Parameter Estimation for the Constrained Gravity Model: A Comparison of Six Methods

Six methods of parameter estimation for the production-constrained gravity model are compared in the context of interurban consumer travel. Monte-Carlo experiments reveal that the nonlinear methods (Batty and Mackie, 1972) are inferior to the linear methods (Nakanishi and Cooper, 1974) when there is specification error present, but that the latter rapidly lose their advantage as sample sizes decrease. Bias in the parameter estimates is a more serious source of error than sampling variation.