Accounting for differences in choice opportunities in analyses of energy expenditure

Zero expenditure poses several challenges when estimating demand systems. Zero expenditure on energy goods occur due to limited opportunity to consume the good or because the household chooses not to use all available equipment (corner solution). In this paper we develop a method to estimate an Almost Ideal Demand System (AIDS) of household energy demand simultaneously using a Maximum Likelihood approach. The multivariate density of energy expenditures depends on the consumption opportunity of the individual household. We model the choice of corner solutions by a stochastic Kuhn-Tucker condition, and distinguish between zero expenditure due to limited consumption opportunities and corner solutions by using a Double Hurdle model. We find that accounting for zero expenditure in the estimation has a significant effect on the estimated parameters. Assuming stochastic interdependence between expenditures on different energy goods within the household, in addition to accounting for zero expenditure, has only a minor effect on the estimated coefficients.