The Effect of Inequality Frames on Support for Redistributive Tax Policies

Method

Seventy-nine U.S. citizens (18–60 years old, mean age = 28.25 years, SD = 11.94) from an online panel participated in the study, which they were told was about people’s views on income inequality within the United States. Participants in the rich-have-more condition were given information on income inequality that included the following passage: “Data from the U.S. Census Bureau (2008) indicates that the top 5 percentile (i.e., the 95th percentile) of wage earners made, on average, $111,000 more than the median wage earner (i.e., the 50th percentile).” Participants in the poor-have-less condition were given information on income inequality that included the following passage: “Data from the U.S. Census Bureau (2008) indicates that the median wage earner (i.e., the 50th percentile) made, on average, $111,000 less than the top 5 percentile (i.e., the 95th percentile).” Participants in the control condition did not receive any information about income inequality.

Next, participants indicated their beliefs about why the rich are wealthy and why the poor are in poverty by completing a variety of internal- and external-attribution measures. Each measure yielded a score for each category of individuals (i.e., the rich and the poor); higher scores indicated a greater extent of external attribution. Participants then indicated their level of support for two redistributive tax policies: one that would create a new tax bracket for people with incomes over $1 million and one that would create a new tax bracket for people with incomes over $5 million. Finally, participants indicated their level of political conservatism and reported their household income (for details on participant recruitment and items used in our measures of attribution, support for redistributive tax policies, conservatism, and income, see Supplemental Methods in the Supplemental Material available online).

Results and Discussion

Means, standard deviations, and correlations among the study variables are shown in Table S1 in the Supplemental Material. Income levels among participants in our sample were similar to those of the U.S. population (see Table S2 in the Supplemental Material). All results controlled for income, but the pattern of results did not change when this control was excluded. An analysis of variance on conservatism across the three conditions indicated that the inequality-frame manipulation did not affect participants’ conservatism, F < 1.

Regression analyses (Table S3 in the Supplemental Material) revealed that conservatism was negatively associated with support for redistributive policies among participants who either received no information about income inequality (i.e., participants in the control condition), β = −0.94, t(72) = −5.14, p < .001, or were told that the poor make less than the rich (i.e., participants in the poor-have-less condition), β = −0.37, t(72) = −2.45, p < .05. That is, in both of these conditions, the more conservative participants were, the lower was their support for the redistributive policies. However, among participants who were told that the rich make more than the poor (i.e., participants in the rich-have-more condition), there was no relationship between conservatism and support for the redistributive policies, β = 0.14, t < 1, a result indicating that conservatives’ opposition to raising taxes on the rich had been reduced. That is, in the rich-have-more condition, liberals and conservatives were equally likely to support raising taxes on the wealthy (see Fig. 1, left panel).

OPEN IN VIEWER

Fig. 1.

Support for redistributive tax policies (left) and external attributions for wealth (right) as a function of condition and conservatism. Scores on a measure of political conservatism were used to divide participants into conservative (1 SD above the mean) and liberal (1 SD below the mean) groups.

We expected that conservatism would be negatively associated with external attributions for poverty and wealth, but that framing inequality as the rich making more than the poor would attenuate these relationships. Regression analyses of attributions for wealth revealed that conservatism was negatively associated with external attributions for the success of the wealthy among participants who were given no information about inequality (i.e., participants in the control condition), β = −0.55, t(72) = −3.78, p < .001, or who were told that the poor make less than the rich (i.e., participants in the poor-have-less condition), β = −0.59, t(72) = −2.87, p = .005. However, among participants who were told that the rich make more than the poor (i.e., participants in the rich-have-more condition), there was no such relationship between conservatism and external attributions for wealth, β = −0.01, t < 1 (see Fig. 1, right panel). Similar results were obtained from regression analyses of attributions for poverty.

Finally, a mediation analysis revealed that the interactive effect of condition and conservatism on redistributive-policy support was partially mediated by attributions for wealth, z = 1.83, p = .07, but not by attributions for poverty.

Conclusion

People’s explanations for poverty and wealth, and their corresponding levels of support for redistributive tax policies, are influenced by how inequality is described. When income inequality is framed as the rich making more than the poor (as opposed to the poor making less than the rich), conservatives are more willing to support redistributive policies, in part because framing inequality in this way makes conservatives more likely to question whether the wealthy are responsible for their own success. This willingness to support redistributive policies is in stark contrast to conservatives’ seemingly default opposition to such policies and strong convictions that the wealthy deserve their wealth. For researchers and policymakers, these findings suggest one simple approach that can influence how individuals think about and respond to income inequality: Change the language used to describe it.