Abstract
Policymakers and scholars are increasingly looking to cities to address challenges including income inequality. No existing research, however, directly and systematically measures local political elites’ preferences for redistribution. We interview and survey 72 American mayors—including many from the nation’s largest cities—and collect public statements and policy programs to measure when and why mayors prioritize redistribution. While many of the mayors’ responses are consistent with being constrained by economic imperatives, a sizable minority prioritize redistributive programs. Moving beyond the question of whether mayors support redistribution, we find that partisanship explains much of the variation in a mayor’s propensity for redistribution. Moreover, the impact of partisanship very rarely varies with institutional and economic contexts. These findings suggest that national political debates may be shaping local priorities in ways contrary to conventional views, and that they may matter even more than other recent findings conclude.
Many politicians, policymakers, and academics, dissatisfied with federal and state government, have increasingly pointed to cities as venues for addressing socioeconomic challenges. As former Philadelphia Mayor Michael Nutter succinctly summarizes, “Cities are incubators of change and innovation, and mayors are at the forefront of it all—we get things done” (Mathis 2014). This optimism in cities includes redistributive policy, an arena that influential scholarship (e.g., Peterson 1981) claims cities are constrained from pursuing. For example, New York Mayor Bill de Blasio made redistributive initiatives a centerpiece of his 2013 campaign. Moreover, at the 2014 U.S. Conference of Mayors (USCM) meetings, he joined with several other mayors to form the “Cities of Opportunity Task Force” to investigate cities’ options for implementing equity-oriented policies (Taub 2014).
Urban politics scholarship has long considered cities’ pursuit, or lack thereof, of these types of redistributive initiatives. While some research argues that economic forces induce city leaders to eschew redistributive policy to pursue growth and focus on their tax bases (Peterson 1981), a body of more contemporary work has moved beyond asking whether city-level redistribution occurs to investigate the conditions that affect its likelihood. These studies suggest that urban elites may, in fact, focus on redistribution when institutions (Carr 2015), competitive pressures (Jimenez 2014; Minkoff 2009), and/or public opinion (Einstein and Kogan 2016; Hajnal and Trounstine 2010; Tausanovitch and Warshaw 2014) favor such policies.
We contribute to this contested literature by providing insight into whether and when city leaders prioritize redistributive policy. We do so using new data from survey-interviews of U.S. mayors alongside public statements and policy programs. These data permit us to test these varying competing explanations simultaneously. Although our approach has its own downsides, collecting information directly from city leaders allows us to complement and supplement prior work and avoid some of its limitations (Fenno 1978; Gerber 2013; Gerber, Henry, and Lubell 2013; Perry 1994). It does so in part because (1) none of the available observational data cleanly and directly speak to the issues, and (2) the pertinent theory is as much about elite policy agendas and priorities as it is about policy outcomes.
Specifically, we asked a representative group of approximately 70 U.S. mayors a battery of policy and leadership questions. Our questions require mayors to make explicit and pertinent trade-offs and/or incur opportunity costs to take proredistribution positions. This approach contributes to a growing body of scholarship that uses elite interviews and surveys to explore local policy-making agendas (Gerber 2013; Gerber, Henry, and Lubell 2013). To bolster the survey findings and assuage concerns that they are “cheap talk,” we also collected and coded public statements and policies related to redistribution for all mayors in the sample.
Our findings reveal that redistributive policy is relatively prominent on mayors’ agenda. Interestingly, its prominence varies with mayors’ partisan affiliations and it does so almost irrespective of cities’ institutional configurations and competitive pressures. As in many other realms of American politics, party affiliation matters and indeed dominates other potentially important factors. Thus, our results build upon a nascent body of scholarship which argues that national partisan identification matters in local politics (Einstein and Kogan 2016; Hajnal and Trounstine 2010; Tausanovitch and Warshaw 2014) and indicate that mayors may not be quite as sensitive to economic and institutional constraints as prior scholarship suggests.
Theoretical Expectations
Before we provide a more detailed description of our measures, we begin with a conceptual discussion of redistribution in cities. Following Peterson (1981), we take redistributive policies to be initiatives that “help the needy and unfortunate . . . [and] provide reasonably equal citizen access to public services” (p. 43). This definition encompasses both policies that explicitly redistribute income (e.g., progressive taxation) as well as initiatives targeting poverty (e.g., subsidized housing), whose effects on income inequality are more implicit.
Peterson’s (1981) influential research argues that competition from neighboring cities, along with state and federal regulatory power, makes city leaders unlikely to pursue redistributive policies. In a modified version of this economic primacy argument, Stone’s (1989) regime theory allows for responsiveness to constituent interests, while still emphasizing the predominance of businesses and wealthy residents. All this research leads us to the following hypothesis:
Despite the prominence and influence of this perspective, other work has challenged it and offered reasons to expect at least conditional local redistribution. Thus, while we begin by exploring if city leaders prioritize redistribution, we mostly focus on when they support such policies.
One potential source of variation in redistributive inclinations is political attitudes and/or local preferences. At the national level, the influence of partisanship on voting behavior (e.g., Campbell et al. 1980; Green, Palmquist, and Schickler 2004) and elite preferences and policy choices (Abramowitz 2010; Fiorina, Abrams, and Pope 2005) is well known. Its impact on local politics, however, is hotly contested. Studies of mayoral partisanship have argued that all else equal, electing a Democrat or a Republican mayor will have little effect on policy outcomes (Ferreira and Gyourko 2009; Gerber and Hopkins 2011). 1 These studies attribute the disconnect between mayoral partisanship and city spending to the constraints facing mayors. Despite their use of a sophisticated regression-discontinuity design, there is reason to doubt that these studies actually demonstrate that mayoral partisanship has no effect. For example, large standard errors due to coarse spending data may explain an ostensible null finding. 2
We therefore believe that the question of partisanship’s effect on mayoral preferences is very much an open one. This view is furthered by recent findings suggesting that mass partisanship and ideology have an impact on local spending patterns—a relationship that exists, at least in part, because of public opinion’s impact on elite behavior (Einstein and Kogan 2016; Hajnal and Trounstine 2010; Tausanovitch and Warshaw 2014). Thus, we derive our second prediction:
The extant research on mass preferences and urban policy outcomes uses either presidential vote returns (Einstein and Kogan 2016; Hajnal and Trounstine 2010) or ideological preferences scaled on the national liberal-conservative dimension (Tausanovitch and Warshaw 2014) to assess the connection between public opinion and urban policy outcomes. One implication we might draw from the independent variables used in these studies is that we should anticipate a sharper partisan divide in mayors’ attitudes on policies that are more clearly connected to national policy debates. For example, opinions on progressive taxes might be split by partisanship, while mayors’ views on gentrification—a more localized issue—might be less linked with partisan views. This suggests the following hypothesis:
Importantly, the ability of mayors to pursue their own and/or their constituents’ partisan preferences may be contingent on structural and institutional factors. Building from Peterson’s insights about horizontal constraints, one line of research contends that cities’ propensity to redistribute is shaped by competition from surrounding municipalities. Leaders of cities facing less competition might be more inclined to promulgate redistributive initiatives or at least have the freedom to pursue those policies if they meet elite and/or constituent preferences (Jimenez 2014; Karuppusamy and Carr 2012; Minkoff 2009). Cities can be insulated from (or vulnerable to) competition in a variety of ways. Mayors of cities with many neighbors, for example, might perceive greater interjurisdictional competition and pursue more developmental initiatives irrespective of their ideological inclinations (Craw 2003; Jimenez 2014). Smaller populations (Minkoff 2009) and tax bases (Jimenez 2014; Minkoff 2009) might generate similar results. Indeed, mayors of larger and more economically developed cities might believe they are sufficiently insulated to enact preferred redistributive policies. This possibility leads us to the following hypothesis:
Similarly, mayors’ capacity to act according to their partisan views might also be shaped by cities’ institutional configurations (see Carr 2015, for a detailed review of the importance of municipal institutional form). Lineberry and Fowler’s (1967) seminal work reveals that council-manager cities spend and tax less than nonreform cities, with more recent research confirming that institutional form at a minimum shapes city spending and financing (Carr 2015; Feiock, Jeong, and Kim 2003; Wong 1988). Several studies have also found that the effect of interjurisdictional competition on local public spending and finance is contingent upon municipal structure, with strong mayor cities more susceptible to shifts in interjurisdictional competition than their council manager and weak mayor counterparts (Karuppusamy and Carr 2012). We take these lines of scholarship together to arrive at the following hypothesis:
There are, of course, a myriad of other considerations that might shape mayors’ propensity for redistribution. When possible, we attempt to consider and/or control for them. For example, if local policy is a function of local needs (e.g., Feiock and West 1993; Lineberry 1977), we might expect the mayors of less wealthy cities to prioritize redistribution. Racial dynamics may similarly affect demand for redistribution. In particular, a more diverse population seemingly dampens public support for welfare spending as individuals are reluctant to endorse spending they expect to benefit other racial groups (Alesina and Glaeser 2004; Gilens 1999; Hopkins 2009, though see Hopkins 2011; Rugh and Trounstine 2011). Finally, a growing body of research suggests that the size of a community shapes its politics in a variety of ways, including electoral behavior, elite powers, and constituent preferences (Judd and Swanstrom 1994; Oliver and Ha 2007; Oliver, Ha, and Callen 2012). Specific to the question of redistribution, large cities’ more disadvantaged populations might point their mayors toward more redistribution. Similarly, arguments about interjurisdictional competition militate in favor of the mayors of larger cities redistributing relatively more.
Data and Methods: Original Survey-Interviews of Mayors
In contrast to previous studies of local redistribution—which have focused on spending outcomes—we gathered most of our data directly from mayors. We did so by conducting a set of original hybrid survey-interviews. We argue that mayoral preferences are, at a minimum, important as a consequence of their agenda-setting power. While mayors certainly face an array of constraints when trying to implement redistributive policies (Elkin 1987; Logan and Molotch 2007; Peterson 1981; Stone 1989), as chief executives, they are nevertheless uniquely positioned to put these issues on the agenda and shepherd programs through. Indeed, influence over a city’s budget represents one among many forms of mayoral influence over the urban policy-making process. Studies of budgets may not capture, for example, the impact of mayoral agenda setting on levers of power such as permitting, zoning variances, and negotiations with community groups. Moreover, because our questions tap into constrained preferences (we elaborate more on this below), we believe that we capture true policy priorities rather than unrealistic dreams or socially desirable position taking. For these reasons, we suggest that mayoral agenda-setting comprises an important quantity of interest separate from other (equally important) outcome measures, such as city budgetary allocations.
About half of our observations were collected via in-person or phone interviews in which we walked though the survey questionnaire directly with a mayor, collecting closed-ended data, open-ended responses, and additional elaborations. Each of these conversations lasted between 15 and 30 minutes. The other observations were collected via an online version of the questionnaire, which captured answers to the same open- and closed-ended questions posed in in-person interviews. As we discuss below, the varied methods through which we collected data are indicative of our extensive efforts to connect with a hard to reach elite population. The in-person and phone interviews, and even some of the online responses, often required multiple correspondences with mayoral staff. We offered the mayors maximum flexibility by doing everything from offering an online version to attending one of their major conferences.
The data we use in this article comprise two different groups that were recruited in slightly different ways: (1) mayors of big cities (population greater than 400,000) and (2) mayors of smaller and mid-sized cities. We aggressively (and personally) targeted the entire population of large city mayors (we describe these procedures in greater depth below). Conversely, our recruitment of the smaller and mid-sized cities centered on a generic email. Although this mixed-sampling strategy would be irregular in the context of a mass opinion survey, collecting preferences from elites, such as mayors, necessitates a mix of systematic and convenience sampling. Although we combine these two samples in this article, we also control for population (and indeed have interaction models with population) to ensure that our results are not driven by population skews. Most importantly, as we elaborate below, our sample closely matches the national population of cities and mayors on key indicators.
We devoted more energy toward recruiting and accommodating big city mayors for both substantive and practical reasons. First, large cities with hundreds of thousands of residents are often the subject of prominent urban politics case studies (Kaufmann 2004; Mollenkopf 1994; Sonenshein 1993) and generally have unique policy priorities and powers (Judd and Swanstrom 1994). The behavior and preferences of their mayors may therefore be of particular interest to urban politics scholars, especially because they are more likely to be able to engage in independent policy making. Second, more informally, these are the types of places many people tend to think about when discussing city government and policy. Third, and perhaps most importantly, these cities are also quite scarce in the broader universe of American cities. For example, large cities are a very small percentage of the membership in the USCM, a large professional association: a mere 3% of USCM members have more than 400,000 residents (indeed, only 20% have populations above 100,000). Because these cities are scarce and have the busiest and hardest to access mayors, we made special efforts to recruit them to ensure enough observations from this special group. We went to the summer meeting of the USCM to offer an in-person interview option to the mayors (especially the big city mayors) that attended. 3 Mayors of the 50 largest cities by population and 15 other large city mayors who were registered for the conference received an email invitation that included a scanned personally addressed letter from Thomas M. Menino, the former mayor of Boston, inviting them to schedule an in-person interview with us at the conference or to schedule a phone interview. Moreover, our research team obtained the contact information for all these mayors’ schedulers and/or assistants to ensure that invitations and follow ups were seen by pertinent people in mayors’ offices and that they did not get lost at a mayor’s generic public access email account.
Importantly, however, we acknowledge that the elections and politics of large and small cities differ in a myriad of important ways that may shape our results. In particular, because the politics of large cities tend to be more ideological (Oliver, Ha, and Callen 2012), we may be more likely to find support for H2 than we would have had our sample focused on the kinds of small cities featured in other recent surveys of local elected officials (Butler et al. 2015). In other words, we do not necessarily expect the findings of our survey to fully generalize to every type of city and town; better understanding the behavior of the large- and medium-sized cities that comprise a disproportionate share of our sample, however, will yield valuable insights.
Our data include 16 of the 46 mayors of cities above 400,000 in the United States. Overall, more than one-third of the large cities that received the full-fledged recruitment participated, yielding a sizable and representative (see below) sample of hard to reach big city leaders.
Of course, America’s largest cities contribute only a fraction of the country’s important urban policy making. Therefore, as part of the broader project, we reached out to a much wider array of cities using a less intensive approach. We sent an email invitation to all mayors in the 2014 USCM database. This list includes all the large cities, hundreds of small cities, and everything in between. We opted to recruit broadly and used membership in the association as our survey frame. In essence, we included all cities that see themselves as policy-making cities (regardless of governing structure) as indicated by their membership in the association. 4 All the mayors/cities that belong to the association received a more generic email invitation (to their official but not necessarily direct or personal accounts) and a similarly generic follow-up. We offered them the same wide range of options for participating, and most of the smaller city mayors participated online or over the phone. 5
In sum, the data we analyze below come from two closely related samples: (1) an intensively recruited group of all the large cities in which we had approximately a 33% response rate and (2) a much more passively recruited group of “all cities” in which the response rate was significantly lower (5%). Because we are studying elites in their professional capacity and asking them questions about their in-office preferences, we believe the most important place to check for representativeness is in the traits of the cities the mayors lead, just as one would check the demographics of Congressional Districts to evaluate the representativeness of a sample of legislators that focused on their priorities and voting. The participating mayors hail from 30 different states and all regions of the country. 6 Table 1 uses 2012 demographic data from the U.S. Census’s American Community Survey 7 to illustrate how our sample demographics align with those of all the nation’s cities. We split the demographic comparison, and some of our analysis below, into “big cities” and “small cities” using 400,000 as our cut-point. These demographic comparisons demonstrate that despite some minor population count skews, the cities that responded generally look like American cities as a whole. Although the participating cities are slightly Whiter and less Hispanic, these differences are minor. Most importantly, given our focus on redistributive policy, our sample’s economic characteristics almost perfectly match those in the full set of cities. Thus, we can discount some of the most obvious and problematic potential skews. Notably, the in-sample mayors do not represent constituencies with abnormal needs for redistributive policies.
Comparison of Average Traits of Cities in Our Sample to all Cities.
Note. Some numbers are rounded. Not all mayors answered all questions. We included all mayors that completed the open-ended priorities and challenges section of the survey in these demographics. All data are from the 2012 American Community Survey and the Office of Management and Budget (we use the Office of Management and Budget’s 2013 list of principal cities for classification). Cities with less than 30,000 people are excluded. (Our smallest is approximately 28,000 people). TELs = Tax and Expenditure Limits.
A second obvious area of concern would be partisanship. We used a couple of different metrics to ensure that our sample did not have a partisan skew—a particularly important check given our focus on partisanship and the fact that a former Democratic mayor participated in recruitment. First, we compared the proportion of our sample that was Democratic with the overall national share using data from Gerber and Hopkins (2011). The two-party partisan split in our data is 65% Democrat. This is virtually identical to the figure included in the appendix in Gerber and Hopkins (67%). Second, we measured the mass partisanship in our sample relative to cities across the country using 2008 Democratic presidential vote share from Einstein and Kogan (2016). The average partisan composition of our sample is virtually identical to that of cities as a whole (the comparison is displayed in Table 1). To ensure that these average comparisons did not mask a bias toward political extremism, we also compared the distribution of the Democratic vote share in our sample relative to cities nationally. Again, we found remarkable similarity: The percentage Democrat at the 25th and 75th percentiles of our data never differed by more than three percentage points from their counterparts in the national data. Participating mayors, thus, lead cities that are politically representative of country as whole. They are not, for example, from a mix of ideologically extreme places that cancel each other out in aggregate statistics.
Third, we also check for institutional representativeness. Using data from the International City/County Management Association (ICMA 2011) and Strong Mayor Council Institute (Strong Mayor Council Institute 2011), we find that in-sample cities are remarkably representative in their institutional configurations. Although our sample features a slightly larger proportion of mayor council cities—unsurprisingly, given our targeting of large cities—the proportion of cities that are council manager systems is identical to that in the country as a whole.
Fourth, we investigate whether our cities are representative in the state legal contexts they face. Using data from the National League of Cities (Hoene and Pagano 2015), we explore the proportion of cities that are located in states with (1) no Tax and Expenditure Limits (TELs), (2) less binding property tax limits, (3) potentially binding property tax limits, and (4) binding property tax limits and general limits. The National League of Cities categorized states as having “less binding” limits if solitary limits are easily bypassed; for example, “a rate limit alone might be circumvented by raising assessments, or an assessment limit alone might be circumvented by raising the property tax rate” (p. 13). A “potentially binding” limit is one in which limits are less easily bypassed: “there is either a levy limit . . . or some combination of rate and assessment limits together, thereby negating the ability of localities to circumvent limits” (p. 13). The National League of Cities also classifies general revenue and spending limits in isolation as “potentially binding.” States that have both binding property tax limits and general revenue and spending limits are considered “binding.” The NLC’s approach is rooted in public administration scholarship (Mullins and Wallin 2004).
Once again, participating sample cities are generally quite representative of those in the country as a whole. While our sample exhibits some small deviations—with a slightly higher percentage falling into the no TELs and binding TELs categories—in general, it largely mirrors cities nationwide.
As we noted above, the most important areas to test for representativeness are those that comprise city, constituent, and/or partisan traits that could directly speak to needs or preferences for redistribution. Nevertheless, it is also possible that we obtained a skewed sample of mayors that is masked by a representative sample of cities. Therefore, we also used biographies on city websites supplemented with Google searches to collect data about the mayors themselves. We collected these data for all cities in the United States with more than 400,000 people and for a random sample of 50 smaller cities. We focused on factors (in addition to partisanship) that relate to (1) the propensity to participate given our recruitment tactics and (2) the propensity to endorse redistributive policies. Recruiting participants at the USCM meeting using a letter from former Boston Mayor Thomas Menino could induce two types of bias. One possibility is that we ended up with an unusual sample of mayors who were close with Mayor Menino. For example, our sample might comprise older mayors with whom Mayor Menino worked for years. This was not the case. The ages of participants closely mirror the broader populations. In fact, if anything, the large city mayors were slightly younger as a group. We also did not obtain a sample dominated by mayors from the northeast; instead, our participating mayors were geographically representative of the country as a whole. Finally, as we elaborated above, we also did not get an unusual partisan skew that one might expect if we obtained a sample dominated by Mayor Menino’s former Democratic allies. A second possibility is that using the conference would result in a sample of extraordinarily well-networked and/or ambitious mayors. This concern would most apply to the smaller cities since smaller city mayors who attend the national conference may be especially different from those who do not. Because attendance at the 2014 conference was endogenous by default, we use attendance at the 2015 summer conference as an indicator of networkedness. We find no differences within the critical smaller cities group (37% vs. 34%, χ2 p = .76). A higher fraction of the big city mayors we spoke with attended the 2015 conference, but this difference is also not statistically significant (p = .23). Indeed, because of the small number of observations, if only two large city participants switched behaviors, the ostensible difference would disappear.
One final possibility, given our focus on redistribution versus development trade-offs, is that mayors with business backgrounds could have different views. Thus, we coded whether a mayor included a job like “businessman” in his or her biography. Both larger and smaller city mayors in our sample were slightly more likely (but not significantly so, p = .29 and p = .22) to have business backgrounds than the corresponding comparison groups. Although not a statistically significant result, we are attentive to the possibility that this slight skew toward business backgrounds might bias our results in favor of H1, with mayors from the business community more inclined toward development in lieu of redistributive policy. Last, but perhaps most importantly, we reemphasize the fact that the survey was pitched as a general survey about city leadership. It was not publicized as a survey about inequality or redistribution or even economic policy. Thus, it is very unlikely mayors’ participation choices were driven by their views on the issues we report on in this article.
Measuring Constrained Redistributive Preferences
Eliciting meaningful responses is critical to addressing the questions we seek to answer. We thus paid close attention to question wording and design. Rather than attempt to devise one perfect way to capture redistribution preferences, we adopt a triangulation strategy in which we rely on different styles of questions and analysis. Most critically, we tried to design questions to capture mayors’ professional constrained preferences. As we noted earlier, simply asking if mayors believe inequality is a problem, or asking them about federal programs would not be very informative. Instead, we aim for the constrained preferences at the heart of the arguments that cities do not redistribute and that mayoral partisanship is inconsequential.
Perhaps the most direct way we measure mayors’ preferences is by asking them two open-ended questions about their agendas. In one, we simply ask, “What are your current top two policy priorities?” The second related question taps into willingness to expend political capital on contentious policy initiatives: “In the next year, on what two issues do you plan to expend the most political capital?” We coded the answers, however expansive, into a manageable set of categories, for example, “education,” “economic development.” In this article, we are primarily interested in responses that fell into our “Socioeconomic Issues” category, which includes priorities related to poverty, inequality, and affordable housing. We include a full list of answers (anonymized) that fell into this category in the appendix.
One important strength of these questions is that they do not force respondents to name or discuss redistribution. They assess whether inequality and redistribution are top-of-the-head considerations for mayors in comparison with other priorities. Second, these lists already have various institutional constraints baked into them. Although some mayors may place controversial items on their lists, it is less likely that they will include items that they are not serious about or that they have no chance of advancing. In our experience, most of the programs and ideas they discussed were already works in progress. Finally, an additional strength of these questions relates to one of the limitations in prior studies that use spending data. Spending data (primarily collected using the Census of Governments) are necessarily provided in coarse categories. They, therefore, require scholars to make tough choices about what exactly constitutes redistributive spending; these broad categories necessarily miss swaths of redistributive spending happening in other policy arenas, like transit and development, and they do not capture the variation that occurs within categories.
To supplement these open-ended questions, we also analyze responses to two questions about policy trade-offs that are likely relevant to many mayors. These questions explicitly capture constrained attitudes toward inequality. In each (full wording below where we report the results), we pose a trade-off and ask mayors how strongly they agree or disagree. One pits fighting inequality against the possibility that doing so will adversely affect the tax base. The other juxtaposes rising property values against the displacement of some lower-income current residents. The first of these trade-offs focuses on income inequality, a prominent and partisan national issue. The second, concerns gentrification and taps a more local set of redistributive concerns. Combined, they help us evaluate H3 (National Politics). 8
Although we believe these questions elicit constrained preferences and not simply cheap talk, we also collected data to provide two more “objective” or verifiable measures based on the ideas and programs mayors are touting. First, using mayors’ and cities’ official websites, we collected all public statements from mayors in our sample endorsing redistributive policies in the year after our survey was conducted (June 2014–June 2015). These statements include press releases and public proclamations/addresses (such as State of City speeches) included on mayoral and city official websites. Second, again using mayors’ and cities’ websites, we investigated whether our each mayor in the sample helped to implement any concrete programs targeting inequality (or, at least, advertised the implementation of these programs on their websites). All press releases include the mayor explicitly endorsing a policy proposal and/or discussing a program she or he is implementing (public proclamations/addresses are authored by the mayor and thus already contain these clear links between mayor and policy). Below, we begin reporting results with these data before turning to the survey.
Measuring Independent Variables
To measure our key independent variable—mayoral partisanship—we asked mayors on the survey for their partisan identification, regardless of whether they run with party labels. For those who did not provide this information, we searched online for any records of party labels or connections to party politics. Specifically, we conducted two separate searches per mayor; one with the mayor’s name and “Democrat,” the other with the mayor’s name and “Republican.” For each search, we looked for evidence of: (1) party endorsements of the mayor, (2) mayoral attendance at party events, and (3) mayoral endorsements of party figures. If mayors evinced party associations for one party, we classified them as a member of that party (none of the mayors in our sample had connections with both parties). Those mayors who did not appear to have party links based on these searches remained unclassified by party. To assess economic pressures, we use three different measures: (1) the number of per capita general purpose local governments in a city’s surrounding metropolitan area, 9 (2) city population, and (3) city median property values. 10 To capture city institutional form, we include a dichotomous measure coded 1 if a city is governed under a strong mayor system and 0 if not. This simple distinction is widely used in urban politics and public administration research (Carr 2015).
In our statistical models, we also include a number of controls. First, to address the potentially confounding impact of mass partisanship, we include 2008 city presidential vote share from Einstein and Kogan (2016), the largest available data set on vote share at the municipal level. Unfortunately, such models cannot neatly parse mass partisan effects from a mayor’s personal affiliation. As we noted earlier, mass partisanship likely contributes significantly to mayoral partisanship and is thus subject to posttreatment bias. These issues would be more problematic if our central goal was to separately identify the effects of mass and mayoral partisanship on mayoral preferences. Instead, we are simply making an argument that mayors’ professional views about redistributive initiatives are filtered through a national partisan lens. Whether that partisanship stems from mass or elite divisions is beyond the scope of our analysis. It is fruitful ground for future research.
Results
Although the survey results comprise the bulk of the analysis, we begin with the data we collected on public statements and actual policies. We coded (see above) press releases and other articles from mayors’ and cities’ websites to investigate (1) whether a mayor publicly endorsed policies redressing income inequality/poverty in the year after we conducted our survey and (2) whether a mayor helped to implement concrete policies targeting these issues in the year after our survey (or, at least, advertised the implementation of such a policy). These data serve two related purposes. For one, they provide strong support for the partisan split in municipal redistributive priorities on which much of the survey analysis focuses. In fact, they portray the strongest partisan divide of any metric we use in our triangulation strategy. Second, in doing so, they provide a preemptive robustness check on the survey results that follow. Despite our best efforts to elicit real and constrained priorities, we acknowledge that stated preferences in a survey may not fully reflect a mayor’s true commitments or her ability to actually promulgate policies. The overlap between these observational data and the survey data strongly suggests that the survey responses are not just “cheap talk” and indeed capture real and constrained preferences.
Both the statements and programs dependent variables (summarized in Figure 1) contradict H1 and support H2. Almost 30% of mayors made a statement supporting redistributive initiatives, and 20% actually implemented such programs, suggesting that a significant proportion of mayors promulgate policies that conflict with a strict economic imperatives perspective (H1). Turning to the partisan split at the center of this article (H2), 50% of Democratic mayors made statements endorsing programs targeting income inequality on their websites, compared with only 5% of Republicans. Similarly, 35% of Democratic mayors helped implement programs targeting these issues (and included these programs on their websites), while only 5% of Republicans did the same. Both differences are highly statistically significant (p < .001 and p < .05, respectively). These large partisan effects hold when controlling for potentially confounding factors such as mass partisanship, city institutional traits, and economic distress (regression results displayed in Figure 2 and Table A3). Interestingly, the gap between the proportion of mayors who endorsed redistributive programs and those who actually implemented such initiatives suggests that while mayors are often able to implement preferred policies, they do face important limitations. Moreover, mayors in bigger cities appear more likely to endorse and promulgate these programs, whereas mayors in communities with large numbers of local governments are less likely—consistent with arguments that more competitively insulated locales are more apt to prefer redistribution.
Proportion making public statements and pursuing programs by mayoral partisanship.
Marginal effects (from logit models) with 95% confidence intervals for the “public statements” (left) and “programs” (right) dependent variables.
To test H4 and H5 while dealing with the empty cells that such a strong partisan effect creates, we estimate a series of bivariate logit regressions (Democrat vs. Republican for each context subgroup) exploring whether the effect of partisanship varies by institutional and economic contexts (Figure 3). It is possible that an ostensible strong partisan effect is the result of a very strong effect in one subgroup and none in another. Or, alternatively, in some of the models we explore below, it is possible that a middling overall partisan effect is the result of a robust partisan effect among strong mayors (or large cities etc.) and no partisan effect among weak mayors (small cities). Basically, the question is whether subgroup variation is obscuring main effects, creating false positive main effects, or neither.
Marginal effects (from logit models) with 95% confidence intervals for the “inequality statement” (left) and “inequality program” (right) dependent variables.
With substantially more data, we could estimate models with the requisite interactions in them to address these issues. Because we have relatively few data points, and, as important, very few (or no) mayors making statements supporting redistributive policies and/or implementing such programs in some subgroups (attesting to the strength of the party differences), we cannot put much faith in such an approach. Instead, we evaluate the partisan effect in each subgroup of interest separately. Figure 3 displays our subgroup analyses with confidence intervals. Most of the estimates are from simple bivariate logit estimates that allow us to estimate the marginal partisan effect (Democrat relative to Republican) for each subgroup (e.g., strong mayors). In some cases—namely, when there were no strong mayor Republican mayors making public statements endorsing redistribution or implementing inequality-oriented programs—the figure reports the difference (i.e., the percent of weak mayor vs. strong mayor Democrats) with 95% confidence intervals (of the difference estimate) from the Stata Proportion Test function. On these plots, the horizontal lines are 95% confidence intervals around the estimates of the marginal Democratic effect. The light vertical line indicates the “main effect” (all Ds vs. all Rs) to make it easy to see when subgroup effects are significantly different than the baseline effect.
In contrast with H4 and H5, we generally find little evidence that the effect of partisanship varies by institutional and economic contexts. The sole exception is in city population size. Consistent with H4, Democratic mayors in larger cities were more likely to make public statements endorsing redistribution and (especially) implement redistributive policy programs than their counterparts in smaller cities.
Survey Results
We turn now toward exploring the survey results. We begin with perhaps our toughest test by exploring whether mayors cite inequality and/or redistributive concerns as one of their top two (open-ended) policy priorities or political capital expenditures. We use the label socioeconomic equality to refer to redistributive policies. This category encompasses all policies related to inequality, race, and housing. A full list of anonymized policies is available in Tables A1 and A2 in the appendix.
The policies that fell into this category are varied. For example, one mayor described his top priority as an overall focus on “equity.” He worried about not just economic inequality but also incarceration, racial inequality, and “inequality in access to government [and] trust in government.” He contended that “inequality is . . . about the people being estranged from government.” He linked these concerns with concrete policy priorities such as affordable housing, child care, job training, and access to transit for lower-income residents. Another mayor said that one of his top two priorities was “addressing chronic homelessness by moving people to permanent housing.” A third’s comments both highlight the efforts that mayors are making toward redistributive policy and even their willingness to work against economic constraints. He said one of his top priorities was a “collective impact model to address health, education, and financial security” and that one of his two biggest political capital expenditures would go toward generating “business community support for his poverty initiatives.” Other examples of redistributive efforts include one mayor’s initiative to study and address Black male achievement and others’ focus on “living wage jobs.”
Eighteen percent of all mayors offered socioeconomic inequality as one of their top two policy priorities, and 19% did the same for political capital. Comparisons with other policy areas we might expect mayors to mention help provide context. Thirty-three percent cite economic development as a priority and 21% mention infrastructure. Similarly, 26% list economic development as a political capital expenditure, and 24% cite infrastructure. Equity-oriented policies are thus somewhat—but not dramatically—less likely to appear as a top-of-the-head consideration than issues we would expect to find at the top of urban agendas. In contrast with the economic imperatives perspective (H1), almost one-fifth of mayors listed an inequality issue—amid the many policies they could have selected—as one of their top two policy priorities and political capital expenditures. These results provide preliminary support against H1—some mayors do indeed appear to prioritize inequality and redistribution at least as strongly as economic development.
Moreover, consistent with H2, these data provide some preliminary evidence that Democratic mayors are more inclined to prioritize equality-oriented initiatives. Twenty-five percent of Democratic mayors selected a redistributive policy as a top priority, compared with only 9% of Republicans (p value of difference = .13). More starkly, 28% of Democratic mayors chose a redistributive initiative as a top political capital expenditure, while only 5% of Republicans did the same (p value of difference = .03). Regression analyses largely bolster these cross-tabulations. Figure 4 plots the estimated marginal effects (from a logit model) for the partisan variables alongside other control variables which one might expect to affect the likelihood of redistributive policy. 11 In neither of these models is the main effect of a mayor’s party ID significant. It is, however, positively signed in both and substantially stronger in the political capital variable. In sum, the responses offer suggestive evidence of a partisan effect that manifests strongly in the bivariate relationships but becomes more muted when controlling for other variables, which may both affect redistributive propensities and be correlated with the likelihood of having a Democratic mayor.
Marginal effects (from logit models) with 95% confidence intervals for the “redistributive policy priority” (left) and “redistributive political capital expenditure” (right) dependent variables.
Figure 5 turns to exploring whether the impact of mayoral partisanship is conditional on economic pressures (H4) or city institutional form (H5). As with Figure 3, it does so using marginal effects from bivariate logit models estimating the impact of mayoral partisanship in differing institutional and economic contexts (Figure A1 provides simple cross-tabulations exploring these same interactive effects for these models and all subsequent survey models in this article). The left-hand panel shows no strong relationships in the policy priority variable. As described above, overall, Democrats were about 16 percentage points (.16 marginal effect) more likely than Republicans to name a socioeconomic policy priority when looking at the bivariate relationship. This relationship does not quite achieve conventional significance. More importantly, this plot shows that there are no significant subgroup effects. When looking at the policy priority variable, none of the subgroup effects is significantly different from zero. Moreover, none is significantly or substantively different from the main effect or from the companion subgroup.
Marginal effects (from logit models) with 95% confidence intervals for the “policy priority” (left) and “political capital expenditure” (right) dependent variables.
The right-hand panel digs into the interactions in the political capital variable. Here, the bivariate main effect (22 percentage points) is substantial and significant. As the figure makes clear, this main effect is not driven by any large subgroup effects. None is significantly different than the overall effect, and none is significantly or substantively different than its companion subgroup. Directionally, we do see stronger partisan effects in large cities (H4) and strong mayor cities (H5). Although these differences are not significant, they are still noteworthy given theoretical expectations and existing work. Regardless, the overall story here is the extent to which there is a partisan effect in the political capital expenditure models.
Inequality and Gentrification Trade-Offs
We turn now to our second set of measures of mayoral redistributive preferences, closed-ended questions that forced mayors to make difficult trade-offs concerning inequality and gentrification. The first of these presents mayors with a trade-off between reducing inequality and harming the interests of businesses and wealthier residents. Specifically, we asked mayors how much they agreed or disagreed with the following statement:
Cities should try to reduce income inequality, even if doing so comes at the expense of businesses and/or wealthy residents.
As with the responses to the open-ended questions, responses to this first trade-off question largely contradict H1. A significant number of mayors do, in fact, prioritize redistribution even when weighed against economic development and tax-base considerations. Just below one-third of mayors agreed—a sizable number in light of the economic imperatives arguments. Although we certainly do not want to understate the predictive power of the economic imperatives literature—55% of mayors opposed the trade-off—the fact that any mayors, let alone one-third, are willing to sacrifice important components of their cities’ tax bases to ameliorate income inequality is striking. Moreover, the variation in responses helps to validate our claim that the question taps into real trade-offs in a meaningful way. Mayors did not all cluster on what some might consider the politically correct answer. One mayor of a mid-sized city said of addressing income inequality locally, “It is hard. Our city is not that big. It is really important but the city has limited capacity.” Another who took a position against making the trade-off nevertheless observed: “I do not think cities should try to get inside people’s pocketbook . . . but we need a more progressive tax structure for city services such as water rates.”
The inequality trade-off does, however, provide evidence for H2: 53% of Democratic mayors agreed with the trade-off, compared with only 6% of Republicans. Figure 6 uses regressions to explore whether these partisan differences hold when we control for other plausible drivers of mayoral attitudes toward redistribution. All models are ordinary least squares (treating the underlying 5-point scale as continuous) 12 with coefficient estimates and 95% confidence intervals illustrated in Figure 6 and available in table form in the appendix (Table A5).
Coefficient estimates for full Ordinary Least Squares models (with 95% confidence intervals) with inequality trade-off (left) and redistributive gentrification trade-off (right) as dependent variables.
The inequality trade-off results in the left panel provide powerful evidence for H2: the coefficient on mayoral partisanship is positive and highly statistically significant, revealing that Democratic mayors were much more likely to support redressing income inequality, even if it came at the expense of wealthy taxpayers and businesses. The effect is substantively large. It is approximately one point on a 5-point scale. Moreover, mayoral partisanship is the only noteworthy effect in this model.
As above, we now check whether the partisan effect is simply a strong main effect, or whether it is driven by subgroups with very large partisan effects. Analogously to Figure 5, we assess the marginal partisan effect (Democrat relative to Republican) in each subgroup by looking at the bivariate relationship (once again, cell size issues prevent us from confidently estimating full interactive models). The left panel of Figure 7 shows these effects for the inequality trade-off. As before, we show the overall Democratic effect alongside subgroup effects. This plot depicts a strong overall partisan effect (more than one point on a 5-point scale)—consistent with H2—but no evidence of party effects that are conditional on institutional structures or economic power, in contrast with H4 and H5. None of the subgroup effects are substantially or significantly different from the main effect. Indeed, as with the main effect, they are all positive and statistically significant from zero. Regardless of city institutional form, wealth, size, and competitive context, Democratic mayors are more likely to endorse the inequality trade-off to similar degrees.
Marginal effects (from Ordinary Least Squares models) with 95% confidence intervals for the “inequality trade-off” (left) and “gentrification trade-off” (right) dependent variables.
We now turn to the second policy trade-off which, as we indicated above, concerns gentrification. This issue also speaks to economic inequality and a city’s tax base, but in different ways than the other trade-off. It does not map onto national partisan divides as neatly as more general questions of redistribution and inequality. Rather, it is more of a local issue. The exact wording of this second trade-off statement is,
It is good for a neighborhood when it experiences rising property values, even if it means that some current residents might have to move out.
13
Here, unlike with the inequality trade-off, the more redistribution-oriented position is disagreeing rather than agreeing with the statement.
The cross-tabulations from this question reveal that—consistent with H3—national partisan alignments are not associated with preferences on this more local redistributive issue. In general, mayors are more evenly divided. Approximately 40% agree with the gentrification trade-off, 30% disagree, and 30% neither agree nor disagree. Indeed, marginal effects displayed in the right-hand panel of Figure 6 confirm the non-relationship between mayoral partisanship and preferences for gentrification.
Since the main effect on this trade-off was essentially nil, we turn to exploring whether the lack of main effect is a consequence of off-setting subgroup effects. As the right panel of Figure 7 shows, there are almost no noteworthy partisan effects in any subgroup. Democrats and Republicans are essentially evenly split on this issue whether they are in big or small cities, strong or weak mayor systems, and irrespective of at least some economic conditions and threats. The only variable that appears to shape the relationship between partisanship and views on this trade-off is median housing prices. Here, in contrast with our predictions in H4, Democratic mayors in wealthier cities are more likely to agree with the trade-off, thus taking the less redistributive position. Combined with the other trade-off question, these results generally suggest that there are significant differences in how mayors think about the trade-offs inherent in addressing income inequality and gentrification.
Summary
We began by announcing a “triangulation” strategy. Having reported results from six different measures, we can now put it all back together. Overall, we find strong evidence of general partisan differences in redistributive preferences and little evidence of conditional party differences. Both objective measures and two of the four survey measures point substantially toward partisan differences. A fifth measure, the policy priorities one, is directionally consistent. The only measure that does not evince even suggestive evidence of party differences is the gentrification trade-off. This is exactly the place we would least expect a partisan gap because it is a purely local issue. Moreover, when we do find support for party effects, these effects are generally independent of institutional or economic factors.
Conclusion
Our analyses represent the first attempt to systematically assess mayors’ constrained preferences and priorities on pressing local issues. The ability to observe city leaders’ views offers unique leverage for exploring the mechanisms undergirding local redistribution. While our data have drawbacks—like any social scientific method—they provide a complementary lens through which to view mayoral responses to structural constraints. Indeed, our survey-interviews blend breadth and depth to provide a generalizable but nuanced portrait of mayoral policy preferences. They also offer a starting point for a number of future research studies; specifically, going forward, we hope to expand upon our analysis to further link mayors’ preferences with public statements and campaign rhetoric, exploring when these important quantities of interest converge and diverge.
Moreover, our findings offer evidence of a broader story about the nationalization of local politics. A significant segment of mayors are actively promoting initiatives in a salient policy arena previously thought to be outside their purview. What’s more, their preferences for initiatives in this sphere are consistent with the national parties’ positions. Local politics therefore may encompass a wider array of policies than scholars have explored and may prove to be fertile ground for evaluating many hotly contested political science theories concerning national politics.
Footnotes
Appendix
Acknowledgements
The authors thank Boston University’s Initiative on Cities, specifically Katharine Lusk, Conor LeBlanc, Tom Menino, and Graham Wilson, for their collaboration and support in the design and implementation of the 21st Century Mayors Survey. They also thank Paul Lewis, Cathie Jo Martin, Kris-Stella Trump, and Chris Warshaw for very helpful comments on the manuscript and Robert Pressel and Ramya Ravindrababu for outstanding research assistance.
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) received no financial support for the research, authorship, and/or publication of this article.