Polarization and the Politics of Education: What Moves Partisan Opinion?

Abstract

This study explores the conditions under which polarization and depolarization occur with respect to public opinion on education issues. I analyze experimental and nonexperimental survey data to gauge the effects of informational and experiential stimuli on education policy preferences. I focus on heterogeneous effects by political party identification that result in increased or decreased polarization. I find evidence that the provision of education spending data has depolarizing consequences, but the effects of ideologically moderate policy endorsements vary by issue and year. I also find tentative evidence in favor of a link between having children in the household and reduced polarization.

Keywords

education policy politics of education public opinion polarization survey experiments

Perhaps the single most prominent trend in contemporary American politics has been the surge in polarization over the past two decades (Abramowitz, 2010; Bafumi & Shapiro, 2009; Fiorina, Abrams, & Pope, 2010; McCarty, Poole, & Rosenthal, 2006; Pew Research Center, 2014). Across an array of issues, Democrats and Republicans have grown increasingly distinct. Compromise across party lines is often seen as a betrayal of ideological principles and antithetical to future electoral success. The typical portrayal of federal policymaking is one of excessive gridlock and dysfunction. Yet, it is in this context that two of the most significant federal education laws, the No Child Left Behind Act (NCLB) of 2001 and the Every Student Succeeds Act (ESSA) of 2015, were passed in an overwhelmingly bipartisan fashion (Henig, Houston, & Lyon, 2017; McGuinn, 2006). Among ordinary citizens, there are notable differences in public opinion on various education issues between Democrats and Republicans, but these differences tend to be smaller than the partisan gaps on issues in other high-profile policy domains (Peterson, Henderson, & West, 2014; Shapiro, Kilibarda, Sinozich, & McClellan, 2016). Although public feuds over education issues may be rhetorically red in tooth and claw, federal education policy, thus far, has largely escaped the pattern of sharp partisan polarization that has embroiled lawmaking in other arenas such as healthcare, gun control, reproductive rights, the environment, and immigration.

This article explores the conditions under which polarization and depolarization occur with respect to public opinion on education issues. I examine the effects of (a) the availability of education spending data, (b) the provision of ideologically moderate policy endorsements from political elites, and (c) having children in the household, focusing on heterogeneous effects by party identification that result in either increased or decreased polarization. I analyze 13 survey experiments from 2008 to 2014 to identify the polarizing or depolarizing effects of providing education spending data (average districtwide per pupil expenditures and average statewide teacher salaries). I analyze five survey experiments from 2009 to 2010 to identify the polarizing or depolarizing effects of providing ideologically moderate policy endorsements (former President Obama’s positions on charter schools, merit-based teacher pay, school accountability reform, and standardized testing). I also conduct an analysis on nonexperimental survey data from 2008 to 2014 to capture the observational relationship between having children in the household and polarization. My objective is to identify the kinds of informational and experiential stimuli that are capable of affecting polarization on matters related to the American K-12 public school system.

Literature Review

Education Policy and Political Polarization

Debates on topics that involve kids and schools have always been heated and can often be quite acrimonious. Battles over education policy inevitably invoke some of the great fault lines of American politics: economic inequality, racial injustice, the role of the federal government, global competitiveness, and even national security (one of the first federal forays into education policy took the form of the National Defense Education Act of 1958, which directed funds to state and local school systems to expand math, science, and foreign language instruction in response to Sputnik and the growing apprehension that the Soviet Union might win the space race). However, the primary federal K-12 education law, the Elementary and Secondary Education Act, was passed in 1965 and reauthorized in 1968, 1974, 1978, 1981, 1988, 1994, 2001, and 2015—the last two reauthorizations are better known as NCLB and ESSA—with large bipartisan margins under both Democratic and Republican presidents (McGuinn, 2017).

This is not to suggest that we live in an age of widespread consensus on issues of education. Rather, the disagreements that define contemporary education policy debates tend to cut across rather than between the two major political parties. For example, in the negotiations that ultimately resulted in the passage of NCLB, the business community on the political right and a subset of the civil rights movement on the political left became unlikely allies in the pursuit of an accountability system that would raise educational standards across the board along with increased pressure to reduce long-standing racial achievement gaps. On the other side, states’ rights activists who resisted the expansion of the federal role in education found unexpected partners in the teachers’ unions, who held that the rise in standardized testing stifled and undermined the teaching profession (McGuinn, 2006; Rhodes, 2012). Moreover, the rhetorical shift in federal education policy from “resources” and “equity” in the 1960s and 1970s to “accountability” and “excellence” in the 1980s and 1990s created new opportunities for provisional partnerships between Democrats and Republicans (Wolbrecht & Hartney, 2014). These scrambled allegiances provided political cover for partisans across the political spectrum to support legislation that left their constituencies divided but not wholly opposed (Henig et al., 2017).

Mechanisms for Polarization and Depolarization

In this study, I explore whether partisan polarization on education issues is affected by the availability of policy information (data on education spending levels), the provision of ideologically moderate policy endorsements from political elites (a subset of Obama’s positions on education), and the presence of children in the household.

Policy information

There is some evidence generated by both political scientists and psychologists that higher levels of policy knowledge can increase polarization between Democrats and Republicans. These researchers suggest that partisans tend to learn new information about policies and politics in politically conditioned ways. In their classic text, The American Voter, Campbell, Converse, Miller, and Stokes (1960) suggest that “identification with a party raises a perceptual screen through which the individual tends to see what is favorable to his partisan orientation. The stronger the party bond, the more exaggerated the process of selection and perceptual distortion will be” (p. 133). Years later, Zaller (1992) echoes this sentiment, arguing that politically engaged individuals develop a form of “partisan resistance” (p. 241) that allows them to filter out evidence that contrasts with their prior ideological commitments, generating two distinct political camps that diverge in response to new information. In this section, I briefly consider some of the academic literature on so-called biased information processing and its effects on polarization.

Lord, Ross, and Lepper’s (1979) canonical study on biased information processing describes a demonstration wherein participants with differing views on the deterrent effects of the death penalty are provided evidence both for and against deterrence. Participants tend to rate evidence that aligns with their preexisting positions as more rigorous and convincing, and they tend to apply heightened scrutiny to evidence that contradicts their prior beliefs. After assessing the available evidence, the opposing camps’ views on the deterrent effects of the death penalty diverge even further. Rather than producing consensus, exposure to a mixed body of evidence on a topic actually increases polarization. Later laboratory experiments replicate this general finding: participants with differing prior beliefs on a given topic can become more entrenched in their views upon being exposed to new information on that topic, producing even more severe polarization (Taber, Cann, & Kucsova, 2009; Taber & Lodge, 2006). Even when partisan opinions move in the same direction, Democrats and Republicans tend to incorporate new information into their belief systems at different rates, depending on the extent to which the information conforms with their political predispositions. For example, both Democrats and Republicans acquire new information upon reading relevant newspaper articles, but partisans are more likely to answer factual questions correctly when the information comes in the form of an article that portrays their party in a positive light (Jerit & Barabas, 2012).

In contrast to the laboratory experiments that document the polarizing effects of policy information, longitudinal public opinion survey data from the mid-20th century often reveal roughly parallel partisan responses to events and new information. Evaluations of the president, party competence, and the strength of the economy typically track changes in economic conditions, and these evaluations tend to shift in the same direction for both Democrats and Republicans (Gerber & Green, 1999; Green, Schickler, & Palmquist, 2002). However, this pattern of seemingly unbiased information processing observable in longitudinal surveys may be changing in the context of our increasingly polarized politics. In their analysis of aggregate public opinion trends from the 1940s to the 1980s, Page and Shapiro (1992) document parallel movement among most social and economic groups (e.g., although Whites and Blacks may differ on average with respect to many policies, support for those policies tends to rise and fall for both groups at the same time—a phenomenon that Page and Shapiro refer to as “parallel publics”). In the last few decades, however, there now appears to be one exception to these parallel publics. On many issues, public opinion among Democrats and Republicans has begun to diverge over time, potentially indicating differential partisan responses to the ongoing flow of events and information (Shapiro & Bloch-Elkon, 2006).

In the field of education, Reckhow, Grossman, and Evans (2015) conduct a series of survey experiments in which they provide additional information about Michigan charter schools to subsets of their participant pool. They note that self-identified liberals and conservatives respond in similar ways to information about some charters’ for-profit status and information about universities’ roles as charter authorizers. However, they also note that conservatives are even more likely to support charter schools upon learning about charters’ tendency to hire nonunion teachers, thereby increasing polarization on the issue (although that piece of information may be unusually likely to evoke ideological considerations). Schueler and West (2016) conduct a set of survey experiments on the effects of providing current education spending data on education spending preferences. They find that information on current spending levels tends to reduce participants’ support for increased education spending, but their study does not report heterogeneous treatment effects by party identification. It is not clear whether education spending preferences diverge in response to new information on education spending levels, or if better informed partisans are more likely to converge toward the same position.

Ideologically moderate policy endorsements

Political scientists have long recognized the effects of cues such as party memberships, political group affiliations, and endorsements on individuals’ attitudes and behaviors (e.g., Downs, 1957; Lupia & McCubbins, 1998; Popkin, 1991). Traditional scholarship typically holds up heuristics and cue-giving as boons to democracy, making democratic responsiveness possible even in the absence of a highly active and informed citizenry. However, more recent scholars have expressed concern over the extent to which citizens’ positions on questions of government policy are being driven primarily by partisan cues and political endorsements rather than by information on the features and potential consequences of the policies themselves. In a series of lab experiments, Cohen (2003) demonstrates that, in the absence of an endorsement by the Democratic or Republican party, participants tend to evaluate a policy based on its content and their personal ideological beliefs; however, when provided a partisan endorsement for or against the policy, participants tend to forgo an equivalent evaluation and merely adopt the views of their party. The power of such cues may be even more severe in highly polarized contexts. Druckman, Peterson, and Slothuus (2013) randomly assign party endorsements on two contested issues (drilling for oil and gas and immigration status for undocumented children brought to the United States in their youth) as well as statements about whether the parties are starkly split on these issues or not. They find that the effects of party endorsements are larger when the debates are described as more polarized. In contrast, there may be limits to the extent to which party cues can influence individuals’ policy preferences. Bullock (2011) argues that the provision of information that signals the ideological content of a policy can have a larger effect on attitudes toward that policy than the provision of the two major parties’ positions.

These studies suggest that party cues can be a powerful and potentially misleading source of information about the merits of specific policies. The implication here is that elite signals often serve to reinforce ideological dogmatism and partisan division at the expense of more substantive and pragmatic evaluations of policies. But what if, as in the case of education policy, party lines are less rigid, and elite actors occasionally adopt positions at odds with their political bases? The internal divisions within parties over education policy—between the business community and states’ rights activists on the right and between some civil rights groups and teachers’ unions on the left—often result in prominent partisans taking positions that are ideologically atypical, such as George W. Bush’s embrace of a strong federal role in education or Barack Obama’s support for charter schools and merit-based teacher salaries (McGuinn, 2006; Rhodes, 2012). For partisans following the policy cues of educational centrists such as Bush and Obama, these elite signals may produce less calcified party lines on the issues.

Children in the household

Public opinion surveys have consistently demonstrated that individuals with closer ties to the public schools tend to have different attitudes toward them. These differences could plausibly influence the degree of partisan polarization on education issues. The Phi Delta Kappa/Gallup Poll of the Public’s Attitudes Toward the Public Schools and the Harvard University Education Next Survey reliably show that satisfaction with public education increases along with familiarity with the schools themselves. Parents tend to be far more enthusiastic than the general public about the quality of local schools, and parents tend to grade their own children’s schools more generously still (Bali, 2016; Bushaw & Lopez, 2011; Cannon & Barham, 1993; Elam, 1995; Loveless, 1997; Peterson et al., 2014). Moreover, having children of your own not only affects perceptions of school quality, it also appears to be linked to a stronger relationship between perceived quality and official measures of school performance. Parents are considerably more likely than their counterparts without school-aged children to hold opinions of their local schools that reflect recent school accountability grades (Chingos, Henderson, & West, 2012). The presence of children at home also likely alters individuals’ ideas about self-interest, increasing the importance they place on educational resources for themselves and their families. Newly salient concerns about the structure and quality of local schooling may outweigh competing partisan or ideological considerations.

Method

In this study, I analyze experimental and nonexperimental survey data to gauge the effects of (a) the availability of policy information (data on education spending levels), (b) the presence of ideologically moderate policy endorsements (a subset of Obama’s positions on education), and (c) having children in the household on the level of political polarization on education issues.

The proposed link between the availability of education spending data and polarization lends itself to a straightforward empirical test: a survey experiment in which information on education spending is randomly assigned prior to the polling of attitudes on education issues. This approach would allow the researcher to identify whether or not there are heterogeneous effects of such information on policy preferences by party identification, revealing its polarizing or depolarizing consequences.

To test whether or not partisans converge on similar positions after being exposed to ideologically moderate policy endorsements, one could employ another survey experiment in which the provision of endorsements by centrist politicians would be randomly assigned prior to the measurement of individuals’ education policy preferences. Much like the exploration of the effects of providing education spending data described above, the researcher would again be interested to observe whether or not these policy endorsements produce heterogeneous treatment effects that would attenuate or exacerbate the partisan gaps on a series of education issues.

With respect to the proposed link between having children at home and polarization, there is no simple and inexpensive experiment that is able to randomly assign kids to households. However, there is a great deal of observable variation with respect to children in the home. As a first step, a researcher could identify a correlational link between having children in the household and attitudes toward education issues. The analytic approach would be quite similar to the experiments sketched above, even if the findings would not authoritatively capture a causal relationship. The researcher would be primarily interested in the interaction between party identification and having children at home. The question is whether the relationships between children in the household and attitudes toward education issues vary by political party, resulting in smaller partisan gaps on those issues among individuals with kids.

Research Questions

Research Question 1: Does the availability of education spending data affect political polarization on education issues?

Research Question 2: Do ideologically moderate policy endorsements affect political polarization on education issues?

Research Question 3: Does having children in the household affect political polarization on education issues?

Data Sources

My data come from the 2008-2014 editions of the Education Next Survey, an annual online poll of education policy preferences designed by researchers at the Harvard University Program on Education Policy and Governance and conducted by Knowledge Networks. Knowledge Networks maintains a nationally representative panel of adults, employing sampling frames that include more than 99% of the U.S. population. Participants are recruited via list-assisted random digit dialing techniques. To contact potential participants, Knowledge Networks sends out an initial mailer and follows up with up to 15 dial attempts. For participants without Internet access, Knowledge Network provides free Internet and a WebTV device. The participant pools in each year are stratified, nationally representative samples of American adults. Full question wordings and other survey details including demographic breakdowns are available in Appendix A and B. I analyze 13 survey experiments to estimate the effects of providing education spending data on polarization. I analyze five survey experiments to estimate the effects of providing ideologically moderate policy endorsements on polarization. I conduct a nonexperimental analysis on the same survey data to determine whether or not there is a link between having children in the household and polarization.

To explore the extent to which the provision of education spending data affects polarization on education issues, I analyze data from survey experiments administered in 2008-2014 in which information on per pupil expenditures and teacher salaries is provided to randomly selected subsets of the participant pool alongside questions about these issues. With respect to per pupil expenditures, a random subset of the participant pool receives current estimates of average per pupil expenditures in their school district. All participants are asked, “Do you think that government funding for public schools in your district should increase, decrease, or stay about the same?” I dichotomize responses such that 1 equals “Increase” (0 otherwise). I exclude the 2010 per pupil expenditure experiment due to a significantly different question wording. For teacher salaries, a random subset of the participant pool receives current estimates of average teacher salaries in their state. All participants are asked, “Do you think that teacher salaries in your state should increase, decrease, or stay about the same?” I dichotomize responses such that 1 equals “Increase” (0 otherwise).

To explore the extent to which the provision of ideologically moderate policy endorsements affects polarization on education issues, I analyze data from survey experiments administered in 2009 and 2010 in which Barack Obama’s positions on charter schools, merit-based teacher pay, school accountability, and standardized testing are provided to randomly selected subsets of the participant pool alongside questions about these issues. With respect to charter schools, a random subset of the participant pool receives Obama’s position: “President Barack Obama has expressed support for charter schools.” All participants are asked, “Many states permit the formation of charter schools, which are publicly funded but are not managed by the local school board. These schools are expected to meet promised objectives, but are exempt from many state regulations. Do you support or oppose the formation of charter schools?” I dichotomize responses such that 1 equals “Support” (0 otherwise). For merit-based teacher pay, a random subset of the participant pool receives Obama’s position: “President Barack Obama has expressed support for the policy of basing teachers’ salaries, in part, on their students’ academic progress on tests.” All participants are asked, “Do you favor or oppose basing a teacher’s salary, in part, on his or her students’ academic progress on state tests?” I dichotomize responses such that 1 equals “Favor” (0 otherwise). For school accountability, a random subset of the participant pool receives Obama’s position: “As you may know, this year Congress is expected to take action on the federal school accountability law. President Barack Obama has proposed to maintain the current requirement that all students be tested in math and reading each year in grades 3-8 and once in high school.” All participants are asked, “Do you support or oppose this proposal?” I dichotomize responses such that 1 equals “Support” (0 otherwise). For standardized testing, a random subset of the participant pool receives Obama’s position: “President Obama has proposed that states be required to toughen the standards and tests used to evaluate student performance.” All participants are asked, “Do you support or oppose this proposal?” I dichotomize responses such that 1 equals “Support” (0 otherwise).

The Education Next Survey does not contain an experiment that randomly assigns children to households (nor is such an experiment easily conceivable). To tackle the question of whether the relationships between having kids at home and education policy preferences vary by party identification, I pool the survey responses from 2008 to 2014 and conduct an analysis on all available data. A participant with at least one child under the age of 18 in his or her household is coded as 1 (0 otherwise). For dependent variables, I use the nonexperimental versions of the education spending, teacher salaries, charter schools, and merit-based pay questions as they appear in the various survey administrations. This allows me to employ similar (often identical) question wordings throughout for each survey question. The sample size for each sub-analysis varies according to the number of years in which the question was asked and the size of the subset of the participant pool that receives the same version of the question. I chose these issues for this analysis because they appear in at least five survey administrations, generating sufficient sample sizes for the precise estimation of party identification interactions. For each education policy preference question, I dichotomize answers as described in the previous two paragraphs.

A small number of participants did not answer some of the questions on education policy preferences. For each case, I code their response as “Do Not Increase,” “Do Not Support,” or “Do Not Favor,” depending on the issue. Because nonresponse is not necessarily equally distributed across experimental conditions, dropping these cases could bias the estimates of the average and heterogeneous treatment effects. Retaining and recoding these participants allows me to focus on the various probabilities of support for a range of education issues among Democrats and Republicans while maintaining the integrity of the original random assignments that lend the experiments their internal validity. Participants who decline to answer these questions may have complex and not exclusively negative attitudes toward these education issues, but I understand their nonresponse as a signal that they do not actively support the policy at hand.

The Education Next Survey also collects extensive demographic information on its respondents. For my analyses, I use measures of participants’ age, educational attainment, race/ethnicity, gender, income, U.S. Census region, the presence of children under the age of 18 in the household, political ideology (coded as “Conservative,” “Liberal,” or “Moderate/Undecided/Other”), and party identification (coded as “Republican/Leans Republican,” “Democrat/Leans Democrat,” or “Independent/Undecided/Other”). As is typical, all analyses in this study incorporate so-called “party leaners” into party identifications (Keith, Magleby, Nelson, Orr, & Westlye, 1992).

There are only a few instances of missing data for these demographic characteristics. Some respondents did not answer the political ideology or party identification questions. For political ideology, I code missing values as “Moderate/Undecided/Other.” For party identification, I code missing values as “Independent/Undecided/Other.” This study is interested in the differences between those who actively identify as Democrats and Republicans. Participants with missing political identity data can be recoded as “Moderate/Undecided/Other” or “Independent/Undecided/Other” without affecting the comparison of interest. Moreover, by recoding these participants into the existing “Other” categories rather than creating new “Missing” categories, I am able to employ a simpler model specification.

Analytic Approach

The analytic approach for each test of the three polarization mechanisms proceeds in the same fashion. I fit a series of ordinary least squares (OLS) regression equations:

Y_{i} = β_{0} + β_{1} T_{i} + β_{2} R e p_{i} + β_{3} O t h e r_{i} + β_{4} T_{i} \times R e p_{i} + β_{5} T_{i} \times O t h e r_{i} + β_{6} X_{i} + ϵ_{i},

where Y is an indicator of support for each education issue; T is an indicator for treatment (the provision of education spending data, the provision of Obama’s position, or the presence of children under the age of 18 in the household); Rep is an indicator for “Republican/Leans Republican”; Other is an indicator for “Independent/Undecided/Other” (leaving “Democrat/Leans Democrat” as the comparison group); X is a vector of covariates for age, educational attainment, race/ethnicity, gender, income, U.S. Census region, political ideology, and survey year (when data are pooled from multiple survey administrations); and ϵ is the error term for individual i. Because of the dichotomous dependent variables, I also conduct these analyses using logistic regression (see Appendix C). The results are substantively equivalent. For simplicity and ease of interpretation, I display only the OLS values in the main text.

In each model, the coefficients of interest are the same. β1 provides the average treatment effect for Democrats. β2 provides the average gap in support for each education issue between Democrats and Republicans in the control group. The sum of β1 and β4 provides the average treatment effect for Republicans. The combination of these values reveals whether the treatment increases or decreases polarization relative to the partisan gap in the control group. In the tables that follow, I display the values of β₁, β₂, and β₄ . Other values are omitted for simplicity. To help visualize the treatment effect heterogeneity, the average effects for Democrats (β₁) and the average effects for Republicans (β₁ + β₄ ) are displayed in Figures 1, 2, and 3.

Findings

Education Spending Data

Table 1 displays the results of the six survey experiments conducted between 2008 and 2014 on the effects of the provision of average districtwide per pupil expenditures on support for increased education spending (see also the first six graphs in Figure 1). Across all survey administrations, Republicans are less likely than Democrats to support increased education spending. Among Democrats, the provision of information about current average districtwide per pupil expenditures reduces the probability of supporting increased education spending by between 11 and 22 percentage points. To calculate the average treatment effect among Republicans, take the sum of the Per Pupil Expenditures coefficient and the Per Pupil Expenditures × Rep coefficient. The provision of this information tends to have a slightly smaller negative effect among Republicans, resulting in an overall depolarization of opinion. These effect size differences between parties tend to be minor (between 2 and 11 percentage points), and they are only statistically significant in a few cases (2012 and 2014). However, the pattern is quite uniform across every experiment. This consistency across multiple iterations of the same study suggests a small but persistent difference in the way Democrats and Republicans respond to the provision of per pupil expenditure data.

Table 1.

Effects of Providing Per Pupil Expenditure Data on Support for Increased Education Spending.

	2008		2009		2011		2012		2013		2014
Per Pupil Expenditures	−0.13***	−0.13***	−0.12***	−0.11***	−0.17***	−0.17***	−0.21***	−0.21***	−0.17***	−0.18***	−0.21***	−0.22***
Per Pupil Expenditures	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.03)	(0.03)	(0.02)	(0.02)	(0.02)	(0.02)
Republican	−0.23***	−0.12***	−0.22***	−0.14***	−0.27***	−0.21***	−0.22***	−0.15***	−0.22***	−0.15***	−0.29***	−0.19***
Republican	(0.02)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.04)	(0.04)	(0.02)	(0.02)	(0.02)	(0.03)
Per Pupil Expenditures × Rep	0.03	0.03	0.04	0.04	0.02	0.03	0.09^†	0.11*	0.03	0.04	0.09**	0.09**
Per Pupil Expenditures × Rep	(0.03)	(0.03)	(0.04)	(0.04)	(0.04)	(0.04)	(0.05)	(0.05)	(0.03)	(0.03)	(0.03)	(0.03)
Covariates		Yes		Yes		Yes		Yes		Yes		Yes
N	3,228	3,228	3,251	3,251	2,632	2,632	1,519	1,519	5,569	5,569	3,491	3,491

Source. Education Next Survey, Harvard Program on Education Policy and Governance.

Note. Values are linear probability model coefficients with standard errors in parentheses; Republican compared with Democrat (Independent/Undecided/Other not shown).

†

p < .1. *p < .05. **p < .01. ***p < .001.

Table 2 displays the results of seven survey experiments conducted between 2008 and 2014 on the effects of the provision of average statewide teacher salaries on support for increased teacher salaries (see also the next seven graphs in Figure 1). Across all survey administrations, Republicans are less likely than Democrats to support increased teacher salaries. Among Democrats, the provision of information about current average statewide teacher salaries reduces the probability of supporting increased teacher salaries by between 9 and 24 percentage points. Beginning in 2012, the provision of this information tends to have a slightly smaller negative effect among Republicans, resulting in an overall depolarization of opinion. These effect size differences between parties tend to be minor (between 4 and 9 percentage points), and they are only statistically significant at the conventional 5% level in 2013. However, like the per pupil expenditure experiments, this consistency across multiple iterations of the same study suggests a small but persistent difference in the way Democrats and Republicans respond to the provision of teacher salary data.

Table 2.

Effects of Providing Teacher Salary Data on Support for Increased Teacher Salaries.

	2008		2009		2010		2011		2012		2013		2014
Teacher Salaries	−0.13***	−0.12***	−0.12***	−0.12***	−0.12***	−0.11***	−0.09***	−0.10***	−0.24***	−0.24***	−0.23***	−0.23***	−0.23***	−0.23***
Teacher Salaries	(0.02)	(0.02)	(0.02)	(0.02)	(0.03)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Republican	−0.15***	−0.07**	−0.10***	−0.04	−0.13***	−0.08**	−0.22***	−0.17***	−0.18***	−0.13***	−0.20***	−0.17***	−0.19***	−0.12***
Republican	(0.02)	(0.03)	(0.03)	(0.03)	(0.02)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.02)	(0.02)	(0.02)	(0.02)
Salaries × Rep	0.02	0.01	0.00	0.00	−0.01	0.00	0.00	0.02	0.06	0.06^†	0.09**	0.09***	0.04^†	0.04
Salaries × Rep	(0.03)	(0.03)	(0.04)	(0.04)	(0.04)	(0.04)	(0.04)	(0.04)	(0.04)	(0.04)	(0.03)	(0.03)	(0.03)	(0.03)
Covariates		Yes		Yes		Yes		Yes		Yes		Yes		Yes
N	3,234	3,234	3,251	3,251	2,776	2,776	2,632	2,632	2,993	2,993	5,569	5,569	5,266	5,266

Source. Education Next Survey, Harvard Program on Education Policy and Governance.

Note. Values are linear probability model coefficients with standard errors in parentheses; Republican compared with Democrat (Independent/Undecided/Other not shown).

†

p < .1. *p < .05. **p < .01. ***p < .001.

To increase the statistical power available to identify this treatment effect heterogeneity, I pool the data from all available years and analyze them as if they were two large experiments. Table 3 and the final two graphs in Figure 1 display the results. Overall, Republicans are less likely to support increased education spending than their peers on the other side of the political spectrum. Among Democrats, there is a moderately large negative effect of the provision of education spending data on education spending preferences: a 17 percentage point decline in the probability of supporting increased education spending in general and an 18 percentage point decline in the probability of supporting increased teacher salaries in particular. The negative effects are 4 to 5 percentage points smaller for Republicans. This suggests that the provision of education spending data can reduce the partisan gap in support for related policies.

Table 3.

Effects of Providing Education Spending Data on Education Spending Preferences (All Years).

	More spending		Teacher salaries
Spending Data	−0.17***	−0.17***	−0.18***	−0.18***
Spending Data	(0.01)	(0.01)	(0.01)	(0.01)
Republican	−0.24***	−0.16***	−0.17***	−0.12***
Republican	(0.01)	(0.01)	(0.01)	(0.01)
Spending Data × Rep	0.05***	0.05***	0.04**	0.04**
Spending Data × Rep	(0.01)	(0.01)	(0.01)	(0.01)
Covariates		Yes		Yes
N	19,690	19,690	25,721	25,721

Source. Education Next Survey, Harvard Program on Education Policy and Governance.

Note. Values are linear probability model coefficients with standard errors in parentheses; Republican compared with Democrat (Independent/Undecided/Other not shown); all models hold survey year constant.

†

p < .1. *p < .05. **p < .01. ***p < .001.

This is not the pattern of responses we would expect from the perspective of the biased information processing literature. In general, the provision of education spending data tends to induce more conservative attitudes toward education spending in general and teacher salaries in particular. If reactions were shaped primarily by party identification, then we would expect to see smaller effects for Democrats, who presumably would be more likely to ignore the spending data, subject it to heightened scrutiny, or conclude that current spending levels are still inadequate. Instead, Democrats make a larger conservative shift in response to the spending data, resulting in an overall convergence in opinion. Partisans seem to be responding to the information itself and not to the ways in which that information would facilitate political goals.

Ideologically Moderate Policy Endorsements

Tables 4 and 5 display the results of the five survey experiments conducted in 2009 and 2010 on the effects of the provision of ideologically moderate policy endorsements from former President Obama on a range of education policies (see also Figure 2). There is a sharp distinction between the results of the 2009 experiments (Table 4) and the 2010 experiments (Table 5). In 2009, Republicans in the control group are more supportive of charter schools than Democrats in the control group. There is a modest positive effect of Obama’s endorsement of charter schools on Democrats’ support for charters (about a 9 percentage point increase in the probability of support). Republicans, however, remain unmoved by Obama’s endorsement. As a result, the partisan gap on the issue is essentially reduced to zero. A similar pattern appears in the experiment on merit-based teacher pay, but the treatment effect heterogeneity fails to meet the threshold of statistical significance.

Table 4.

Effects of Providing Obama’s Position on Attitudes Toward Education Issues (2009).

	Charter schools		Merit pay
Obama’s Position	0.09**	0.09**	0.16***	0.16***
Obama’s Position	(0.03)	(0.03)	(0.03)	(0.03)
Republican	0.11***	0.09*	0.04	0.02
Republican	(0.03)	(0.04)	(0.03)	(0.04)
Obama’s Position × Rep	−0.09^†	−0.08^†	−0.06	−0.06
Obama’s Position × Rep	(0.05)	(0.05)	(0.05)	(0.04)
Covariates		Yes		Yes
N	2,164	2,164	2,142	2,142

Source. Education Next Survey, Harvard Program on Education Policy and Governance.

Note. Values are linear probability model coefficients with standard errors in parentheses; Republican compared with Democrat (Independent/Undecided/Other not shown).

†

p < .1. *p < .05. **p < .01. ***p < .001.

Table 5.

Effects of Providing Obama’s Position on Attitudes Toward Education Issues (2010).

	Accountability		Testing		Merit pay
Obama’s Position	0.10***	0.10***	0.14***	0.13***	0.05^†	0.05*
Obama’s Position	(0.02)	(0.02)	(0.03)	(0.03)	(0.03)	(0.02)
Republican	−0.01	0.02	0.02	0.02	0.03	0.03
Republican	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)
Obama’s Position × Rep	−0.14***	−0.15***	−0.16***	−0.15***	−0.14***	−0.13***
Obama’s Position × Rep	(0.04)	(0.04)	(0.04)	(0.04)	(0.04)	(0.04)
Covariates		Yes		Yes		Yes
N	2,776	2,776	2,776	2,776	2,776	2,776

Source. Education Next Survey, Harvard Program on Education Policy and Governance.

Note. Values are linear probability model coefficients with standard errors in parentheses; Republican compared with Democrat (Independent/Undecided/Other not shown).

†

p < .1. *p < .05. **p < .01. ***p < .001.

In 2010, however, a dramatically different result occurs. Democrats have a predictable response to Obama’s endorsements: they become somewhat more supportive of those policies. Republicans, however, respond negatively to Obama’s endorsements, despite the fact that his positions are consistent with the modal Republican stance. In the treatment group, the probability that Democrats support the policies that Obama endorses increases by between 5 and 14 percentage points, but the probability that Republicans support those same policies decreases by between 2 and 9 percentage points. By contrast, in the control groups, Democrats and Republicans only diverge by 3 percentage points or less. In all three 2010 experiments, the provision of an ideologically moderate policy endorsement actually serves to increase polarization.

As a result, a consistent interpretation of the data remains elusive. It may be the case that, in a less heated political context, such signals can have depolarizing consequences. However, extreme resistance to the politician providing the signal may alter the underlying dynamic. Alternatively, perhaps 2009 is the aberrant case, captured at the tail end of a president’s honeymoon period at the beginning of the first term. In more typical political contexts, the negative partisanship apparent in 2010 may be the norm. The depolarizing consequences of the 2009 charter school experiment may shed some light on the political success of charter school policies over the last decade. Obama’s opinion leadership on this issue appears capable of reducing or even eliminating the partisan gap on charter schools, potentially facilitating the passage of charter-friendly policies. Unfortunately, the charter school experiment does not recur in the 2010 survey administration, limiting our ability to understand how Republicans’ changing reactions to Obama affect the relationship between his opinion leadership and polarization on this issue.

Children in the Household

Table 6 displays the results of a series of regressions of education policy preferences on the presence of children under the age of 18 in the household (see also Figure 3). Compared with their peers without children at home, Democrats and Republicans with kids have smaller partisan gaps in support for increased education spending, support for increased teacher salaries, and support for charter schools. Both Democrats and Republicans with children at home are more likely to support increased education spending than their childfree co-partisans, but the difference is larger among Republicans. As a result, the gap between partisans with kids is about 9 percentage points smaller than the gap between partisans without kids. There is no difference between Democrats with and without children in support for increased teacher salaries. However, Republicans with kids are more likely to support increased teacher salaries than their childfree co-partisans. As a result, the gap between partisans with children is about 6 percentage points smaller than the gap between partisans without children. Democrats and Republicans with children in the household also have a smaller partisan gap in support for charter schools. Democrats with children at home are slightly more likely to support charters than their childfree co-partisans, while Republicans with children at home are slightly less likely to support charters than their childfree co-partisans. As a result, the gap between Democrats and Republicans with kids is about 5 percentage points smaller than the gap between Democrats and Republicans without kids. The partisan gap on merit-based teacher pay does not appear to vary by the presence of children in the household.

Table 6.

Relationships Between Children in the Household and Attitudes Toward Education Issues.

	2008-2009, 2011-2014		2008-2014		2008-2014		2008-2011, 2013
	More spending		Teacher salaries		Charter schools		Merit pay
Child	0.05***	0.04*	0.01	−0.01	0.02*	0.04***	0.01	0.00
Child	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Republican	−0.27***	−0.18***	−0.20***	−0.12***	0.15***	0.13***	0.07***	0.06***
Republican	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Child × Rep	0.09***	0.08***	0.06***	0.04*	−0.05**	−0.05**	−0.03	0.00
Child × Rep	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Covariates		Yes		Yes		Yes		Yes
N	9,925	9,925	12,787	12,787	17,228	17,228	12,239	12,239

Source. Education Next Survey, Harvard Program on Education Policy and Governance.

†

p < .1. *p < .05. **p < .01. ***p < .001.

These findings are robust to the inclusion of a large range of covariates, such as age, educational attainment, race, gender, income, U.S. Census region, and political ideology. However, the presence of children in the household is not random, and the possibility of selection bias remains. It is reasonable to assume that individuals who currently have children at home may have systematically different attitudes toward issues of education than those whose households do not contain or no longer contain children—for reasons unrelated to the presence or absence of the kids themselves. The kinds of individuals who choose to have children may be a self-selecting group with somewhat different education policy preferences. These analyses, therefore, should not be read as definitive evidence in favor of the argument that having children reduces political polarization. Rather, this should be viewed as tentative evidence that invites future investigations into the issue.

Discussion

This study explores the conditions under which partisan polarization and depolarization occur with respect to public opinion on education issues. I investigate the extent to which (a) the availability of education spending data, (b) the provision of ideologically moderate policy endorsements from political elites, and (c) having children in the household cause partisan attitudes toward education issues to converge or diverge. I find consistent evidence that the provision of education spending data has depolarizing consequences, but the effects of ideologically moderate political endorsements on polarization vary by issue and year. I also find tentative evidence in favor of a link between having children at home and reduced polarization on education issues.

To test the effects of the availability of education spending data on polarization, I analyze 13 survey experiments that randomly assign information on current education spending levels before polling respondents on their attitudes toward increased education spending in general and increased teacher salaries in particular. Although the differences in effect sizes by party identification are small, the provision of spending data tends to reduce the partisan gap on education spending issues. While this evidence for treatment effect heterogeneity only occasionally meets the conventional threshold for statistical significance, the pattern is remarkably consistent across over a dozen essentially identical experiments—arguably a better indicator of validity. These results are somewhat surprising from the perspective of the literature on biased information processing. In this case, both Democrats and Republicans shift toward a more conventionally conservative position, with Democrats exhibiting a larger overall change. Partisans’ reactions to the provision of education spending data do not appear to be shaped by the expectations of their party affiliations. This analysis is not able to identify why most of the opinion change occurs among Democrats. It may be the case that Democrats are more willing to update their beliefs upon receiving new information, less committed to their positions, or some other explanation. Additional research is necessary to tease out these patterns.

To test the effects of ideologically moderate policy endorsements on polarization, I analyze five survey experiments that randomly assign former President Obama’s positions on a range of education issues before polling respondents on their attitudes toward those issues. The results of the experiments conducted in 2009 are considerably different from the results of the experiments conducted in 2010. In 2009, the provision of Obama’s positions tends to shift Democrats toward conventionally Republican positions while Republicans are generally unmoved, effectively reducing or even eliminating the partisan gaps. In these cases, partisans react in politically predictable ways, but the ideologically atypical content of the policy endorsement has depolarizing consequences. In 2010, however, Republicans respond quite negatively to Obama’s positions, increasing polarization on those issues. Obama’s 2010 endorsements appear to ignite Republicans’ partisan attachments, adding political heat to previously lukewarm debates.

Finally, to understand the link between having children in the household and polarization, I conduct a nonexperimental data analysis that investigates the relationships between the presence of kids at home and a range of education policy preferences as they vary by party identification. I find evidence that partisans with children at home are less divided on education spending, teacher salaries, and charter schools. However, with the current analysis, it is not possible to argue definitively that these instances of partisan convergence are directly caused by the presence of children. Indeed, there may be another factor, such as self-selection among parents, that drives this heterogeneity.

These results also provide some insight for those engaging in contemporary education policy debates in an increasingly polarized political landscape. It appears that partisans are capable of responding to some pieces of education policy information (specifically, spending data) in ways that are unconditioned by their partisan identities, potentially facilitating more open and productive debate. A similar dynamic may follow the information accrued and the experiences gleaned from having children. However, the power of party affiliations can reemerge if prominent political elites take an explicit position in an ongoing debate. In some cases, partisans appear to remobilize in accordance with or in opposition to that position. If political actors and ordinary citizens are interested in bipartisan solutions to education issues, the path forward may lie in the conversations and decisions that occur at less politically elevated levels, in which the blessings and curses of famously polarizing politicians can be kept safely at bay.

Supplemental Material

Online_appendices – Supplemental material for Polarization and the Politics of Education: What Moves Partisan Opinion?

Supplemental material, Online_appendices for Polarization and the Politics of Education: What Moves Partisan Opinion? by David M. Houston in Educational Policy

Footnotes

Acknowledgements

Author(s) would like to thank Jeffrey Henig, Priscilla Wohlstetter, Sarah Cohodes, Justin Phillips, Donald Green, Robert Shapiro, and Robert Erikson for their guidance and thoughtful feedback. Also, they are grateful to Paul Peterson, Martin West, Michael Henderson, and the Harvard Program on Education Policy and Governance for their extensive assistance and for generously sharing the Education Next Survey data. This research is supported by the National Academy of Education/ Spencer Dissertation Fellowship and the Teachers College Research Dissertation Fellowship. The code for this study can be found on the Open Science Framework at .

Author’s Note

David M. Houston is now affiliated with George Mason University.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is supported by the National Academy of Education/Spencer Dissertation Fellowship and the Teachers College Research Dissertation Fellowship.

Supplemental Material

Supplemental material for this article is available online.

Author Biography

David M. Houston is a postdoctoral research fellow in Education Policy and Governance at Harvard University, Kennedy School of Government. His research interests include education policy and public opinion on education issues.

References

Abramowitz

A. I.

(2010). The disappearing center: Engaged citizens, polarization, and American democracy. New Haven, CT: Yale University Press.

Bafumi

Shapiro

R. Y.

(2009). A new partisan voter. The Journal of Politics, 71, 1-24.

Bali

V. A.

(2016). Evolving trends in public opinion on the quality of local schools. Educational Policy, 30, 688-720.

Bullock

J. G.

(2011). Elite influence on public opinion in an informed electorate. American Political Science Review, 105, 496-515.

Bushaw

W. J.

Lopez

S. J.

(2011). Betting on teachers: The 43rd annual Phi Delta Kappa/Gallup Poll of the public’s attitudes towards the public schools. Bloomington, IN: Phi Delta Kappa Educational Foundation.

Campbell

Converse

P. E.

Miller

W. E.

Stokes

D. E.

(1960). The American voter. Chicago, IL: The University of Chicago Press.

Cannon

C. L.

Barham

F. E.

(1993). Are you and your public polls apart? What those national poll results really mean. Executive Educator, 15(10), 41-42.

Chingos

M. M.

Henderson

M. B.

West

M. R.

(2012). Citizen perceptions of government service quality: Evidence from public schools. Quarterly Journal of Political Science, 7, 411-445.

Cohen

G. L.

(2003). Party over policy: The dominating impact of group influence on political beliefs. Journal of Personality and Social Psychology, 85, 808-822.

10.

Downs

(1957). An economic theory of democracy. New York, NY: Harper & Row.

11.

Druckman

J. N.

Peterson

Slothuus

(2013). How elite partisan polarization affects public opinion formation. American Political Science Review, 107, 57-79.

12.

Elam

S. M.

(1995). How America views its schools: The PDK/Gallup Polls, 1969-1994. Bloomington, IN: Phi Delta Kappa Educational Foundation.

13.

Fiorina

M. P.

Abrams

S. J.

Pope

J. C.

(2010). Culture war? The myth of a polarized America (3rd ed.). New York, NY: Longman.

14.

Gerber

Green

D. P.

(1999). Misperceptions about perceptual bias. Annual Review of Political Science, 2, 189-210.

15.

Green

D. P.

Schickler

Palmquist

(2002). Partisan hearts and minds: Political parties and the social identities of voters. New Haven, CT: Yale University Press.

16.

Henig

J. R.

Houston

D. M.

Lyon

M. A.

(2017). From NCLB to ESSA: Lessons learned or politics reaffirmed? In Hess

F. M.

Eden

(Eds.), The Every Student Succeeds Act: What it means for America’s schools, systems, and states (pp. 29-42). Cambridge, MA: Harvard Education Press.

17.

Jerit

Barabas

(2012). Partisan perceptual bias and the information environment. The Journal of Politics, 74, 672-684.

18.

Keith

B. E.

Magleby

D. B.

Nelson

C. J.

Orr

Westlye

M. C.

(1992). The myth of the independent voter. Berkeley: University of California Press.

19.

Lord

C. G.

Ross

Lepper

M. R.

(1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37, 2098-2109.

20.

Loveless

(1997). The structure of public confidence in education. American Journal of Education, 105, 127-159.

21.

Lupia

McCubbins

M. D.

(1998). The democratic dilemma: Can citizens learn what they need to know? Cambridge, UK: Cambridge University Press.

22.

McCarty

Poole

K. T.

Rosenthal

(2006). Polarized America: The dance of ideology and unequal riches. Cambridge, MA: MIT Press.

23.

McGuinn

(2006). No Child Left Behind and the transformation of federal education policy, 1965-2005. Lawrence: University Press of Kansas.

24.

McGuinn

(2017). From ESEA to NCLB: The growth of the federal role and the shift to accountability. In Hess

F. M.

Eden

(Eds.), The Every Student Succeeds Act: What it means for America’s schools, systems, and states (pp. 13-28). Cambridge, MA: Harvard Education Press.

25.

Page

B. I.

Shapiro

R. Y.

(1992). The rational public: Fifty years of trends in Americans’ policy preferences. Chicago, IL: The University of Chicago Press.

26.

Peterson

P. E.

Henderson

West

M. R.

(2014). Teachers versus the public: What Americans think about schools and how to fix them. Washington, DC: Brookings Institution Press.

27.

Pew Research Center. (2014). Political polarization in the American public: How increasing ideological uniformity and partisan antipathy affect politics, compromise, and everyday life. Retrieved from http://www.pewresearch.org/wp-content/uploads/sites/4/2014/06/6-12-2014-Political-Polarization-Release.pdf

28.

Popkin

S. L.

(1991). The reasoning voter: Communication and persuasion in presidential campaigns. Chicago, IL: The University of Chicago Press.

29.

Reckhow

Grossman

Evans

B. C.

(2015). Policy cues and ideology in attitudes toward charter schools. Policy Studies Journal, 43, 207-227.

30.

Rhodes

J. H.

(2012). An education in politics: The origins and evolution of No Child Left Behind. Ithaca, NY: Cornell University Press.

31.

Schueler

B. E.

West

M. R.

(2016). Sticker shock: How information affects citizen support for public school funding. Public Opinion Quarterly, 80, 90-113.

32.

Shapiro

R. Y.

Bloch-Elkon

(2006, May). Political polarization and the rational public. Paper presented at the Annual Meeting of the American Association for Public Opinion Research, Montreal, Québec, Canada.

33.

Shapiro

R. Y.

Kilibarda

Sinozich

McClellan

(2016, May). American public opinion and partisan conflict: Education’s exceptionalism? Paper presented at Harvard University, Kennedy School of Government for “The Politics of Education Policy: An International Perspective.” Retrieved from https://sites.hks.harvard.edu/pepg/conferences/politics-of-education-2016/home.html

34.

Taber

C. S.

Cann

Kucsova

(2009). The motivated processing of political arguments. Political Behavior, 31, 137-155.

35.

Taber

C. S.

Lodge

(2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political Science, 50, 755-769.

36.

Wolbrecht

Hartney

M. T.

(2014). “Ideas about interests”: Explaining the changing partisan politics of education. Perspectives on Politics, 12, 603-630.

37.

Zaller

J. R.

(1992). The nature and origins of mass opinion. New York, NY: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.44 MB