Abstract
The gravity model of interregional trade pattern or of interregional transaction seems to have established itself as a choice behavior model in the last decade. The background for this understanding is briefly summerized in section 1.
Notwithstanding the remarkable development of theoretical studies of the gravity models, existing numerous examples of estimated gravity models are still far from proving worthy of our trust. This seems to result mainly from thinking of the statistical behaviors or characteristics of the stochastic gravity models which can be variously specified.
The purpose of this paper is to throw light on the importance of comprehending the statistical behavior of a stochastic gravity model, using the logit-type gravity model of interregional trade coefficient.
Two types of stochastic models, the polynomial distribution model (P-model) and the normal distribution model (N-model), are discussed in section 2. The never experienced superiority of the logit-type gravity model of inter regional trade coefficient is empirically verified from a numerical example in section 3.
Calling to mind that other ways of difining interregional trade pattern than the coefficient discussed in this paper are possible, we should say it leaves room for further studies in order to derive a more generalized conclusion. However, it may certainly be said the crucial point in such a study for generalization is to clarify the statistical behavior of any other relevant model.
