An Empirical Analysis of Adult Cigarette Demand

This paper examines the magnitude of prediction bias when employing a traditional two-part model of adult cigarette demand. In particular, this paper is the first econometric study to investigate the bias associated with naively assuming that error retransformation is homoscedastic in price. The traditional two part model that is estimated in this paper models the propensity to smoke and the intensity of smoking separately. A modified two-part model that allows the error retransformation to be heteroscedastic in price is estimated. The results from this research imply that the price of cigarettes is a very important determinant of adult cigarette demand. The estimates imply that policies to increase the price of cigarettes will reduce not only prevalence of smoking, but will also decrease the number of cigarettes smoked by smokers. However, the price coefficients obtained from the conditional demand equations are biased away from zero when using the traditional log-transformed dependent variable. The price elasticities are biased. A generalized linear model yielded conditional price elasticities less than one-half of those obtained using the traditional method.