Initially, respondents classified themselves as current smokers, and met all of the inclusion criteria. Data were col- lected in the latter part of , approxi- mately 7 months after the antismoking cam- paign first began airing. The primary independent variables were antismoking advertisingβbased beliefs in the addictiveness of smoking 1 item , the harm- fulness of ETS 2 averaged items , and the de- ceptiveness of the tobacco companies in their advertising practices 2 averaged items.
Each item was measured via cued recalls of the specific advertisements and their respective belief themes. Choose any number from 0 to 10, where 0 means the ad made you feel the tobacco industry is not at all de- ceptive and 10 means the ad made you feel the tobacco industry is very deceptive.
As suggested by our hypotheses, a number of control variables also were taken into ac- count. The belief literature and recent evi- dence on adolescent smoking suggest that existing general beliefs about an action can have pronounced effects on that action.
Data on these existing general beliefs were col- lected near the beginning of the survey, be- fore the cued recall and evaluations of adver- tising-based beliefs. As such, the responses to the existing general beliefs measures were not influenced by the cued advertisement re- call procedures. Also, standard discriminant validity tests showed that existing general beliefs were distinct from the corresponding advertising-based beliefs. Within each of the themes, scores on items were summed and then aver- aged to form an overall theme composite.
Table 1 presents summary statistics and correlations between the study variables. Given the dichotomous dependent variable and the nature of our hypotheses, we used multiple logistic regression analyses with in- teraction terms to test our models. All of these analyses were conducted via the SPSS logistic regression algorithm. We mean cen- tered all independent variables covariates before estimating the 3 models 18 , 19 shown in Table 2.
