demonstrated that the study results depended on which of three alternative empirical measures of advertising was used (Saffer and Chaloupka 2000). Most of the studies finding that advertising was not an important predictor of cigarette demand used annual or quarterly national aggregate expenditure data. The investigators argue that these studies lacked statistical power and were thus likely to find insignificant results because national expenditures lose variance because of aggregation effects and measure advertising where the marginal effect of advertising is near zero. In contrast, studies using cross-sectional data (typically measured at the local level for periods of less than a year) have greater variation in the advertising data and greater statistical power and thus are more likely to identify a positive relationship between advertising and consumption. Finally, studies that measure advertising on the basis of advertising bans produced various results that depended on the scope of advertising restrictions, leading the investigators to conclude that comprehensive advertising bans can reduce tobacco consumption but that a limited set of advertising bans will have little or no effect.

Saffer and Chaloupka caution that attempts to restrict advertising must be sufficiently comprehensive to eliminate the possibility that tobacco companies will simply substitute the remaining legal forms of advertising and promotion (Saffer and Chaloupka 2000). Advertising bans achieve the greatest success when they eliminate a wide range of media outlets, which diminishes opportunities for substitution, and which defeats industry efforts to replace advertising in the banned media with advertising in alternative channels. For example, the ban on outdoor advertising required by the MSA may have little effect on consumption because other forms of promotion, including print advertising, point-of-sale advertising, sponsorships, and other forms of retail promotion, will not be prohibited.

From the standpoint of the initiation of smoking by youth, the most important feature of tobacco advertising is its noninformational characteristics. The most compelling data are those that link positive feelings toward smoking with exposure to tobacco advertising and to ownership of commodities with tobacco company logos and paraphernalia.

The very purpose of noninformational tobacco advertising is to associate smoking with positive attributes and consequences and to create a positive affect toward smoking and people who smoke. In addition, advertising in magazines and retail displays creates the impression that smoking is a widespread and normal social practice and that tobacco is a normal consumer product. The images used in tobacco marketing associate smoking with lifestyles and experiences that appeal to young people, and these positive associations tend to displace or override risk information in adolescent decision making. The evidence clearly shows that youth exposure to images that create a positive association with smoking is associated with a