Modeling Corner Solutions with Panel Data: Application to Industrial Energy Demand in France

This paper is providing an initial empirical application of Lee and Pitt's approach to the problem of corner solutions with panel data. This approach deals with corner solutions in a manner consistent with behavioral theory. Furthermore it allows the use of flexible form cost functions and general error structure. In this model energy demand, at industrial plant level, is the result of a discrete choice of type of energy to consume and a continuous choice to define the demand level. The econometric model is essentially an endogenous switching regime model which require the evaluation of multivariate probability integrals. We estimate the random effect model by maximum likelihood using a panel of industrial French plants. We verify that estimations predict globally well the model and we simulate the effects of prices variations and a CO2 tax on energy demand.