For much of the past four decades, the study of urban and local government has been largely separated from the field of American politics, due to important differences in theoretical and methodological approaches (Peterson, Reference Peterson1981; Judd, Reference Judd2005; Trounstine, Reference Trounstine2009; Anzia, Reference Anzia2020). In recent years, however, American politics scholars with an interest in the study of representation have sought to exploit the multiplicity and diversity of local governments to investigate how well, or poorly, these governments represent their constituents (e.g., Ferreira and Gyourko, Reference Ferreira and Gyourko2009; Tausanovitch and Warshaw, Reference Tausanovitch and Warshaw2013, Reference Tausanovitch and Warshaw2014; de Benedictis-Kessner and Warshaw, Reference de Benedictis-Kessner and Warshaw2016, Reference de Benedictis-Kessner and Warshaw2020a, Reference de Benedictis-Kessner and Warshaw2020b; Einstein and Kogan, Reference Einstein and Kogan2016; Schaffner et al., Reference Schaffner, Rhodes and La Raja2020). Drawing on new data sources, and using innovative statistical techniques, many of these studies have highlighted the role of residents' ideology in the process of representation in local politics (Tausanovitch and Warshaw, Reference Tausanovitch and Warshaw2013, Reference Tausanovitch and Warshaw2014; Einstein and Kogan, Reference Einstein and Kogan2016; Schaffner et al., Reference Schaffner, Rhodes and La Raja2020).
However, the new scholarly emphasis on the importance of ideology in representation in local government is not without controversy. Recent research typically uses or interprets measures of ideology as revealing residents' preferences on local issues. But because these measures are usually derived from nationally oriented sources such as presidential vote tallies or national public opinion surveys, it is unclear whether they are measuring preferences on distinctively local issues such as local economic development, zoning, police and fire protection, or sewers and roads (Anzia, Reference Anzia2020, 135). Because the precise relationship between residents' preferences on local issues and their preferences on national issues is poorly understood, studies that rely on nationally oriented measures of ideology may be making unwarranted inferences about the role of ideology toward local policy issues in representation in local government.
In this paper, we directly evaluate whether and how individuals' attitudes on distinctively local issues are related to their attitudes on national issues, and thereby advance theory and method in the study of ideological representation in local government. Using a pre-registered analysis of a survey of American adults with a demographically and politically balanced set of respondents that includes a host of questions about both local issues and national issues, and employing a variety of statistical techniques, we comprehensively examine the relationship between residents' preferences on local issues and their preferences on national issues.Footnote 1
The results of our analyses suggest that a coherent liberal-conservative ideology structures attitudes toward many (though not all) local issues; and that this ideology is closely related to the familiar liberal-conservative ideology that shapes individuals' attitudes toward national issues. While attitudes toward some local issues, particularly those relating to business tax breaks, local policing, and local budgets do seem to be distinctive, attitudes toward numerous other local issues align predictably with attitudes toward national issues such as affirmative action, gun control, and immigration. Interestingly, we find that some (though not all) attitudes relating to local land use—often considered to be the prototypical local issue (e.g., Cullingworth, Reference Cullingworth1993)—appear to be structured by the same underlying factor as that of attitudes toward many national issues. A second (non-preregistered) study using nationally representative data from the 2021 Cooperative Election Study (CES) largely confirms the results of our main analysis, by showing that an issue-based measure of national policy ideology is highly correlated with an issue-based measure of local policy ideology.
Our results suggest that, far from being distinctive to municipal politics, attitudes toward many local issues are closely related to attitudes toward national issues. An important methodological implication of our findings is that scholars may use measures of national ideology to shed light on individuals' preferences toward many local issues.
1. The debate about the role of ideology in US local politics
Although scholars of urban and local government in the United States recognize that ideology may play some role in local politics (Abrajano and Alvarez, Reference Abrajano and Alvarez2005; Oliver and Ha, Reference Oliver and Ha2007; Ramakrishnan and Wong, Reference Ramakrishnan, Wong and Varsanyi2010; Tausanovitch and Warshaw, Reference Tausanovitch and Warshaw2014; Bucchianeri, Reference Bucchianeri2020), many also note important reasons why the role of ideology in representation in municipal politics may be limited. Some scholars argue that, because municipalities must compete to maintain the tax base, attract employers, and entice skilled labor, they are likely to enact similar bundles of growth-friendly policies such as relatively low corporate taxes and good schools. According to this view, competition-based convergence on similar policies limits the ability of municipalities to entertain the redistribution of resources and opportunities from the affluent to the poor which is typically the primary source of left-right ideological conflict (Peterson, Reference Peterson1981). Some also argue that, because the federal and state governments impose considerable legal constraints on local policymaking authority, opportunities for local governments to respond to the ideologies of residents are constrained (Yinger and Ladd, Reference Yinger and Ladd1989; Peterson, Reference Peterson1995; Nivola, Reference Nivola2002; Oliver et al., Reference Oliver, Ha and Callen2012). Finally, institutional characteristics peculiar to (many) municipal governments—formally non-partisan elections, elections held off-cycle from federal and state contests, and at-large systems of representation—may disrupt ideology-based responsiveness to resident demands (Wood, Reference Wood2002; Anzia, Reference Anzia2014; Schaffner et al., Reference Schaffner, Rhodes and La Raja2020).
Much research on urban and local politics has also emphasized cleavages other than ideology, including those based on homeownership (Fischel, Reference Fischel2001; Oliver et al., Reference Oliver, Ha and Callen2012; Hankinson, Reference Hankinson2018; Einstein et al., Reference Einstein, Glick and Palmer2019), race and ethnicity (Barreto, Reference Barreto2007; Hajnal, Reference Hajnal2009; Hajnal and Trounstine, Reference Hajnal and Trounstine2014; Trounstine, Reference Trounstine2019, Reference Trounstine2020), age differences (Kogan et al., Reference Kogan, Lavertu and Peskowitz2018; Anzia, Reference Anzia2019), and public sector employees versus taxpayers (Moe, Reference Moe2011; Anzia and Moe, Reference Anzia and Moe2015; DiSalvo, Reference DiSalvo2015).
Finally, some central issues faced by local governments, such as zoning and land-use policies, seem to be different in kind from the issues that are the focus of higher levels of government, perhaps making ideology less relevant, at least on these matters (Einstein et al., Reference Einstein, Glick and Palmer2019).
An additional reason why scholars of urban and local politics often have not emphasized the role of ideology in representation in local politics is that the necessary data were unavailable. Research on local politics lacked comprehensive data, across many communities, on the key elements needed to study ideological representation in local government: measures of public opinion of residents on policy issues, measures of elites' preferences on those same issues, and measures of local government policy outputs (Anzia, Reference Anzia2020).Footnote 2
Recently, however, scholars have used a variety of innovative approaches to measure ideology at the local level to study ideological representation in municipal politics (Anzia, Reference Anzia2020, 136–137). These include using data on presidential voting at the county or municipal level as a proxy for municipal ideology (Choi et al., Reference Choi, Bae, Kwon and Feiock2010; Einstein and Kogan, Reference Einstein and Kogan2016); estimating municipal ideology scores with multilevel regression with post-stratification on pooled survey samples (Tausanovitch and Warshaw, Reference Tausanovitch and Warshaw2014); and exploiting community-level ideology scores available in a large voter file (Schaffner et al., Reference Schaffner, Rhodes and La Raja2020).
Although these studies make impressive contributions, there is reasonable disagreement whether they actually measure residents' ideologies with respect to distinctively local issues, or instead measure residents' ideologies concerning federal and/or state issues. Anzia (Reference Anzia2020, 139–141) has argued that measures of ideology based on presidential vote shares or national surveys may be measures of local-level preferences on mostly national issues such as Social Security and health care rather than distinctively local issues such as local zoning and policing. Anzia also argues that, because the large public opinion surveys that often undergird ideology measures mostly ask about national policy issues, these measures of ideology primarily tap respondents' preferences on national issues rather than local issues.
Besides Tausanovitch and Warshaw's (Reference Tausanovitch and Warshaw2014) effort, there have been few attempts to directly examine the relationship between individuals' preferences on local issues and their preferences on national issues. One study using a measure of partisan affiliation finds only modest levels of partisan sorting and polarization over local development issues (Jensen et al., Reference Jensen, Marble, Scheve and Slaughter2021) but does not directly measure preferences on national issues. Another study demonstrates that ideology plays a role in shaping homeowners' preferences on local housing development, but its influence was greatest on issues (like housing vouchers, rent controls, and renter tax credits) that imposed minimal, diffuse costs on homeowners (Marble and Nall, Reference Marble and Nall2021). Neither of these studies uses nationally representative samples that cover a broad range of local and national issues.
The uncertainty about the relationship between preferences on local issues and preferences on national issues limits our understanding of how ideology operates across levels of government. In the remainder of this article we directly investigate this relationship. Our research addresses two questions at the heart of the debate over the role of ideology in local politics:
1. Do attitudes toward local policies exhibit a unidimensional, liberal-conservative ideological structure; or are they organized in a more complex, or idiosyncratic, fashion?
2. If attitudes toward local policies exhibit an ideological structure, does this local ideology correlate with attitudes toward national issues (and with a measure of national ideology derived from attitudes toward national issues)?
For this research, we undertook two different studies of public opinion. We describe the data, methods, and results for each study in turn.
2. Study 1
2.1 Data and methods
We fielded a survey of American adults that contained a battery of local and national policy questions, and then analyzed the structure of respondents' preferences over local and national issues using a variety of statistical techniques (details on the demographic and political characteristics of the sample are contained in the online Supplementary materials, Section 1). Our strategy for assessing the relationship between local preferences and national preferences is similar to that employed in Tausanovitch and Warshaw (Reference Tausanovitch and Warshaw2014), except that we used a more diverse array of techniques to study individual preferences, and also pre-registered our design to ensure full transparency in our analysis.
We pre-registered our research design with the Open Science Framework (OSF). The pre-registration, which contains our data analysis plan, Stata code, and survey instrument is contained in the online Supplementary materials, Sections 9, 10 and 11, respectively. As part of our pre-registration, we committed to implementing the data analysis exactly as presented in our data analysis plan. We review the key features of our research design and data analysis plan below.
2.1.1 Survey design and questions
We contracted with Lucid, a firm that provides respondents for online surveys, to field a survey of American adults. Lucid is a desirable source of respondents for this study because the firm provides a demographically and politically diverse set of respondents. In this case, the sample obtained from Lucid was designed to hit quotas to be nationally representative on age, gender, ethnicity, and region. Coppock and McClellan (Reference Coppock and McClellan2019) demonstrate the validity of responses provided by subjects from Lucid. This panel provider has also been used for representative surveys carried out on a variety of topics, such as a Covid-19 tracking poll (Schaffner, Reference Schaffner2020). For this study, we did not weight the sample to be nationally representative since our interest here is in the correlational structure of the data rather than specific point estimates.
The survey was in the field on October 7–8, 2021. After excluding respondents who failed an attention check or completed the survey too quickly (as outlined in our pre-registration), we had a total sample size of 909 American adults. The survey instrument, presented in the online Supplementary materials, Section 11, was designed to query respondents' positions on a wide range of local and national issues, permitting detailed investigation of the structure of preferences on local issues and national issues, respectively, as well as the relationship between them.
The questionnaire included 18 questions measuring respondents' attitudes toward local issues such as land use, recycling, rent controls, and parks and recreation. These questions were derived from items in several surveys fielded by the International City/County Managers Association (ICMA) and used in research by scholars of ideology in local politics (e.g., Tausanovitch and Warshaw, Reference Tausanovitch and Warshaw2014; Schaffner et al., Reference Schaffner, Rhodes and La Raja2020): the 2010 Survey of Government Sustainability, the 2014 Survey of Economic Development, and the 2015 Survey of Government Sustainability.Footnote 3 We used the ICMA surveys to inform our choice of issues because ICMA members deemed these issues important enough to warrant having their national association survey their members about them. Due to space constraints, we could not ask about all the issues surveyed in the ICMA studies, so we sampled questions relating to a wide variety of local issues.Footnote 4
Following the ICMA surveys, our questions ask about general opposition to or support for various local policies. Many of the questions ask about overall preferences for more or less government, though some also refer to policies that involve redistribution within communities (e.g., affordable housing, rent control, transportation subsidies for marginalized groups, etc.). Generally speaking, the question wordings do not explicitly prime consideration of the distributional implications of policies. We frame our questions in this way to focus on respondents' left-right ideology, without the “contaminating” influence of self-interest (e.g., Marbel and Nall, Reference Marble and Nall2021). This approach to the wording of our questions about local policies brings them into alignment with the wording of our questions about attitudes toward national policies, which also ask about general opposition or support.
The survey also included 13 questions gauging respondents' attitudes toward various national issues such as affirmative action, gun control, health care, immigration, welfare, the environment, and national security. The wording for these questions was drawn from the CES. The questionnaire also contained a separate battery of questions querying respondents' demographic characteristics, partisanship, and ideology.
The response options for each policy question in this survey were strongly support, somewhat support, neither support nor oppose, somewhat oppose, and strongly oppose. Each of these response options was assigned a numerical value (1 through 5) and we left these numerical values as they were when conducting our analyses in order to account for both the direction and strength of attitudes.
2.2 Pre-registered analysis
To test whether and to what extent opinions on local and national issues are related, we take three approaches as outlined in our pre-registration. This allows us to avoid the possibility that our conclusions are attributable to the assumptions underlying any particular method, as well as enabling us to triangulate across methods.
2.2.1 Item correlations
We begin by examining a correlation matrix of all the issue-related questions in our survey to understand the average correlations among national issues, among local issues, and between national issues and local issues, respectively. With 31 policy questions in the survey, there are 465 unique pairwise correlations to calculate. Since presenting this many correlation coefficients is unwieldy, we instead calculate the average of the absolute values of the correlations among all pairs of local policy questions, the average among all pairs of national policy questions, and the average among all pairs of one local and one national policy question.
As described in our pre-registration, this investigation has the following implications for our answers to the two questions animating our study. If the average correlations among pairs of local policy questions and among pairs of national policy questions, respectively, are much larger than the average correlation among pairs of one local and one national policy question, this would provide evidence for the proposition that attitudes toward local issues are distinctive from attitudes toward national issues. Conversely, if the average correlation among pairs of one local and one national policy question are similar in magnitude to the average correlations among pairs of local policy questions and among pairs of national policy questions, respectively, this would provide evidence in favor of the argument that attitudes toward local issues are not wholly distinctive from attitudes toward national issues. Table 1 presents these results.
Table 1. Average correlation coefficients between items based on scope, study 1
Note: Entries in the first column are the average value of the absolute value of the correlation coefficient between each unique pair of items and the standard error of that average. Entries in the second column are the weighted average within community correlation coefficients and standard errors for that average. N = 465 unique pairwise correlations.
The average correlation between opinions on local policy items was 0.231; however, as Figure 1 reflects, there was a great deal of variance.Footnote 5 About one-fourth of the pairwise correlations for local policy items were above 0.30, while another 25 percent were less than 0.15. For national policy issues, the average correlation was slightly lower at 0.199. Half of these correlations were below 0.15 while about one-fourth were greater than 0.30. Finally, for issue items that included one local policy question and one national policy question, the average correlation was 0.183. Notably, this is just slightly lower than the average correlation for national policy items. In total, 46 percent of the correlations for local and national question pairs were below 0.15 while 18 percent were above 0.30. Additionally, while these average correlations may seem small, recent work from Broockman finds that the average correlation between national issues is 0.13 (Reference Broockman2016, 193).
Figure 1. Distribution of correlation coefficients.
Note: Plot shows the distribution of correlation coefficients between each unique pair of two local items, one local and one national item, and two national items.
One concern is that analyzing simple correlation coefficients does not account for the fact that the correlational structure of these items might differ across different cities and towns. To test whether this is the case, the second column of entries in Table 1 is the within-community correlations. We estimate these by first using each respondent's zip code to identify the Census place in which they reside. For any Census place with more than one respondent, we then estimate the correlations within that place and then take the weighted mean of those correlations (weighted by the sample size for each place).Footnote 6 Using this approach, the average correlations for each type of pairing are a bit higher, but the general pattern is unchanged. The average correlation among local policy items is only slightly larger than the average correlation between local policy items and national policy items. This observation provides initial evidence that attitudes toward distinctively local issues are not wholly separate from attitudes toward national issues.
2.2.2 Exploratory factor analysis
Next, we present findings from an exploratory factor analysis of all 31 issue items included in the survey. The purpose of this factor analysis is to investigate whether and to what extent responses to national issues and responses to local issues tend to load on the same underlying factor, or on different factors. The exploratory common factor analysis is, at the outset, agnostic as to whether and how attitudes toward local issues are related to attitudes toward national issues: relationships between attitudes toward local issues and attitudes toward national issues, to the degree they exist, are discovered based on their tendency to share common variance and thus load on the same underlying factor.
The factor analysis has the following implications for our inquiry about the relationship between preferences toward local issues and preferences toward national issues. If responses to national policy questions and local policy questions tend to load on the same factor, it would provide evidence of a single underlying ideology linking attitudes at these different levels of government. Alternatively, if responses to questions about national issues and responses to questions about local issues tend to load on different factors, it would provide evidence that attitudes toward national issues and attitudes toward local issues are distinctive.
We estimate our exploratory common factor analysis with maximum-likelihood estimation and limit the initial extraction to six factors to keep the output readable. We then conduct a parallel analysis to determine the number of factors to retain. Parallel analysis involves comparing the eigenvalues from the factor analysis estimated on the actual observed data to eigenvalues from a factor analysis performed on randomly generated data of the same sample size, and thus provides a data-driven basis for determining the number of factors to retain.
Table 2 presents the loadings for the first four factors estimated, as only four factors were deemed to be significant from the parallel analysis.Footnote 7 Shaded cells indicate factor loadings with an absolute value greater than 0.3, the threshold our pre-registration sets for identifying a loading as substantively meaningful. The first factor has an eigenvalue of 6.307 and 19 of the 31 items show significant loadings on this factor. Notably, the items loading on this factor are just as likely to be local issues as national ones. In total, 11 of the 18 local issues (61 percent) and 8 of the 13 national issues (62 percent) have a factor loading above 0.3 (or below −0.3) on the first factor.
Table 2. Factor loadings, study 1
The local issues that load strongly on this factor include some issues that might be expected to align with attitudes toward national issues either because these local issues involve the classic left-right divide over the size of government (pre-education, public transit, increase parking, expand internet access) or the “post-material” cleavage (Inglehart, Reference Inglehart1977) over environmental and cultural issues (benefits for same-sex partners and require recycling) that has folded with the left-right divide in a single dimension of conflict in contemporary American national politics (e.g., Mason, Reference Mason2018).
Notably, certain land-use and zoning issues—affordable housing, rent controls, land-use limits, condemn blighted property, and allow apartment buildings in neighborhood—also load strongly on this first factor. Given that land-use and zoning issues are often considered central to local politics, that these issues do not have obvious analogues at the national level, and that some previous research (e.g., Marble and Nall, Reference Marble and Nall2021) suggests that attitudes toward some land-use policies are structured by the homeowner/renter divide, the loading of attitudes toward certain land-use issues on the first factor alongside attitudes toward many national issues is intriguing. One likely explanation is that, due to national partisan disagreement over how to address today's record-high housing prices and rents (Giorno, Reference Giorno2024), the issue of housing affordability has become linked to national partisan conflict, thus enabling residents to link their preferences toward the issue with their preferences on national issues in an ideologically consistent way.
A more nuanced possible explanation—though not wholly inconsistent with the first—is that our research design and question wording encouraged responses that were more likely to exhibit an ideological (rather than self-interested) orientation. Specifically, in contrast to Marble and Nall's approach, our questions querying respondents' attitudes toward zoning and land-use policies did not prime respondents to think about these matters in relation to their self-interest as homeowners or renters, and thus may have attenuated the influence of self-interest while encouraging more ideological responses, particularly on issues such as the construction of affordable housing or apartment buildings. Of course, given the nature of our design we can only speculate on this matter, and future research should more directly test how the suppression or priming of self-interested motivations may influence the relationship between attitudes toward local issues and attitudes toward national issues.
The local policy items that do not load very strongly on this first factor are those gauging opinions on tax breaks to attract businesses to the community, police funding, cutting municipal employee pensions, cutting local services to ensure a balanced budget, and raising property taxes to ensure a balanced budget. Among the set of national policy items, opinions on abortion restrictions, cuts to domestic spending to reduce the deficit, tariffs on China, the use of the military to destroy a terrorist camp, and voter ID laws all failed to load strongly on this first dimension. There is only one strong negative loading with the first factor—that for opinions on increased border security. The main reason we do not see additional strong negative loadings is that we asked about fewer conservative policies and many of those we did ask about are not as strongly associated with the first factor (e.g., voter ID laws and tariffs on China).
The subsequent factors, which all have much smaller eigenvalues (reflecting the fact that they explain less of the covariance in the items), appear to be a bit more idiosyncratic. The second factor contains a diverse array of items, though virtually all relate to taxation and/or spending on social services. The third factor appears to solely explain responses to the three questions asking about local tax incentives to attract businesses to the community. This factor may relate to respondents' attitudes toward the distinctively local issue of inter-municipal competition for economic development (Peterson, Reference Peterson1981). And the fourth factor appears to pick up attitudes on largely non-economic issues, including voter ID, immigration, abortion, local policing, and condemning blighted properties. This factor, which is comprised of attitudes toward both some local and some national issues, may be related to post-material or symbolic values and beliefs (Inglehart, Reference Inglehart1977), or even underlying racial views on issues that have gained national salience.
Overall, however, the most prominent factor appears to explain opinions on a large number (indeed, a super-majority) of both local and national issues, providing evidence that a single dimension of policy liberalism/conservatism explains attitudes on many issues at both the local and national levels. We note, however, that some local issues highlighted in previous work, particularly those related to local policing, local tax breaks, and local service cuts, do not align on this dimension.
2.2.3 Comparing issue item scales
The third approach called for in our pre-registration is to create separate scales of national and local policy attitudes and then examine the extent to which those scales are associated with each other. Unlike with factor analysis, this approach assumes that national policy attitudes and local policy attitudes are distinct quantities, but then examines the extent to which those attitude scales are related.
We begin by using an item response theory (IRT)-graded response model to scale separately the responses to (1) all of the local items and (2) all of the national items. The graded response model is employed here because the responses to the questions in our survey are categorical. The IRT model is also a useful approach because unlike factor analysis, it does not require case-wise deletion for respondents who did not answer every single item. The resulting scale produced by the IRT function is a standard normal variable which is meant to capture each respondent's value of the underlying latent trait (local or national policy conservatism).
Figure 2 plots each respondent based on his position on both scales. Because both scales are standard normal, the axes can be understood in terms of standard deviations. For example, a respondent with a zero for the local policy scale and a −1 for the national policy scale would have average policy conservatism when it came to local policy questions but would be 1 standard deviation below the mean on national policy items. In other words, this individual would be on the liberal side of the national policy scale and at the center of the local policy scale.
Figure 2. Scatterplot of survey respondents' positions on local and national policy attitude scales, study 1.
Figure 2 shows a strong relationship between these two scales.Footnote 8 In fact, the correlation between the two scales is 0.658. To put that in perspective, the national policy scale is correlated with an ideological self-placement question at 0.556, while the ideological self-placement question has a correlation of 0.281 with the local policy scale. Knowing an individual's positions on local issues is a stronger signal about their national policy positions than where they place themselves on the standard five-point ideology question. This is powerful evidence that local ideology is predictive of national ideology, and vice versa. This finding has important theoretical implications, reinforcing the conclusion from the previous section that, far from being a distinctive dimension of public opinion, attitudes toward local issues, at least at the aggregate level, are nationalized so that they align with attitudes toward national issues.
Our finding that our scaled measure of national policy ideology is strongly correlated with our scaled measure of local policy ideology also has important methodological implications. Specifically, it suggests that researchers may be on safe ground if they use a scaled measure of national policy ideology as a proxy measure of overall local policy ideology (though, as we explain in greater detail below, they must take much greater care in using national policy ideology as a stand-in for preferences on specific local issues).
2.3 Additional preliminary analysis (not pre-registered)
In this section, we further investigate the nature of the relationship between national and local issue attitudes, respectively. First, we assess the correlations between our summary measure of national policy attitudes (i.e., our scaled measure of national policy ideology) and responses to each of the local issue questions in our survey. This exercise allows us to shed light on the types of local issues for which national policy ideology is more (and less) predictive. This part of our analysis was not included in the pre-registration, and therefore should be considered preliminary. Figure 3 plots the correlation coefficient between the national issues scale created in the previous section and each of the individual local policy items in our survey.
Figure 3. Local issue item correlations with national policy scale, study 1.
As can be seen in Figure 3, national policy ideology is moderately correlated with attitudes toward some local issues (affordable housing, same-sex benefits, pre-education, internet access, deficits and taxes, require recycling, rent controls, NIMBY, and public transit); weakly correlated with others (increased parking, land-use limits, and aesthetic impact); and negligibly correlated with still others (business tax breaks, employee pension cuts, cut social services, and increase local police).Footnote 9
These findings are generally consistent with the main results we have already presented, while also adding important caveats. While our scaled measure of national policy ideology is strongly correlated with our scaled measure of local policy ideology, the scaled measure of national policy ideology is noticeably less strongly correlated with certain individual local policy items, and this is especially so for local tax breaks, employee pension cuts, social service cuts, and local policing. This observation resonates with the results of the factor analysis we presented earlier, which similarly highlighted the distinctiveness of attitudes toward this subset of local policy issues. These findings also underscore the important methodological point that scholars must not assume that measures of national policy ideology predict preferences toward specific local issues; and, in particular, that they should not use measures of national policy ideology as proxies for specific attitudes relating to local business tax breaks, employee pensions, cuts to social services, or local policing.
As a second analysis that we did not pre-register, we also explore the degree of partisan polarization for both the national and local issue scales created from our IRT models. Figure 4 presents the average placement of Democrats and Republicans on both scales. We included party identifiers and independents who report that they lean toward one party as partisans. Unsurprisingly, the plot shows that partisans are especially polarized on national issues—the average placement for Democrats is about 1.2 standard deviations to the left of the average placement for Republicans. We also see clear partisan differences on local issues, though not quite as much polarization. For local issues, Democrats are about 0.7 standard deviations more liberal than Republicans. Thus, it seems clear that partisanship is strongly associated with attitudes on local issues, though not quite as much as it is for national issues.
Figure 4. Average placement of Democrats and Republicans on local and national issue scales.
Note: Plot shows the average placement of Democrats and Republicans on the national and local issue scales produced from the IRT models along with 84 percent confidence intervals.
3. Study 2
3.1 Data and methods
To provide further confidence in our main findings from study 1, we undertook study 2 (which was not pre-registered with the OSF, but follows a similar approach as study 1). The data for study 2 come from a module of 1000 respondents on the 2021 CES, which was in the field from November 3rd to December 7th, 2021. We conducted this additional study because the CES is considered a higher-quality academic survey and we wanted to explore whether the findings from study 1 would be robust on this alternative platform. The survey instrument measured respondents' positions on six local issues and 40 national issues. The national issues all came from the common content questionnaire, which can be found at https://doi.org/10.7910/DVN/OPQOCU. The wording for the questions about local issues can be found in the online Supplementary materials, Section 7.
4. Results
Given that the primary purpose of study 2 is confirmatory, we focus on assessing the direction and magnitude of the correlation between local policy ideology and national policy ideology. If the structure of local policy ideology is like that of national policy ideology, and the two quantities are related, there should be a substantial positive correlation between the two measures.
As in study 1, we construct separate measures of local policy ideology and national policy ideology, using an IRT response model to scale separately the responses to (1) all the local items and (2) all the national items in our module. In Figure 5, we plot each respondent in study 2 based on their relation to each scale. Again, because both scales are standard normal, the axes can be understood in terms of standard deviations. For example, a respondent with a zero for the local policy scale and a −1 for the national policy scale would have average policy conservatism when it came to local policy questions but would be 1 standard deviation below the mean on national policy items.
Figure 5. Scatterplot of survey respondents' positions on local and national policy attitude scales, study 2.
Figure 5 shows a moderate overall correlation (0.575) between the two scales. By comparison, our measure of national policy ideology is correlated with the conventional 5-point ideology self-placement measure at 0.727. Of course, it is important to note that there are only six local policy items to scale whereas there are 40 national policy items. So, it is possible that this correlation is low because of the much smaller number of local items. To make more of an equal comparison, we also conducted an analysis where we randomly selected six national policy items 30 different times and then correlated each of those national scales with the local scale and with each other. The results from this analysis are presented in Section 8 of the online Supplementary materials, but the average correlation between the local policy scale and the 30 six-item national policy scales was 0.534. By comparison, the six-item national policy scales correlated with each other at an average of 0.762. Overall, in study 2, national policy ideology is quite predictive of local policy ideology, though the relationship between separately created national policy scales is even stronger.
As can be seen in Figure 6, national policy ideology is moderately correlated with attitudes toward some local issues (affordable housing, rent controls); weakly correlated with others (apartment buildings, condemning blighted areas); and negligibly correlated with still others (granting tax breaks to local businesses, increasing parking).Footnote 10
Figure 6. Local issue item correlations with national policy scale, study 2.
At a general level, the patterns of correlation between specific local preferences and the scaled national ideology measure are similar to those that we observed in study 1. In particular, and consistent with study 1, we note that while attitudes toward several issues relating to land use and zoning (affordable housing, rent controls, and apartment buildings) seem to be related to national policy ideology, attitudes toward tax breaks toward local businesses are quite distinctive.
The results of study 2 generally confirm our findings from study 1 about the relationship between preferences toward local issues and preferences toward national issues, though the strength of this relationship is somewhat stronger in study 1 (possibly because we have more, and more varied, measures of local policy preferences in study 1). Study 2 reaffirms that attitudes toward many local issues appear to align with attitudes toward national issues, suggesting an important ideological dimension to local politics that may reflect the nationalization (Hopkins, Reference Hopkins2018) of some, though not all, aspects of municipal affairs. Study 2 also confirms the important methodological findings from study 1 that (1) a scaled measure of national policy ideology is highly predictive of a scaled measure of local policy ideology; but (2) preferences toward specific local issues vary notably in the strength of their correlation with the national ideology scale.
Admittedly, there are some differences between our findings in study 1 and those in study 2. In study 1 we find that attitudes toward increased parking are positively but weakly correlated with our measure of national ideology; but in study 2 we find that attitudes toward increased parking are negatively, but negligibly, correlated with our national ideology measure. One important task for future research is to assess how the number and content of questions about local and national issues influence both the overall strength of the relationship between local ideology and national ideology and the nature of the relationship between individual local policy attitudes and national ideology.
Another question for future research is whether we observe differences in the relationship based on geography. Where are relationships between local and national ideology strongest or weakest? One might, for example, theorize that the correlations are weaker in urban areas, with more diverse populations and complex policy options. It is also possible that rural areas, with limited resources, may rely more on national media to get news, which encourages residents to view local issues more through the frame of national politics. To evaluate this possibility, we compared the correlations between local and national policy scales among people living in cities, suburban, and rural areas. The differences across them are negligible (see Section 8 in the online Supplementary materials), although we do not rule out the possibility of some issue differences based on structural differences in social and economic life in these different places.
5. Conclusion
In recent years, scholars have exploited new data sources and novel statistical techniques to reexamine the role of ideology in municipal politics, illustrating that liberal-conservative ideology is related to residents' preferences on local issues and influences the behavior of local elected officials. This research, however, suffers from an important vulnerability—because existing work draws on measures of ideology that are oriented toward national issues, it is unclear whether and to what extent these measures gauge individuals' preferences on distinctly local concerns. This vulnerability means that we have little theoretical or empirical insight into whether and how attitudes toward national issues, and attitudes toward local issues, respectively, relate to each other. This uncertainty also raises questions about the substantive interpretation of apparent correlations between measures of ideology and municipal-level outcomes such as councilor representation or policy adoptions (Anzia, Reference Anzia2020).
Our goal in this study was to directly assess the relationship between citizens' attitudes on issues that are salient to local governments and those that are salient at the national level. This exercise sheds light on whether and to what extent local preferences are related to national preferences, providing fresh insights on the relationship between local policy ideology and national policy ideology. Our work also provides a data-driven assessment of whether and to what extent the use of nationally oriented measures of ideology to gauge preferences on local issues—a common tactic, given the dearth of large datasets with batteries of questions measuring preferences on distinctly local issues—is appropriate. To increase the rigor and transparency of our analysis, we pre-registered the analysis plan, survey questionnaire, and code used in study 1; and we undertook a separate (but not pre-registered) study 2 to confirm the findings in study 1.
The pre-registered analyses demonstrate that citizens do have the same structure of attitudes toward many (though not all) local policies as they do for national issues. While this conclusion was supported by each of the pre-registered tests, perhaps the clearest evidence comes from the factor analysis. A single first dimension from the factor analysis generated significant loadings from a super-majority of both the local and national policy items.
Furthermore, when we scaled local and national items separately, the resulting scales were strongly correlated with each other. In the Lucid data, the correlation between the national and local policy scales was 0.658 and in the 2021 CES data it was 0.575. To contextualize the strength of these correlations, consider two benchmarks we can calculate from the 2022 CES common content data. First, the CES asks respondents whether they would like to see their state increase or decreasing spending in five policy domains (welfare, health care, education, law enforcement, and transportation/infrastructure). When we scale these items using an IRT model, that scale correlates at 0.640 with the national policy items scale. Second, the CES also asks a question gauging support or opposition to sending US troops abroad under six different circumstances (e.g., to destroy a terrorist camp, protect US allies, prevent genocide, etc.). We estimated an IRT model on these items to create a foreign interventionism scale and that scale correlated at just 0.182 with the national policy scale. Thus, this exercise allows us to say that a scale of local policy items correlates with a scale of national policy items at about the same strength as a scale derived from a set of state-level spending questions, and at a much stronger magnitude than a scale of attitudes regarding foreign interventionism.
An additional analysis (not pre-registered) shows that national policy ideology is significantly associated with a variety of local issues. Notably, attitudes toward various issues relating to the size of local government and economic redistribution (which seem to fall along the classic left-right dimension of politics) tend to align closely with attitudes toward national issues. We hasten to add, however, that attitudes toward some local issues, including attitudes toward tax breaks for local businesses and local policing, were negligibly correlated with, if not completely orthogonal to, attitudes toward national issues. Thus, local politics may occupy a liminal space that is neither wholly subsumed by, nor completely independent from, national ideological struggles.
Our findings have important methodological implications for the study of local politics. If attitudes toward many local issues align on the same underlying factor as do attitudes toward many national issues, scholars can more readily draw on theories used to study national politics (e.g., theories of representation, preference formation, coalition-building, and voter mobilization) to study local politics. Our findings also imply that conventional measures of national ideology might be used to characterize individuals' ideologies concerning local issues. This is a major methodological gain, since most data sources used to estimate ideology—agglomerations of large surveys, commercial voter files, etc.—typically do not contain large batteries of questions about preferences toward local issues. The linkages between preferences toward national issues and preferences toward local issues might also help advance an understanding of the degree of ideological polarization at all levels of government, particularly during an era when politics appears increasingly nationalized (Hopkins, Reference Hopkins2018). However, scholars must exercise care in using measures of national ideology to study local politics. Recent research cautions that measures of ideology based on aggregations of preferences across issues capture ideological consistency between policy domains rather than views within domains, and therefore cannot be used to characterize attitudes on specific issues. Scholars must avoid mischaracterizing individuals who have moderate scores according to an aggregated measure of ideology as having generally moderate preferences (Broockman, Reference Broockman2016). Our findings in this paper are consistent with this work; and make clear that scholars must not assume that aggregated measures of national ideology are associated with attitudes toward specific local issues, especially those relating to issues such as tax breaks for local businesses, local policing, or service cuts to local government.
Future research should investigate what makes attitudes about these specific local issues different from the many which are highly correlated with attitudes about national level issues. Previous research points to some fruitful areas of inquiry. Peterson (Reference Peterson1981) famously argued that local politics are distinctive in part because municipalities must compete for economic development; this may help explain why attitudes toward tax breaks for local businesses consistently fail to align with attitudes toward national issues (and other local issues) in our surveys. Another explanation is that some issues at the national level have not (yet) become polarized at the local level. At the state level, Grumbach (Reference Grumbach2018) reports that issues concerning education and criminal justice have not polarized in the same way as taxation, health care, and welfare, perhaps because of the ways that they have been framed in partisan debates at the national level. Similarly, certain local issues may remain (for the time being) the exclusive province of local struggles because they have not been incorporated by elected officials into national partisan debates. Future work should explore a more nuanced understanding of which local issues do, and do not, map well onto conventional liberal-conservative measures of ideology.
Future research should also assess whether and to what extent our findings hold across different types of communities—urban versus rural, racially diverse versus racially homogeneous, “red” versus “blue,” and so forth. Although some classic research (e.g., Peterson, Reference Peterson1981; Oliver et al., Reference Oliver, Ha and Callen2012) emphasized how structural factors tended to homogenize municipal politics, more recent work (e.g., Rodden, Reference Rodden2019, Brown and Enos, Reference Brown and Enos2021) has highlighted growing political, cultural, and economic divergences between differently situated communities, pointing to the possibility that local ideology may play a distinctive role in different contexts. Although a rigorous test is beyond the scope of this paper, one plausible hypothesis is that the relationship between local ideology and national ideology may be stronger in municipalities where the social cleavages that structure national politics (race and ethnicity, income, education, etc.) are particularly strong and salient, because the presence and salience of these nationally oriented cleavages may lead municipal residents to think about local issues in more nationalized ways. Future research should investigate these and other hypotheses to advance our understanding of whether, when, and how much ideology influences local politics.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2024.58.
To obtain replication material for this article, https://doi.org/10.7910/DVN/WDEIZA.
Open access
