Anti-Sugar-Sweetened Beverage Messages Elicit Reactance: Effects on Attitudes and Policy Preferences

Abstract

Messages that convey the dangers associated with consumption of sugar-sweetened beverages (SSBs) may be the most effective means of changing attitudes toward consumption and policy preferences. However, there is a risk that this message type also stimulates reactance, a form of resistance to persuasion. A study (N = 618) using messages from the 2012 New York City anti-SSB campaign and a sample of New York City residents showed just such effects. Reactance was heightened by prior message exposure, conservative political orientation and prior consumption of SSBs. The net message effect was still persuasive overall for attitudes, but could be improved by 17% if reactance were eliminated. In contrast, the net message effect on policy preferences was counterpersuasive, due to processes other than reactance. Anti-SSB threat appeals can change attitudes toward one’s own behavior in a more healthful direction, while simultaneously eroding support for more restrictive SSB policies.

Notes

¹ Frequency and quantity were correlated at .75.

² The interaction of political orientation × message condition on threat was significant at p = .08. Partitioning the data on message condition revealed a slightly stronger association between conservatism and threat following exposure to anti-SSB messages versus control messages. The bivariate correlations between the manifest variables were .26, p < .0001 versus .11, p = .07. But, because the difference was small, the interaction was ordinal, and model fit was improved by its elimination, we ultimately settled on a model that did not include this interaction term.

³ We also tested a model in which the location of threat and reactance was reversed (all else equal). The data strongly favored the theoretical model over the alternative: AIC and BIC difference = 12.13.

⁴ Standardized estimates of direct effects of exposure, consumption, and reactance on policy preferences were -.10, -.28, and -.20, respectively.

⁵ Rains (2013) coded his data such that reactance produced a positive coefficient. We added a negative sign to align his results with our study– a purely cosmetic change.