Why children start smoking cigarettes: predictors of onset

TLDR

The authors recommend that future prospective studies be theory‑driven, use reliable and valid measures, report scale properties, apply appropriate statistical methods, and investigate transitions to experimental or regular smoking through multi‑wave designs with careful measurement of behavior changes. They reviewed 27 prospective studies of smoking onset conducted since 1980, highlighting the need for multi‑wave designs and precise measurement of smoking behavior changes. Across nearly 300 predictor measures, 74 % supported theory‑based predictors, with strong evidence for socioeconomic status, peer and school bonding, peer smoking and approval, refusal self‑efficacy, and intentions, while family smoking, bonding, and approval received unexpectedly low support, underscoring both confirmed and uncertain predictors.

Abstract

Abstract We review findings from 27 prospective studies of the onset of cigarette smoking conducted since 1980. Almost 300 measures of predictors of smoking onset were examined, and 74% of them provided multivariate support for predictors of onset derived from theory and previous empirical findings. Expected relationships were strongly supported for (a) socioeconomic status, with students with compromised status being more likely to try smoking; (b) social bonding variables, particularly peer and school bonding, with less support for family bonding; (c) social learning variables, especially peer smoking and approval, prevalence estimates, and offers/availability, with less consistent support for parent smoking and approval; (d) refusal skills self efficacy; (e) knowledge, attitudes and intentions, with the expected stronger predictions from intentions than from attitudes than from knowledge; and (f) broad indicators of self‐esteem. The few investigators who analyzed their data separately by age, gender, or ethnicity found many differences by these factors, though there were too few of them to detect any pattern with confidence. Though the 27 studies are far from perfect, we believe that they confirm the importance of many well‐accepted predictors and raise some questions about others. In particular, family smoking, bonding and approval each received unexpectedly low support, ft is not clear whether this lack of support reflects reality as it has always been, is due to a changing reality, reflects developmental changes, either in the age of subjects or the stage of onset, or is due to poor measurement and too few tests. Future prospective studies need to be theory‐driven, use measures of known reliability and validity, report analyses of scale properties, and use statistical methods appropriate to the hypotheses or theories under study. Finally, we encourage more investigations of the potentially different predictors of transitions to experimental or regular cigarette smoking. This will require multi‐wave studies and careful measurement of changes in smoking behavior.