College Students' Estimation and Accuracy of Other Students' Drinking and Believability of Advertisements Featured in a Social Norms Campaign

Abstract

Social norms campaigns, which are based on correcting misperceptions of alcohol consumption, have frequently been applied to reduce college students' alcohol consumption. This study examined estimation and accuracy of normative perceptions for students during everyday drinking occasions. Students who reported having 4 or fewer drinks underestimated the percentage of other students who had 4 or fewer drinks, while those who drank 5 or more drinks overestimated the percentage of other students who had 5 or more drinks. Believability of advertisements featured in social norms campaigns also played a crucial role in this process. Those who believed the ad more closely estimated alcohol consumption by their peers while ad believability moderated the relation between drinking behaviors and accuracy.

Acknowledgments

This research was supported by a grant from the Social Norms Research Center to Dennis Martell and Sandi W. Smith, principal investigators.

Notes

¹Because this study used two types of norm advertisements, there is a possibility that differential believability of descriptive norm and injunctive norm advertisements could have effects on overestimation/underestimation and accuracy. When ad norm type was included in the main analyses, the results showed that ad norm type did not have any substantial effects.

When a 2 (drinking amount at the last party: zero to four drinks versus five or more drinks) × 2 (ad believability: yes versus no) × 2 (ad type: descriptive norms versus injunctive norms) between-subjects analysis of variance was run on estimation, the main effect for ad type was not significant, F(1, 1058) = 0.02, p = .89, η² = 00. The interaction between drinking amount and ad type was not significant, F(1, 1058) = 0.23, p = .63, η² = .00. The interaction between ad believability and ad type was not significant, F(1, 1058) = 0.20, p = .66, η² = .00. The three-way interaction among drinking amount, believability, and ad type was not significant, F(1, 1058) = 0.47, p = .49, η² = .00.

When a 2 (drinking amount at the last party: zero to four drinks versus five or more drinks) × 2 (ad believability: yes versus no) × 2 (ad type: descriptive norms versus injunctive norms) between-subject analysis of variance was run on accuracy, the main effect for ad type was not significant, F(1, 1058) = 0.13, p = .72, η² = 00. The interaction between drinking amount and ad type was not significant, F(1, 1058) = 2.00, p = .16, η² = .003. The interaction between ad believability and ad type was not significant, F(1, 1058) = 1.01, p = .32, η² = .00. The three-way interaction among drinking amount, believability, and ad type was not significant, F(1, 1058) = 0.09, p = .77, η² = .00.

²Participants were asked to indicate how many times they had seen the advertisements. For descriptive norms ads, the majority of the participants reported having seen the advertisements once (23.00%), twice (25.15%), or three times (23.55%). The rest indicated four times (12.55%), five times (6.48%), and six or more times (9.27%). For injunctive norms ads, the majority of the participants reported to have seen the advertisements once (35.20%) or twice (35.61%). The rest indicated four times (11.62%), five times (3.05%), and six or more times (14.53%). When the frequency of seeing the advertisements was included in any of the main analyses, they did not affect any of the findings. For example, the correlation between the frequencies and estimation was not significant, r(285) = .05, p = .39 for descriptive norm ads, and r(207) = −.02, p = .83 for injunctive norm ads. Correlations between the frequencies and accuracy was not significant, r(285) = −.02, p = .79 for the descriptive norm ad, and r(207) = .11, p = .11 for the injunctive norm ad.

³An anonymous reviewer pointed out that the large standard deviation for the underestimation scores of individuals who drink zero to four drinks indicated a good portion of the scores could be overestimation. We dichotomized estimation scores, depending on whether the scores were above or below zero. The distribution showed that 46.3% of individuals who drink zero to four drinks showed estimation scores greater than zero and 53.7% of them showed estimation scores smaller than zero. Alternatively, 74.4% of individuals who drink five or more showed estimation scores greater than zero and 25.6% of them showed estimation scores smaller than zero. When chi-square was calculated on these percentages, it was significant, χ² = 85.18, p < .001, ϕ = .28.

⁴Research has shown that male students are more likely to be heavy drinkers than are female students (Wechsler et al., 2002) and that male students, compared to female students, were more likely to overestimate other students' alcohol consumption amount (Miley & Frank, 2006). Thus, gender was included as a factor in the analyses. The results showed that gender did not have a significant main effect on overestimation/underestimation, F(1, 1050) = 3.43, p = .06, η² = .003. However, gender had a significant interaction with drinking amount, F(1, 1050) = 11.66, p = .001, η² = .009. Among students who reported having zero to four drinks, men (M = −3.99, SD = 22.13) and women (M = −2.25, SD = 20.57) did not differ in their estimation of the percentage of others who had zero to four drinks, t(600) = 0.85, p = .40. Among students who reported having five or more drinks, however, men (M = 16.93, SD = 21.48) overestimated more strongly the percentage of others who had five or more drinks than did women (M = 10.44, SD = 18.79), t(462) = 3.31, p = .001.

For accuracy, the results showed that gender had a significant main effect, F(1, 1050) = 29.03, p < .001, η² = .03. Women (M = 17.15, SD = 12.03) were more accurate than men (M = 21.09, SD = 13.78) in estimating the percentage of others who had the same amount of drinks as theirs. Gender had a significant interaction with drinking amount, F(1, 1050) = 6.11, p = .014, η² = .006. For women, those who reported having zero to four drinks (M = 16.99, SD = 11.78) and those who reported having five or more drinks (M = 17.46, SD = 12.51) did not differ, t(570) = 0.41, p = .68. For men, however, those who reported having zero to four drinks (M = 18.70, SD = 12.40) were more accurate than those who reported having five or more drinks (M = 23.13, SD = 14.57), t(493) = 3.55, p < .001. Gender also had a significant interaction with ad believability, F(1, 1050) = 15.56, p < .001, η² = .014. For women, those who believed the ads (M = 17.09, SD = 11.65) and those who did not believe the ads (M = 17.28, SD = 12.81) did not differ in their accuracy, t(570) = 0.16, p = .88. On the other hand, for men, those who believed the ads (M = 18.59, SD = 12.55) were more accurate than those who did not believe the ads (M = 26.47, SD = 14.78), t(493) = 6.20, p < .001.