How Media Campaigns Influence Children's Physical Activity: Expanding the Normative Mechanisms of the Theory of Planned Behavior

Abstract

This study explicates mechanisms of media campaign effectiveness in the context of children's physical activity. The authors’ model expands the theory of planned behavior by integrating injunctive and descriptive norms into its normative mechanism. Analysis of a 3-wave nationally representative evaluation survey among 1,623 tweens indicates that campaign exposure is significantly related, but only indirectly, to both physical activity intention and physical activity behavior. Instead, campaign exposure seems more strongly related to perceived behavioral control and attitudes toward physical activity. By contrast, perceived behavioral control and descriptive norms are strongly related to behavioral intention. The findings suggest that integrating normative mechanisms with the theory of planned behavior can improve efforts to predict and explain a health behavior.

Acknowledgments

This work was supported by research funds provided by Hanyang University (HY-2012-N) to the first author.

Notes

^a Physical activity behavior (measured in 2005) was constructed through the following steps. First, children reported types of physical activities they did during the week before the interview. Second, for the first five activities, they identified whether the activity was done as part of an organized group or in their free or discretionary time. Third, they reported the number of days on which they had done the activity during the past 7 days; if appropriate, separate questions were asked about an activity done both as an organized activity and during free time. The number of organized physical activity sessions or free-time sessions was calculated as the sum of the number of days the children did all organized or free-time activities. Last, the number of organized and free-time sessions was summed to form a physical activity behavior index.

^b Perceived weight status was measured by asking how tweens would describe themselves on a 5-point scale ranging from 1 (very underweight) to 5 (very overweight). Existing physical activity behavior was constructed with the same process that was used for the physical activity behavior but with the 2003 data to control preexisting behavior.

^c Existing physical activity behavior was measured in 2003, using the same procedure as physical activity behavior. The remaining control variables—sex, age, race, urban dwelling, and perceived weight status—were drawn from the 2004 data.

¹Correlation coefficients among all the variables and gamma estimates are not included due to space limitations but available upon request.

*p < .05. **p < .01. ***p < .001.

²This seems to be a case of “negative suppression” (Conger, 1974; Darlington, 1968; also labeled as “net” suppression by Cohen & Cohen, 1975), because the positive association between a predictor (a suppressor) and a dependent variable became negative after controlling for other predictors. Negative suppression usually occurs when the suppressor is highly correlated with other predictors, all of which are also correlated with the dependent variable. In this study's case, because of the injunctive norm's moderately high correlations with subjective norm (r = .40) and descriptive norm (r = .41), its positive relation with physical activity intention at a bivariate level might become negative after controlling for the other norm variables.

³It should be noted that, based on the analytic procedure used by Anderson and Gerbing (1988) and Kline (1998), the high correlation among the three types of perceived norms does not threaten the discriminant and convergent validity.

⁴Peer proximity literature indicates differential effects of different (distal vs. proximal) peers on individuals’ own behavior (Paek, 2009; also see societal vs. personal norms by Park & Smith, 2007). However, this study was not able to distinguish the different levels of injunctive and descriptive peer norms. The reason is that the alternative confirmatory factor analysis model with the two levels of perceived peer norms separated did not achieve reasonable discriminant and convergent validity. That is, there was a higher correlation between the two levels of injunctive peer norm items (i.e., “My friends think that doing physical activities is fun” and “Kids my age think that doing physical activities is fun”) than there was within the items constructing a similar latent construct (e.g., “My friends think that doing physical activities is fun” and “My friends think that doing physical activities is important”; interitem correlation =.51 vs .35, respectively).