Quitting smoking: The importance of non-smoker identity in predicting smoking behaviour and responses to a smoking ban

Abstract

Objective: We examined how ‘smoker’ and ‘non-smoker’ self- and group-identities and socio-economic status (SES) may predict smoking behaviour and responses to antismoking measures (i.e. the Dutch smoking ban in hospitality venues). We validated a measure of responses to the smoking ban.

Design: Longitudinal online survey study with one-year follow-up (N = 623 at T1 in 2011; N = 188 at T2 in 2012) among daily smokers.

Main outcome measures: Intention to quit, quit attempts and ‘rejecting’, ‘victimizing’, ‘socially conscious smoking’ and ‘active quitting’ responses to the smoking ban.

Results: Non-smoker identities are more important than smoker identities in predicting intention to quit, quit attempts and responses to the smoking ban, even when controlling for other important predictors such as nicotine dependence. Smokers with stronger non-smoker identities had stronger intentions to quit, were more likely to attempt to quit between measurements, and showed less negative and more positive responses to the smoking ban. The association between non-smoker self-identity and intention to quit was stronger among smokers with lower than higher SES.

Conclusion: Antismoking measures might be more effective if they would focus also on the identity of smokers, and help smokers to increase identification with non-smoking and non-smokers.

Keywords:

• identity
• socio-economic status/educational level
• smoking cessation
• responses
• antismoking measures
• smoking ban

Acknowledgement

We would like to thank Sander van der Heiden for his help with data collection at T1.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Although for SES χ² was significant, no standardised residuals larger than 1.96 were found for specific cells, indicating absence of significant deviations from the expected counts.

2. In addition, in the context of these other variables a suppression effect was found, leading the smoker group-identity × SES (higher vs. lower) interaction to take on an unusual form. Specifically, smokers with lower SES had a stronger intention to quit when smoker group-identity was stronger, whereas smoker group-identity was unrelated to intention to quit among higher SES smokers, F(1,169) = 3.24, p = .07, ΔR² = .01. This contrary effect became nonsignificant when the analysis was repeated with only the smoker group-identity × SES interaction as predictor of intention to quit in Step 3 (controlled for gender, SES, age at smoking onset, years smoked, nicotine dependence, and identity variables), F(1,176) = .43, p = .51, ΔR² < .01. Further, regression coefficients for simple slopes became nonsignificant (ps > .10).

3. In addition, in the context of these other variables a suppression effect was found, leading the non-smoker group-identity × SES (average vs. lower) interaction to take on an unusual form. Specifically, smokers with average SES had a weaker intention to quit when non-smoker group-identity was stronger, whereas non-smoker group-identity was unrelated to intention to quit among lower SES smokers, F(1,169) = 4.34, p = .04, ΔR² = .02. This contrary effect became marginally significant when the analysis was repeated with only the non-smoker group-identity × SES interaction as predictor of intention to quit in Step 3 (controlled for gender, SES, age at smoking onset, years smoked, nicotine dependence, and identity variables), F(1,176) = 2.78, p = .097, ΔR² = .01. Also, the zero-order correlation between non-smoker group-identity and intention to quit is positive among lower SES smokers (r = .34, p < .01) and nonsignificant among average SES smokers (p > .99).

4. In addition, in the context of these other variables a suppression effect was found, suggesting that older age at smoking onset marginally predicts more rejecting responses. However, the zero-order correlation between age at smoking onset and rejecting is small and nonsignificant (r = .08, p = .29).

5. In addition, in the context of these other variables a suppression effect was found, suggesting that smokers with a stronger smoker self-identity showed more active quitting responses to the smoking ban. This contrary effect changed into the expected direction and became nonsignificant when the analysis was repeated with only smoker self-identity as predictor of active quitting (controlled for control variables and SES): smoker self-identity b = −.08, p = .69. Further, the zero-order correlation between smoker self-identity and active quitting is in the expected direction (r = −.13).

6. In addition, in the context of these other variables two suppression effects were found, leading the smoker group-identity × SES (average vs. lower) and smoker group-identity × SES (higher vs. lower) interactions to take on unusual forms. Specifically, whereas the smoker group-identity × SES (average vs. lower) interaction effect was significant (F(1,169) = 4.02, p = .047, ΔR² = .02), simple slopes among lower and average SES smokers were nonsignificant (ps > .10). The interaction effect became nonsignificant when the analysis was repeated with only the smoker group-identity × SES interaction as predictor of intention to quit in Step 3 (controlled for gender, SES, age at smoking onset, years smoked, nicotine dependence, and identity variables), F(1,176) = .65, p = .42, ΔR² < .01. Also, whereas the smoker group-identity × SES (higher vs. lower) interaction effect was significant (F(1,169) = 4.37, p = .04, ΔR² = .02), the simple slope among higher SES smokers was only marginally significant (b = −.38, p = .099) and the simple slope among lower SES smokers was nonsignificant (p > .10). The interaction effect became nonsignificant when the analysis was repeated with only the smoker group-identity × SES interaction as predictor of intention to quit in Step 3 (controlled for gender, SES, age at smoking onset, years smoked, nicotine dependence, and identity variables), F(1,176) = .97, p = .33, ΔR² < .01.

Additional information

Funding

This work was supported by MIRO/CAPHRI/Pfizer [grant number MIROKPJ-09-090814]. The views expressed are those of the authors and not necessarily of the funders.