Which Citizens Do Elected Officials Target With Distributive Spending? A Survey Experiment on U.S. Municipal Officials

Abstract

Research is mixed as to whether politicians target swing voters or core supporters with distributive spending and whether citizens’ turnout affects this strategy. I use a novel data set and research design to examine this—a survey experiment on elected municipal officials. Respondents indicated which of two neighborhoods to target with a local project. I find that local officials, on average, target swing neighborhoods over core ones because they believe that swing voters are more likely than core voters to electorally punish politicians for targeting other groups. Yet, a large proportion still target core voters but not for reasons consistent with extant theory. Officials generally target high turnout neighborhoods over low turnout ones but under certain conditions are also willing to target lower turnout citizens. These findings point to the need for ongoing work to identify the conditions under which officials will target core or swing voters.

Keywords

distributive politics local politics survey experiment

The central premise of distributive politics is that politicians target citizens with public spending to improve their electoral prospects. But given that politicians do not have unlimited access to such distributive goods, which types of citizens do they target? Do they focus their efforts on core supporters, those whose support for the politician is already the strongest? Or do they use distributive spending to try to woo swing voters, those whose loyalties are unclear? If politicians believe they can count on the support of their core voters, then the obvious vote-maximizing strategy is to target swing voters (Lindbeck & Weibull, 1987). However, if politicians are risk averse and uncertain as to how distributive spending will affect swing voters’ electoral behavior, targeting core voters may be their safest bet (Cox & McCubbins, 1986).

A complicating factor for a candidate deciding whether to target core or swing voters is whether these citizens will actually turn out to vote. If distributive goods only buy support and do not affect turnout, then politicians should target citizens who are more likely to vote to minimize wasting effort on nonvoters. However, if politicians believe that distributive spending has a mobilizing effect (Chen, 2013; De La O, 2013; Matsubayashi & Jun-Deh, 2012), then they have an additional incentive to target core supporters, particularly core supporters who are less likely to turn out but have the potential to do so if mobilized (Nichter, 2008).

The existing literature does not provide a clear answer to these questions. Foundational theories make contradictory predictions, while empirical results are mixed and rely on research designs that are unable to identify intended targeting strategies from observable budget allocations.

It is on this latter point that I hope to contribute to this literature. To do so, I use a novel approach for the study of legislative behavior: a survey experiment on elected municipal officials from across the United States. This approach directly measures the perceptions of the population of interest and provides causal evidence of politicians’ strategic decision-making. In the survey experiment, respondents read a vignette about a city councilor who must decide which of the two neighborhoods to target with a local road repair project. Respondents only know how the neighborhoods differ in terms of two factors that are randomized across neighborhoods: (a) the neighborhood’s support for the incumbent and (b) its residents’ turnout propensity. The respondents are instructed to advise the city councilor on which neighborhood to target. In a follow-up experiment, respondents predict what the electoral ramifications would be if the city councilor chose one of the neighborhoods over the other. This setup allows for an analysis of which types of voters policymakers believe are the most electorally advantageous to target and why.

Overall, I find that policymakers target swing neighborhoods over core ones and high turnout neighborhoods over lower turnout ones. Although policymakers believe that distributive spending is a net benefit for an incumbent, regardless of which neighborhood is chosen, they target the swing neighborhood over the core one because they believe that swing voters, relative to core voters, are much more likely to electorally punish incumbents for directing spending to other groups. In short, policymakers believe they can take the support of core voters for granted. At the same time, support for this swing voter strategy is not overwhelming. Nearly 43% of respondents still targeted core supporters over swing voters, suggesting that current theory, which often argues for one targeting strategy, is not sufficient for explaining elected officials’ behavior and perceptions on this front.

With regard to turnout, I find that local policymakers overwhelmingly believe that targeting citizens who have a high propensity to vote over those with a low chance of doing so is the vote-maximizing strategy. In addition, policymakers are uncertain as to whether the distributive spending will boost turnout in the targeted neighborhood. As such, they would rather target certain voters than risk wasting distributive goods on an attempt to mobilize those with a lower propensity to turnout. Although officials favor targeting swing over core and high turnout over low turnout, the interaction of voters’ turnout and support has some effect on policymakers’ targeting strategy. All else equal, officials are slightly more likely to target a neighborhood if it consists of high turnout swing voters or low turnout core voters. In a second study where the turnout between citizens is not as drastic, I find that officials favor the higher turnout residents when choosing between swing voters but favor the lower turnout residents when choosing between core supporters. Thus, the size of the difference in turnout affects the extent to which officials consider the interaction of constituents’ support and turnout.

This article makes several contributions to the distributive politics literature. The first is its use of a novel data set and research design to help adjudicate between competing theories on an important question in distributive politics. The responses are from actual elected officials who make distributive choices in the real world and whose motivations are in line with relevant theory. Second, the analysis not only measures whom politicians would target but also examines why they would target one type of citizen over another. I primarily focus on respondents’ choice between core and swing voters and fail to find evidence supporting the assumptions underlying Cox and McCubbins’ (1986) core voter model. This suggests that conditional theories (e.g., Fleck, 1999; Hirano et al., 2009) might provide a better explanation of why some policymakers believe targeting core voters is more electorally advantageous. Third, the finding that high turnout voters are rewarded for their participation (see also Martin, 2003) has potentially negative implications for representation at the local level, given the skew in who participates in local elections (Anzia, 2013; although see Oliver et al., 2012).

Whom Should Policymakers Target and Why?

In this section, I lay out the main predictions of the theories between which this article helps adjudicate. I begin with the swing voter models. In these models (e.g., Dixit & Londregan, 1996¹; Lindbeck & Weibull, 1987; but see Stokes, 2005 for a slight alternative), citizens’ support or ideology is conceptualized as their affinity for two opposing candidates in an open-seat race independent of any distributive promises made by either candidate. Candidates are assumed to have no chance of winning over their opponents’ supporters, so they have to decide whether to target a group of their core voters (i.e., citizens who ideologically favor the candidate over the other one) or a group of swing voters (i.e., citizens who are ideologically indifferent between the two candidates). The prediction that politicians will target swing over core stems from the argument that core voters “cannot credibly threaten to punish their favored party if it withholds rewards,” but swing voters can. Thus, politicians “should not waste rewards on” their core supporters (Stokes, 2005, p. 317). These models make the following hypotheses:

Hypothesis 1 (H1): Target Swing Voters: All else equal, policymakers prefer to allocate distributive benefits to swing voters over core voters.

Hypothesis 1.1 (H1.1): Loyal Core Voters: Policymakers believe that core voters will punish them less than swing voters for targeting benefits to other voters.

Others (Cox et al., 1984; Cox & McCubbins, 1986) argue that policymakers actually have a stronger incentive to target core voters over swing voters, given the following two assumptions: (a) politicians are risk averse and (b) citizens’ support for the candidate correlates with the politicians’ familiarity with those citizens. In this framework, core voters are “well-known quantities” to the candidate and have consistently supported her in the past (Cox & McCubbins, 1986, p. 378). She knows how they will react to the allocation of distributive goods. Swing voters, on the other hand, are less familiar or “unattached” to either candidate. They make for “riskier investments” both because the politician is uncertain how distributive benefits will affect swing voters’ electoral behavior and because swing voters may be targeted by the other candidate. As such, risk-averse politicians will “over-invest” in core voters. This leads to the following hypotheses:

Hypothesis 2 (H2): Target Core Voters: All else equal, policymakers prefer to allocate distributive benefits to core voters over swing voters.

Hypothesis 2.1 (H2.1): Risk Aversion: Policymakers’ willingness to allocate distributive spending to core voters increases in their aversion to risk.

Hypothesis 2.2 (H2.2): Uncertain about Swing Voters: Policymakers are less certain about how swing voters react to the allocation of distributive spending than how core voters do.

Subsequent models have attempted to generalize when politicians should target core or swing voters by considering the effects of other factors such as the “leaky bucket” of government transfers (Dixit & Londregan, 1996) or primary elections (Hirano et al., 2009). In this article, I examine how citizens’ propensity to turnout affects politicians’ targeting strategy. If policymakers believe that distributive spending also has a mobilizing effect—and some research suggests that distributing spending does have such an effect (Chen, 2013; De La O, 2013; Matsubayashi & Jun-Deh, 2012)—then politicians have another reason to target core supporters, and specifically, those core supporters who have the potential to vote but would be unlikely to do so without being mobilized. By the same logic, targeting core voters who have a high propensity to turnout would be a waste of resources because the policymaker can already rely on their vote (Dunning & Stokes, 2008; Nichter, 2008²). Among swing voters, policymakers should only target those who are certain to turnout in an attempt to buy their support. Low turnout swing voters, on the other hand, should be avoided because their support is uncertain and politicians would prefer that they stay home on Election Day (Dunning & Stokes, 2008). Based on this logic, I propose the following:

Hypothesis 3 (H3): Target High Turnout Swing and Low Turnout Core: All else equal, politicians prefer to allocate distributive benefits to core voters who have the potential to vote if mobilized and swing voters who are certain to vote.

Hypothesis 3.1 (H3.1): Mobilizing Effect: Policymakers believe that distributive spending increases turnout among the recipients of that spending.

However, policymakers may not believe that distributive spending has a mobilizing effect, or at the very least, they may believe that the mobilizing effect is too small to compensate for the benefits of targeting likely voters. As such, they should target high turnout voters to maximize their reelection chances (Fleck, 1999; Key, 1950; Martin, 2003), leading to the following:

Hypothesis 4 (H4): Target High Turnout: All else equal, politicians prefer to allocate distributive benefits to voters who have a higher propensity to turnout.

Hypothesis 4.1 (H4.1): No Mobilizing Effect: Policymakers do not believe that distributive spending increases turnout among the recipients of that spending.

Empirical Evidence is Lacking

The extant empirical literature³ on the targeting strategies of elected officials does not adjudicate between the competing hypotheses derived above. The evidence is decidedly mixed (Golden & Min, 2013). Some of this ambiguity results from a mismatch between the question of interest and the empirics. For example, many studies examine the allocation of spending across districts when the question is specifically about the allocation within districts (see Cox, 2006). Another mismatch occurs in comparative studies that use data from developing democracies with clientelistic parties (e.g., Calvo & Murillo, 2004; Stokes, 2005). These include studies that test how the interaction of citizens’ support and turnout affect politicians’ distributive strategies (Dunning & Stokes, 2008; Nichter, 2008). As Stokes et al. (2013) argue, the local party workers who affect the distribution of goods in these polities operate under different incentives than the elected officials who are modeled in the theories. Ambiguity also exists among studies in comparative politics that examine the intradistrict allocation of distributive spending in democracies where elected officials, and not party workers, determine those allocations. For example, Case (2001) finds that swing voters in Albania are targeted, while Dahlberg and Johansson (2002) find that core municipalities within legislative districts in Sweden are. One potential source for these mixed results is the difficulty of causal identification in these observational studies. Another, which I will return to again at the end of the article, is that reelection-minded policymakers may use other targeting strategies besides those that have been the focus of major theoretical work (Golden & Min, 2013).

There is less work directly testing these theories in U.S. local politics, which is the context for this article’s empirical test. The most relevant work is from the literature on the distribution of urban services, which finds evidence that suggests that local officials favor their core supporters with benefits. The early consensus in this literature argued that political considerations had no detectable influence on the distribution of urban services even in cities, such as Chicago, where political influence was assumed to be greatest. Instead, other factors like bureaucratic decision-making (e.g., Antunes & Plumlee, 1977; Levy et al., 1974; Lineberry, 1977; Lipsky, 1969; Mladenka, 1980) and economic considerations (e.g., Peterson, 1981; Tiebout, 1956) explained the allocation of government outlays. More recent work overturns this consensus on both theoretical and empirical grounds (e.g., Hajnal & Trounstine, 2010; Meier et al., 1991; Tausanovitch & Warshaw, 2014). Among these studies is a subset that argues that core voters benefit from this political influence (Cingranelli, 1981; Einstein & Kogan, 2016; Koehler & Wrightson, 1987; Miranda & Tunyavong, 1994; Trounstine, 2006). These findings, however, have several limitations. The first concerns their generalizability. All four studies focus on urban cities, and only Trounstine (2006) examines outlays in more than one city (nine, in fact). Whether the results would hold across a broader spectrum of municipalities is unclear. Second, the focus of the studies is to identify whether political variables are associated with the distribution of services and not to identify the targeting strategies examined here; thus, they do not try to test whether officials target core voters over swing voters, all else equal.

An overarching limitation of these and other empirical studies is the difficulty of identifying politicians’ targeting strategy from observable budget outcomes when so many factors affect the budget allocation process. In short, this amounts to a problem of omitted variable bias. Knowing the observational equilibrium outcome of budget allocations does not necessarily reveal politicians’ utility function. Instead, the researcher needs to be able to manipulate the politicians’ choice options, which is the approach taken in this analysis. Given the difficulty of randomly assigning a neighborhood’s support for a politician or turnout propensity, a survey experiment is a natural research strategy for examining politicians’ targeting strategies and overcoming barriers to causal identification. This is the approach used by Stokes et al. (2013) to further examine the targeting strategies of party officials and is the one I use on policymakers.

Survey of Elected Municipal Officials

For this article, I embedded a vignette-style experiment in a large, nationwide survey of elected municipal officials, the 2012 American Municipal Official Survey (AMOS), which was administered by Daniel Butler and the author and included questions for multiple projects from a variety of scholars. The 2012 AMOS was conducted between July and October 2012. To gather the list of municipal officials, research assistants began with the U.S. Census Bureau’s list of 26,566 U.S. municipalities⁴ and then conducted an exhaustive online search for each of these municipalities’ websites to gather the title, name, and email address of the municipality’s elected officials. In most cases, this consists of legislators (e.g., aldermen, city councilors, selectmen, or supervisors) and elected executives (e.g., mayors). Throughout the article, I refer to them collectively as local, city, or municipal policymakers, politicians, or officials. The search for these municipal websites was conducted in random order and resulted in a list of 26,531 elected officials’ email addresses from 5,024 municipalities.⁵

The 2012 AMOS was conducted in five rounds with each elected official randomly assigned to be invited by email to participate in one of the rounds. The questions for this analysis were included in the fourth round of the survey, which was conducted in September 2012. The response rate was around 23%,⁶ on par with recent surveys on elites of this nature (e.g., Fisher & Herrick, 2013; Harden, 2013) and double the typical response rate for contemporary telephone surveys of the mass public. In each survey round, invitees received three email invitations over the course of several weeks. The email invitations contained a link to the survey, which was conducted online using Qualtrics. To keep the survey length to a minimum (around 15 minutes), the questions and vignettes in this analysis were designed to be as brief as possible.

Table A1 in the Supplemental Appendix presents summary data about the cities in the sample. The cities fall under one of the three categories: (a) those where none of the email addresses of the city’s elected officials was found; (b) those where emails were found but none of the officials took the survey; and (c) those where at least one of the officials from that city answered a question in the survey. The mean population of cities in Category 1 (3,127) is much smaller than those in Categories 2 (17,635) or 3 (36,304), which indicates that larger cities were more likely to have websites with emails and their elected officials were more likely to respond. Although the 2,989 cities with responses represent only 11.2% of total cities, they contain 108.5 million inhabitants or 51.2% of the population in the Census Bureau’s list of cities. As Figure A1 in the Supplemental Appendix illustrates, the cities with respondents are also relatively evenly dispersed across the United States.

One important consideration is the extent to which I should anticipate these theories to apply to elected municipal officials because they are often modeled as being less concerned about reelection than their counterparts at higher levels of government.⁷ For example, 90% to 95% of members of Congress run for reelection each year. Around 70% of state legislators do. This number drops to about 45% among municipal officials (International City/County Management Association [ICMA], 2006; Trounstine, 2013). Although this is significantly lower, it is still a sizable portion of these officials. In addition, about 13% report a strong likelihood of running for higher office (Dynes et al., 2019). As I show in Section D of the Supplemental Appendix, the local officials examined in this article express similar levels of ambition as state legislators (Maestas, 2003) for their current office or higher office. The percent who have competitive elections is also similar.⁸ Moreover, research on how different institutions affect local politics (such as having at-large seats instead of districts) regularly posit that these institutional effects are driven by officials adapting their behavior for electoral benefits (e.g., Bradbury & Stephenson, 2003; Langbein et al., 1996; Meier et al., 2005). And though local officials from small towns receive few extrinsic benefits and lack realistic chances of securing higher office, it does not necessarily mean that they do not have ambitions to stay in office (Lascher, 1993; Sokolow, 1989). Overall, past work suggests that many local officials have incentives to strategically target politically important constituents, but to further allay these concerns, in the robustness checks section, I also examine the targeting strategies of the most ambitious municipal officials in our sample.

Vignette and Treatment Conditions

For this study, survey respondents were presented with a vignette-style survey experiment that has two parts. Part 1 sets up the hypothetical scenario and tests whom local policymakers would target with a distributive good. Part 2 tests why policymakers would choose one type of voter over another by examining policymakers’ beliefs about how citizens would respond to different distributive choices made by a hypothetical city councilor.

The text of the vignette used in Part 1 is presented in Box 1. The vignette asks respondents to imagine that they are the campaign manager for a hypothetical city councilor named Mr. Smith, who has to choose between two neighborhoods for a local road project. The vignette explains that the city councilors were deciding the transportation budget and had room for one more project. The next two on the priority list happened to be in Mr. Smith’s district. The demand and need for the projects are the same in both neighborhoods. Unsure which project to support, Mr. Smith asks for advice from his campaign manager who has electoral data about the two neighborhoods. This information is presented in a two-by-two table that displays two pieces of information about each neighborhood: (a) the neighborhood’s support for Mr. Smith (core vs. swing voters) and (b) the turnout propensity of the residents in each neighborhood (high turnout vs. low turnout voters). These characteristics of the neighborhoods are experimentally manipulated and discussed in more detail below. At the bottom of the vignette, the survey asks respondents to indicate which neighborhood project they think Mr. Smith should support. The general framework of this survey experiment is similar to one used by Stokes et al. (2013) to test the targeting strategies of local party workers or “brokers.”

Box 1.

Text of Vignette and Survey Question in Part 1 of the Survey.

Scenario 1, Part 1: Imagine that you are the campaign manager for a city councilor, Mr. Smith, who [barely / easily] won his last election and [will face a high quality challenger in the upcoming elections / expects to do well in the upcoming elections]. In this municipality, the city council sets the transportation budget and is currently deciding which local road repair projects to fund. After allocating most of the budget, the municipality has sufficient funds for one more project. The next two projects with the highest priority are in two different neighborhoods on opposite ends of Mr. Smith’s district. Both projects involve roads that are primarily used by residents in that neighborhood. Both have equal merit and need–he has been contacted by residents in both neighborhoods about repairing the road. Mr. Smith mentions the projects to you and the difficulty he’s having in deciding which one to support. Below is your best guess about the composition of the two neighborhoods based on campaign work, mail-in surveys, voter registration files, census data, etc.—in sum, you’re a very savvy campaign manager.
	Neighborhood 1	Neighborhood 2
Support for Mr. Smith currently % of residents who support Mr. Smith or are undecided about either candidate. (The figures are the same for both voters and non-voters.)	[IF CORE:] 70% support Mr. Smith. 15% are undecided.[IF SWING:] 15% support Mr. Smith. 70% are undecided.	[CORE/SWING]
Expected Voter Turnout % of residents who will definitely vote or might vote. (The figures are the same for both Mr. Smith’s supporters and the undecided residents.)	[IF HIGH:] 65% will definitely vote. 10% could potentially vote if mobilized by a campaign.[IF LOW:] 10% will definitely vote. 65% could potentially vote if mobilized by a campaign.	[HIGH/LOW]
As his political adviser, which neighborhood do you think Mr. Smith should support?• Mr. Smith should push for the project in Neighborhood 1.• Mr. Smith should push for the project in Neighborhood 2.

Before describing the treatment conditions in more detail, I want to explain a few key aspects of the vignette⁹ beginning with our use of a road repair project in the vignette. I did so because it clearly meets the requirements of a distributive good and is the most common service provided by municipalities based on U.S. Census data (U.S. Census Bureau, 2008).¹⁰ As Cox and McCubbins (1986) point out, “capital goods do not easily meet the basic requirements of [their] model” except “when geographic and political groups coincide” (384), which they do in the vignette. Distributive goods should also be finely targetable, which is why the projects are on opposite ends of Mr. Smith’s district involving roads used by local traffic. Projects like a library or park, which are also not as commonly provided by municipalities as road repairs, would not meet these criteria because they benefit constituents beyond the neighborhood in which they reside.

Another important aspect of the vignette is asking respondents to provide campaign advice rather than asking them how they would behave in this scenario. The theories I am testing assume that all else is equal except for the recipients’ support for a candidate and their turnout propensity. Thus, I only presented respondents with the electoral characteristics of the two neighborhoods. However, I worried that some respondents would balk at being asked to make such a politically calculated choice. Asking respondents to provide campaign advice allowed me to structure the vignette in a way that would naturally make sense to our subjects why I only presented them with electorally relevant characteristics about the two neighborhoods. A review of respondents’ open-ended feedback at the end of the survey suggests that this structure was successful. (See Section D and Table A4 in the Supplemental Appendix.)

Part 1 of the survey experiment has three experimental elements that are manipulated. The first is Mr. Smith’s electoral vulnerability, which is mentioned at the beginning of the vignette. This variable was included to test the possibility that Mr. Smith’s electoral vulnerability would affect politicians’ targeting strategy. However, the treatment does not have any discernible effects on respondents’ answers. For the sake of brevity, I ignore it in this analysis.

The other two experimental elements are the neighborhoods’ support for Mr. Smith and the turnout propensity of voters in each neighborhood. Support for Mr. Smith is described in terms of the percent of residents in the neighborhood who currently support Mr. Smith (i.e., core voters) or are undecided between Mr. Smith and his opponent (i.e., swing voters).¹¹ In core neighborhoods, “70% [of residents] support Mr. Smith” and “15% are undecided.” In swing neighborhoods, the numbers are switched: “15% support Mr. Smith” and “70% are undecided.” To avoid any bias that might result from number preferences among respondents, I used the same values (15% and 70%) in both conditions. I operationalize support this way for a couple of reasons. First, this is similar to prior empirical work, which measures citizens’ support based on either their partisan identity; their vote choice in the most recent election (e.g., Dahlberg & Johansson, 2002); their stated support for one party or candidate over another (e.g., Stokes, 2005); or officials’ perception of whether potential voters are supporters or undecided (Stokes et al., 2013). Second, this measure of support likely mimics how elected officials conceptualize voter support across the neighborhoods in their city or district. As Fenno notes in his interactions with members of Congress, elected officials think of their supporters (i.e., their reelection constituency) as those who vote for them in the general election (Fenno, 1977).

The neighborhoods’ turnout propensity is presented as the “% of residents who will definitely vote or might vote.” In high turnout neighborhoods, “65% will definitely vote,”¹² while “10% could potentially vote if mobilized by a campaign.” In low turnout neighborhoods, the numbers are switched: “10% will definitely vote,”¹³ while “65% could potentially vote if mobilized by a campaign.” In describing the low turnout neighborhoods, I emphasized that these voters could turn out in much higher numbers if they were mobilized. According to the turnout propensity model, politicians target core voters who have the potential to vote but would unlikely do so absent being mobilized by a campaign. It is important that this idea is made clear in the descriptions (Stokes et al., 2013). In the robustness checks section, I examine whether the differences in turnout was potentially too large to ever lead officials to target the low turnout neighborhood over the high turnout one.

Subjects were randomly assigned to one of the four treatment conditions. Each one displays a different pairwise comparison of the possible descriptions of the two neighborhoods in the vignette.¹⁴ I also randomized which neighborhood in each cell was the swing or high turnout neighborhood. These four pairwise comparisons were the following:

A swing neighborhood and a core neighborhood that both have high turnout;

A swing neighborhood and a core neighborhood that both have low turnout;

A high turnout neighborhood and a low turnout neighborhood that are both swing;

A high turnout neighborhood and a low turnout neighborhood that are both core.

Comparisons 1 and 2 present respondents with a swing neighborhood versus a core neighborhood. The turnout between the two neighborhoods is fixed. In Comparison 1, both neighborhoods have high turnout. In Comparison 2, both have low turnout. Pooling the results from the respondents assigned to Comparisons 1 and 2 allows me test whether policymakers favor swing voters over core voters (H1: Target Swing) or vice versa (H2: Target Core). Respondents assigned to Comparisons 3 and 4, on the other hand, must choose between a high turnout neighborhood and a low turnout one. In Comparison 3, both neighborhoods are swing, while in Comparison 4, both are core. If H4 (Target High Turnout) is correct, then respondents should choose the high turnout neighborhood in both Comparisons 3 and 4, but if H3 (Target Low Turnout Core and High Turnout Swing) is correct, then respondents should choose the high turnout swing neighborhoods in comparisons 1 and 3 and the low turnout core neighborhoods in Comparisons 2 and 4.

Policymakers Target Swing and High Turnout Voters

The results, displayed in Figure 1, suggest that policymakers target swing voters over core voters and high turnout voters over low turnout voters. As displayed in panel A of Figure 1, 57% of respondents chose the swing neighborhood over the core neighborhood. This percent is statistically significant from 50% at the 0.01 level. When the results in Panel A are split up based on the neighborhoods’ turnout propensity, respondents were slightly more likely to choose the swing (core) neighborhood when turnout was high (low) as predicted by H3 (Target Low Turnout Core and High Turnout Swing). However, this 6-point difference is not statistically significant. Furthermore, respondents did not choose the core neighborhood over the swing neighborhood as H3 predicts should occur when turnout is low in both neighborhoods. Even though the overall results support the swing voter hypothesis (H1) over the core voter hypothesis (H2), there is substantial heterogeneity in their selection, with a sizable portion (43%) choosing the core neighborhood over the swing one. In Part 2 of the survey experiment, I explore why this might be.

Figure 1.

Which types of voters do policymakers target with distributive spending?

Turning to panel B, we see that a large majority of respondents (82%; p = .000) chose the high turnout neighborhood over the low turnout one. When the results are split up based on the neighborhoods’ support for Mr. Smith, a similar pattern to the one displayed in Panel A emerges. The percent of respondents choosing the high turnout neighborhood decreases when both are core compared to when both are swing, but the 7-point difference is not very large (p = .09), and the majority of respondents do not choose the low turnout neighborhood over the high turnout neighborhood as H3 predicts should occur when both neighborhoods are core. Overall, the results in Panel B are more supportive of H4 (Target High Turnout), which predicts that high turnout voters will be targeted, regardless of their support for the incumbent.

To further test how the interaction of constituents’ support and turnout propensity affects officials’ targeting strategy (H3), I can pool the results across Panels A and Panel B and identify any time an official chooses either the High Turnout Swing neighborhood or the Low Turnout Core neighborhood over the other options. When I do so, I find that the likelihood that an official chooses a neighborhood increases 12 percentage points (p = .002) if that neighborhood is full of either High Turnout Swing voters or Low Turnout Core voters. Thus, the logic of H3 still has an impact on officials, but overall, H1 (Target Swing) and H4 (Target High Turnout) are more dominant.

Why Are Swing and High Turnout Voters Targeted?

Part 2 of the experiment is a continuation of the same hypothetical scenario from Part 1 (see Box 2 for the exact wording) and tests why respondents chose one neighborhood over another in Party 1 of the survey experiment. In Part 2, which appeared on a new screen directly following Part 1, respondents are asked to predict how citizens in the two neighborhoods would respond if Mr. Smith decided to target one of the two neighborhoods in the vignette. The neighborhood chosen by Mr. Smith randomly varies. Thus, subjects who were randomly assigned in the first part to see a swing neighborhood versus a core neighborhood are randomly assigned in the second part to one of the two conditions: (a) where Mr. Smith chooses swing over core or (b) where Mr. Smith chooses core over swing. Subjects who saw a high turnout neighborhood versus a low turnout one are similarly assigned to one of the two conditions: (a) where Mr. Smith chooses high turnout over low turnout or (b) where Mr. Smith chooses low turnout over high turnout.

Box 2.

Text of Vignette and Survey Question in Part 2 of the Survey.

Scenario 1, Part 2: What would happen if Mr. Smith had decided to push for road repairs in Neighborhood [1/2], and the project in that neighborhood was mostly completed before the next election?Based on the information provided, please indicate how likely you think each of the following statements would ultimately be true if Mr. Smith had decided to push for road repairs in Neighborhood [1/2] (with 0% meaning never, 50% meaning a complete toss-up, and 100% meaning certain to happen – you can choose any number between 0% and 100%):[STATEMENTS, DISPLAYED IN RANDOM ORDER] [1] The residents in neighborhood [1/2] will be more likely to vote for Mr. Smith in this election. [2] Residents in neighborhood [2/1] will be less likely to vote for Mr. Smith if they find out he chose another neighborhood over theirs. [3] The residents in neighborhood [2/1] will find out that their roads could have been repaired but Mr. Smith chose another neighborhood over theirs. [4] The residents in neighborhood [1/2] will be more likely to turn out to vote in this election. [5] Mr. Smith’s decision will have a positive impact on his re-election chances.

The respondents are then shown a list of five statements describing possible political outcomes resulting from Mr. Smith’s choice. They are asked to rate on a scale from 0% to 100% the likelihood that each of the statements would ultimately be true, given Mr. Smith’s choice. The five statements measure whether respondents agree that Mr. Smith’s choice of one neighborhood (the recipient) over the other (the non-recipient) would

Increase the vote for Mr. Smith in the recipient neighborhood;¹⁵

Decrease the vote for Mr. Smith in the nonrecipient neighborhood, assuming they discover his choice;

Be discovered by the nonrecipient neighborhood;

Increase turnout in the recipient neighborhood;

Have a positive impact on his reelection.

These statements allow me to test all the remaining hypotheses, except for H2.2, which concerns respondents’ risk aversion. How each statement relates to these hypotheses is discussed below in the presentation of the results from Part 2 of the survey experiment.

Officials Believe Swing Voters Are More Likely to Punish

As displayed in Figure 2, the results from Part 2 of the survey support H1.1 (Loyal Core Voters), which predicts that respondents choose the swing neighborhood over the core one because they believe that swing voters are more likely than core voters to punish Mr. Smith for targeting other groups. Figure 2 displays the mean responses of subjects who were assigned to either the “swing” condition (gray bars), in which Mr. Smith chooses the swing neighborhood over the core neighborhood, or the “core” condition (white bars), in which Mr. Smith chooses the core neighborhood instead.

Figure 2.

Why do local policymakers target swing voters?

Although respondents believe that the distributive spending is slightly more likely to increase the vote for Mr. Smith in the recipient neighborhood when the core neighborhood is targeted (mean = 56%, Statement 1) than when the swing neighborhood is (mean = 53%; diff. = −3, p = .284), they believe that swing voters are much more likely to punish Mr. Smith when they are the nonrecipients. According to Statement 2, respondents predict that there is a 59% probability that a nonrecipient swing neighborhood will be less likely to vote for Mr. Smith if they find out that another neighborhood was targeted over theirs. This probability drops to 47% (diff. = 12; p = .000) when the nonrecipient is a core neighborhood. In short, respondents believe they have a bit of leeway to take their core supporters for granted. This 12-point difference in responses to Statement 2 appears to be the main driver across the five statements of why local politicians, on average, target swing over core.

To vote against a politician for targeting other groups, voters must discover the politician’s targeting strategy. As demonstrated in Statement 3, respondents believe that residents in both neighborhoods have an equal and slightly likely chance of discovering that Mr. Smith targeted another neighborhood over theirs (mean = 57% for swing; 56% for core). This finding somewhat counters the assumption in the core voter model that politicians are less familiar with swing voters (H2.1). At the very least, officials believe that swing voters are familiar enough with their officials to know when they’ve been overlooked for a project.

The responses to Statement 4 in Figure 2 also provide evidence in favor of H4.1 (No Mobilizing Effect) and against H3.1 (Mobilizing Effect). Respondents believe that the project is somewhat more likely to increase turnout in the recipient neighborhood when the core neighborhood is chosen (mean = 50%) than when the swing one is (mean = 44%). This difference (p = .026) might explain why respondents were slightly more likely to choose the core neighborhood when both neighborhoods had lower turnout. Regardless, respondents, on average, believe that the distributive spending is more likely to not have a mobilizing effect than it is to have one, which explains why they overwhelmingly targeted the high turnout neighborhood.

Why Did 43% Choose Core?

Although a majority of respondents chose the swing neighborhood, a substantial portion (43%) still chose the core. In this section, I examine possible explanations for this heterogeneity in targeting strategies, beginning with a comparison (displayed in Figure 3) of how the responses in Part 2 differ based on which neighborhood the respondents targeted in Part 1. Even though the respondents’ choice in Part 1 was not experimentally manipulated, examining how their beliefs are moderated by that choice is an initial step in developing a more robust theory of local policymakers’ targeting strategies. One finding that emerges from the difference-in-differences across these two groups (right column of Figure 3) is that respondents’ choice of whom to target results from distinct beliefs about the political ramifications of different targeting strategies. All of the difference-in-differences are statistically significant at the 0.05 level. For those who chose swing in Part 1 (left column of Figure 3), the likelihood that swing voters will punish incumbents for targeting other groups appears to motivate their targeting strategy. Those who targeted core (middle column of Figure 3), on the other hand, think swing and core neighborhoods are just as likely to punish. Their targeting strategy appears to be driven by a belief that distributive spending is much more likely to increase support and turnout among targeted core voters.

Figure 3.

How do responses in Figure 2 differ based on the neighborhood targeted by respondents in Part 1 of the survey experiment?

What explains these different perceptions about the behavior of targeted citizens? The hypotheses derived from the core voter model provide two potential explanations. The first is that those who chose core are more risk averse than those who chose swing (H2.1: Risk Aversion). In an earlier section of the survey, I measured respondents’ risk aversion by asking them to rate their willingness to take risks on an 11-point sliding scale, where 0 means they are “not at all willing to take risks” and 10 means they are “very willing to take risks.”¹⁶ As displayed in Figure 4, respondents’ risk aversion does not correlate with their choice of neighborhood; risk-averse politicians are just as likely to choose the swing neighborhood as risk-accepting politicians. When control variables are included (Table A6 in the Supplemental Appendix), the correlation between risk aversion and choosing the core neighborhood strengthens to the extent that a 1 standard deviation change in risk aversion from below the mean to a standard deviation above predicts that an official would be 15 percentage points more likely to choose core over swing. However, the results fail to reach statistical significance (p = .181). In sum, I fail to find strong support for the risk aversion hypothesis (H2.1)

Figure 4.

Policymakers’ risk aversion does not predict which neighborhood they target (N = 307).

Another possible explanation stemming from the core voter model is that politicians who chose the core neighborhood are less certain about how swing voters respond to receiving distributive spending than core voters (H2.2: Uncertain about Swing Voters). As a rough measurement of this uncertainty, I identify whether respondents skipped a statement in Part 2 of the survey experiment or indicated that a statement had a 50% chance of being true because this value was labeled in the survey as being a “complete toss-up” and was the de-facto “not sure” response. If H2.2 is correct, then respondents who are randomly assigned in Part 2 of the survey experiment to evaluate the behavior of a swing neighborhood should be more likely to provide an “uncertain” answer than those assigned to evaluate the behavior of a core neighborhood. To account for the possibility that H2.2 only applies to respondents who chose the core neighborhood in Part 1, I interact the treatment assignment with respondents’ targeting strategy. The results in Table A7 in the Supplemental Appendix, however, fail to reject the null hypothesis that respondents, and especially respondents who chose the core neighborhood, are no more uncertain about swing voters’ response to distributive spending than they are about core voters. Although the coefficient on the interaction variable is always in the right direction (positive), it is only statistically significant at the 0.1 level in one instance (Model 2).

In sum, I do not find evidence for the mechanisms of the core voter hypotheses even when trying to explain the behavior of local policymakers who chose the core neighborhood over the swing one. In addition, I fail to find evidence that other characteristics, such as electoral vulnerability or years in office, systematically predict respondents’ targeting strategies. (See Table A6 in the Supplemental Appendix.) These findings validate ongoing theoretical work that seeks to identify the conditions that affect whether policymakers target core or swing groups.

Distributive Spending Does Not Increase Turnout Enough

Figure 5 examines local politicians’ targeting strategies with regard to voters’ turnout propensity. The results suggest that politicians targeted the high turnout neighborhood over the low turnout one because they believe that the mobilizing effect of distributive spending is insufficient to justify targeting spending to mobilize low turnout core supporters. The gray bars in Figure 5 display the mean responses of subjects assigned to the “high turnout” condition in which Mr. Smith chooses the high turnout neighborhood over the low turnout neighborhood. The white bars indicate the mean responses of subjects assigned to the “low turnout” condition, in which Mr. Smith chooses the low turnout neighborhood instead.

Figure 5.

Why do local policymakers target high turnout voters?

According to the responses to Statement 1, local politicians believe that distributive spending is more likely to increase the vote for Mr. Smith in the recipient neighborhood when that neighborhood is full of high turnout voters (mean = 62%) than when it is full of low turnout ones (mean = 53%). This 9-point difference is statistically significant at the 0.01 level. Neither type of neighborhood appears to be more likely to punish incumbents for targeting other groups—although low turnout neighborhoods are predicted to be slightly more likely to punish (diff. = 2; p = .551), high turnout neighborhoods are predicted to be slightly more likely to find out that they were overlooked (diff. = −5; p = 0.069).

According to the responses to Statement 4, local politicians are unsure as to whether distributive spending has a mobilizing effect. The mean response across both conditions is 50%, which was labeled in the surveys as “a complete toss-up.” Further decreasing their incentive to target low turnout core voters is the respondents’ belief that distributive spending is more likely to increase turnout in the high turnout neighborhood (mean = 53%) than in the low turnout one (mean = 47%; diff. = 6; p = .037). This is consistent with voter mobilization research that finds that mobilization efforts are more effective with high turnout propensity voters in low-salience elections (Arceneaux & Nickerson, 2009).

Robustness Checks and Second Study

In this section, I examine the robustness of our results and address several potential concerns about the analyses. More detailed examinations of these and related issues are presented in Sections C, D, and E of the Supplemental Appendix. As shown there, the main results from Figure 1 concerning which neighborhood is targeted hold when controlling for a variety of individual- and municipal-level variables. (See Tables A8–A12.)

A potential concern mentioned earlier in the article is that these theories may not apply well to local officials because they have lower political ambitions than their counterparts at higher levels of government. To examine this further, I identify the more ambitious officials in our sample and examine whether they systematically respond differently than their less ambitious colleagues. I use several metrics to measure ambition. The first, by Maestas (2003), labels officials as ambitious if they plan to run for higher office or stay in office for three or more terms. Another consideration is the size of respondents’ city because Oliver et al. (2012) argue that officials from larger cities (pop. around 100,000 or higher) are more ambitious. Finally, in some of the specifications, I also account for officials’ length of service because some officials may not plan to serve much longer because they have already been in office for some time, which also indicates more ambition.

Across several measures of ambition using these different metrics (Figures A15–A20), I consistently find that more ambitious officials are similar to less ambitious officials in terms of targeting swing versus core voters (Panel A of Figure 1). They also target high turnout voters at the same rate in the pooled results (Panel B of Figure 1). However, when the results in Panel B are broken down by whether both neighborhoods are swing or core, ambitious officials behave somewhat differently than less ambitious ones. While the less ambitious officials target the high turnout neighborhood at the same rate (about 80%), regardless of whether both neighborhoods are core or swing, the ambitious officials are even more likely (about 95%) to target the high turnout neighborhood when both are swing. But when both neighborhoods are core supporters, the percent of ambitious officials targeting the high turnout neighborhood drops significantly to about 70%.¹⁷ Although it is still the case that ambitious officials on average favor high turnout neighborhoods over low turnout ones (consistent with H4), the logic of not wasting distributive goods on likely voters who already support the candidate (H3) appears to have a bigger influence on more ambitious officials than less ambitious ones.¹⁸ Overall, the results in Figures A15 to A20 suggest that the general findings from this article apply to the more ambitious officials whose motives are probably more in line with theory and reelection-minded politicians in general.

Another potential concern with our analysis is the large difference in turnout between the two neighborhoods in the vignette. It is possible that more officials would have targeted the low turnout neighborhood when both were core (Panel B of Figure 1) if the difference in turnout was less drastic. More generally, our results may hinge on particular aspects of how I set up the experiment. To examine this possibility, I turn to an alternative version of the survey experiment that was administered to a small subset of respondents (N = 201). The structure of the vignette was exactly the same except that the treatment conditions differed slightly and were described with text. In this second version of the experiment, the neighborhood with lower turnout was described as having average turnout, rather than quite low turnout relative to the higher turnout treatment. The exact language was as follows:

High Turnout: “About the highest in the city.”

Average Turnout: “Average, but could be increased through campaign efforts.”

Swing: “Swing voters, not strong supporters of either Smith or his opponent.”

Core: “Strong supporters of Smith, and have been in the past, too.”

The results from this additional survey experiment, shown in Figure 6, are quite similar to those in the first experiment (Figure 1) when officials are choosing between a swing or core neighborhood (Panel A of both figures). However, the results are quite different when officials are choosing between a high and low turnout neighborhood (Panel B). When both neighborhoods are swing, 76% choose the high turnout neighborhood. However, when both neighborhoods are core, only 39% do so (diff. = 37; p = .00). The results in Panel B of Figure 6 are more consistent with the hypothesis that officials should target high turnout swing voters and lower turnout core voters (H3). Given the low N, the results from Part 2 of the survey experiment are too noisy to identify any effects. Nonetheless, the results in Panel A of Figure 6 generally confirm those from Panel A of Figure 1 even with a different presentation of the treatment conditions. At the same time, the results in Panel B of Figure 6 demonstrate that the decision to target either a higher or lower turnout neighborhood is more sensitive to how the treatment conditions are described, suggesting that the incentive to target a lower turnout core neighborhood increases when their turnout is not so drastically low relative to the higher turnout core neighborhood.

Figure 6.

Which types of voters do policymakers target with distributive spending? (Text).

Discussion and Conclusion

Whom do politicians target with public spending? To address this central question in the distributive politics literature, I use a novel research design for studies of legislative behavior: survey experiments on a sample of elected municipal officials. Across two experiments, I find that nearly 60% of officeholders target swing voters over core voters because they believe that swing voters are more likely than core voters to electorally punish incumbents for targeting other groups. In general, these findings support hypotheses derived from swing voter models (e.g., Lindbeck & Weibull, 1987, 1993; See H1 in the theory section). Even though a sizable minority of respondents believes that the vote-maximizing strategy is to target core supporters, I fail to find evidence that hypotheses derived from the core voter model (Cox & McCubbins, 1986) explain these respondents’ distributive choice (H2). Other factors appear to be at play, which is something that deserves additional attention in future work.

With regard to targeting citizens based on their propensity to vote, I find that politicians are unsure as to whether distributive spending has a mobilizing effect on the recipients of that spending, especially if their turnout is quite low (as in Figure 1). In these situations, they overwhelmingly (at 82% of respondents) target very likely voters over citizens with a much lower propensity to vote (H4). These results, however, are not as clear cut as our initial analysis (Figure 1) suggests for two reasons. First, more ambitious local officials consider voters’ support when choosing whether to target a high turnout or low turnout neighborhood. Although they are still more likely to target high turnout neighborhoods overall (at 85%) and especially when both neighborhoods are swing (at 95%), the percent targeting high turnout neighborhoods drops significantly (to 72%) when the low turnout neighborhood and high turnout neighborhood are both full of their core supporters, which is somewhat consistent with H3. Second, in an alternative survey experiment where turnout was high in one neighborhood but average in the other (Figure 6), I find that officials are less likely to target the high turnout neighborhood (39%) over one with average turnout when both are full of core supporters. Overall, our results suggest that targeting high turnout voters over lower turnout ones is the more dominant strategy (H4) but is still conditioned to some degree on the degree to which turnout propensity differs and whether those voters are swing or core supporters (H3).

A key benefit of this analysis is that it directly manipulates politicians’ choice options. However, the research design is also subject to its own limitations, especially in terms of external validity and generalizability. As is generally the case with survey experiments, respondents’ choices may not reflect their actual behavior. Moreover, their choice of which neighborhood to target may be influenced by other factors that correlate with a neighborhood’s support for the incumbent and its turnout propensity (e.g., wealth). On the other hand, policymakers’ targeting strategies are consistent with their perceptions of how different types of voters would respond to the allocation of distributive spending. This finding, combined with the heterogeneity in policymakers’ responses, suggests that there is not an overwhelmingly dominant answer to whether municipal officials target core supporters or undecided constituents. This highlights the need for ongoing theoretical development on politicians’ targeting strategies, in line with Golden and Min’s (2013, p. 82) observation in their review of this literature.

Although the theories motivating this research are quite general and have been applied to a broad spectrum of political contexts, municipal legislatures are distinct from other legislative bodies in the United States on several dimensions (e.g., Oliver et al., 2012; Trounstine, 2009). About 75% run in nonpartisan elections and about 60% of municipalities have completely at-large city councils (Svara, 2003).¹⁹ Municipal legislatures are also composed of fewer members, on average, than legislative bodies at higher levels of government.²⁰ All these differences should be carefully considered before generalizing the findings of this project to other contexts. At the same time, the responses of the more ambitious officials in our sample, which were not extremely different from our overall findings, are probably more applicable to politicians in general.

Concerns about generalizability also highlight paths for future empirical work. For example, the research design from this article could also be used to test the targeting strategies of not just politicians but also candidates and party officials in other countries and levels of government, including other forms of local government, such as counties and school boards. There is also the question of whether officials’ targeting strategies would differ with other types of distributive goods or with programmatic policies. The logic of targeting strategies may also differ significantly with a class of land use policy that is particularly relevant in local politics: the placement of locally unwanted land-uses (or “LULU’s”; Langbein et al., 1996) like affordable or high-density housing (Marble & Nall, 2018), drug treatment centers (De, Benedictis-Kessner, & Hankinson, 2019), or water treatment plants.

The findings of this study also have important normative implications. The first concerns politicians’ strategy of targeting swing voters. If local policymakers believe that they can gain the support of swing voters through distributive spending, they may use this additional leeway to implement programmatic policies that favor their preferences at the expense of the median citizen’s policy preferences. Although the findings from this study do not fully address these implications, they do weaken the claim that distributive spending simply buys support. Recall that local policymakers target swing voters not because they think it will overwhelmingly increase the swing voters’ support for the incumbent (Statement 1 of Figure 2), but because they fear swing voters’ reaction when other groups are targeted (Statement 2 of Figure 2).

The second set of normative implications concerns policymakers’ strategy of targeting high turnout voters. If carried out with programmatic decisions in addition to distributive ones, this strategy could shift policy outcomes away from the median constituent’s preferences to the extent that the preferences of high turnout voters differ from those of low turnout voters and nonvoters (Anzia, 2013; Oliver et al., 2012). Indeed, the findings in this article may help explain why local outcomes are biased toward the elderly (e.g., Kogan et al., 2018) and wealthy (e.g., Rhodes et al., 2016), two groups who are more likely to participate in local elections (Kogan et al., 2018; Oliver & Ha, 2007).

Supplemental Material

APR_who-is-targeted_appendix_revisions-2 – Supplemental material for Which Citizens Do Elected Officials Target With Distributive Spending?

Supplemental material, APR_who-is-targeted_appendix_revisions-2 for Which Citizens Do Elected Officials Target With Distributive Spending? by Adam M. Dynes in American Politics Research

Footnotes

Acknowledgements

I would like to thank Brittni Anderson, Leslie Bull, David Clove, Charlotte Dillon, Allison Douglis, Jason Guss, Walter Hsiang, Josh Kalla, Raphael Leung, Diana Li, Yusu Liu, McKell McIntyre, Shahla Naimi, Cameron Rotblat, Joyce Shi, and Christopher Vazquez for research assistance; Daniel Butler, Alan Gerber, Gregory Huber, David Mayhew, Michael Ting, Jessica Trounstine, and seminar participants at Yale University (2013) for useful feedback; and Daniel Butler, Jacob Hacker, Gregory Huber, and the Institution for Social and Policy Studies at Yale University for funding the survey.

Author’s Note

An earlier version of this article was presented at the 2013 Annual Meeting of the American Political Science Association.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors recieved financial support from Gregory Huber, Jacob Hacker, Daniel Butler, and the Institution for Social and Policy Studies at Yale University to fund the survey for the research.

ORCID iD

Adam M. Dynes

Supplemental Material

Supplemental material for this article is available online. Replication data are available at the author’s website: .

Notes

Author Biography

Adam M. Dynes is an assistant professor of political science at Brigham Young University. He researches representation and legislative behavior at the national, state, and local levels in the US.

References

Antunes

G. E.

Plumlee

J. P.

(1977). The distribution of an urban public service ethnicity, socioeconomic status, and bureaucracy as determinants of the quality of neighborhood streets. Urban Affairs Review, 12(3), 313–332.

Anzia

S. F.

(2013). Timing and turnout: How off-cycle elections favor organized groups. The University of Chicago Press.

Arceneaux

Nickerson

D. W.

(2009). Who is mobilized to vote? A re-analysis of 11 field experiments. American Journal of Political Science, 53(1), 1–16.

Bradbury

J. C.

Stephenson

E. F.

(2003). Local government structure and public expenditures. Public Choice, 115(1–2), 185–198.

Calvo

Murillo

M. V.

(2004). Who delivers? Partisan clients in the Argentine electoral market. American Journal of Political Science, 48(4), 742–757.

Case

(2001). Election goals and income redistribution: Recent evidence from Albania. European Economic Review, 45, 405–423.

Chen

(2013). Voter partisanship and the effect of distributive spending on political participation. American Journal of Political Science, 57(1), 200–217.

Cingranelli

(1981). Race, politics, and elites: Testing alternative assumptions of municipal service delivery. American Journal of Political Science, 25, 664–692.

Cox

G. W.

(2006, October). Swing voters, core voters and distributive politics [Conference session]. Representation and Popular Rule, Yale University, New Haven, CT, United States.

10.

Cox

G. W.

McCubbins

M. D.

(1986). Electoral politics as a redistributive game. The Journal of Politics, 48(2), 370–389.

11.

Cox

G. W.

McCubbins

M. D.

Sullivan

(1984). Policy choice as an electoral investment. Social Choice and Welfare, 1(3), 231–242.

12.

Dahlberg

Johansson

(2002). On the vote-purchasing behavior of incumbent governments. American Political Science Review, 96, 27–40.

13.

De Benedictis-Kessner

Hankinson

. (2019). Concentrated burdens: How self-interest and partisanship shape opinion on opioid treatment policy. American Political Science Review, 113(4), 1078–1084.

14.

De La O

A. L

. (2013). Do conditional cash transfers affect electoral behavior? Evidence from a randomized experiment in Mexico. American Journal of Political Science, 57(1), 1–14.

15.

Diaz-Cayeros

Estévez

Magaloni

(2006). Poverty and vote buying: The politics of social transfers in Mexico, 1989-2006 [Unpublished book manuscript]. Stanford University.

16.

Dixit

Londregan

(1996). The determinants of success of special interests in redistributive politics. Journal of Politics, 58, 1132–1155.

17.

Dohmen

Huffman

Sunde

Schupp

Wagner

G. G.

(2011). Individual risk attitudes: Measurement, determinants, and behavioral consequences. Journal of the European Economic Association, 9(3), 522–550.

18.

Dunning

Stokes

(2008, August). Clientelism as persuasion and as mobilization [Paper presentation]. American Political Science Association Annual Meeting, Boston, MA, United States.

19.

Dynes

A. M.

Hassell

H. J.

Miles

M. R.

(2019). The personality of the politically ambitious. Political Behavior, 41(2), 309–336.

20.

Ehrlich

Maestas

(2010). Risk orientation, risk exposure, and policy opinions: The case of free trade. Political Psychology, 31(5), 657–684.

21.

Einstein

K. L.

Kogan

(2016). Pushing the City Limits: Policy Responsiveness in Municipal Government. Urban Affairs Review, 52(1), 33–65.

22.

Fenno

R. F.

Jr. (1977). US House members in their constituencies: An exploration. American Political Science Review, 71(3), 883–917.

23.

Fisher

S. H.

III Herrick

(2013). Old versus new: The comparative efficiency of mail and internet surveys of state legislators. State Politics & Policy Quarterly, 13(2), 147–163.

24.

Fleck

R. K.

(1999). The value of the vote: A model and test of the effects of turnout on distributive policy. Economic Inquiry, 37(4), 609–623.

25.

Golden

Min

(2013). Distributive politics around the world. Annual Review of Political Science, 16, 73–99.

26.

Hajnal

Z. L.

Lewis

P. G.

(2003). Municipal institutions and voter turnout in local elections. Urban Affairs Review, 38(5), 645–668.

27.

Hajnal

Z. L.

Trounstine

(2010). Who or what governs? The effects of economics, politics, institutions, and needs on local spending. American Politics Research, 38(6), 1130–1163.

28.

Harden

J. J.

(2013). Multidimensional responsiveness: The determinants of legislators’ representational priorities. Legislative Studies Quarterly, 38(2), 155–184.

29.

Hirano

Snyder

J. M.

Jr. Ting

M. M.

(2009). Distributive politics with primaries. The Journal of Politics, 71(4), 1467–1480.

30.

International City/County Management Association. (2006). Municipal form of government survey.

31.

Key

V. O.

Jr. (1950). Southern politics. Knopf.

32.

Koehler

D. H.

Wrightson

M. T.

(1987). Inequality in the delivery of urban services: A reconsideration of the Chicago parks. Journal of Politics, 49, 80–99.

33.

Kogan

Lavertu

Peskowitz

(2018). Election timing, electorate composition, and policy outcomes: Evidence from school districts. American Journal of Political Science, 62, 637–651.

34.

Langbein

L. I.

Crewson

Brasher

C. N.

(1996). Rethinking ward and at-large elections in cities: Total spending, the number of locations of selected city services, and policy types. Public Choice, 88(3–4), 275–293.

35.

Lascher

E. L.

Jr. (1993). Explaining the appeal of local legislative office. State & Local Government Review, 25(1), 28–38.

36.

Levy

Meltsner

A. J.

Wildavsky

A. B.

(1974). Urban outcomes: Schools, streets, and libraries. University of California Press.

37.

Lindbeck

Weibull

J. W.

(1987). Balanced budget redistribution and the outcome of political competition. Public Choice, 52, 273–297.

38.

Lindbeck

Weibull

J. W.

(1993). A model of political equilibrium in a representative democracy. Journal of Public Economics, 51(2), 195–209.

39.

Lineberry

R. L.

(1977). Introduction on the politics and economics of urban services. Urban Affairs Review, 12(3), 267–272.

40.

Lipsky

(1969, September). Toward a theory of street-level bureaucracy. Institute for Research on Poverty, University of Wisconsin.

41.

Maestas

C. D.

(2003). The incentive to listen: Progressive ambition, resources, and opinion monitoring among state legislators. The Journal of Politics, 65(2), 439–456.

42.

Maestas

C. D.

Pollock

W. M.

(2010). Measuring generalized risk orientation with a single survey item (Working Paper No. 1599867). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1599867

43.

Marble

Nall

(2018). Where self-interest trumps ideology: Liberal homeowners and local opposition to housing development [Working paper]. http://www.nallresearch.com/uploads/7/9/1/7/7917910/interest.3.3.pdf

44.

Martin

P. S.

(2003). Voting’s rewards: Voter turnout, attentive publics, and congressional allocation of federal money. American Journal of Political Science, 47(1), 110–127.

45.

Matsubayashi

Jun-Deh

(2012). Distributive politics and voter turnout. Journal of Elections, Public Opinion and Parties, 22(2), 167–185.

46.

Meier

K. J.

Juenke

E. G.

Wrinkle

R. D.

Polinard

A. J.

(2005). Structural choices and representational biases: The post-election color of representation. American Journal of Political Science, 49(4), 758–768.

47.

Meier

K. J.

Stewart

England

R. E.

(1991). The politics of bureaucratic discretion: Educational access as an urban service. American Journal of Political Science, 35(1), 155–177.

48.

Miranda

R. A.

Tunyavong

(1994). Patterned inequality? Reexamining the role of distributive politics in urban service delivery. Urban Affairs Review, 29(4), 509–534.

49.

Mladenka

K. R.

(1980). The urban bureaucracy and the Chicago political machine: Who gets what and the limits to political control. American Political Science Review, 74(4), 991–998.

50.

Nichter

(2008). Vote buying or turnout buying? Machine politics and the secret ballot. American Political Science Review, 102(1), 19–31.

51.

Oliver

J. E.

S. E.

(2007). Vote choice in suburban elections. American Political Science Review, 101, 393–408.

52.

Oliver

J. E.

S. E.

Callen

(2012). Local elections and the politics of small-scale democracy. Princeton University Press.

53.

Peterson

P. E.

(1981). City limits. University of Chicago Press.

54.

Rhodes

J. H.

Schaffner

B. F.

La Raja

R. J.

(2016). Race- and class-based inequality and representation in local government [Paper presentation]. Annual Meeting of the American Political Science Association, September, Philadelphia, PA, United States.

55.

Sokolow

A. D.

(1989). Legislators without ambition: Why small-town citizens seek public office. State & Local Government Review, 21, 23–30.

56.

Stokes

S. C.

(2005). Perverse accountability: A formal model of machine politics with evidence from Argentina. American Political Science Review, 99(3), 315–325.

57.

Stokes

S. C.

Dunning

Nazareno

Brusco

(2013). Brokers, voters, and clientelism: The puzzle of distributive politics. Cambridge University Press.

58.

Svara

J. H.

(2003). Two decades of continuity and change in American city councils. National League of Cities.

59.

Tausanovitch

Warshaw

(2014). Representation in municipal government. American Political Science Review, 108(3), 605–641.

60.

Tiebout

C. M.

(1956). A pure theory of local expenditures. Journal of Political Economy, 64(5), 416–424.

61.

Trounstine

(2006). Dominant regimes and the demise of urban democracy. Journal of Politics, 68(4), 879–893.

62.

Trounstine

(2009). All politics is local: The reemergence of the study of city politics. Perspectives on Politics, 7(3), 611–618.

63.

Trounstine

(2013). Turnout and incumbency in local elections. Urban Affairs Review, 49(2), 167–189.

64.

Van Houweling

R. P.

Tomz

(2009). The Electoral Implications of Candidate Ambiguity. American Political Science Review, 103(1), 83–98.

65.

U.S. Census Bureau. (2008). City government finances. U.S. Department of Commerce. https://www.census.gov/programs-surveys/gov-finances.html

Supplementary Material

Please find the following supplemental material available below.

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

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

0.00 MB

5.47 MB