Education Higher Ed, Affirmative Action
February 16th, 2023 23 Minute Read Issue Brief by Renu Mukherjee

Stacking the Deck in Favor of Affirmative Action? How “Framing” Affects Polling on One of America’s Most Controversial Policies


In a plurality opinion in the 1978 Supreme Court case Regents of the University of California v. Bakke, Justice Lewis Powell held that colleges and universities could consider an applicant’s race in the admissions process in order to attain a diverse student body.[1] In a pair of cases that will be decided in the current term, the Supreme Court has been asked to reconsider that precedent, sparking a reinvigorated public debate about the merits of affirmative action. In these policy debates, much has been written about the attitudes of blacks, whites, and Hispanics toward racial preferences. The views of Asian Americans, however—who are often portrayed as aggrieved victims of this policy alongside whites—have received scant attention.[2]

In a set of companion cases pending before the Supreme Court—Students for Fair Admissions v. President and Fellows of Harvard College and Students for Fair Admissions v. University of North Carolina—the plaintiffs have focused on the way in which those schools’ policies penalize Asian-American students. According to statistics published by an organization called AAPI Data, however, over two-thirds of Asian Americans are supposedly in favor of affirmative action. The claim has gone viral, appearing in popular media outlets such as NBC News, USA Today, Vox, and The Nation.[3] We are told by AAPI Data and these outlets that most Asian Americans are in favor of race-conscious college admissions and that those behind the lawsuits are nothing but a “vocal Asian-American minority railing against affirmative action.”[4]

However, a closer look at the survey questions AAPI Data has used to measure Asian-American attitudes toward affirmative action over the years reveals that Asian Americans may not be as enthusiastic about this policy as the public has been led to believe. Taken from AAPI Data’s biennial Asian American Voter Survey, these questions appear to be designed to elicit a favorable view of affirmative action from respondents. Other survey questions on affirmative action from organizations like Pew Research Center, Gallup, and Quinnipiac, which ask respondents about this policy in a more neutral way, elicit, on average, a much less positive view.

All of these sources, taken together, show that survey questions on affirmative action that define the policy as providing preferential treatment to its beneficiaries—which is how affirmative action programs actually work—produce much less support for the policy than those that don’t. Moreover, survey questions that portray women as beneficiaries of the policy produce more support than those that do not.

The question on which AAPI Data relies inaccurately describes affirmative action as “designed to provide blacks, women, and other minorities with greater access to higher education.” By contrast, in the 2016 and 2018 Asian American Voter Survey, AAPI Data included a second question on affirmative action that asked respondents about their views on the policy after explaining it is “designed to increase the number of black and minority students on college campuses”—which more closely reflects real and existing affirmative action programs.

Recent claims that broad swaths of Asian Americans and the general public support affirmative action are inaccurate. They are based on biased survey questions that do not reflect how affirmative action programs really work. When respondents are informed of how affirmative action works in practice, support for the policy is much lower.


AAPI Data and Its Affirmative Action Questions

The statistic of interest here—that over two-thirds of Asian Americans favor affirmative action—comes from the group’s Asian American Voter Survey (AAVS), a national survey of Indian-, Vietnamese-, Filipino-, Korean-, Japanese-, and Chinese-American registered voters conducted every other year since 2016. (These are the six largest Asian-origin groups in the United States.)

AAPI Data’s question on affirmative action in the most recent (2022) AAVS read: “Do you favor or oppose affirmative action programs designed to help Black people, women, and other minorities get better access to higher education?”[5] This question yielded results that were overwhelmingly positive. Not only did 69% of respondents say they favor affirmative action programs, but a majority of each Asian-origin group included in the survey also said that they favored such programs (Table 1).

The 2022 results are virtually indistinguishable from those of the 2020 AAVS (also shown in Table 1), when 70% of respondents said they favored the policy. There was only a small fluctuation in support for affirmative action among the Asian subgroups between the 2022 and 2020 AAVS. This is not surprising, given that the wording of the question was almost identical in both surveys.[6] (The only difference was that in 2020, half of respondents were told that affirmative action is “designed to help Black people” while the other half of respondents were told it is “designed to help Blacks.”)

Table 1

Unlike the 2022 and 2020 surveys, however, the 2018 and 2016 iterations of the AAVS contained two different questions about affirmative action that were quite different in wording from one another.[7] One of them was the same question that was asked on the 2022 and 2020 AAVS. It read (in both 2018 and 2016): “Do you favor or oppose affirmative action programs designed to help blacks, women, and other minorities get better access to higher education?” Support for the policy in these years (64% and 66%, respectively) was only slightly lower than in 2020 and 2022 (Table 2). Asian-American support for affirmative action was on the rise slightly until the 2022 AAVS, at which point it dropped 1 percentage point from the previous AAVS.

Table 2

But in those same years, 2016 and 2018, AAPI Data also included another question about affirmative action, which read: “In general, do you think affirmative action programs designed to increase the number of black and minority students on college campuses are a good thing or a bad thing?” The answers to that question tell a different story entirely.

Table 3

As shown in Table 3, when respondents were told that affirmative action would “increase the number” of specific groups, the response was much more negative. In 2018, support for affirmative action among Asian Americans as a whole was 8 percentage points lower (58%) in response to this question than in response to the “better access” question (66%). In 2016, there was a 12 percentage-point gap between the two questions—64% support on the “better access” question, but only 52% on the “increase the number” question. What’s more, in 2016, nearly two-thirds of Chinese-American respondents (63%) responded to the alternate question by saying affirmative action was a “bad thing,” the only instance in which a majority in any AAVS has disapproved of the policy.

Survey Question Framing and Policy Image

What explains the divergence in response to these two questions? One explanation is that the questions describe, or “frame,” affirmative action policies differently, thus conveying a very different image to the respondent.

In 2006, political scientist Terry Moe explained how framing of a similar policy over time can change levels of support.[8] He evaluates the validity of a claim long made by Phi Delta Kappa (PDK), the country’s premier association of educational professionals. Based on the results of its annual poll, which includes questions on school vouchers, PDK says that support for school vouchers is declining. But as with AAPI Data’s affirmative action questions in the AAVS, PDK’s school voucher measures have changed over the years. From the 1970s until 1991, the organization defined vouchers in its annual poll as “a government-funded program allowing parents to choose among public, private, and parochial schools.” Its question thus read: “In some nations, the government allots a certain amount of money for each child’s education. The parents can then send the child to any public, parochial, or private school they choose. This is called the ‘voucher system.’ Would you like to see such an idea adopted in this country?” In PDK’s 1991 poll, support for school vouchers rose to an all-time high of 50% (with 39% opposed) in response to this question.

Then in 1993, PDK replaced its previous school vouchers question with a new one, which read: “Do you favor or oppose allowing students or parents to choose a private school to attend at public expense?” The results were strikingly more negative; only 24% of respondents said they supported school vouchers, a drop of 26 percentage points in just two years. Soon after this second question on school vouchers was released, PDK introduced a more neutrally worded question, to be asked alongside the “public expense” question, that read: “A proposal has been made that would allow parents to send their school-age children to any public, private, or church-related school they choose. For those parents choosing nonpublic schools, the government would pay all or part of the tuition. Would you favor or oppose this proposal in your state?” The results of the more neutral one showed greater support for vouchers than the results of the “public expense” one. In 1996, for example, PDK’s public expense question produced a level of support for vouchers of 36%, while the other question produced a level of support of 43%; in 1999, these results were 41% and 51%, respectively, and in 2001, they were 34% and 44%.

As Moe explains, most Americans have little information and poorly formed views on specific policy proposals—including school vouchers—and will therefore be sensitive to information conveyed in the question’s specific wording. As Moe notes, serious researchers know that “information also needs to be balanced, so that respondents are not pushed to see the issue in a positive or a negative light.” The goal of a survey after all, is supposed to be measuring public opinion, not shaping it.

In their book Agendas and Instability in American Politics, political scientists Frank Baumgartner and Bryan Jones introduce a similar concept, which they call policy image—that is, how a policy is portrayed to and subsequently understood by the public.[9] Baumgartner and Jones begin from the same premise as Moe’s essay on PDK’s school vouchers measure: Most Americans tend to be uninformed of or apathetic about complex matters of public policy. For this reason, policymakers must devise a way to bring a previously uninformed or apathetic American public into the fray when building a coalition of support for or against a given policy. Policymakers can most readily do this, they conclude, by reducing the policy to a symbolic and easy-to-understand image.

A policy’s image, however, can change in the minds of the public—indeed, swaying public opinion often depends on changing a certain policy’s image. Baumgartner and Jones point to the changes in public opinion on nuclear power as an illustrative case study. In its early years, nuclear power enjoyed widespread support; those who developed the technology had successfully cultivated for it an image of progress: “atoms for peace; electricity too cheap to meter; a clean, high-tech technology; low-cost source of energy for the future.” However, when opponents of nuclear power presented to the public a competing image of danger and environmental degradation, Americans’ support for the policy plummeted. This negative image continues to dominate the debate around nuclear power. As recently as March 2022, Pew reported that only 35% of U.S. adults say that the federal government should encourage production of nuclear power.[10]

Competing Images of Affirmative Action

To understand why the responses to AAPI Data’s two questions on affirmative action were so divergent, then, we have to consider the competing images of the policy that they present.

Image 1: Equality of Opportunity and Non-Discrimination in College Admissions

The affirmative action question that AAPI Data has used in every AAVS reads: “Do you favor or oppose affirmative action programs designed to help blacks, women, and other minorities get better access to higher education?” This question, particularly the word “access,” conveys an image of equal opportunity and non-discrimination in college admissions; it suggests that affirmative action simply helps minority students apply for and (if accepted) attend an institution of higher education. An overwhelming majority of Americans support equality of opportunity, so it is unsurprising that AAPI Data’s “better access” question on affirmative action yields an average net favorability of nearly 50%.[11]

Image 2: Equality of Outcome and Quota Systems in College Admissions

AAPI Data’s alternate question on affirmative action, which was dispensed with after the 2018 AAVS, read: “In general, do you think affirmative action programs designed to increase the number of black and minority students on college campuses are a good thing or bad thing?”[12] The phrase “increase the number of ” conveys an image of a de facto quota system, like those used in the 1970s to further equal outcomes between blacks and whites in areas like employment and education, which were eventually outlawed by the Supreme Court. Quota systems were unpopular when they were first introduced, and they appear to remain unpopular today. A July 2022 poll from the State Policy Network found that 78% of Americans agreed with the statement “We should aim for equality of opportunities, not equality of outcomes.”[13] Not surprisingly, this question yielded an average net favorability for affirmative action of just 30%.

In a Twitter thread published five days after the Supreme Court heard oral arguments in the Students for Fair Admissions cases, AAPI Data codirector Janelle Wong defended the framing of her organization’s “better access” question and criticized that of a question from Pew Research Center in a March 2022 survey.[14] Pew’s question read: “Here are some factors colleges and universities may consider when making decisions about student admissions. Do you think each of the following should be a major factor, minor factor, or not a factor in college admissions?”[15] The factors listed were: high school grades, standardized test scores, community service involvement, being first in one’s family to go to college (first-generation status), athletic ability, whether one’s relative attended the school (legacy status), gender, and race or ethnicity. In response to Pew’s question, 74% of survey respondents said that an applicant’s race or ethnicity should not be a factor in college admissions decisions (compared to 26% who said that an applicant’s race or ethnicity should be a factor). Among Asian-American respondents, 63% said that an applicant’s race or ethnicity should not be a factor in college admissions decisions (compared to 37% who said it should be). Asian Americans were 32 percentage points less supportive of affirmative action in response to Pew’s question than they were in response to AAPI Data’s 2022 “better access” question.

Wong suggests that the difference can be attributed to roughly three aspects of how each of these questions is framed. First, Pew’s question, she claims, is written in a way that suggests race is the “only or primary factor” colleges consider under a race-conscious admissions system, as opposed to being just one factor among many. Second, because Pew’s question does not mention affirmative action by name, “some may not even know they are being asked about” affirmative action. Lastly, Pew’s question does not say who benefits from affirmative action. Because AAPI Data’s “better access” question “asks DIRECTLY about ‘affirmative action’—which is how this debate is framed, provides context—names policy & beneficiaries,” she concludes that it better represents the policy, as well as Asian-American attitudes toward it.

None of these aspects of AAPI Data’s “better access” question, however, reflect how affirmative action works in practice today. Affirmative action provides applicants from certain racial minority groups with an admissions “tip,” thereby increasing their chances of admission to a given college or university. Because there are a limited number of spots, the chances for whites and unpreferred racial minorities (who don’t receive such a tip) are decreased. For this reason, affirmative action is more akin to a de facto quota system than to one which guarantees equal opportunity and non-discrimination for all applicants in the admissions process.

Contrary to what Wong argues, Pew’s March 2022 affirmative action question does not suggest that race is the only factor colleges consider under a race-conscious admissions scheme; it specifically asks survey respondents whether race should be a factor—as opposed to “the factor”—and it lists seven other factors. And in reality, an applicant’s race or ethnicity can be the deciding factor in their admissions decision. At oral argument in the Supreme Court, Harvard’s own attorney Seth Waxman conceded as much, saying “Race in some—for some highly qualified applicants can be the determinative factor.” In fact, the district court, despite ruling for Harvard, found that race is determinative for 45% of black and Hispanic applicants who are accepted to Harvard.[16]

Wong’s criticism of Pew for not expressly calling the policy “affirmative action” is similarly unconvincing. More important than a policy’s name—which can often be a contentious issue in itself—is an honest explanation of how the policy functions, and Pew’s question does this by explaining to survey respondents, in the simplest of terms, that the policy at issue involves using race as part of the criteria for evaluating students. AAPI Data’s question, by contrast, suggests that affirmative action simply provides “Blacks, women, and other minorities” with “better access” to higher education without specifying how this is achieved. As far as the respondent knows, better access might simply mean encouraging more students from these groups to apply, rather than using their race as a factor in admissions. As we have noted, average people tend to be less informed about matters of policy than wonks, and thus to honestly gauge what they think about affirmative action, a pollster ought to explain clearly, like Pew does, what this policy entails.

Wong’s third criticism of Pew’s framing—that their question does not expressly say who benefits from affirmative action—actually highlights another flaw in AAPI Data’s “better access” question, which obscures who the true beneficiaries of the policy today are. AAPI Data’s question suggests that “blacks,” “women,” and “other minorities” benefit equally from affirmative action. The inclusion of “women” and “other minorities,” is particularly misleading, as it may suggest to some Asian-American respondents that they, or their children, benefit from the policy. In reality, the largest beneficiaries are “underrepresented” minorities, namely black and Hispanic applicants. Duke economist Peter Arcidiacono has found that while an Asian-American applicant in the top academic decile has a 12.7% chance of getting admitted to Harvard, and a white applicant in the same decile has a 15.3% chance, an African-American applicant in the top academic decile has a 56.1% chance, while a Hispanic applicant has a 31.3% chance.[17] (As one might expect, Harvard’s expert witness in the Students for Fair Admissions case, Berkeley economist David Card, disputes Arcidiacono’s findings. His own preferred admissions model used to estimate the effects of Harvard’s race-conscious admissions policy shows no penalty against Asian-American applicants.)[18]

For these reasons, Pew’s March 2022 question on affirmative action provides a more realistic depiction of this policy than AAPI Data’s 2022 “better access” question, which fails to explain to survey respondents what affirmative action programs are or how they operate. The same can be said of AAPI Data’s “increase the number of ” question in comparison to its “better access” question.

Stacking the Deck in Favor of Affirmative Action?

The difference in support for affirmative action between AAPI Data’s “better access” question and its “increase the number of ” question, as well as the difference in support for affirmative action between the organization’s “better access” question and Pew’s March 2022 affirmative action question, suggests that Americans’ support of this policy will be at its lowest when the policy is in some way described as giving preferential treatment to a group on account of the group’s race or ethnicity (i.e., when survey respondents are told what affirmative action, in effect, does). Interestingly, Wong, along with her fellow AAPI Data senior researcher, Jennifer Lee, and CUNY sociology professor Van Tran, admitted as much in a little-noticed but consequential October 2018 blog post published on AAPI Data’s website.[19] The post discussed five affirmative action questions across three separate 2016 national surveys: the Asian American Voter Survey (AAVS), Collaborative Multiracial Post-Election Survey (CMPS), and Post-Election National Asian American Survey (NAAS) (Table 4).

Table 4

The authors note that:

Questions that refer to ‘increasing numbers’ of black and minority students’ rather than ‘increasing access to’ higher education seem to negatively influence overall rates of support, as do questions that mention ‘considering race’ rather than simply referring to affirmative action.

When affirmative action is framed as increasing access to under-represented groups, it garners higher levels of support. When the policy is framed as increasing the number of under-represented groups, it draws higher levels of opposition, perhaps because it invokes the idea of quotas, which have been ruled unconstitutional since 1978.

The three researchers also concluded that support for affirmative action as a policy tends to be lower among survey respondents “when ‘women’ are not included as potential beneficiaries in the question” than when they are included.

To see if these patterns held among a larger sample of survey questions on affirmative action, I turned to those amassed by the Roper Center for Public Opinion Research, the world’s oldest archive of social science data and the largest specializing in data from public opinion surveys. Roper has 68 questions that asked survey respondents about their views on affirmative action in a manner similar to or exactly like (in terms of wording and structure) AAPI Data’s two affirmative action questions, along with the accompanying results. I sorted these 68 questions and results into five categories and then calculated the average net favorability of affirmative action for each category (Table 5).

Table 5

The first category (Describes Preferential Treatment) includes questions that asked survey respondents about their views on affirmative action after describing the policy as in some way providing its beneficiaries with preferential treatment. These questions yield an average net favorability for affirmative action of –10%. By contrast, the second category of questions (Does Not Describe Preferential Treatment) asked survey respondents about their views on affirmative action only using the benign phrase “affirmative action” (i.e., without explaining how the policy works). This category yielded an average net favorability for affirmative action of 16–26 percentage points higher than that of the first.

The third category includes questions that asked survey respondents about their views on affirmative action after specifying that quotas would not be used. These survey questions are from the 1980s and 1990s, with none being asked after 1995; they yielded an average net favorability for affirmative action of 42%. I chose to include this category of questions in my overall analysis because it appears to be evidence of the fact that affirmative action can, in fact, convey two different images: one of providing minorities with the equal opportunity to succeed (no preferential treatment) and one of furthering equal outcomes among the racial groups (through systems of preferential treatment for certain racial minorities, like quotas). The results suggest that while Americans are more than supportive of the former (42%), they are unequivocally against the latter (–10%).

The final two categories of questions have to do with whether women are included as beneficiaries of affirmative action. I found that when women are included as benefitting from this policy, its average net favorability is 26%. However, when they are not included, the average net favorability for affirmative action drops to 14%.

Recommendations for Polling on Affirmative Action and Beyond

When affirmative action programs are described for what they are in survey questions—racial preferences geared toward increasing the number of underrepresented minorities (excluding women) on college campuses and, in certain cases, employment—support for them is much lower among Americans than if no such description is provided.

Thus, if researchers of public opinion wish to report on Americans’ views of affirmative action in an honest way, then they ought to do two things:

  • First, write survey questions that explain to respondents how affirmative action functions today, that is, by giving preferential treatment to underrepresented racial minorities in an effort to bolster their numbers on college campuses, as well as in certain jobs. They ought to frame affirmative action, in other words, as seeking to further equal outcomes among the various racial groups rather than simply giving all groups equal opportunity to succeed. Merely asking survey respondents if they support “affirmative action,” or describing affirmative action as providing minorities with greater “access,” is not enough to gauge a true view of what Americans think about this policy.
  • Second, specify in their survey questions who the primary beneficiaries of racial preferences in education and employment are: blacks and Hispanics. Given that both college admissions and hiring are zero-sum games, respondents should be made aware of this because providing one applicant with a tip for the applicant’s race inevitably results in some other applicant who did not receive such a tip being penalized. If this step is not taken, then survey respondents may be led to believe that they, too, will benefit from affirmative action if they fall under the vague category of “minority,” or that affirmative action involves no trade-offs, when in reality, Asian Americans and whites have been shown to be adversely impacted by this policy.


This dearth of information on Asian Americans’ attitudes toward affirmative action has made it possible for the results of one misleading set of survey questions—suggesting that over two-thirds of Asian Americans support this policy—to gain an outsized role in the debate over the issue. Americans have been told, by AAPI Data and much of the mainstream media, that any Asian American who took part in or supports Students for Fair Admissions’ lawsuits against Harvard and UNC is merely part of a “vocal minority.”

But the survey question behind that viral statistic seems designed to elicit a favorable response from respondents about affirmative action. It obscures the reality of affirmative action (a racial preference in admissions decisions) and instead suggests that the program simply promotes “better access,” or more equal opportunity, in admissions. Indeed, when the same survey respondents were asked about affirmative action after being told that such programs were designed to “increase the number” of various groups, support was far lower. Similarly, when Pew described how affirmative action policies work and asked whether race should be one factor among many, 63% of Asian-American respondents said it should not be considered.

I don’t wish to accuse AAPI Data of intentionally stacking the deck in favor of affirmative action. It is curious, however, that in 2018, the organization shared a blog post in which two of its senior researchers noted that one way to increase support for affirmative action is to word questions that frame the policy as “increasing access” and benefiting women. In 2020, the organization opted to use only this very type of question—which certainly raises the possibility that they were more interested in obtaining inflated support for the policy than in a genuine measurement of public opinion.

Asian Americans are the fastest-growing racial or ethnic group in the U.S. today.[20] More effort ought to be put into understanding their views, on a variety of policies, in an unbiased way.



Please see Endnotes in PDF

Photo: TanawatPontchour/iStock


Are you interested in supporting the Manhattan Institute’s public-interest research and journalism? As a 501(c)(3) nonprofit, donations in support of MI and its scholars’ work are fully tax-deductible as provided by law (EIN #13-2912529).