of grid size as a source of error, but highlights the possibility that the costs associated with smaller grid divisions may be out of proportion to any increase in predictive accuracy which may or may not accrue.Use of the facilities of the Computing Centre at the University of British Columbia is gratefully acknowledged.