Predicting parents’ intentions not to smoke indoors in the presence of their children using an extended version of the theory of planned behaviour

Abstract

The present study examined whether the theory of planned behaviour (TPB) could predict parents’ intentions not to smoke indoors in the presence of their children. Moral norms, smoker identity, parent identity, and positive and negative anticipated affect were included as additional predictors. A questionnaire was sent to a representative sample of 1000 households in Norway with children born in 1998, and was completed by 612 respondents. The TPB variables and the extension variables were measured among the smokers in the sample (101 women and 61 men). Intentions were predicted by subjective norms and perceived behavioural control. The data were supportive of the inclusion of smoker identity and moral norms in the TPB. We also found a significant interaction effect between parent identity and smoker identity, and between negative anticipated affect and moral norms. Furthermore, we identified significant interactions between gender and three of the predictors of behavioural intentions. The TPB components explained 56% of the variance in intentions, while the additional variables explained a further 19%.

Keywords:

• Theory of planned behaviour
• moral norms
• smoker identity
• parent identity
• positive and negative anticipated affect
• interaction effects
• gender differences

Acknowledgements

The financial support of the Research Council of Norway is gratefully acknowledged. We would also like to thank Elin K. Bye, Karl Erik Lund and Ellen Amundsen at the Norwegian Institute for Alcohol and Drug Research, and two anonymous reviewers for helpful comments on the earlier drafts of this article.

Notes

We succeeded in identifying one theoretically based study (or rather an outline of a research project) concerning this issue (see Borrelli et al., 2002). The results from this study have however not been published yet.

The present study was conducted among parents’ who had children born in 1998. Thus they had at least one child. We cannot, however, know whether they have more than one child, but we must assume that many of them do. Therefore we write “intentions not to smoke indoors in the presence of their children”.

Theoretically self-identity, moral norms and anticipated emotions should be treated as distinct concepts. It can, however, be useful to examine whether they also operate as differentiated concepts empirically. We therefore performed a principal component analysis of all the items employed to measure the independent variables in the present study. The results from the analysis are presented in the Method section (Table I).

The Norwegian Cancer Society funded the data collection in the present study, and the third author (M. Andersen) was responsible for collecting the data.

Three women were later excluded from the data set (see footnote 7).

Since this is an unusual finding, we did a more detailed analysis of the data to explore whether (i) there was any coding or entry problem, and (ii) whether the data supported the use of parametric statistics (see footnote 7). We failed to identify the source of the low correlation between attitude and the other variables for women.

Royston's (1982) extension of the Shapiro and Wilk's W statistic was used to test whether the residuals were normally distributed. A Shapiro–Wilk score which is not significantly different from 1 indicate normality. The analysis showed that the residuals from the regression analysis for men were normally distributed (Shapiro–Wilk score: 0.999), while the women's residuals were skewed. Three respondents located outside 3 standard deviations were identified. Except the fact that the three respondents scored low on intention (2–3) and high on all other variables (6–7), nothing remarkable was found. The results from the regression analysis performed without the three outliers revealed normally distributed residuals (Shapiro–Wilk score: 0.987). We also tested whether the residuals were heteroscedastic (i.e. whether the variance in the residuals were associated with the predicted value) by making a scatterplot of the standardized predicted value of intention and the standardized residuals for women and men. The plots revealed that residuals were homoscedastic. Thus, the results supported use of parametric statistics (Hankins et al., 2000).

We would like to thank a reviewer of an earlier version of this article for suggesting to test these interaction effects.