Simulation-Based Estimation of Spatial Price Equilibrium Models and Market Integration

Abstract

This dissertation explores the applicability of recently developed simulation-based econometric methods to the analysis of spatial price determination and integration of markets. As such a measure of market integration is developed within the context of well-known point-location competitive price equilibrium model. Two markets are said to be integrated to the extent that an excess demand shock arising in one region is transmitted to the other region. This model imposes the market efficiency (i.e., the law of one price) and explicitly accounts for the nonlinear relationship between prices and underlying variables. A critical distinction from most of the current literature is that market integration is defined and estimated as a degree rather than a binary outcome. Simulation-based estimation approach to spatial market integration can be outlined as follows: summarize the information contained in prices in an auxiliary (reduced) model and match the parameters of the auxiliary model obtained from observed data to the parameters of the same model obtained from simulated data. The particular estimation framework used is known as the extit[indirect inference] methodology. The auxiliary model is chosen as a finite order Gaussian vector auto-regression for prices. The underlying variables, autarky prices and transaction costs, are modelled as low-order vector auto-regression. Transaction costs on each link is modelled as a function of a small number of common factors. After the structural parameter estimates are obtained the measure of market integration is approximated by simple Monte Carlo integration methods. The results from a set of controlled experiments indicate that the estimation strategy works reasonably well. It is also shown that simulation-based estimation methods can be useful for the two-location switching regime models with serially correlated underlying variables.