Introduction
One of the earliest victories of the civil rights movement was the desegregation of public schools. The year 2024 marks seventy years since the landmark Brown v. Board of Education (1954) Supreme Court decision that ended the Jim Crow era of “separate but equal” and opened the door to public schools and educational equality for Black children. School integration met substantial resistance from White racial conservatives determined to maintain the Jim Crow status quo. The image of six-year-old Ruby Bridges, federal agents in tow, defiantly passing through a White mob and climbing the steps of an all-White Louisiana school has become an iconic symbol of the promise that racial progress and equality are on the horizon.
Battles over education access and equity continued since the image of Ruby Bridges became iconic, creating a puzzle for political scientists. While support for public schooling remains strong (Plutzer and Berkman Reference Plutzer, Berkman, Plutzer and Berkman2010), and a majority of White Americans are willing to invest more in schooling programs for Black children, as our data spanning the period from 1990 to 2021 show, a substantial minority of White Americans resist increases in funding for programs targeting Black children. This is concerning because school segregation is on the increase, and graduation rates for Black kids remain lower than for White children. Furthermore, for years, there has been a strong push for vouchers and other policies that serve to weaken public education and exacerbate racial and economic inequality in American society (Quinn and Cheuk Reference Quinn and Cheuk2018; Sattin-Bajaj and Roda Reference Sattin-Bajaj and Roda2020). At the same time, proposals for universal pre-K and other measures thought to level the education playing field for Black children have stalled. Most recently, funding for universal pre-K was dropped from President Biden’s budget (Mayn Reference Mayn2022). Yet, studies show that if more resources were channeled to majority Black schools, these children would make significant educational gains (Owens Reference Owens2020).
Why does a substantial minority of White Americans continue to resist funding increases for educational programs for Black children when such programs can have significant social benefits? One answer comes from the voluminous literature on prejudice (Sears et al. Reference Sears, Hetts, Sidanius, Bobo, DO, Sidanius and Bobo2000; Henry and Sears Reference Henry and Sears2002; Kinder and Sanders Reference Kinder and Sanders1996). The study of racially targeted policies, including in education, has long focused on the role of racial motivations and specifically racial prejudice in shaping Whites’ attitudes toward racial policy (Sears, Hensler, and Speer Reference Sears, Hensler and Speer1979; Chudy Reference Chudy2021). Furthermore, research on explicit and implicit racial priming has suggested that the effects of racial prejudice on Whites’ political preferences are bounded by norms of racial equality and may emerge only in specific contexts, depending on how racial information is communicated (Mendelberg Reference Mendelberg2001; Hurwitz and Peffley Reference Hurwitz and Peffley2005); however, scholars have qualified this model (Huber and Lapinski Reference Huber and Lapinski2006; Valentino, Neuner, and Vandenbroek Reference Valentino, Neuner and Vandenbroek2018).
Social identity theory provides another explanation for why many White Americans do not support increases in funding for the education of Black children. Scholars have shown that ingroup membership automatically induces people to favor their own groups over rival groups, even if they do not harbor negative attitudes toward outgroups. From this perspective, ingroup bias and outgroup prejudice are distinct psychological phenomena and ingroup bias, even absent prejudice, can lead to discrimination (Brewer Reference Brewer1999; Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986). Ingroup favoritism is a factor in economic decisions (Dickinson, Masclet, and Peterle Reference Dickinson, Masclet and Peterle2018) and in redistribution preferences (Magni Reference Magni2021). In the context of education, ingroup favoritism may be an automated response (Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986), or a calculated one (Bobo and Kluegel Reference Bobo and Kluegel1993). Many Whites, even those who display low levels of racial prejudice, may act as “opportunity hoarders” (Lewis and Diamond Reference Lewis and Diamond2015), simultaneously supporting public funding for education programs that may benefit their ingroup, and resisting increases in programs that do not.
In this study, we consider each of these explanations using six survey experiments with non-Hispanic White respondents embedded in three surveys that together span more than three decades. These split-ballot experiments measured people’s support for increases in public spending for various education programs by varying the description of the target group to represent a nonracial category that includes Whites and a racially targeted group that mentioned Black people either explicitly (Black children) or implicitly (e.g., urban, big city). The nonracial control condition was “public schools,” “economically disadvantaged children,” or “poor children,” the latter two cutting across racial groups and focused on income rather than race. The first two experiments were embedded in the 1990 General Social Survey (GSS), and another two were embedded in the 2004 ANES panel. We replicated and extended the design in two experiments fielded in 2021. These experiments, unlike the GSS and ANES ones, allowed us to directly compare the effects of implicit and explicit cues to a nonracial control.Footnote 1 Across all experiments, we find that White respondents were more likely to support increased spending when in the nonracial control condition than when in the experimental condition. We also find no evidence of racial priming across all experiments as all interactions with racial prejudice are null.
Our study makes several important contributions to the literature. First, it shows that although majorities of White Americans support increased funding for education programs targeting Black children, on average support for increases is higher for programs that are targeted to nonracial categories of children that include the ingroup. This has not changed over three decades with attendant normative and substantive implications for social equality and mobility. Second, we present evidence that in the domain of education, White Americans’ discriminatory preferences are likely attributable to ingroup favoritism and not necessarily to racial prejudice. Interactions with stated prejudice are null in all our models. Finally, we demonstrate that neither implicit nor explicit priming produce differential effects by levels of racial priors. The temporal dimension is crucial here. Contrary to the extant theory which suggests that racial priming (especially implicit priming) might have been effective in earlier times, but not in the post-Obama era (Mendelberg Reference Mendelberg2001; Valentino, Neuner, and Vandenbroek Reference Valentino, Neuner and Vandenbroek2018), we find that neither implicit nor explicit priming works in this context.
American Public Education in Black and White
Public education has long been recognized by activists and political leaders as a tool for social mobility and, thus, racial equality. Since limiting African Americans’ access to quality education was the key to maintaining a system of White supremacy and racial subordination, segregation in public schools was a main target for the Civil Rights Movement. The landmark 1954 Brown v. Board decision outlawed segregation and nullified the White supremacist principle of “separate but equal.” Although Whites grudgingly came to accept the principle of racial equality (Kinder and Sanders Reference Kinder and Sanders1996; Schuman et al. Reference Schuman, Steeh, Bobo and Krysan1997), the court decision led to a protracted battle over the means through which school integration and racial equality were to be achieved. Two policies drew great resistance from many Whites: busing and affirmative action. National surveys conducted in the 1970s show that 85% of Whites opposed busing (Sears, Hensler, and Speer Reference Sears, Hensler and Speer1979). Resistance to integration through busing led to declines in white support for public schools as many moved their children to private “White academies.” Furthermore, Whites were also not keen to spend more public funds on majority-Black schools. Decades later, White opposition to increased spending on schools in Black neighborhoods, especially for preschool and early education remained high (Bobo and Kluegel Reference Bobo and Kluegel1993). Current debates that involve race in educational settings and that have brought White resistance into view have to do with diversity, equity, and inclusion (DEI) efforts in colleges and the teaching of the nation’s racial history in secondary and post-secondary classrooms (Kaiser et al. Reference Kaiser, Dover, Payton Small, Brady and Major2022; Sailor and Kissel Reference Sailor and Kissel2021; Monroe Reference Monroe2022; Shah, Weinschenk, and Yiannias Reference Shah, Weinschenk and Yiannias2023; Filindra and Burnett Reference Filindra and Burnett2022).
Recent research on people’s views about education has focused on policy evaluations of specific programs as implemented in schools (Schneider, Teske, and Marschall Reference Schneider, Teske and Marschall2002). However, new policies such as government-funded early education (colloquially known as universal pre-k), college loan forgiveness, and free college tuition have entered the public agenda, triggering responses from the American public. Much like free public education, these newer alternatives have the potential to serve broad constituencies and, unlike busing, are not what scholars and policymakers refer to as “race-targeted policies.” At the same time, these programs can help level the playing field for racial and ethnic minority students.
The early literature suggests that White voters may be less inclined to support increases in public funding for majority Black schools (Bobo and Kluegel Reference Bobo and Kluegel1993), but less is known about contemporary White support for raising funding for public schools and early education programs targeted to Black children. Studies in areas such as public or affordable housing (Pearson-Merkowitz and Lang Reference Pearson-Merkowitz and Lang2020) and welfare policy (e.g., Gilens Reference Gilens1999) underscore the limits of White support for liberal policies meant to alleviate racial inequalities. Specifically, there is significant evidence that White Americans are more interested in paying more for programs that help their group than race-targeted programs from which their group does not directly benefit.
The Role of Racial Prejudice
Classic work in political science and sociology has attributed lagging White support for a variety of racial and racialized policies to racial prejudice. This literature argues that racially conservative Whites, that is, people who score high on measures of racial prejudice, are less likely to support race-targeted policies or policies perceived to differentially benefit minorities over Whites. Scholars also documented the detrimental effect of racial prejudice on White opposition to racial policies in the domain of education, especially busing (Bobo Reference Bobo1983; Sears, Hensler, and Speer Reference Sears, Hensler and Speer1979; Sears and Kinder Reference Sears and Kinder1985). Later studies showed similar results for affirmative action in higher education (Sidanius et al. Reference Sidanius, Singh, Hetts, Federico, DO, Sidanius and Bobo2000), and scholarship programs aimed specifically at Black students (Bobo and Kluegel Reference Bobo and Kluegel1993; Feldman and Huddy Reference Feldman and Huddy2005). Recent work documents the persistent negative effect of racial prejudice, and specifically racial resentment, on support for increased spending on educational opportunities for Black children (Chudy Reference Chudy2021) as well as other racially targeted policies (Wetts and Willer Reference Wetts and Willer2018). Other research has documented the effect of racial prejudice on White Americans’ policy preferences in a variety of policy domains (e.g., Bloeser and Williams Reference Bloeser and Williams2022; Tesler Reference Tesler2016; Filindra and Kaplan Reference Filindra and Kaplan2016), candidate choice (Buyuker et al. Reference Buyuker, Jadidi D’Urso, Filindra and Kaplan2021; Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2019), and confidence in election processes (Appleby and Federico Reference Appleby and Federico2018; Filindra, Kaplan, and Buyuker Reference Filindra, Kaplan and Buyuker2022).
Later scholarship on race and political communication claims that the effects of underlying prejudice do not manifest equally across all contexts. Specifically, when it comes to the communication of racial content and, thus, how we solicit support for or strengthen opposition to racial policies, language matters. The implicit/explicit racial priming theory argues that in the post-civil rights era, new norms of racial parity have led to direct appeals to race being proscribed. When confronted with overt racial messaging, White racial conservatives are likely to feel bound by equality norms and express political attitudes consistent with such norms (Mendelberg Reference Mendelberg2001). In the context of education, this means that racial conservatives are likely to behave similarly to racial liberals when asked whether they support increases in funding for schools in “Black” neighborhoods. However, racial content can also be implicitly communicated through coded or “dog whistle” language. When this type of language is used, White racial conservatives are more likely to express their true attitudes about racial policies (Hurwitz and Peffley Reference Hurwitz and Peffley2005; Valentino, Hutchings, and White Reference Valentino, Hutchings and White2002). In the education context, therefore, racial conservatives may be more likely to oppose increases in education funding for “urban” or “inner city” schools, dog whistle terms for majority Black schools.
Although several studies have found evidence in support of the theory that implicit communication of racial content can “activate” racial priors resulting in significant differences in attitudes between those with low and high racial priors (e.g., Tesler Reference Tesler2015; Filindra and Kaplan Reference Filindra and Kaplan2016), other studies suggest that, especially in the post-Obama era, both forms of priming can work equally well or not at all (Huber and Lapinski Reference Huber and Lapinski2006; Valentino, Neuner, and Vandenbroek Reference Valentino, Neuner and Vandenbroek2018).
The Role of Ingroup Identity
Another explanation for differences in support of various policies among White Americans draws on social identity theory. This paradigm argues that group memberships powerfully shape people’s attitudes and behaviors. As social beings, we are strongly predisposed to favor our ingroups, even when these group identities are not meaningful or consequential. This is because people’s self-esteem is tied to these memberships (Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986).
Scholars have developed various measures of ingroup identity attachment for Whites (Jardina Reference Jardina2019; Filindra, Buyuker, and Kaplan Reference Filindra, Buyuker and Kaplan2023). However, since ingroup favoritism emerges automatically even in contexts where people do not have strong prior attachments to the ingroup, as is the case in the minimal group experiments, it is not necessary to show that differences in response by levels of ingroup attachment to infer ingroup favoritism as a mechanism. It is possible that those with strong prior attachments to the ingroup may be more strongly primed by the experiment, but this is not necessary from the perspective of the theory.
Ingroup favoritism can lead to discriminatory behavior against outgroups, even in the absence of outgroup prejudice. People do not need to have negative attitudes toward other groups in order to promote an unequal distribution of resources; ingroup favoritism is sufficient to produce such results (Brewer Reference Brewer1999). Racial ingroup favoritism may be weakly correlated with racial prejudice (Jardina Reference Jardina2019). In the education context, this suggests that White Americans, independent of racial priors, may be strongly predisposed to favor an unequal distribution of education resources in a way that favors their ingroup and discriminates against Black people.
Based on these theoretical perspectives, this study tests two hypotheses:
-
1. SIT Hypothesis: Non-Hispanic Whites are more likely to support increased spending on education programs targeting a nonracial category that consists largely of ingroup members than increases for programs targeting Black children.
-
2. Racial Priming Hypothesis: The effect of the racial treatment (implicit or explicit) on support for spending for education programs should be stronger among Whites with high racial priors than those with low racial priors.
We test these hypotheses using six split-ballot survey experiments embedded in three distinct surveys from three different decades: 1990, 2004, and 2021. We proceed with our analysis in chronological order with the earliest experiments first and concluding with the most recently fielded experiments.
Study 1: Support for Increased Spending on Neighborhood Schools-Explicit Cue Experiment (1990 GSS)
The GSS is a nationally representative survey of Americans conducted every two years since 1972.Footnote 2 The 1990 survey was administered to 1,118 non-Hispanic Whites and included two split-ballot experiments related to education programs for Black children. The control, nonracial, condition in the first experiment asked respondents: “Here are several things that the government in Washington might do to deal with the problems of poverty and unemployment. I would like you to tell me if you favor or oppose them. Would you say that you strongly favor it, favor it, neither favor nor oppose it, oppose, or strongly oppose it? Spending more money on the schools in poor neighborhoods especially for pre-school and early education programs.” In the experimental condition, instead of “poor neighborhoods,” people were asked about “Black neighborhoods.”Footnote 3 This very simple priming experiment is meant to focus respondents on the racial cue in the education context without adding any extraneous cognitive overload (Hurwitz and Peffley Reference Hurwitz and Peffley2005). This is the case for all experiments we present in the article. This is an explicit racial cue because it directly references race (i.e., Black), rather than use a coded term (i.e., urban). Although many poor people are members of minority communities, most poor people are Whites, and therefore, many White Americans are likely to be sympathetic with this nonracial group even if they do not fully identify with the group (Chudy Reference Chudy2021). The data are weighted to match the national population. We recoded the dependent variable as an ordinal measure scaled from 0 to 1. Descriptive statistics are in online Appendix Table A1.
First, it is important to note that 68% of White Americans favored increased spending for schools in Black neighborhoods, compared to 87% who favored increased spending in poor neighborhoods (Top-2 Box: Strongly favor/ favor). That is a 19ppt difference between the two groups. We used ordinary least squares (OLS) regression to estimate the effect of the treatment on respondents’ support for increased funding for preschool and early education programs. OLS is preferred relative to ordered logistic models because of the ease of interpreting the coefficients. Especially since in subsequent experiments we are modeling binary dependent variables in models with interaction terms, best practices suggest the use of linear modeling specifications (Mood Reference Mood2010).
Figure 1 shows the difference in means between the two conditions. The figure shows the probability of support for increased spending. There was a statistically significant decline in mean support for such programs when associated with Black neighborhoods (b = −.110, p < .001). The results are similar whether weights are used or not (b = −.108; p < .001 if unweighted) (online Appendix Table A2). These results suggest that support for increased funding of early education and preschool programs in 1990 was much stronger when it was described to target poor neighborhoods, a majority-White category, than Black neighborhoods, an explicitly racial category.
Treatment effect of support for increased spending for neighborhood schools (1990 GSS).
The survey does not include the racial resentment measure, which had only recently been introduced and used in the development of racial priming theory (Mendelberg Reference Mendelberg2001),Footnote 4 but it does include two measures of racial stereotypes. These are the “intelligent” and “violent” stereotypes, and they are on 7-point anchored scales (for exact wording see online Appendix A). A factor analysis and reliability statistics show that the two items do not load on the same dimension and do not form a reliable additive index (a = .370). We specified models with interactions between the treatment and each of these two items. We also specified interaction models with ideology and partisanship. Consistent with best practices (Kam and Trussler Reference Kam and Trussler2017), our models include demographic controls. However, models without demographic controls produce similar results. As shown in online Appendix Table A2, interactions with racial prejudice are null at conventional levels of statistical significance, indicating that those with stronger racial priors are not more likely to be primed by an explicit prime (“Black”). The same is the case for the interactions with either ideology or partisanship (online Appendix Table A2). Models with alternate specifications (ordered logistic) produce similar results (online Appendix Table A3).
Study 1 provides evidence consistent with social identity theory, suggesting that ingroup favoritism may influence White Americans’ differential support for increased funding for educational programs. However, we find no evidence to support our second hypothesis that priming may be in effect. This is consistent with the original expectations of the I/E model (Mendelberg Reference Mendelberg2001) which suggests that explicit priming should not activate racial priors. Additionally, interactions with ideology, partisanship, or income are also null. This suggests that resistance to increased spending on educational opportunities in Black neighborhoods was not different across partisan or ideological lines in 1990. This is likely because the partisan realignment that started with the New Deal and intensified after the Civil Rights Movement was still incomplete in 1990 (Carmines and Stimson Reference Carmines and Stimson1989; Schickler Reference Schickler2016).
Study 2: Support for Increased Spending on Special College Scholarships-Explicit Cue Experiment (1990 GSS)
The 1990 GSS included a second split-ballot experiment and 1,125 non-Hispanic Whites participated in this one. The control condition in the second experiment asked respondents: “Here are several things that the government in Washington might do to deal with the problems of poverty and unemployment. I would like you to tell me if you favor or oppose them. Would you say that you strongly favor it, favor it, neither favor nor oppose it, oppose, or strongly oppose it? Provide special college scholarships for children from economically disadvantaged backgrounds who maintain good grades.” In the experimental condition, instead of “economically disadvantaged backgrounds,” people were asked about “Black children.” This is another explicit reference to race (Mendelberg Reference Mendelberg2001). The data are weighted to match the national population. We recoded the dependent variables as an ordinal measure scaled from 0 to 1. Descriptive statistics are in online Appendix Table A1.
Once again, the data show that a two-thirds majority of White Americans (69%) supported increased funding for scholarship programs for Black children; however, support for scholarship programs for poor children – a category that includes many ingroup members – was substantially higher (92%) (Top-2 Box: Strongly favor/ favor), evidence of ingroup favorability.
We used OLS regression to estimate the effect of the treatment on respondents’ support for increased funding for scholarship programs. Figure 2 shows the probability of support for increased expenditures. As Figure 2 shows, on average, there was a statistically significant decline in support for scholarship programs when associated with Black children (b = −.151, p < .001). The results are similar regardless of whether weights are used or not (b = −.147; p < .01; without weights) (online Appendix Table A4). As was the case with Study 1, all interaction models with racial prejudice, ideology, and partisanship are null (online Appendix A, Table A4). Models with alternate specifications (ordered logistic) produce similar results (online Appendix Table A5). Once again, the results show that the use of an explicit prime does not differentially activate prejudicial responses among those who have high or low racial priors. Therefore, there is no evidence of priming effects. This is consistent with the original expectations of the I/E model (Mendelberg Reference Mendelberg2001) which suggests that explicit racial priming should not activate racial priors. However, in both experiments, we lack an implicit racial treatment to make more direct comparisons. Study 2 yields supporting evidence only for our first hypothesis that ingroup favoritism may be influencing Whites’ education policy preferences.
Treatment effect on support for increased funding for scholarships (1990 GSS).
Study 3: Support for Spending on Pre-School Programs for Black Children –Explicit Racial Cue Experiment (2004 ANES Panel)
The ANES is a nationally representative survey conducted during presidential election years. The 2004 Panel Study was conducted between November 3 and December 20, 2004. The study consisted of a post-election survey of 717 non-Hispanic white respondents who had previously been interviewed in both the 2000 ANES study and the 2002 ANES study. Of those, 706 responded to this first experiment. The data are weighted to match the national population.
Respondents were asked a question about early education spending and were randomly assigned to one of two versions of the question. The first version asked, “Should federal spending on preschool and early education for poor children be increased, decreased, or stay the same?” The second version substituted “poor” for “Black.” This is very similar to the GSS experiment but with different response options. Specifically, instead of asking whether the respondent “favors” increased spending, it asks whether spending “should be increased.” We recoded the variable so that “1” corresponds to support for increased spending for schools and “0” corresponds to decreased spending, and .5 corresponds to “stay the same.” Question-wording and summary statistics for the experiment are in Appendix B.
A total of 52% of White respondents said that funding should be increased for Black children compared to 69% who supported increases for “poor” children. To the degree that the two questions are comparable when we compare the 1990 GSS and the 2004 ANES, there appears to be a decline in overall support for increased spending on preschool and early education as well as a decline in support for increased spending for programs targeted to Black children in the span of about a decade and a half.
We specify OLS models with and without demographic controls. We find a negative and significant main effect of the treatment on increased support for early education spending (b = −.182; p < .01). The results suggest that those exposed to the “Black children” treatment are significantly less likely to support increased spending for early education compared to the control condition (Figure 3). Including demographic controls does not change the size of the coefficient substantively. The results remain robust. Also, the results do not change if we use unweighted data (online Appendix Table B2). We also test whether the effects are conditional on racial priors using the racial resentment measure as our moderator. We find that the interaction is null. Alternate models using the anti-Black stereotypes as our moderator are also null, so the results are not sensitive to the measure of prejudice used. The same is the case for interactions with partisanship, ideology, and income (See online Appendix Table B2). Furthermore, interactions with the White feeling thermometer are also null suggesting that the ingroup favoritism effect is automatic and not conditional on the strength of affinity for the ingroup (online Appendix Tables B6 and B7). Ordered logistic models confirm the results (online Appendix Tables B3 and B7).Footnote 5 Once again, the results are consistent with the original expectations of the I/E model (Mendelberg Reference Mendelberg2001) which suggests that explicit racial priming should not activate racial priors. Instead, consistent with our first hypothesis, White respondents show favoritism for the category that includes many members of the ingroup (poor).
Treatment effect on support for funding increase for early education (2004 ANES panel).
Study 4: Funding for “Big City” Schools-Implicit Priming Experiment (2004 ANES Panel Study)
The fourth study was also embedded in the 2004 ANES Panel. A total of 710 non-Hispanic White Americans received the second experimental question. In this wording experiment, respondents were randomly divided into two conditions. Roughly half of the respondents were asked, “Should federal spending on public schools be increased, decreased, or stay the same?” while for the other half, the term “public” was replaced with “big city.” Extant research suggests that terms such as “urban” or “big city” are racially coded terms; that is, their invocation brings up racial considerations in respondents’ minds below the level of consciousness (Hurwitz and Peffley Reference Hurwitz and Peffley2005). Once again, we recoded the variable so that where “1” corresponds to support for increased spending for schools and “0” corresponds to decreased spending and .5 corresponds to “stay the same.” The 1990 GSS did not include a similar implicit prime so there is no direct comparison between the two surveys on this question.
Question-wording and summary statistics for the experiment are in online Appendix B. Data crosstabulations show that in 2004, 48% of White Americans, a little less than half, supported increases in federal spending on “big city” schools, compared to 67% for “public schools. This is a difference of 19 percentage points. Again, we specified OLS models (and ordered logistic models in the online Appendix). As Figure 4 presents, the results show a negative and significant relationship between the treatment and support for increased school spending (b = −.198; p < .01). This suggests that Whites exposed to the “big city schools” treatment were significantly less likely to support increases in school spending compared to the “public schools” control.Footnote 6 Once again, White Americans show favoritism for the category that they tend to associate with their in-group (“public schools”) than the one most frequently associated with minority groups (“big city”).
Treatment effect on support for increased funding for schools (2004 ANES panel).
Like the explicit model of Studies 1–3, all interaction results show no significant effects. This once again suggests that the effect of the treatment is not conditional on racial resentment, anti-Black stereotypes, the White thermometer, ideology, or partisanship (online Appendix Tables B8–B13). The three-way interaction is also not significant.
The results of Study 4 and Study 3 are not directly comparable but they both suggest that in the early 21st century White Americans were less supportive of increases in funding for the education of African American children, and it didn’t matter whether the information was presented to them with an explicit or a coded cue. Taken together, these experiments show no evidence that explicit or implicit cues differentially affect the attitudes of Whites with high and low racial priors, that is, there is no evidence of priming effects as expected by the implicit/explicit model (Mendelberg Reference Mendelberg2001). This is especially important because these experiments were conducted before the Obama era which is purported to have introduced a change in the operation of norms of racial equality (Valentino, Neuner, and Vandenbroek Reference Valentino, Neuner and Vandenbroek2018). As with the previous experiments we only find evidence of ingroup favoritism consistent with social identity theory.
Study 5: Public School Funding 2021 Experiment
The two 1990 GSS and the two 2004 ANES experiments provide evidence that in the 1990s and the early 21st century Whites’ support for increased funding for education opportunities for Black children lagged their support for public education overall, most likely because of ingroup favoritism. However, there is no evidence of priming effects in either dataset. At the same time, both the GSS and the ANES experiments were limited because they did not include implicit and explicit cues in a single experiment thus not allowing direct comparisons between the two, and the comparison groups were not always symmetrical, as in some experiences one group was income based and the other race-based.
As a next step, we sought to replicate and extend the experiments almost two decades later to address these limitations. The third experiment was fielded by Lucid between May 29 and June 1, 2021. Lucid matches samples to Census demographics to approximate national representativeness. Gender within age quotas were also employed.Footnote 7 The study included 878 respondents who self-identified as non-Hispanic Whites. The mean duration of the interview was 23 minutes. The survey included multiple attention checks, and 152 people failed at least one attention check. However, regression analyses show that the findings do not change if inattentive respondents are included or excluded. To be conservative, we include the entire sample. Experiments 5 and 6 were included in the same survey but in randomized order. We control for the effects of exposure to the other treatments in our models. These effects are null and have no impact on our findings.
Respondents were randomly assigned to one of three conditions. They were asked, “Should funding for [public/Black/urban] schools be increased a lot, increased somewhat, kept the same, decreased somewhat, or decreased a lot?” This allows us to directly compare an explicit and an implicit prime to each other and to a control condition. This was not possible with any of the previous experiments.
As with the previous experiments, our variables were re-coded on a zero-to-one scale consistent with the original nature of the variable. The data are weighted using raking weights. Descriptive statistics are in online Appendix Table C1. Balance tests are in online Appendix Table C13. First, in 2021, 70% support increases in the control condition, compared to 54% in the Black condition, and 63% in the urban condition. Again, we see a 16-percentage point difference between the control and the Black condition.
We use OLS regression to test the relationship between the dependent variable and the treatments. Figure 5 shows the average main effects of the three treatments. The results remain robust even after we include demographic controls (online Appendix Table C2). First, support for public school funding in the control condition is significantly higher than in either of the other two conditions. Consistent with Study 1 and Study 3 (early education funding), we find that exposure to the explicit treatment (“Black schools”) reduces support for public school funding (b = −.068; p < .001). The same is true for the “racially coded” (“urban schools”) treatment (b = −.390; p < .05, one-tailed). The difference between the explicit and “coded” treatments is not statistically significant (F = 1.92; p = .166). The results hold if ordered logistic specifications are used (online Appendix C Table C3).Footnote 8 As far as education policy is concerned, White Americans are more resistant to increased funding programs targeting Black children. This is largely consistent with results from the 1990s and 2004, if we account for differences in question wording. The results here suggest ingroup favoritism as was the case in the earlier experiments.
Support for increased funding for schools (2021 Lucid).
Turning to the interaction models, again we find no statistically significant effect for the interaction between the treatments and racial resentment. Contrary to the implicit/explicit priming model, we find no evidence that neither the explicit nor the implicit prime activates racial priors in this context. Interactions with ideology are also null. The interaction with partisanship is negative and significant only for the racially coded treatment, which suggests that White Republicans more so than Democrats are less likely to support increases in spending for “urban schools” (online Appendix Table C2). The results are robust to ordered logistic specification (online Appendix C Table C3). A three-way interaction between the treatment, partisanship, and racial resentment is null (online Appendix C Table C5). Furthermore, models using the anti-Black stereotype measure in the interaction are also null.Footnote 9 Moreover, models using the White identity salience measure show counterintuitive and mixed results.Footnote 10 Overall, the results are consistent with ingroup favoritism but show no evidence of racial priming whether in the implicit or the explicit condition.
Study 6: 2021 Pre-K Experiment
In the same survey as Study 5, we included a second question related to education funding, this time for pre-kindergarten programs. Again, the purpose of this experiment was to replicate and extend the 2004 ANES and 1990 GSS experiments. Here too, respondents were randomly assigned to one of three versions of the question. The survey was also randomized as to the location and order in which respondents answered the two education questions; therefore, we did not find question order effects.
The earlier experiments contrasted between poor-/low-income children and Black children on the expectation that Whites were more likely to sympathize with the broad category of “poor,” which includes many members of the ingroup even if they do not think of themselves as members of the poor category. However, this remains an asymmetric comparison between an income-based category and a race-based category. In this experiment, we sought to introduce more symmetry by categorizing all groups as “poor,” and thus controlling for income status of the target group. The experiment asked: “Do you favor or oppose expanding funding for pre-kindergarten programs so that it is available for [poor/poor Black/poor-inner city] children nationwide? The $24 billion a year cost would be paid for by higher taxes.”Footnote 11 Like “urban,” “inner-city” is another implicit reference to African American populations. Furthermore, by using the term “poor” in all conditions, we can control for the differences connoted by socioeconomic status in the GSS experiments. The response options were “strongly favor, somewhat favor, somewhat oppose, strongly oppose” (4pt scale).
Since all respondents received the information about the cost of the program, we do not expect that this influenced our results. Question wording and descriptive statistics are given in online Appendix C. Balance tests are in online Appendix Table C14. As was the case with public school funding, here too, 79% favored increased spending in the control condition, compared to 72% for either of the other two conditions. Support for pre-K in 2021 appears stronger than in 2004, but the two samples are not directly comparable, and this may also be attributable to the further specification here that the funds would be targeted to poor kids.
Again, our variables were recorded on 0–1 scales consistent with the original nature of the variable. The data are weighted using raking weights. As Figure 6 shows, the effect of the explicit treatment is null (b = −.031; p = .281), but the effect of the “racially coded” treatment is negative and statistically significant (b = −.062; p < .05).Footnote 12 However, given that there is no statistically significant difference between the two experimental conditions (F = 1.34; p = .246), the overall point here is that, once again we see evidence of ingroup favoritism, to the degree that White respondents recognize “poor kids” as part of the ingroup. The results hold even when we use ordered logistic regression (online Appendix Table C7). We thus take our results to be broadly consistent as showing ingroup favoritism, albeit the effect is weaker in this experiment where we control for low-income status of the target group.
Effect of treatment on support for increased funding for Pre-K (2021 Lucid).
Once again, the interactions between the treatments and racial resentment are null (see online Appendix Table C6), indicating no priming effects.Footnote 13 Interactions with ideology and partisanship are also null (online Appendix Table C6), as is a three-way interaction between the treatment, racial resentment, and partisanship (online Appendix Table C7). These results suggest that White Americans are reluctant to support increased spending on pre-K programs targeted to poor Black children relative to poor children, a category that mostly includes the ingroup.
Discussion
In this article, we use six survey experiments with non-Hispanic White respondents conducted between 1990 and 2021 to test whether White Americans are resistant to increasing expenditures earmarked for the education of African American children and whether racial prejudice or ingroup favoritism better explains such attitudes. Our experiments lead to important conclusions. First, average White American support for increased funding for various education programs is higher when associated with non-racial group cues, categories which can reflect the ingroup, than when associated with Black children. Across different comparison groups and specifications, as well as across time, we document significant differences in support for increases between the non-racial control condition and the racially targeted experimental conditions. This is consistent with the expectations of social identity theory (Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986): White Americans show favoritism for the non-racial category over the category that is explicitly or implicitly associated with Black children. Furthermore, White Americans may show greater sympathy for non-Black categories even if they do not identify with the specific group (Chudy Reference Chudy2021). Importantly, ingroup favorability can emerge automatically and does not require strong ingroup priors. This was best demonstrated in the minimal group experiments (Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986). Our analyses show null effects for the interactions between the treatment and ingroup priors.
Second, contrary to the expectations of the racial priming theory (Mendelberg Reference Mendelberg2001), when primed with racial cues, Whites who score high on racial resentment or anti-Black stereotypes are not less likely to support public school and early education funding for Black children than those who score low on these measures. Across all six experiments, we find no evidence of either explicit or implicit priming effects. There is no evidence in any of our experiments that exposure to a racial prime of either type may activate racial priors and thus influence attitudes via such a pathway. This is important because scholars have argued that the effect of implicit priming has dissipated since the Obama era and in the Trump years, both types of primes can work equally well or not at all. However, we find no priming effects in experiments that predate the Obama era. This does not necessarily mean that racial prejudice is not a factor in White Americans’ education policy preferences and attitudes, but rather that exposure to primes related to school spending fail to activate racial priors as would be expected by the I/E model (Carbone, Harell, and Soroka Reference Carbone, Harell and Soroka2024; Chudy Reference Chudy2021; Cullen, Butler, and Graham Reference Cullen, Butler and Graham2021; Filindra and Burnett Reference Filindra and Burnett2022).
The question is why does the I/E model fail in this case? We offer two potential explanations that we do not test herein: one focused on the policy itself and the other focused on the policy target. Education is central to the American liberal mythos as the key path to social and economic mobility. Therefore, education and schooling carry a positive valence. This may not be enough to counter the ingroup favorability effect but enough to counter the negative affective valence associated with the racial cues, leading to a null interaction effect, although this is not sufficient to neutralize ingroup favoritism. Another related possibility is that the target group in this policy domain is not adults who are expected to be responsible and self-sufficient, but children. Social construction theories in political science and psychology suggest that children are positively constructed as likable and worthy of assistance (Kreitzer and Smith Reference Kreitzer and Smith2018; Schneider and Ingram Reference Schneider and Ingram1993). It is thus possible that the positive valence associated with the category of young children may counter the negative affect associated with racial cues, leading to a null interaction. If this is the case, this may account for some other null effects in the literature where scholars have used images of positively constructed groups and especially children (Huber and Lapinski Reference Huber and Lapinski2006): the negative effect associated with Black people may be counteracted by the positive affect associated with children, thus producing null results.
Overall, we suggest that when a policy domain or a target group is positively construed (e.g., education and schoolchildren), racial priming may not work as expected because the negatively valenced racial stereotypes that would otherwise be activated by implicit or explicit cues are counteracted by norms associated with education, or positive stereotypes and norms of protection for the young. In other words, the cognitive dissonance created by contrasting stereotypes may render priming ineffective. When a group is negatively constructed, racial priming may be more likely to work as the implicit/explicit model predicts because it affectively links two sets of negatively signed cognitions. For example, priming may work in the domain of crime policy because Whites hold negative stereotypes about both Black people and criminals. Extant studies of counter-stereotypical cues typically frame Whites as policy targets of negatively valenced policies, such as Medicaid (Valentino, Hutchings, and White Reference Valentino, Hutchings and White2002) and welfare (Huber and Lapinski Reference Huber and Lapinski2006; Mendelberg Reference Mendelberg2001). Others place Black people in a positive light—as veterans, waiving the American flag (Valentino, Hutchings, and White Reference Valentino, Hutchings and White2002) or pictured on the campaign trail with political elites (Stephens-Dougan Reference Stephens-Dougan2016). Yet, these studies have not systematically varied the valence of the policy and the policy target group along with the race cues. Furthermore, the existing research on counter-stereotypical cues offers mixed results, with some claiming such cues fail to prime racial attitudes or even reduce the impact of racial considerations on candidate preferences (Valentino, Hutchings, and White Reference Valentino, Hutchings and White2002), while others show the opposite (Stephens-Dougan Reference Stephens-Dougan2016).
Although we offer robust results indicating that racial priming is ineffective at activating racial resentment when it comes to White preferences for increases in public school and early education spending, further research is necessary to determine whether this is about norms related to the treatment of children, norms related to education, or a combination of both. Future studies should manipulate both the social construction of target groups and the policy domain to better understand when and under what conditions implicit/explicit priming may not work.
What our results show is that seven decades after Brown, White Americans are not against increased spending on public education or early education programs—if these programs are perceived beneficial to the ingroup. The very high average support in our nonracial control groups attests to that. However, our finding, which support for increased funding drops so drastically when the question is framed to advantage Black students, suggests White Americans ingroup favoritism may turn them into “opportunity hoarders”(Lewis and Diamond Reference Lewis and Diamond2015), protective of school funding and programming for White children even if this comes at the expense of Black children. Our data suggest that an automatic process of ingroup favoritism leads to discriminatory effects even when the motivation is not racial prejudice.
We also need to caution about the limitations of these experiments. First, the GSS and ANES experiments did not include implicit and explicit treatments in the same study, so we have no way of knowing if there were any differences between explicit and explicit primes in the 1990s or in 2004. The picture remains incomplete. Second, the early experiments were contrasting between a broad income-based category (“poor”) and a race-based category (Black/big city) making for a somewhat asymmetric design. We sought to correct for this limitation with our 2021 experiments, but we have no way of knowing if a different design would have produced other results in the earlier eras. Third, statistical power may be a concern across all experiments when it comes to interaction effects especially if the effect is very small. That said, in this context, a very small statistical effect, even if significant, may not be substantively important.
Conclusion
With the 70th anniversary of Brown v. Board of Education this year, we need to take stock of racial progress in the United States. The hope of the civil rights movement and the activists who fought for school desegregation was that education would be the door to social and economic progress for the next generations of African Americans. However, in much of America, the purse strings along with the authority to make decisions about curriculum and other school policies are held by White voters who continue to be the majority of voters in many places. Yet, as our data make clear and contemporary debates over education and race further substantiate (Kaplan and Owings Reference Kaplan and Owings2021; Sailor and Kissel Reference Sailor and Kissel2021), White Americans are significantly more eager to invest more resources in public education programs that are likely to benefit their own racial group, than they are programs targeted to Black children. Even though this disparity is due to ingroup favorability and not racial prejudice, it still translates to discriminatory outcomes for Black children who continue to experience high and increasing levels of unequal and segregated schooling (Owens Reference Owens2020). The arc of the moral universe is indeed very long, much longer than seven decades.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/rep.2024.15
Availability of data and material
Data will be available through the Harvard dataverse.
Code availability
Code will be available in Stata.
Funding statement
The study was supported by research funds provided by the University of Illinois at Chicago.
Competing interests
No conflicts.
Authors’ contributions
The authors claim equal credit for this study.
Ethics approval
Studies 5 and 6 was approved by the UIC (Protocol #2013-0959). The approval was granted through expedited review. We certify that the studies were performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.
Consent to participate
All respondents were informed about the purposes of the study, the risks and benefits, confidentiality and privacy. Since the studies were conducted on the internet, signatures were waived and informed consent was obtained through selecting to participate in the survey. Participants were told that they could end their participation at any time and withdraw their data if they so wished.
Consent for publication
The IRB approved consent form either implicitly or explicitly includes consent for publication. In all cases, respondents are told that the researchers are only collecting de-identified data. All data used in the analyses stem from de-identified and aggregated data so there is no way people’s individual responses can be tracked from our analysis or the replication code we provide.
Open access
