B.E. Journals in Economic Analysis & Policy
Using panel data for a cross-section of countries, several previous studies estimate the effect of advertising bans on cigarette consumption. These studies suffer from three general problems: (1) structural change in cigarette demand functions; (2) endogeneity of advertising bans; and (3) non-stationarity of cigarette consumption data. Using annual data for 20 OECD countries, the present study tests for unit roots. I show that growth rates of consumption (log differences) are stationary, but levels data are not. I estimate single-equation panel models for 1970-1995 and test formally for structural change. The tests and recursive coefficient estimates confirm a regime change beginning in 1985. Results for different time periods are reported for the effects of price, income, health warnings, country fixed-effects, and moderate and strong advertising bans. The study also considers the possibility of endogenous advertising bans. A public-choice model in the form of a two-equation model of advertising legislation and cigarette demand. The adoption of advertising bans is modeled as a Poisson count regression, and fitted values for the number of banned media are used as instruments in the demand equation. The results in the paper fail to demonstrate that advertising bans reduce aggregate cigarette consumption. Empirical results in previous studies are not robust to use of stationary data; refinements in model specification; different time periods; and endogeneity of advertising bans. Due to a decline in smoking prevalence, especially among males, there was a change in the political climate in favor of stronger restrictions on cigarette advertising. Overall, advertising bans have had no effect on cigarette consumption, regardless of the time period considered or the severity of the bans.