Quantile regression analysis of the rational addiction model: investigating heterogeneity in forward-looking behavior.

The time path of consumption from a rational addiction (RA) model contains information about an individual's tendency to be forward looking. In this paper, we use quantile regression (QR) techniques to investigate whether the tendency to be forward looking varies systematically with the level of consumption of cigarettes. Using panel data, we find that the forward-looking effect is strongest relative to the addiction effect in the lower quantiles of cigarette consumption, and that the forward-looking effect declines and the addiction effect increases as we move toward the upper quantiles. The results indicate that QR can be used to illuminate the heterogeneity in individuals' tendency to be forward looking even after controlling for factors such as education. QR also gives useful information about the differential impact of policy variables, most notably workplace smoking restrictions, on light and heavy smokers.