A dynamic disequilibrium model for panel data: An application in housing market

In this dissertation, a new simulated maximum likelihood estimation method for dynamic disequilibrium panel data model is proposed. Disequilibrium is specified as a situation where the observed quantity equals the minimum of the quantities demanded and supplied, whereas in equilibrium prices always clear the market. The new method uses simulation to handle the multiple integrals in the likelihood function. I used a Markov structure in which the demand and supply equations depend on their own lagged latent-variables to reduce the computational complexity in the simulation. The new method is the first one designed for dynamic disequilibrium panel data models with unknown sample separation. I also apply the proposed method to the regional panel data on U.S. housing markets. In contrast to previous empirical studies, I conduct an analysis of housing demand and supply that controls for heterogeneity and the serial correlation of the regional housing markets. Using two different set of data, I found that price and income both have significant impacts on the quantity of houses demanded, while price and construction costs have significant negative impacts on quantity of houses demanded. Estimates of disequilibrium models with the Case-Schiller index used as price suggest that most regions experienced excess demand during the sample periods. Finally, a Hausman-type test is employed to ensure that the estimated parameter values correspond to global maximums of the likelihood functions.