Competing Frames for a Public Health Issue and Their Effects on Public Opinion

Abstract

This study examines the effects of goal framing on opinion about a public health issue. A content analysis of newspaper coverage regarding a New York City trans fat ban identified four frames, each of which invoked a policy goal (promoting public health or protecting business). An experiment tested the effects of goal framing on support for banning trans fat, as well as the effects of competition between frames invoking the same goal and competition between frames invoking different goals. The findings suggest that goal framing can shape opinion about public health issues but that competitive framing can undermine these effects.

Notes

¹Much of this research, in turn, has focused on framing in terms of political values, though Hoffman and Slater (2007) explored the use of human values (openness to change vs. conservation and self-enhancement vs. self-transcendence) in the framing of health issues.

²We used two-tailed tests here given that we did not have clear directional hypotheses.

³We also found cases where mutually reinforcing frames appeared together.

Note. Table entries are means; standard errors are in parentheses. Means differed significantly across conditions in a one-way analysis of variance, F(5, 190) = 2.94, p = .01.

⁴In an additional manipulation, participants who received the pro-ban public health frame were exposed to a version of this frame that either included a passage establishing knowingly created risk (see Lawrence, 2004) or did not include this passage (“The Board of Health vote comes a year after it conducted an unsuccessful campaign to persuade restaurants to eliminate trans fats…about half the city's restaurants continued to serve trans fats, about the same as before the campaign”). The inclusion or exclusion of this passage did not affect the results, implying that the presence or absence of knowingly created risk was not responsible for the effect that this frame produced. Accordingly, we combined participants who received the two versions into one condition.

⁵We used this measure as the dependent variable in our analysis because numerous studies of issue framing effects (e.g., Druckman, 2001a; Druckman & Nelson, 2003; Nelson, Clawson, et al., 1997; Nelson, 2004) have treated overall policy opinion as the ultimate outcome variable of interest. The same studies have typically treated other variables that may be influenced by framing (e.g., value or goal priorities) as mediators of framing effects on overall policy opinion. Our purpose in this study was not to replicate previous research regarding the psychological processes governing framing effects but to test the effects of competitive goal framing on policy opinion; thus, our analysis did not focus on potential mediators of such effects.

⁶We used one-tailed tests here given that we proposed directional hypotheses.

⁷Apart from our test of H1, our hypothesis tests revolved not around capturing the relative effects of exposure to one rival frame or another (as in the standard framing experiment design previously described) but around capturing the effects of (a) exposure to a single frame versus exposure to no frame, (b) exposure to a single frame versus exposure to competitive framing, or (c) exposure to one form of competitive framing versus exposure to another form of competitive framing. The standard two-condition framing manipulation essentially captures the maximum possible effect of framing, which makes it relatively easy to find a statistically significant difference across conditions. Most of our tests, in contrast, focused on subtler—and, thus, more difficult to find—differences across conditions. Given this, we report differences that were significant at the .10 level.

⁸We used a two-tailed test for RQ1, which did not posit expected differences across conditions.

⁹In light of Haider-Markel and Joslyn's (2001) finding that the impact of exposure to one rival frame or another can vary with receivers' political predispositions, we used a pair of regression analyses (support = condition + predisposition + (condition × predisposition)) to test whether the relative impact of exposure to the pro-ban public health frame or the anti-ban business frame (see H1) varied across ideology or partisanship. Neither the key multiplicative term nor the predisposition term (ideology or partisanship) attained significance in either model. Looking across all six conditions, the pattern of means among liberals paralleled the pattern among moderates and conservatives; likewise, the pattern of means among Democrats paralleled the pattern among independents and Republicans. Still, it is important to note that Haider-Markel and Joslyn's tests for variation in framing effects across predispositions used an experiment in survey question wording that allowed for a much larger total sample size per condition.

Additional information

Notes on contributors

David Wise

David Wise (M.A., University of Wisconsin–Milwaukee, 2008) is a Ph.D. student in Journalism and Mass Communication, University of Wisconsin–Milwaukee. His research interests include the effects of mediated communication on public opinion in the areas of politics, science, and public health.

Paul R. Brewer

Paul R. Brewer (Ph.D., University of North Carolina–Chapel Hill, 1999) is a Professor in Journalism and Mass Communication, University of Wisconsin–Milwaukee. His research interests include public opinion, political communication, and science communication.