The Racializing Influence of Romney’s Welfare Ad

This is a guest post by Brown University political scientist Michael Tesler.  For more on his work, see this Slate piece.


Race has recently taken center stage in the presidential campaign.  From Joe Biden’s suggestion that a Romney-Ryan presidency would re-enslave African-Americans, to some liberal commentators’ contentions that the Romney campaign is using racial code words like “welfare” and “anger” to mobilize anti-black sentiments against President Obama, charges and counter-charges of playing the race card now abound.

Part of this racialized turn in the campaign involves Romney’s welfare ad earlier this month—an ad that questionably accused Obama of ending welfare for work requirements.  While that charge may seem race-neutral, there is a long-standing and strong association in white Americans’ minds between welfare and “undeserving” African-Americans (see here and here).  According to Jonathan Chait, then, “the political punch of this messaging derives from the fact that white middle-class Americans understand messages about redistribution from the hard-working middle-class to the lazy underclass in highly racialized terms.”  An extensive body of social science research described as racial priming seems to support Chait’s contention.  That research shows that such code words as “welfare” and “inner-city,” especially when combined with racial imagery (e.g., the hardworking whites in Romney’s ad), can make racial attitudes a more central determinant of political evaluations (see: 1, 2, 3, 4, 5).  One might therefore expect the welfare ad to activate racial attitudes in public opinion.

We can test that expectation thanks to some unique experimental data collected last week by YouGov.  The survey randomly assigned half of its 1,000 respondents to view the Romney welfare ad (see above) while the remaining half of the sample did not see the ad.  Respondents then answered a series of questions to discern whether and how the ad affected their opinions.  Unfortunately, these follow-up questions did not include vote choice or candidate favorability, which were asked earlier in the survey.  We did, however, ask respondents how well Mitt Romney and Barack Obama’s policies would benefit the following groups in society: the poor, the middle class, the wealthy, African-Americans and white Americans.  Answers were then recoded to range from 0 (“hurt them a great deal”) to 100 (“help them a great deal”).

The welfare ad did not appear to affect people’s overall answers to those questions.  However, it did make attitudes toward blacks a stronger predictor of respondents’ views about the consequences of Romney’s policies for the poor, the middle class, and African-Americans.  To measure attitudes toward blacks, we use a scale called “racial resentment” in the scholarly literature.  For respondents to this survey, we actually assessed racial resentment much earlier, when these respondents were first interviewed in a December 2011 survey.  The four questions that make up this measure are here.

The figure below shows that there was almost no relationship between racial resentment and the opinions of people who did not see the ad.  But among those who saw it, racial resentment affected whether people thought Romney will help the poor, the middle class, and African-Americans.  Moreover, seeing the ad did not activate other attitudes, such as party or ideological self-identification.  It only primed racial resentment.

(Note: Predicted values were calculated from OLS coefficients by setting partisanship, ideology, and race to their sample means.  Source: YouGov Survey, August 2012)

At the same time, the ad failed to “racialize” views of whether Romney’s policies would benefit whites and the wealthy.  This likely stems from the fact that Romney favorability ratings are strongly related to thinking his policies will help the poor, the middle class, and blacks, but only weakly related to believing he’d help whites and the wealthy.

Interestingly, the ad did not appear to further racialize the perceived consequences of Obama’s policies, either.  This is probably because racial attitudes are already linked to Obama, and a single political ad isn’t enough to significantly strengthen an already strong relationship.

Nevertheless, the results from our experiment suggest that ads like the one in this post may well contribute to the growing polarization of public opinion by racial attitudes beyond the voting booth in the age of Obama.

[I thank Brendan Nyhan for suggesting a study of this topic, and John Sides and Lynn Vavreck for help in designing the survey questions.]

[Cross-posted at Model Politics.]

17 Responses to The Racializing Influence of Romney’s Welfare Ad

  1. RobC August 20, 2012 at 3:00 pm #

    “Racial resentment” seems like an odd descriptor for the four YouGov questions, two of which failed to garner a majority of affirmative answers even from African-American respondents.

    Let me address just one question (which did receive 64% agreement by blacks): “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class.” Agreement to this question requires agreement to two components: (1) it is difficult for blacks to work their way out of the lower class, and (2) the reason for this is generations of slavery and discrimination. So someone who agrees with the first premise but attributes it to the effect of dysfunctional family structures or pre-1996 misguided welfare policies or inadequate inner city educational opportunities is, it appears, guilty of racial resentment. Let me try to put this in polite terms: Balderdash! Bushwa! Poppycock! In fact, allow me to go so far as to say: Horsefeathers! Piffle! Tommyrot!

    • Michael Tesler August 20, 2012 at 4:24 pm #

      The questions in the scale, according to one of its creators (Donald Kinder), “distinguish between those who whites who are sympathetic towards blacks from those who are unsympathetic, who resent the failure of blacks, as they see it, to demonstrate the virtues of self-reliance and hard work.” To be sure, this does not mean that everyone who scores high on this measure is unsympathetic to African-Americans or racially resentful. On average, though, individuals who score high will be much more resentful towards African-Americans than those scoring low.

      Moreover, the same racializing pattern in this experiment was produced when using a more direct racial attitude measure–the difference between respondents’ thermometer ratings of blacks and whites. The comparability in effects between these two measures suggest that the racial resentment scale is tapping into something distinctly racial.

      • RobC August 20, 2012 at 4:57 pm #

        Kinder’s comment is helpful. Perhaps the questions do distinguish between those who are and are not “sympathetic” towards blacks, at least if “sympathetic” means acceptance of victimhood rather than a more nuanced desire to identify real reasons black Americans may not be succeeding as often as one would hope. But lack of sympathy or willingness to treat a group as victims is different from resentment. To elide those two concepts seems like putting the rabbit in the hat.

        May I return to the second question, the one I quoted above? If you were to present that question to a social scientist for agreement or disagreement, I’d hope he or she would first want to know whether it is more difficult for blacks to work their way out of the lower class than it is for whites or Hispanics or Asians. If not, one would have to consider the very real possibility that the same factors that make it difficult for whites to work their way out of the lower class are also holding back blacks. If there is a differential in the ability of different races and ethnic groups to work their way out of the lower class, the next step in the analysis would be to examine all the possible reasons for that differential. My point is that a serious social scientist who was being honest would probably answer the question as “don’t know” until he or she had examined the data and considered the possible explanations.

        To be sure, an unserious social scientist might give the off-the-cuff knee-jerk answer that the researchers clearly favor: yes, blacks are held back by generations of slavery and discrimination. But it would be a darn shame if rigorous, analytical social scientists were to find themselves branded as guilty of racial resentment simply because they were unwilling to jump to the conclusions that supposedly demonstrate how sympathetic they are.

        • PBR August 20, 2012 at 8:15 pm #

          RobC — I share some of your reservations about the survey questions. They are not perfect measurements of people’s attitudes.

          That said, the findings comport well with some other facts about this ad. As the post notes, the claim that Obama is ending work requirements has been proven wildly wrong (see link below). Moreover, given what social scientists know about the history of racial-partisan conflict, and what they know about the racial connotations of terms like “welfare queen” and “inner-city,” the conclusion isn’t that far-fetched at all. In fact, they corroborate an *entire body of research literature*.

          Social science is hard, and ideal experiments are not always feasible. But that doesn’t mean every academic who comes up with a conclusion you dislike is biased, or that an imperfect experiment tells us nothing.

          • RobC August 20, 2012 at 8:48 pm #

            PBR, I have no problem with the conclusions, only the scale that claims they’re related to racial resentment, because that’s not what those questions are measuring. We’ve seen this sort of thing before, where for example opposition to affirmative action is treated as a measure of racial animus. That’s bad science because it makes logical leaps that are invalid. It begs the question in the correct sense of that phrase: it assumes its conclusion.

            Science should not be polemics. We need to be scrupulous about rigor and precision and not drawing unwarranted conclusions from the evidence. I’m sorry to be tiresome about this, but well-intentioned bad science is still bad science.

            • John Sides August 20, 2012 at 9:30 pm #

              RobC: The measurement of racial resentment is still not without controversy, but it has evolved long beyond issues like using views of affirmative action to measure attitudes toward blacks. See this piece:


              Also, one aspect of Michael T.’s comment has not yet gotten its due: the same results emerged when he used a completely different measure of racial attitudes — one that simply asked respondents how favorable or unfavorable their feelings towards blacks and whites were. So while it’s fine to ask questions about measures — especially since racial attitudes are not always easy to measure — in this case the measures don’t much affect the point of the post.

              • RobC August 20, 2012 at 10:21 pm #

                Thanks for the citation to the article, John, although candor compels me to admit it made my teeth hurt. I continue to find much of the nomenclature unnecessarily provocative. Even if as the authors contend the construct referred to as symbolic racism is a real belief system, is it racism, symbolic or otherwise? As I wrote earlier, doesn’t that put the rabbit in the hat?

                I note in that connection that one of the NES and LACSS responses that denote denial of continuing discrimination and hence symbolic racism is an affirmative answer to the question, “Has there been a lot of real change in the position of black people in the past few years?” I suppose it’s open to question what “the past few years” means, but I thought it was incontrovertible that recent decades have seen very significant growth of the black middle class and even upper middle class. Am I wrong about that?

                Before leaving the measure of racial resentment, let’s note a quirk in the measure. Tessler and Sears describe a measure similar to Kinder’s:

                Since 1987, the Pew Research Center and its predecessor Times Mirror, in their series of surveys on American values, have regularly asked a battery of four race-related questions that approximates the
                content of Kinder and Sanders’s (1996) racial resentment battery.
                These questions gauge the extent of discrimination against African Americans, the group’s societal advancement, whether we should do everything we can to help blacks and other minorities even if it means giving them special preferences, and whether the country has gone too far in pushing for equal rights (see appendix for question wordings). The items do not form quite as reliable a scale as the racial
                resentment battery (α = .54 across survey years compared to about .75 for the racial resentment scale). They are also specially unreliable for African Americans (α = .29 across survey years), who as a result are excluded from our Pew analyses.

                That the four questions yield unexpected results for African Americans is especially telling. Doesn’t that strongly suggest that, whatever agreement with those propositions may signify, it isn’t a basis for judging racial resentment?

                • John Sides August 21, 2012 at 9:48 am #

                  RobC: Yes, the nomenclature “symbolic racism” has also attracted criticism.

                  Re: racial resentment and blacks. Just because racial resentment isn’t a reliable measure among blacks doesn’t ipso facto lessen its usefulness among whites. Perhaps the architecture of whites’ attitudes toward blacks and blacks’ attitudes toward blacks is just different. It’s the same challenge that arises when you try to measure a particular construct in two different cultures. Trying to measure “affinity for sports” in the US might entail asking questions about basketball and football, but in the UK it might entail asking about soccer and cricket.

                  And again, the particular measure of racial attitudes doesn’t appear to affect the conclusions of this post.

  2. CC August 20, 2012 at 8:12 pm #

    The bigger problem here, which goes back to another recent post on The MonkeyCage, is that researchers are asking questions and interpreting responses one way while the respondents may have had very different thought processes that the researchers don’t consider.

    The researchers are the ones bringing up racial group considerations in the first place in their questions and rigging the “results” of the scale to conform to their own worldviews of race. In the Kinder “racial resentment scale,” the researchers interpret certain choices as racist (which is what the “racial resentment” measure is invariably used as a proxy for) rather than actually viewing any example of racist behavior by individuals and towards individuals.

    The real unaddressed debate here is what the researchers think makes someone a racist in the first place. If you think that affirmative action is unfair and a net harm to society, is that racist? If you think that the cultures fostered by certain ethnic groups matter in whether or not they succeed, is that racist?

    From my readings of the literature, it appears that the Kinder et al. crowd seem to think that holding the above positions make one a racist. If that’s the case, then the researchers are the ones who need to examine their own biases (as RobC notes above) before using race as a cudgel in one-sided debates.

    • PBR August 20, 2012 at 8:28 pm #

      It’s not just about the question wording. As the author explains in the comments:

      “…the same racializing pattern in this experiment was produced when using a more direct racial attitude measure–the difference between respondents’ thermometer ratings of blacks and whites. The comparability in effects between these two measures suggest that the racial resentment scale is tapping into something distinctly racial.”

      So the ad hominem attacks in your comment seem misplaced.

  3. IOKIYAR(ight-wing) August 21, 2012 at 2:40 am #

    Romney’s racist ads are specifically designed to tap into Republican’s “Southern Strategy” of using racial resentment for political purposes.

    Republicans are the Party of Lee Atwater, Roger Ailes, and Frank Luntz, masters at manipulating emotions for political power.

    Romney and his team know exactly what they are doing with provoking racial resentment.

    Romney’s racist ads are malevolent, manipulative, and divisive.

  4. Nadia Hassan August 21, 2012 at 10:37 am #

    How beneficial might this be to the Romney campaign? On the one hand, it seems like it could help by disposing racially resentful white voters more favorably towards Romney. I do question, though, how easily the Romney team can build an entire campaign around it. Gilens noted in the book and to me that welfare is the exception rather than the rule on antipoverty policy, and Romney goes after a lot of popular programs, middle class entitlements, and cuts taxes for the wealthy (cue the estate tax paper!).

    Moreover, Prof. Sides mentioned that ads seem to have rapid decay. They also do not exist in isolation. How much would these ads help Romney in cycles where news features gaffes, debates, discussions of policy proposals, etc. ? It seems like this could improve Romney’s favorability and perceptions of class-related attributes via racial polarization, but it’s an open question as to how strong the effects would be.

    • Michael Tesler August 21, 2012 at 11:20 am #

      For your first question, the results suggest that the ad will not do much good for Romney–at least not nationwide. In fact, those who saw the ad were actually slightly less favorable towards Romney than the control group, although that difference wasn’t significant. Of course, the Romney camp will probably be better at targeting individuals receptive to this message–racially resentful whites–than our national sample.

      For your second question, it’s difficult to generalize results from this highly unrealistic setting, whereby respondents see only one side’s ad and then have their opinions assessed immediately afterwards. The experiment is good for causally establishing how implicitly racial ads can activate racial attitudes in political evaluations, but it’s indeed likely that any real-world effects of the ad will decay when other events and communications make new considerations salient.

  5. J Severs August 21, 2012 at 8:03 pm #

    Romney ad defended here: Opening quote: “I had some serious doubts about Mitt Romney’s ad attacking Obama’s welfare “waivers”–until I read the New York Times editorial denouncing it. Now I know Romney’s ad isn’t as accurate as I’d thought. It’s much more accurate.”

  6. Ted Brader August 22, 2012 at 5:24 pm #

    Michael and John, kudos on the interesting post and discussion. A few quick observations and questions.

    1. It is not surprising that some readers have questions about racial resentment, since many scholars have wrestled with this measure as well (as John notes). It almost certainly captures differences in sympathy/resentment toward blacks as well as adherence to economic individualism/self-reliance (it was designed by Kinder and Sears to pick up both!). That can make it “complicated” to interpret. In fact, research by Stanley Feldman and Leonie Huddy shows that it apparently captures these aspects differently for liberals and conservatives (capturing racial aspects more for the former, and ideological aspects more for the latter). It is thus important as you noted to compare to other measures and to note (as Michael and John have) that the ad apparently didn’t prime ideology and partisanship. This doesn’t settle the matter, but it increases our confidence it was the racial component being primed. One question out of curiosity since the numbers aren’t shown here: was the change (increase) in the weight of racial resentment across conditions was statistically larger than the increase in ideology/partisanship across conditions?

    2. The substantive plotting of results is as always helpful, and I wonder if it provides some clues to how we might interpret the data, especially seeing how the questions about how Romney will affect these groups. To the extent there is any asymmetry, it looks like there’s somewhat more “action” (from exposure to the ad) on the low end of the racial resentment scale. Thus, one might make the case that the “racialization” provoked by this ad primarily takes the form of convincing racially sympathetic voters that Romney is bad for blacks, the poor, and the middle class. Put differently, this ad seems to cause racial liberals to become even more pessimistic about Romney. So not exactly a Willie Horton moment!

    3. Other scholars have found it harder to observe much racial priming in recent studies, suspecting that Obama’s election has somehow racialized many issues or our national politics generally — and thus racial attitudes are quite active most of the time. From that perspective, it is interesting, Michael, that you were able to find evidence of racial priming where others have not. I wonder if the effects were observable because they focus on Romney and his policies specifically, as opposed to Obama and/or a national policy under debate? You sort of allude to this near the end, I think.

    • Michael Tesler August 22, 2012 at 7:45 pm #

      Thanks for these comments, Ted. For your first question, the increased effects of resentment came at the expense of ideological self-placement’s influence. To ease interpretation, I constructed an index of these five items weighted by their effect on Romney favorability in the control group. The interactive effect of resentment*ad on that Romney index is .27 (se = .09), while the coefficient on conservative ideology*ad = – .40 (.12). The impact of party id is slightly stronger in the ad condition, but taken together increased effect of resentment on the Romney index is significantly stronger than the combined increase of PID and ideology.

      You are absolutely right in noticing that Romney the ad activated racially liberal opposition to Romney. So much so, in fact, that the net effect of the ad was negative on Romney, although that negative effect was not significant.

      For your third point, I think that the choice of DVs here definitely made it easier to obtain significant priming results. Sears and I speculated in Obama’s Race that it should be difficult to either deactivate or prime resentment’s influence since race is so accessible in Obama evaluations; and we were familiar with some of the experimental studies you alluded to that further confirm that point. So, everyone involved in designing the experiment expected null findings going ahead of time. That’s precisely what we got for Obama’s policy consequences too. The baseline negative effect of resentment on the Obama index was quite large -.29(.07), making it hard for the ad to significantly racialize him further (resentment effect in ad condition = -.37(.08)). Meanwhile the baseline effect of resentment on the Romney index was 0 in the control group. Had we been able to ask about actual Romney favorability in the post-test, whose baseline resentment effect is now much higher than zero in YouGov polls, we may not have gotten significant results.

      As for policy preferences,I’ll show in a future Model Politics posting that the ad failed to further racialize public opinion about welfare and unemployment, both of whose baseline resentment effects were already huge, but it did increase the positive effects of resentment on social security and Medicare preferences, ala

  7. Acompanhantes September 2, 2012 at 5:10 pm #

    What’s up friends, its enormous post concerning teachingand fully explained, keep it up all the time.