The Politics of Achievement Gaps

Abstract

For decades, researchers have documented large differences in average test scores between minority and White students and between poor and wealthy students. These gaps are a focal point of reformers’ and policymakers’ efforts to address educational inequities. However, the U.S. public’s views on achievement gaps have received little attention from researchers, despite playing an important role in shaping policymakers’ behaviors. Drawing on randomized experiments with a nationally representative sample of adults, we explore the public’s beliefs about test score gaps and its support for gap-closing initiatives. We find that Americans are more concerned about—and more supportive of proposals to close—wealth-based achievement gaps than Black-White or Hispanic-White gaps. Americans also explain the causes of wealth-based gaps more readily.

Keywords

achievement gap educational policy equity experimental research politics poverty race survey research

In education, like other areas of public policy, what the public believes about an issue can affect how that issue is addressed. Publicly elected officials have incentives to serve their constituents, and public opinion can shape policymaking agendas and policymakers’ decisions (Burstein, 2003; Erikson, Wright, & McIver, 1993; Kingdon, 1984; Page & Shapiro, 1983). Public opinion is not the sole determinant of elected officials’ behaviors—interest groups, for example, have considerable influence—but elected officials tend to address issues to their constituents’ liking, particularly when those issues are salient to the public (Burstein, 2003).

One issue that many education researchers have urged policymakers to prioritize is the “achievement gap,” or the difference in academic outcomes between historically advantaged and disadvantaged groups.¹ Since at least the Coleman Report (Coleman, 1966), researchers have devoted careers to measuring, explaining, and eliminating disparities in students’ opportunities and outcomes. These gaps are also well known to education reformers and even the American public, which has heard calls for education reform from its last two presidents motivated by these large, persistent differences in average test scores (National Constitution Center, 2009; White House, 2008).

The achievement gap’s ubiquity in policy discourse and implications for American society make it important to understand the public’s beliefs about it. Many proposals for closing gaps require action from policymakers, and policymakers’ actions depend on the public’s views. Yet despite the import of public opinion, there have been few attempts to assess and compare what Americans believe about today’s gaps between students of different races, ethnicities, and economic statuses.

This study examines what the U.S. public believes about three distinct test score gaps: Black-White, Hispanic-White, and poor-wealthy. We test for differences by randomly assigning survey respondents to answer questions about one gap and then comparing the responses across groups. This allows us to explore whether policymakers might find different levels of public support for addressing different types of gaps. We use a large, nationally representative sample, enabling us to draw inferences about the beliefs of the full U.S. adult population and notable subgroups within it.

Our findings reveal strong, consistent evidence that the American public is more concerned about wealth-based test score gaps than race- or ethnicity-based gaps. For example, 64% of American adults say it is essential or a high priority to close the poor-wealthy test score gap, whereas only 36% and 31% say the same about the Black-White gap and the Hispanic-White gap, respectively. Respondents were also more supportive of proposals to narrow wealth-based rather than race- or ethnicity-based gaps, and they more readily attributed these gaps to the explanations we provided.

Background and Literature

Test score gaps between poor and wealthy, Black and White, and Hispanic and White children can all be seen as symptomatic of policies and practices that have long provided unequal educational opportunities to members of different social classes, races, and ethnic groups. However, these gaps have different origins, patterns, and implications.

Early American aspirations for an educated citizenry largely excluded African Americans. Slave states provided virtually no opportunities for formal education, and many states passed anti-literacy laws to suppress organized rebellion (Williams, 2005). The education system that emerged after Reconstruction left most Black children in segregated schools with minimal resources. Black students across the country confronted incongruence between their aspirations and career options, with educators grappling with how much and what type of education to provide Black students with little access to white-collar jobs (Tyack, 1974). Progress toward desegregating and improving schools has been slow since the U.S. Supreme Court’s Brown v. Board of Education decision for reasons including active resistance to school desegregation and residence-based school assignment in the context of severe residential segregation (Orfield & Eaton, 1996).

Hispanic and Latino Americans also have a long history of placement in segregated schools or classrooms. In general, this was less rooted in strict de jure segregation and more a part of appeasing White parents who wanted to keep Spanish-speaking children out of their own children’s classrooms (MacDonald, 2013). Despite some legal and political progress in the 1960s and 1970s, these struggles have continued over issues like bilingual education and immigrant education and have become increasingly complex as the number of Latino immigrants has grown (Stepler & Brown, 2015).

Today, most Black and Hispanic students attend schools where the majority of students are Black and Hispanic. These schools generally have fewer resources than schools where the majority of students are White, including teachers with less experience and formal education, lower licensure scores, and lower value-added scores (Adamson & Darling-Hammond, 2012; Goldhaber, Lavery, & Theobald, 2015; Ladson-Billings, 2013; Lankford, Loeb, & Wyckoff, 2002). Many of these differences in resources are also evident between poor and wealthy students, with high-poverty schools tending to have lower instructional staff, teacher, and nonpersonnel expenditures than other schools in their districts (Heuer & Stullich, 2011).

Wealth- and race-based test score gaps are interconnected due to the high correlations between race and economic status in the United States. However, the gaps should not be conflated. Although both minority and poor children confront challenges in their treatment, expectations, and opportunities, some of the specific challenges confronting these groups of children are similar whereas others are different (Wilson, 2012). Moreover, these test score gaps do not move in unison. When Reardon (2011) examined long-term trends in income- and race-based gaps in math and reading, he found a substantial widening of the test score gap between poor and wealthy children and a narrowing of the gap between Black and White children. Whereas the Black-White gap was substantially larger than the poor-wealthy gap (10th/90th income percentiles) for cohorts born in the 1940s through 1960s, the poor-wealthy gap has been substantially larger for recent cohorts.

Another reason to distinguish between wealth- and race-based test score gaps is that they may have different political implications. If the American public does not feel similarly about wealth- and race-based gaps, then policymakers might have different incentives, strategies, and opportunities for addressing these gaps. And there are psychological, sociological, and political reasons to expect this could be the case. Psychologists have extensively documented the tendency for individuals to favor the “in-groups” with which they identify over the “out-groups” with which they do not (for a review, see Hewstone, Rubin, & Willis, 2002). However, what one sees as one’s in-group depends, in part, on the perceived permeability of these groups (Ellemers, van Knippenberg, & Wilke, 1990; Huddy, 2013). In the context of test scores, it may be that White Americans—who constitute about two thirds of the U.S. population—will be more inclined to support proposals to eliminate wealth-based achievement gaps because they can more easily conceive of themselves as poor than as Black or Hispanic. Many American adults, in fact, have experienced variation in their financial wealth over time. People could see the boundary between rich and poor as more permeable than the boundary between White and Black or Hispanic, resulting in greater identification with those of different incomes.

A related sociological phenomenon that could lead to divergent reactions to wealth-based and race-based gaps is homophily, or the tendency for people to prefer developing social ties with those who are similar over those who are dissimilar (McPherson, Smith-Lovin, & Cook, 2001; Mollica, Gray, & Treviño, 2003). If the majority of Whites see those of a different economic class as more similar to them relative to those of a different race, then one might expect greater connection and concern between classes than races. This, in turn, could influence how important a particular achievement gap is perceived to be and what one is willing to do about it.

More generally, the politics of race and class in the United States could yield different attitudes toward efforts to close race- and wealth-based test score gaps. Surveys have shown almost universal support for abstract principles of racial equality but lackluster support for related policies (Krysan, 2000). This is evident in a 2012 Phi Delta Kappan/Gallup poll in which 97% of respondents called it very or fairly important to improve urban schools but only 62% would pay higher taxes to accomplish it (Bushaw & Lopez, 2012). There is some evidence that Americans prefer race-neutral programs that target benefits to the poor to race-conscious programs that target benefits to racial minorities. Feldman and Huddy (2005), for example, found that class-based affirmative action policies were more popular than race-based policies. The political reality that race-conscious policies are unpopular in the United States has contributed to calls to focus on class-based policies (Kahlenberg, 1996) or build multiracial political coalitions that focus on shared economic interests rather than racial divisions (Wilson, 1999).

We draw on these insights to examine K-12 test score gaps and the types of government actions that may narrow them, using randomized survey experiments to compare Americans’ beliefs about poor-wealthy, Black-White, and Hispanic-White gaps. The paper proceeds with a description of our sample and data, followed by details about our survey items. We then describe our findings, which we present both for the full sample and disaggregated by key subgroups. We conclude by discussing implications of this study, paying particular attention to policymaking and educational equity.

Data

This study’s data come from a survey administered to members of YouGov’s online respondent panel. YouGov is a firm specializing in academic survey research and online political polling, with an opt-in panel of approximately 1.8 million U.S. residents who complete surveys in exchange for modest prizes (YouGov, 2015). The items described here appeared as part of a broader omnibus survey constructed by an interdisciplinary group of Stanford University researchers through Stanford’s Laboratory for the Study of American Values.

The sampling process was designed to produce a weighted sample that is approximately representative of the U.S. adult population. YouGov used demographic and political variables to match members of its panel to a random sample of the U.S. Census Bureau’s American Community Survey (ACS). It then invited matched panelists to complete our omnibus survey, before settling on a final sample of 1,000 respondents. YouGov generated respondent-specific weights to further improve the fit between the survey sample and the matched ACS sample, and we weighted our sample accordingly for all analyses.²

How well this sample represents the U.S. adult population depends on how well the variables used for matching and weighting produced an analytic sample that reflects the broader public’s beliefs about achievement gaps. The variables used for matching and weighting consist of respondents’ gender, age, race, education, party identification, political ideology, and political interest. Assessing the sample’s true representativeness is difficult, but YouGov has performed well in predicting election outcomes using similar methods (Rivers, 2012; Silver, 2012) and partnered with The New York Times and CBS News for election polling (Cohn, 2014). At minimum, this sample should be approximately representative of the U.S. adult population, demographically and politically. Indeed, the descriptive statistics presented in Table 1 indicate that the weighted YouGov sample closely mirrors recent statistics issued by the U.S. Census Bureau (2013a, 2013b, 2013c) and Gallup (Jones, 2014).

Table 1

Demographics of Sample and U.S. Adult Population

Variable	U.S. Population (Ages 18+)	YouGov Sample (Weighted)	YouGov Sample (Unweighted)
Race/ethnicity
Black, non-Hispanic	11.9%	11.0%	7.8%
Hispanic	15.0%	12.0%	6.6%
White, non-Hispanic	65.7%	70.4%	78.2%
Other race	7.4%	6.6%	7.3%
Family income
Less than $30,000	20.8%	30.2%	27.9%
$30,000 to $99,999	50.7%	48.0%	47.9%
$100,000+	28.5%	11.1%	12.0%
Prefer not to say	—	10.7%	12.1%
Gender
Female	51.3%	52.1%	57.0%
Male	48.7%	47.9%	43.0%
Age
Mean age (years)	46.9	47.6	48.2
Educational attainment
No college	42.1%	45.7%	37.4%
Some college/2-year degree	29.0%	30.8%	33.9%
Completed 4-year degree	28.9%	23.4%	28.6%
Political identification
Leans Democrat	47%	43.8%	46.2%
Leans Republican	41%	35.8%	35.5%
Independent (no lean)	—	16.3%	15.2%
Not reported	—	4.1%	3.0%
No. of observations	N/A	996	996

Note. All subsequent analyses incorporate respondent weights. Data on race/ethnicity, family income, gender, age, and educational attainment for the U.S. adult population come from the U.S. Census Bureau (2013a, 2013b, 2013c). Family income is reported at the household level for the U.S. population (the percentage of households in each income bracket) and the respondent level for the YouGov samples (the percentage of respondents from households in each income bracket). Data on political affiliation for the U.S. adult population come from Gallup (Jones, 2014). For the U.S. population data, respondents were asked to identify themselves as Democrat, Republican, or Independent. Those who identified as Independent were then asked whether they lean to the Democratic Party or Republican Party and have been classified according to their reported lean. For the YouGov data, respondents were classified as “leans Democrat” if they reported a political affiliation of “strong Democrat,” “not very strong Democrat,” or “lean Democrat” and classified as “leans Republican” if they reported “strong Republican,” “not very strong Republican,” or “lean Republican.”

Methods

Our basic strategy was to randomly assign survey respondents to a treatment condition and then compare their responses across conditions, drawing causal inferences from the differences. These types of survey experiments are increasingly common in the social sciences (Mutz, 2011). For the past several years, Education Next and the Program on Education Policy and Governance (PEPG) have jointly conducted annual public opinion surveys that use randomization (e.g., Howell, West, & Peterson, 2007; Henderson, Peterson, & West, 2016). Peterson, Henderson, and West (2014) report on several experiments along these lines, focusing especially on teachers’ and the public’s views of U.S. education policy. Other survey experiments in education have tested how Tennesseans’ opinions of educational institutions and policies change in response to receiving student performance data (Clinton & Grissom, 2015) and how the format of school report cards affects Americans’ opinions of school quality (Jacobsen, Snyder, & Saultz, 2014).

We used two sets of experimental manipulations. First, we randomly assigned respondents to one of three “gap” groups. Approximately one third of respondents answered questions about poor-wealthy test score gaps, one third about Black-White gaps, and one third about Hispanic-White gaps. This treatment assignment persisted throughout the survey such that every item we presented to a respondent referred to that person’s assigned test score gap. Second, for one item, we randomly assigned respondents to one of two “proximity” groups. Approximately half of respondents answered a question about a policy that would affect their local school district and the other half about a policy that would affect districts elsewhere in the country.

The survey contained one question about how to prioritize closing test score gaps, four questions about support for specific gap-closing proposals, and four questions about why the gaps exist. The first item asked respondents to what extent closing their assigned test score gap should be a national priority. The item appeared as follows, with respondents seeing only the text relevant to their assigned treatment condition³:

In the United States today, [White students/students from wealthy families/ White students ] usually score higher than [Black students/students from poor families/ Hispanic students ] on math and reading tests.

Thinking about all of the important issues facing the country today, how much of a priority do you think it is to close this achievement gap between [White students and Black students/students from wealthy families and students from poor families/ White students and Hispanic students ]?

We use this item to compare the importance that Americans assign, in abstract terms, to closing these test score gaps. There were five possible responses: 1 = not a priority, 2 = low priority, 3 = medium priority, 4 = high priority, 5 = essential. We asked respondents to think about test score gaps in the context of other important issues in order to communicate a sense of the potential trade-offs involved in closing the gap and to discourage respondents from overstating their concern about the issue.

The second set of items asked respondents about their support for three specific proposals that might reduce the size of test score gaps. Here, too, we communicated that there could be trade-offs involved. The item began with the following text:

Here are three proposals to reduce the test score gap between [White students and Black students/students from wealthy families and students from poor families/ White students and Hispanic students ]. Each proposal may require public funding from higher taxes.

Please show whether you would support or oppose each proposal.

We looked for actual proposals that are widely understood by the American public and would seem plausible as gap-closing strategies. The proposals were as follows:

Offer experienced teachers bonus money if they work in schools with mostly [Black students/students from poor families/ Hispanic students ].

Provide government funds to help [Black students/students from poor families/ Hispanic students ] pay for private school tuition.

Create summer school programs that [Black students/students from poor families/ Hispanic students ] can sign up to attend for free.

For each proposal, respondents expressed their support or opposition using the following 5-point scale: 1 = strongly oppose, 2 = somewhat oppose, 3 = neither support nor oppose, 4 = somewhat support, 5 = strongly support. Every respondent saw these three proposals simultaneously.

The next item also asked about a specific proposal—and used the same response options—but randomized respondents along an additional dimension. We asked some respondents about a hypothetical proposal for their own local school districts, and we asked others about the same proposal for districts elsewhere in the country. For those assigned to the local district condition, the item appeared as follows:

Imagine that your local school district is considering a proposal that might narrow or close the test score gap between [White students and Black students/students from wealthy families and students from poor families/ White students and Hispanic students ]. Under this proposal, more [Black students/students from poor families/ Hispanic students ] in your district would be assigned to the district’s best teachers. As a result, more [White students/students from wealthy families/ White students ] in your district would be assigned to the district’s worst teachers.

Would you support or oppose this proposal?

Respondents asked about nonlocal districts saw “some school districts in other parts of the country are” in place of “your local school district is” and “these districts” in place of “your district.” Otherwise, the item was identical across conditions except for grammatical adjustments. This enabled us to test for differences in how supportive respondents were depending on whether they were asked about nearby or distant schools, in addition to whether they were asked about poor-wealthy, Black-White, or Hispanic-White gaps. Among advantaged groups, greater support for these proposals when enacted elsewhere could suggest a form of the “not-in-my-backyard” (NIMBY) syndrome: People desire a good—like more equitable educational opportunities—but want other communities, rather than their own, to incur the costs of acquiring it. Peterson et al. (2014) used similar survey experiments to assess the existence of NIMBY in U.S. education politics, finding only mild and mixed evidence of its existence, perhaps strongest among affluent Americans.

Our final set of items asked respondents to explain the causes of their assigned test score gap. We adapted items used by Feldman and Huddy (2010) to study how White Americans explain Black-White disparities in economic and educational outcomes. They adapted their items from similar questions that appeared in the General Social Survey. We used very similar language and identical response options to Feldman and Huddy, modifying our items to incorporate questions about poor-wealthy and Hispanic-White gaps. We combined their items about discrimination and injustice, and made other more subtle changes, to facilitate coherent comparisons across our wealth and race groups.

Our questions correspond to attributing the gap to (a) discrimination and injustice, (b) student motivation, (c) parenting, and (d) genetic differences. However, these concepts are interconnected. For example, one might say that parenting differences across groups account for part of a test score gap but believe that those parenting differences are rooted in societal injustices. The item asked how much of the difference in test scores could be attributed to each cause, with four response options: 1 = none, 2 = a little, 3 = some, 4 = a great deal.⁴ We did not ask respondents to rank the explanations partly because of their inherent interconnectedness and partly because we wanted to examine variation across gap conditions in the extent to which respondents attributed gaps to any of the explanations provided.

Analytically, our highest priority is comparing responses across treatment conditions. Randomization makes these comparisons straightforward since, in expectation, the members of these treatment groups should be similar except for their exposure to treatment-specific survey language. For interpretability and transparency, we present our main findings in terms of simple comparisons of means across treatment conditions. For example, we tested whether respondents assigned to the poor-wealthy condition expressed greater support for the summer school proposal (on the 5-point Likert scale) than respondents assigned to the Black-White condition. These are essentially t tests, using the scales described above, that we ran using ordinary least squares (OLS) regression without covariates. Where sample size allows, we report results disaggregated by respondents’ race, ethnicity, and family income, focusing on the groups identified in the test score gap conditions (Black, Hispanic, White, poor, and wealthy).

There are two basic concerns with this approach. First, if the randomization process yields dissimilar groups, then differences across treatment conditions could result from differences in the types of people assigned to each group rather than the impact of the treatment language itself. Second, by converting these survey responses to numerical values, we handle ordinal scales as if they are interval, potentially distorting the measured differences.

We have taken precautions to ensure that neither issue distorts our findings. Appendix Table A1 shows descriptive statistics for each of our “gap” groups, in addition to the results of tests for statistically significant differences between them. Appendix Table A2 shows the same for our “proximity” groups. In general, respondents are well balanced across treatment conditions with respect to their observed characteristics. However, with some chance differences evident (especially in Table A1), we checked the robustness of our findings to the inclusion of covariates. Appendix Table A3 shows our key results from models with and without covariates. These results are very similar, which increases confidence that the differences observed across conditions are attributable to the treatments themselves. This table also shows results from ordinal logistic regression models—with and without covariates—that relax the assumption that our scales are interval. The magnitude of the coefficients is difficult to compare across the OLS and ordered logit models, but the signs and significance are again very similar, suggesting that treating these scales as interval (in order to facilitate interpretation) does not distort the main results.

Results

Our findings reveal clear, consistent patterns. Most notable among them, respondents asked about test score gaps between students from poor and wealthy families were consistently more concerned about these gaps—and more supportive of gap-closing policies—than respondents asked about Black-White and Hispanic-White test score gaps. The differences were generally large in magnitude. Where there were significant differences between the Black-White and Hispanic-White groups, they reflected stronger support for closing the Black-White gap, but these differences were smaller in magnitude.

Table 2 compares how much the poor-wealthy, Black-White, and Hispanic-White groups prioritize closing test score gaps and how strongly they support the three proposals aimed at closing these gaps. This table includes results disaggregated by respondents’ race, ethnicity, and family income. It reports tests comparing responses across gap conditions both for the full sample and within each race, ethnicity, and income subgroup.

Table 2

Importance of Closing Test Score Gaps and Support for Related Proposals by Type of Gap and Respondent Subgroup

	Mean Response by Condition			Test of Differences Across Conditions
Variable	Poor-Wealthy	Black-White	Hispanic-White	Poor-Wealthy Minus Black-White	Poor-Wealthy Minus Hispanic-White	Black-White Minus Hispanic-White
Priority to close gap	3.68	3.03	2.85	0.65***	0.83***	0.18
	(1.13)	(1.31)	(1.21)	[0.12]	[0.12]	[0.12]
Among Blacks	3.90	3.87	3.09	0.04	0.81	0.78*
Among Hispanics	3.92	3.35	3.45	0.57	0.47	-0.09
Among Whites	3.56	2.87	2.73	0.70***	0.84***	0.14
Among <$30,000	4.01	3.06	2.92	0.94***	1.09***	0.15
Among $100,000+	3.58	2.84	2.93	0.74**	0.65**	-0.09
Policies
Teacher bonuses	3.39	2.76	2.53	0.63***	0.86***	0.22*
	(1.12)	(1.36)	(1.27)	[0.12]	[0.12]	[0.12]
Among Blacks	3.68	3.24	2.62	0.44	1.06**	0.62
Among Hispanics	3.45	3.36	2.94	0.09	0.51	0.41
Among Whites	3.30	2.64	2.46	0.66***	0.84***	0.18
Among <$30,000	3.61	2.78	2.68	0.83***	0.94***	0.11
Among $100,000+	3.42	2.87	2.66	0.54*	0.76**	0.22
Vouchers	2.81	2.33	1.95	0.48***	0.86***	0.38***
	(1.31)	(1.39)	(1.11)	[0.13]	[0.12]	[0.11]
Among Blacks	3.78	3.88	2.16	−0.10	1.62***	1.72***
Among Hispanics	2.79	3.15	2.64	−0.36	0.16	0.52
Among Whites	2.61	2.02	1.76	0.60***	0.86***	0.26**
Among <$30,000	3.12	2.66	2.11	0.47*	1.01***	0.54***
Among $100,000+	2.45	2.39	1.66	0.06	0.80***	0.73**
Summer school	3.84	3.44	3.19	0.40***	0.65***	0.25**
	(1.20)	(1.44)	(1.38)	[0.13]	[0.14]	[0.12]
Among Blacks	4.53	4.03	2.92	0.51	1.62***	1.11***
Among Hispanics	3.33	3.57	3.57	−0.24	−0.24	0.00
Among Whites	3.84	3.34	3.15	0.50***	0.69***	0.19
Among <$30,000	4.22	3.54	3.13	0.68***	1.09***	0.41*
Among $100,000+	4.02	3.23	3.75	0.80**	0.27	−0.53
No. of observations	320	354	322	674	642	676

Note. Each coefficient represents the results of a separate regression model run for the treatment groups, sample, and outcome variable specified. Standard deviations appear in parentheses and standard errors appear in brackets. Responses indicating how much of a priority it is to close the gap were given on the following scale: 1 = not a priority, 2 = low priority, 3 = medium priority, 4 = high priority, and 5 = essential. Responses indicating support for each gap-closing policy were given on the following scale: 1 = strongly oppose, 2 = somewhat oppose, 3 = neither support nor oppose, 4 = somewhat support, and 5 = strongly support.

p < .1. **p < .05. ***p < .01.

Respondents said that closing wealth-based test score gaps is a substantially higher priority than closing race- or ethnicity-based gaps. On our 5-point scale, the mean response for the poor-wealthy gap was 3.68, compared to 3.03 for the Black-White gap and 2.85 for the Hispanic-White gap. The poor-wealthy mean is significantly greater than the Black-White and Hispanic-White mean (p < .01 in each case), but the difference between the Black-White and Hispanic-White means is not statistically significant. In terms of the raw survey responses, 63.7% of the poor-wealthy group reported that closing the gap is essential or a high priority, compared to 35.6% of the Black-White group and 31.0% of the Hispanic-White group.⁵

Looking across subgroups, we see results that are alike in some ways and different in others. High-income, low-income, and White respondents all indicated it was a higher priority to close poor-wealthy gaps than to close Black-White and Hispanic-White gaps (p < .05 in each case). There were no significant differences within the Black or Hispanic subgroups in how much they prioritized closing wealth-based versus race- and ethnicity-based gaps. To some extent, this could be a product of statistical imprecision, since relatively small subgroups of respondents were divided across three conditions. Black respondents reported that closing the Black-White gap is a higher priority than closing the Hispanic-White gap (3.87 vs. 3.09, p < .1), but no other subgroup demonstrated a statistically significant difference for this gap.

The rest of Table 2 shows support for three specific gap-closing proposals. Here, too, we see more emphasis on closing the poor-wealthy gap than the other gaps, and we also see more support for targeting the Black-White gap than the Hispanic-White gap. Among the full sample, there are statistically significant differences in every treatment comparison for each proposal, with many of these differences substantively large. Respondents in the poor-wealthy condition more strongly supported the teacher bonus proposal than respondents in both the Black-White condition (3.39 vs. 2.76, p < .01) and Hispanic-White condition (3.39 vs. 2.53, p < .01). This difference between the Black-White and Hispanic-White conditions was marginally significant (p < .1). These levels of support correspond to strongly support or somewhat support responses from 52% of the poor-wealthy group, 31% of the Black-White group, and 27% of the Hispanic-White group. The voucher proposal received greater support from the poor-wealthy group (2.81) than from the Black-White (2.33) and Hispanic-White (1.95) groups, with all of these differences highly significant (p < .01). The summer school proposal also received greater support from the poor-wealthy group (3.84) than from the Black-White (3.44) and Hispanic-White (3.19) groups, with all of these differences significant (p < .05).

We observe stronger support for the summer school proposal than the teacher bonus and voucher proposals but advise caution interpreting these differences. The overall support for proposals like these can be sensitive to the particular wording used. We constructed these items to assess differences between respondents assigned to the poor-wealthy, Black-White, and Hispanic-White conditions, not to assess Americans’ support for these policies in general.

Table 3 shows results for the final proposal, to assign students from lower-scoring groups to a district’s best teachers. We focus on whether support for this type of gap-closing policy differs when one is asked about her local school district versus other school districts. Aggregating across the gap groups, we find marginally significant differences in support (p < .1) depending on whether respondents were asked about their own districts or other districts around the country. The mean response for the local-district group was 1.92, with 11% strongly or somewhat supporting the proposal. The mean response for the other-districts group was 2.10, with 15% strongly or somewhat supporting the proposal. Although, as suggested by Peterson et al. (2014), high-income respondents had a higher mean level of support for this proposal in other districts than in their own, this difference is not statistically significant (2.06 and 1.86, respectively; not displayed in tables).

Table 3

Support for Redistributive Teacher Assignment Proposal by Type of Gap and Whether Asked About Local or Other Districts

	Mean Response by Condition and Tests of Differences Across Local-/Other-District Conditions				Test of Differences Across Gap Conditions
Condition	Total	Poor-Wealthy	Black-White	Hispanic-White	Poor-Wealthy Minus Black-White	Poor-Wealthy Minus Hispanic-White	Black-White Minus Hispanic-White
Local district (n = 515)	1.92	2.22	1.84	1.67	0.38**	0.56***	0.18
	(1.10)	(1.23)	(1.02)	(0.95)	[0.17]	[0.17]	[0.12]
Other districts (n = 480)	2.10	2.46	1.78	2.05	0.68***	0.41**	−0.27
	(1.22)	(1.24)	(1.13)	(1.18)	[0.17]	[0.19]	[0.17]
Local minus other	−0.18*	−0.23	0.07	−0.38**
	[0.10]	[0.20]	[0.13]	[0.16]
No. of observations	995	319	354	322

Note. Each coefficient represents the results of a separate regression model run for the treatment groups specified. Standard deviations appear in parentheses and standard errors appear in brackets. Responses indicating support for this proposal were given on the following scale: 1 = strongly oppose, 2 = somewhat oppose, 3 = neither support nor oppose, 4 = somewhat support, and 5 = strongly support.

p < .1. **p < .05. ***p < .01.

Next, we consider views of the causes of the poor-wealthy, Black-White, and Hispanic-White test score gaps (Table 4 and Appendix Table A4). Two results stand out.

Table 4

Explanations of the Causes of Test Score Gaps by Type of Gap and Respondent Subgroup

	Mean Response by Condition			Test of Differences Across Conditions
Cause	Poor-Wealthy	Black-White	Hispanic-White	Poor-Wealthy Minus Black-White	Poor-Wealthy Minus Hispanic-White	Black-White Minus Hispanic-White
Cause 1: Discrimination/injustice	2.48	1.97	1.95	0.52***	0.54***	0.02
	(1.02)	(1.06)	(1.00)	[0.10]	[0.10]	[0.09]
Among Blacks	3.14	2.38	2.25	0.77**	0.90**	0.13
Among Hispanics	2.50	2.49	2.33	0.02	0.17	0.15
Among Whites	2.38	1.88	1.81	0.50***	0.57***	0.06
Among <$30,000	2.80	1.99	1.88	0.81***	0.92***	0.11
Among $100,000+	2.12	2.09	2.12	0.03	0.01	−0.03
Cause 2: Student motivation	2.66	2.34	2.18	0.33***	0.48***	0.16
	(1.07)	(1.14)	(1.04)	[0.11]	[0.11]	[0.10]
Among Blacks	2.40	1.82	1.33	0.58*	1.06***	0.48**
Among Hispanics	3.31	2.57	1.97	0.74**	1.34***	0.60
Among Whites	2.48	2.40	2.34	0.07	0.13	0.06
Among <$30,000	2.67	2.35	2.26	0.33	0.41*	0.09
Among $100,000+	2.02	2.32	2.03	−0.30	−0.01	0.29
Cause 3: Parenting	2.75	2.48	2.39	0.27***	0.36***	0.09
	(0.97)	(1.18)	(1.14)	[0.10]	[0.11]	[0.11]
Among Blacks	2.17	1.66	1.52	0.51	0.66**	0.14
Among Hispanics	2.88	2.70	2.01	0.19	0.87***	0.69**
Among Whites	2.78	2.60	2.60	0.18*	0.19*	0.01
Among <$30,000	2.54	2.22	2.18	0.31	0.36*	0.04
Among $100,000+	2.83	2.82	2.59	0.02	0.25	0.23
Cause 4: Genetic differences	1.70	1.51	1.52	0.19*	0.18*	−0.01
	(0.96)	(0.89)	(0.86)	[0.10]	[0.10]	[0.08]
Among Blacks	1.72	1.78	1.61	−0.06	0.11	0.17
Among Hispanics	1.91	1.91	1.69	0.00	0.23	0.23
Among Whites	1.61	1.46	1.47	0.14*	0.14	0.00
Among <$30,000	1.89	1.63	1.45	0.26	0.44**	0.18
Among $100,000+	1.33	1.44	1.32	−0.11	0.00	0.12
No. of observations	317	353	319	670	636	672

Note. Each coefficient represents the results of a separate regression model run for the treatment groups, sample, and outcome variable specified. Standard deviations appear in parentheses and standard errors appear in brackets. Responses were given on the following scale: 1 = none, 2 = a little, 3 = some, and 4 = a great deal. The question stem was “How much of the difference in test scores between [students from wealthy families and students from poor families / White students and Black students / White students and Hispanic students] . . . .” For Explanation 1, the item read, “. . . can be explained by discrimination against [the poor / Blacks / Hispanics] or injustices in society?” For Explanation 2, it read, “. . . occurs because most [students from poor families / Black students / Hispanic students] just don’t have the motivation or will power to perform well?” For Explanation 3, it read, “. . . occurs because most [poor people / Blacks / Hispanics] do not teach their children the values and skills that are required to be successful in school?” For Explanation 4, it read, “. . . occurs because of fundamental genetic differences between [students from wealthy families and students from poor families / White students and Black students / White students and Hispanic students]?

p < .1. **p < .05. ***p < .01.

First, respondents in the poor-wealthy group attributed more of the gap to each cause presented than did respondents in the Black-White and Hispanic-White groups. This was true for parenting (2.75 for poor-wealthy, 2.48 for Black-White, and 2.39 for Hispanic-White), student motivation (2.66, 2.34, and 2.18), discrimination and injustice (2.48, 1.97, and 1.95), and genetic differences (1.70, 1.51, and 1.52). All of these differences between the poor-wealthy and Black-White groups and between the poor-wealthy and Hispanic-White groups are significant at least at the 10% level. This could indicate that people find the causes of the poor-wealthy gap less mysterious than the causes of the Black-White and Hispanic-White gaps. Alternatively, it could indicate that they find explanations for the Black-White and Hispanic-White gaps not provided in this survey more plausible, although that was not evident in pretesting.

Second, for each group, there was a tendency to attribute the gaps, in descending order, to parenting, student motivation, discrimination and injustice, and genetic differences. Perhaps most notable, almost half of the Black-White and Hispanic-White groups (44% of each) said that none of the test score gaps could be explained by discrimination or injustice in society. Only about 10% of each group attributed a great deal of the gap to these causes. The responses were more balanced in the poor-wealthy group, with 22% attributing none and 21% attributing a great deal of the poor-wealthy test score gap to discrimination or injustice.

Our final analyses appear in Table 5. Here, we examined predictors of the importance that respondents assigned to closing test score gaps. The first two specifications focus on respondents’ demographic and political characteristics. We see, for example, that Democrats assigned a priority level 0.82 points higher than Republicans to closing test score gaps (p < .01), controlling for the other variables in the model. Hispanic and Black respondents assigned higher priority levels than White respondents. The third and fourth specifications offer our most direct tests of the in-group preferences hypothesis. We tested interactions between respondents’ gap treatment conditions and their membership in the lower-scoring group represented in that treatment condition. We see a substantively large, marginally significant interaction effect for Black respondents asked about the Black-White test score gap. We also see a large, significant interaction effect for low-income respondents asked about the poor-wealthy gap. The interaction between Hispanic and the Hispanic-White treatment condition is not statistically significant, although this could be the product of a small sample and the resulting large standard error. Taken together, these results are suggestive of in-group preferences as they relate to closing test score gaps.

Table 5

Predictors of the Importance of Closing Test Score Gaps

Variable	(1)	(2)	(3)	(4)	(5)	(6)
Treatment conditions
Poor-wealthy	3.49***	3.08***	3.42***	3.04***	2.65***	2.56***
	[0.20]	[0.21]	[0.20]	[0.20]	[0.25]	[0.24]
Black-White	2.87***	2.46***	2.85***	2.44***	2.23***	2.13***
	[0.21]	[0.21]	[0.21]	[0.21]	[0.23]	[0.23]
Hispanic-White	2.66***	2.30***	2.70***	2.33***	2.04***	1.95***
	[0.21]	[0.21]	[0.21]	[0.21]	[0.22]	[0.22]
Covariates
Black, non-Hispanic	0.58***	0.26	0.32	−0.01	0.34*	0.25
	[0.20]	[0.20]	[0.26]	[0.25]	[0.19]	[0.19]
Hispanic	0.56***	0.48***	0.47**	0.35*	0.48***	0.46***
	[0.19]	[0.18]	[0.22]	[0.20]	[0.16]	[0.16]
Other race	0.11	0.10	0.10	0.10	0.04	0.04
	[0.17]	[0.17]	[0.17]	[0.17]	[0.17]	[0.17]
Income: <$30,000	0.23	0.28*	0.01	0.10	0.20	0.23
	[0.16]	[0.15]	[0.18]	[0.17]	[0.15]	[0.15]
Income: $30,000 to $99,999	0.02	0.07	0.02	0.07	0.06	0.08
	[0.14]	[0.13]	[0.14]	[0.13]	[0.13]	[0.12]
Income: Prefer not to say	0.04	0.09	0.03	0.08	0.11	0.13
	[0.17]	[0.16]	[0.17]	[0.16]	[0.15]	[0.15]
Female	0.17*	0.08	0.19**	0.10	0.12	0.09
	[0.09]	[0.09]	[0.09]	[0.09]	[0.09]	[0.09]
Age	0.00	0.00	0.00	0.00	0.00	0.00
	[0.00]	[0.00]	[0.00]	[0.00]	[0.00]	[0.00]
No college	−0.27**	−0.21*	−0.25**	−0.19	−0.14	−0.13
	[0.12]	[0.12]	[0.12]	[0.12]	[0.11]	[0.11]
Some college/2-year degree	−0.15	−0.10	−0.15	−0.11	−0.03	−0.02
	[0.11]	[0.11]	[0.11]	[0.11]	[0.10]	[0.10]
Politics: Leans Democrat		0.82***		0.81***		0.36***
		[0.10]		[0.10]		[0.10]
Politics: Independent (no lean)		0.22*		0.21		0.10
		[0.13]		[0.13]		[0.12]
Politics: Not reported		0.18		0.17		0.02
		[0.38]		[0.32]		[0.38]
Interactions (in-group preferences)
Black × Black-White Condition			0.70*	0.72*
			[0.36]	[0.38]
Hispanic × Hispanic-White Condition			0.32	0.43
			[0.34]	[0.35]
Low Income × Poor-Wealthy Condition			0.51**	0.39*
			[0.21]	[0.20]
Explanations of gap
Discrimination/injustice					0.49***	0.43***
					[0.04]	[0.04]
Student motivation					−0.05	−0.05
					[0.06]	[0.05]
Parenting					−0.04	−0.02
					[0.05]	[0.05]
Genetic differences					−0.16***	−0.15***
					[0.05]	[0.05]
No. of observations	990	990	990	990	976	976

Note. Each column represents results from a separate ordinary least squares regression model. Standard errors appear in brackets. Omitted groups are White/non-Hispanic (for race/ethnicity), $100,000+ (family income), male, completed 4-year degree (education), and leans Republican (political identification). The constant is suppressed, with all treatment conditions included, to facilitate interpretation of the interaction coefficients.

p < .1. **p < .05. ***p < .01.

The last two specifications in Table 5 show the relationship between respondents’ beliefs about the causes of test score gaps and their desires to close those gaps. A 1-point increase in how much respondents attributed a test score gap to discrimination or injustice is associated with rating the importance of closing that gap 0.4 to 0.5 points higher (p < .01). This is true even controlling for the demographic and political variables in the model, although we refrain from interpreting it as a causal relationship. For example, it may be that the types of people who perceive wealth- and race-based injustice in the world are the types who care most about closing achievement gaps and not necessarily that convincing people that gaps result from discrimination or injustice will lead them to care more about closing those gaps.

Discussion

The public’s beliefs about social issues can shape policymakers’ incentives and the policies that result. Understanding the public’s views on education is perhaps especially critical, since focusing events in education are rare. Few events in schools can galvanize the public to demand policy changes the way a plane crash might focus attention on air safety or a terrorist attack on national security. Rather, education ceaselessly competes for policymakers’ attention, adding urgency to identifying the political dynamics that underlie education issues.

This paper examines the politics of one of the most foundational, vexing issues in education today: test score gaps between historically advantaged and disadvantaged populations. Using randomized survey experiments with a nationally representative sample, we find sharp differences in the U.S. public’s attitudes toward different types of gaps. Americans are much more concerned about wealth-based gaps than race- and ethnicity-based gaps, and they express greater support for proposals aimed at narrowing the former.

Our results suggest that there may be greater political will for policies targeting poor-wealthy gaps than Black-White or Hispanic-White gaps. With public opinion being an important determinant of policymakers’ agendas and incentives (Burstein, 2003), it might seem puzzling that education reformers often describe U.S. educational inequity in terms of race-based gaps rather than wealth-based gaps. Although it is tempting to interpret this study’s findings to mean that reformers would enjoy more favorable political conditions if they simply recast, or rename, the Black-White gap as a poor-wealthy gap, we see problems with this interpretation. Chief among them, it neglects that race- and wealth-based gaps have different origins, patterns, and implications, with poverty and racial discrimination presenting distinct challenges. However, it is possible that race-neutral initiatives that disproportionately benefit minority children are more politically viable than initiatives that overtly favor racial or ethnic minorities.

We see evidence of in-group preferences, with respondents especially concerned about the test score gaps confronting the groups to which they belong. This could have long-term political implications, since demographic change might yield changes in U.S. public opinion toward educational equity. Relative to others in our study, Black respondents were particularly concerned about closing the Black-White gap, and low-income respondents were particularly concerned about closing the poor-wealthy gap. Our finding that Americans, in general, are most concerned about wealth-based gaps might also reflect in-group preferences, since the relative permeability of wealth groups could mean that even nonpoor Americans have been—or could imagine becoming—members of economically disadvantaged groups. The homophily of social networks might play a role, as well, with race and ethnicity “clearly the biggest divide in social networks today in the United States” (McPherson et al., 2001, p. 420) and people perhaps more sympathetic to the challenges facing their peers.

However, other explanations are possible. One possibility is that Americans are more willing to prioritize closing wealth-based gaps because they better understand the relationship between wealth and academic performance or because they better understand wealth-based gaps more generally. An intriguing finding from this study is that respondents asked about poor-wealthy gaps more readily attributed those gaps to all of the explanations presented (discrimination and injustice, student motivation, parenting, and genetic differences). Although it is possible that respondents’ explanations for race-based gaps simply did not appear in our survey, it also seems likely that respondents, for whatever reason, believe they have a better understanding of what causes poor-wealthy test score gaps. Believing one better understands the cause of a problem may have implications for how willing one is to do something about it.

Another notable finding is that large percentages of respondents, and especially those asked about Black-White and Hispanic-White gaps, attributed little or none of these gaps to discrimination and injustice. Future work in this area might examine whether support for equity-focused education policies increases if more people are persuaded that discrimination and injustice play prominent roles in creating these gaps.

Illuminating the political dynamics of test score gaps can improve our understanding of the paths available to improve educational equity. Education researchers have studied these gaps carefully and examined promising strategies for closing them. Many of these strategies require action from policymakers, and many of these policymakers are publicly elected officials with incentives to satisfy the desires of their constituents. There are, of course, important considerations besides political feasibility in deciding how the discussion of educational inequality should be approached. But insofar as Americans are more concerned about wealth-based gaps than race- and ethnicity-based gaps, there could be greater political will—or less political resistance—for educational interventions perceived as beneficial to students in poverty.

Footnotes

Appendix

Table A4

Distribution of Responses to Questions About the Causes of Test Score Gaps (in percentages)

	Poor-Wealthy				Black-White				Hispanic-White
Cause	None	A Little	Some	A Great Deal	None	A Little	Some	A Great Deal	None	A Little	Some	A Great Deal
Cause 1: Discrimination/injustice	21.7	29.7	27.3	21.3	44.0	25.2	20.6	10.1	43.8	27.3	19.6	9.4
Among Blacks	12.8	11.9	23.3	52.0	23.5	33.6	24.6	18.3	25.3	42.8	13.8	18.1
Among Hispanics	8.3	50.2	24.7	16.9	32.3	13.4	27.9	26.4	33.1	13.7	40.3	12.9
Among Whites	27.1	26.5	27.6	18.8	47.7	24.8	19.4	8.1	50.0	25.7	17.2	7.1
Among <$30,000	17.1	18.5	32.0	32.4	40.8	26.6	25.2	7.4	53.2	16.3	19.5	11.0
Among $100,000+	44.3	14.8	25.4	15.6	35.3	26.4	32.4	5.9	34.9	31.6	20.5	13.0
Cause 2: Student motivation	21.6	18.5	31.7	28.2	29.9	26.0	24.4	19.7	33.6	27.8	25.4	13.2
Among Blacks	31.9	9.0	46.6	12.5	41.0	39.8	15.9	3.3	77.3	11.8	10.8	0.0
Among Hispanics	10.6	4.1	29.1	56.2	25.0	10.4	47.5	17.2	27.5	48.9	22.7	1.0
Among Whites	24.1	24.9	30.2	20.8	28.1	25.2	25.1	21.6	28.6	25.8	28.5	17.1
Among <$30,000	22.0	19.9	27.0	31.0	28.9	26.2	26.4	18.5	28.5	32.7	23.3	15.5
Among $100,000+	44.2	21.2	23.0	11.7	40.5	12.5	21.0	26.0	40.8	22.7	29.0	7.6
Cause 3: Parenting	13.3	25.7	33.8	27.3	27.4	23.2	23.3	26.1	30.5	22.4	24.4	22.8
Among Blacks	28.0	43.0	12.9	16.2	65.4	13.3	11.3	10.0	73.2	11.8	5.2	9.8
Among Hispanics	3.8	21.4	57.9	16.9	13.5	16.9	56.2	13.4	32.7	41.8	17.3	8.1
Among Whites	14.1	25.6	28.3	32.0	21.5	25.9	23.7	29.0	24.8	19.1	27.8	28.3
Among <$30,000	20.1	25.0	36.2	18.8	34.3	29.2	16.5	20.1	34.7	27.6	22.8	14.9
Among $100,000+	12.8	25.5	27.2	34.5	20.6	16.1	24.3	39.0	26.4	12.4	37.2	24.0
Cause 4: Genetic differences	60.6	16.8	15.0	7.6	69.2	15.0	11.3	4.5	68.4	15.1	12.4	4.1
Among Blacks	48.9	34.3	13.1	3.7	47.8	30.8	17.1	4.3	54.0	34.2	8.9	2.9
Among Hispanics	52.8	17.0	16.3	14.0	52.5	12.2	26.9	8.5	65.3	6.1	23.0	5.6
Among Whites	64.9	15.4	13.8	5.9	72.2	13.6	9.6	4.6	71.6	13.6	11.3	3.6
Among <$30,000	56.7	12.6	15.7	15.0	57.1	26.8	12.2	3.9	72.8	12.4	12.3	2.6
Among $100,000+	78.5	10.2	11.3	0.0	78.5	8.1	4.2	9.2	80.8	7.6	10.3	1.4
No. of observations	317				353				319

Note. Table shows the percentage of respondents providing each answer. See Table 4 for full text of items.

Notes

Authors

JON VALANT, PhD, is a postdoctoral fellow in the Department of Economics at Tulane University, 1555 Poydras Street, Suite 700, New Orleans, LA 70112; jvalant@tulane.edu . His research focuses on K-12 education policy and politics.

DANIEL A. NEWARK, PhD, is an assistant professor in the Strategic Organization Design unit in the Department of Marketing and Management at the University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark; dnew@sam.sdu.dk . His research interests include decision making, identity, and prosocial behavior.

References

Adamson

Darling-Hammond

(2012). Funding disparities and the inequitable distribution of teachers: Evaluating sources and solutions. Education Policy Analysis Archives, 20(37). Retrieved from http://epaa.asu.edu/ojs/article/view/1053

Burstein

(2003). The impact of public opinion on public policy: A review and an agenda. Political Research Quarterly, 56(1), 29–40.

Bushaw

W. J.

Lopez

S. J.

(2012). Public education in the United States: A nation divided. Phi Delta Kappan/Gallup. Retrieved from http://pdkintl.org/wp-content/blogs.dir/5/files/2012-Gallup-poll-full-report.pdf

Clinton

J. D.

Grissom

J. A.

(2015). Public information, public learning and public opinion: Democratic accountability in education policy. Journal of Public Policy, 35(3), 355–385.

Cohn

(2014, July 27). Explaining online panels and the 2014 midterms. The New York Times. Retrieved from http://www.nytimes.com/2014/07/28/upshot/explaining-online-panels-and-the-2014-midterms.html

Coleman

J. S.

(1966). Equality of educational opportunity. Washington, DC: National Center for Educational Statistics.

Ellemers

van Knippenberg

Wilke

(1990). The influence of permeability of group boundaries and stability of group status on strategies of individual mobility and social change. British Journal of Social Psychology, 29(3), 233–246.

Erikson

R. S.

Wright

G. C.

McIver

J. P.

(1993). Statehouse democracy: Public opinion and policy in the American states. Cambridge, UK: Cambridge University Press.

Feldman

Huddy

(2005). Racial resentment and white opposition to race-conscious programs: Principles or prejudice? American Journal of Political Science, 49(1), 168–183.

10.

Feldman

Huddy

(2010). The structure of white racial attitudes. Paper presented at the annual meeting of the American Political Science Association, Washington, DC. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1643879

11.

Goldhaber

Lavery

Theobald

(2015). Uneven playing field? Assessing the teacher quality gap between advantaged and disadvantaged students. Educational Researcher, 44(5), 293–307.

12.

Henderson

M. B.

Peterson

P. E.

West

M. R.

(2016). The 2015 EdNext poll on school reform. Education Next, 16(1), 8–20.

13.

Heuer

Stullich

(2011). Comparability of state and local expenditures among schools within districts: A report from the study of school-level expenditures. Washington, DC: U.S. Department of Education. Retrieved from http://www2.ed.gov/rschstat/eval/title-i/school-level-expenditures/school-level-expenditures.pdf

14.

Hewstone

Rubin

Willis

(2002). Intergroup bias. Annual Review of Psychology, 53(1), 575–604.

15.

Howell

W. G.

West

M. R.

Peterson

P. E.

(2007). What Americans think about their schools: The 2007 Education Next-PEPG survey. Education Next, 7(4), 12–26.

16.

Huddy

(2013). From group identity to political cohesion and commitment. In Huddy

Sears

D. O.

Levy

(Eds.), Oxford handbook of political psychology (pp. 737–773). New York, NY: Oxford University Press.

17.

Jacobsen

Snyder

J. W.

Saultz

(2014). Informing or shaping public opinion? The influence of school accountability data format on public perceptions of school quality. American Journal of Education, 121(1), 1–27.

18.

Jones

J. M.

(2014). Record-high 42% of Americans identify as Independents. Washington, DC: Gallup. Retrieved from http://www.gallup.com/poll/166763/record-high-americans-identify-independents.aspx

19.

Kahlenberg

R. D.

(1996). The remedy: Class, race, and affirmative action. New York, NY: Basic Books.

20.

Kingdon

J. W.

(1984). Agendas, alternatives, and public policies. Boston, MA: Little, Brown.

21.

Krysan

(2000). Prejudice, politics, and public opinion: Understanding the sources of racial policy attitudes. Annual Review of Sociology, 26, 135–168.

22.

Ladson-Billings

(2013). Lack of achievement or loss of opportunity? In Carter

P. L.

Welner

K. G.

(Eds.), Closing the opportunity gap (pp. 11–22). New York, NY: Oxford University Press.

23.

Lankford

Loeb

Wyckoff

(2002). Teacher sorting and the plight of urban schools: A descriptive analysis. Educational Evaluation and Policy Analysis, 24(1), 37–62.

24.

MacDonald

V. M.

(2013). Demanding their rights: The Latino struggle for educational access and equity. In American Latinos and the making of the United States: A theme study. Washington, DC: National Park Service.

25.

McPherson

Smith-Lovin

Cook

J. M.

(2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415–444.

26.

Mollica

K. A.

Gray

Trevino

L. K.

(2003). Racial homophily and its persistence in newcomers’ social networks. Organization Science, 14(2), 123–136.

27.

Mutz

D. C.

(2011). Population-based survey experiments. Princeton, NJ: Princeton University Press.

28.

National Constitution Center. (2009). A more perfect union: A virtual exhibit of Barack Obama’s race speech at the Constitution Center on March 18, 2008. Retrieved from http://constitutioncenter.org/amoreperfectunion/

29.

Orfield

Eaton

S. E.

(1996). Dismantling desegregation: The quiet reversal of Brown v. Board of Education. New York, NY: New Press.

30.

Page

B. I.

Shapiro

R. Y.

(1983). Effects of public opinion on policy. American Political Science Review, 77(01), 175–190.

31.

Peterson

P. E.

Henderson

West

M. R.

(2014). Teachers versus the public. Washington, DC: Brookings Institution Press.

32.

Reardon

S. F.

(2011). The widening academic achievement gap between the rich and the poor: New evidence and possible explanations. In Duncan

G. J.

Murnane

R. J.

(Eds.), Whither opportunity? Rising inequality, schools, and children’s life chances (pp. 91–116). New York, NY: Russell Sage Foundation.

33.

Rivers

(2007). Sampling for web surveys. Paper presented at the 2007 Joint Statistical Meetings, Salt Lake City, UT. Retrieved from http://www.laits.utexas.edu/txp_media/html/poll/files/Rivers_matching.pdf

34.

Rivers

(2012). YouGov poll performance in the 2012 U.S. elections. Retrieved from https://today.yougov.com/news/2012/11/07/yougov-poll-performance-2012-us-elections/

35.

Silver

(2012, November 10). Which polls fared best (and worst) in the 2012 presidential race. The New York Times. Retrieved from http://fivethirtyeight.blogs.nytimes.com/2012/11/10/which-polls-fared-best-and-worst-in-the-2012-presidential-race/

36.

Stepler

Brown

(2015). Statistical portrait of Hispanics in the United States, 1980-2013. Washington, DC: Pew Research Center. Retrieved from http://www.pewhispanic.org/2015/05/12/statistical-portrait-of-hispanics-in-the-united-states-2013-key-charts

37.

Tyack

D. B.

(1974). The one best system: A history of American urban education. Cambridge, MA: Harvard University Press.

38.

U.S. Census Bureau. (2013a). Current population survey 2014: Annual social and economic supplement. Retrieved from http://www.census.gov/hhes/www/cpstables/032014/faminc/finc01_000.htm

39.

U.S. Census Bureau. (2013b). Educational attainment in the United States: 2013. Detailed tables. Retrieved from http://www.census.gov/hhes/socdemo/education/data/cps/2013/tables.html

40.

U.S. Census Bureau. (2013c). National characteristics: Vintage 2013. Retrieved from http://www.census.gov/popest/data/national/asrh/2013/index.html

41.

Welner

K. G.

Carter

P. L.

(2013). Achievement gaps arise from opportunity gaps. In Carter

P. L.

Welner

K. G.

(Eds.), Closing the opportunity gap (pp. 1–10). New York, NY: Oxford University Press.

42.

White House. (2008). Interview excerpts of President Bush and First Lady Laura Bush by Doro Bush Koch for StoryCorps. Retrieved from http://georgewbush-whitehouse.archives.gov/news/releases/2008/11/20081128.html

43.

Williams

H. A.

(2005). Self-taught: African American education in slavery and freedom. Chapel Hill: University of North Carolina Press.

44.

Wilson

W. J.

(1999). The bridge over the racial divide. Berkeley: University of California Press.

45.

Wilson

W. J.

(2012). The truly disadvantaged: The inner city, the underclass, and public policy (2nd ed.). Chicago, IL: University of Chicago Press.

46.

YouGov. (2015). YouGov omnibus. Retrieved from https://today.yougov.com/find-solutions/omnibus/?page=12