Under What Condition Do Voters Have the Greatest Ability to Influence or Steer Policy

Representative commonwealth works all-time when voters can use the ballot to motivate policy makers to correspond their interests faithfully while in office. Given evidence that voters remain woefully ignorant near the functioning of American political institutions and the policy positions of competing candidates on Election 24-hour interval (e.g., Campbell et al. 1960; Delli-Carpini and Keeter 1996), whether ordinary citizens possess the power to do and so remains a contested question at the centre of the study of political beliefs.

Retrospective voting provides one potential machinery to resolve this democratic dilemma, allowing for effective representation with minimal voter effort. Past workout their support for political incumbents on 1 slice of easily attainable data—the observed regime performance—voters can utilise elections to create the political equivalent of Darwinian "natural choice," ensuring that just true-blue agents retain their offices over the long haul (e.g., Barro 1973; Ferejohn 1986). An extended literature examines whether voters acquit in this way in national elections, providing ample back up for the retrospective voting hypothesis (e.m., Lenz 2012).

But a nifty deal of policy making takes place exterior of Washington, DC, as many political issues remain under the purview of state and local governments. Only recently, however, take scholars shifted their attending to examine how these dynamics play out in subnational elections. To date, the research has produced mixed findings. While there is evidence that voters' subjective satisfaction with government operation influences their back up for elected incumbents (Oliver and Ha 2007), it is far from clear whether these subjective evaluations accurately reflect the objective reality on the basis (Healy and Malhotra 2013). For case, while economical conditions announced to bear on both gubernatorial approval and elections, voters practise non e'er correctly apportion responsibleness for economic operation betwixt different levels (Chocolate-brown 2010) and branches (Rudolph 2003a) of government. Moreover, national political considerations—such as approval of the incumbent president—appear to matter much more for land elections than state-level outcomes, such every bit homicide rates, Saturday scores, and the burdens of land taxation (Rogers 2013).

We suspect that the mixed nature of these findings reflect two serious limitations in the existing literature. Offset, data availability rather than clear theory has generally driven the choice of performance outcomes examined in published analyses. In their report of mayoral approval in New York, for example, Arnold and Carnes (2012) rely on the number of city employees every bit a proxy for municipal service quality. When such measures yield null furnishings—as was the case in their study—it is impossible to know whether ane should translate the results as evidence against retrospective voting or simply equally a failure to measure the dimensions of government performance that voters discover most salient.

On the other hand, studies that accept plant bear witness of retrospection in subnational voter behavior largely focus on outcomes over which these governments take limited, if any, direct control. When constituents condition their votes on perceived service quality, nosotros fence it is important to consider whether they correctly attribute credit or blame for observed outcomes to the proper agents in government. When they fail to do so—and hold the wrong office accountable or blame public officials for outcomes beyond their control—retrospective voting might create perverse incentives that pb to worse democratic outcomes, a possibility that the existing literature on local accountability largely overlooks.^ane While a number of recent studies note that voters sometimes misattribute responsibility for performance, peculiarly in federal or multilevel contexts (e.g., Healy and Malhotra 2013; Rudolph 2003b), the consequences of this insight—and the importance of choosing the appropriate policy domain on which to focus ane's inquiry—have not been adequately incorporated into inquiry on local retrospective voting.

In this written report, we seek to address both of these limitations using novel authoritative information on pothole complaints in one of America's largest cities. Our assay focuses on an outcome—the quality of local streets—that nosotros show is both highly salient to voters and conspicuously under the control of the local government. Aside from the substantive focus on this dimension of performance, another central difference between our approach and other recent studies is that we specifically examine within-city variation in both service quality and electoral beliefs. Rather than comparison election outcomes between cities or within cities over time, we leverage neighborhood-level differences in the number of citizen pothole complaints and examine how these differences bear on the vote share won by the same incumbents across precincts. Our approach recognizes that government services and outputs probable vary significantly within cities, and nosotros demonstrate that such variation can explain neighborhood-level differences in support for incumbents from both major political parties in two political offices—mayor and city quango—across several balloter cycles.

In the residue of the newspaper, nosotros proceed by providing an overview of the existing theoretical and empirical literature on voter beliefs in urban center elections. In doing then, we situate the available research on local retrospective voting within the broader debate about the constraints facing local governments and the dynamics of local republic. Next, we introduce our case—the city of San Diego—and describe how we combine geocoded municipal records on street repair requests with a panel of precinct-level data on ballot outcomes, allowing us to examine how road quality influences voter behavior in San Diego city elections. Subsequently describing the unusual political context during the years included in our analysis—a context that provides a specially shut fit between theory and our data—we present the results of our analysis. The concluding section examines the implications of our findings for democratic accountability in local government. We argue that, in normative terms, the dynamics we document could encourage ii sets of behaviors among public officials. On the one paw, accountability pressures may pb these officials to prioritize and invest in the services that citizens value the most, encouraging faithful democratic responsiveness exactly as envisioned by normative theory. On the other hand, strategic politicians may also respond to the resulting balloter incentives through a diversity of gaming behaviors—such equally shifting funding from less to more than salient services and linking budgetary outlays to electoral calendars—that, over the long term, may reduce social welfare.

Retrospective Voting in Local Elections

Applying theoretical models built to explain beliefs in national elections to local democracy requires acknowledging important differences between these 2 levels of government. For example, while all federal elections are partisan, approximately 80% of local government positions are filled through nonpartisan contests (Kaufmann 2004). In addition, local governments generally attract much less media coverage and candidate spending, among many other important differences. Are these differences meaningful enough to bear on voting beliefs and, in plow, the nature and quality of democratic representation?

A large literature in urban politics argues that cities are indeed unique. Local governments face a number of constraints—including competitive pressures from nearby jurisdictions, interference from college-level governments, and binding tax-and-expenditure limitations—that greatly limit local discretion and the menu of bachelor policy options.² As a result of these constraints, political contest at the metropolis level oftentimes focuses on the question of how to divide a fixed corporeality of resource, pitting broad groups of citizens against i another. As Kaufmann (2004, 18–19) summarizes:

Local politics—and the kinds of issues that dominate local elections—are often more proximate and more discrete than are the larger symbolic issues of national elections. Local governments, while sometimes lawmakers, are principally service providers. … To most Americans, the role of their local government is to maintain or heighten their immediate quality of life, to provide necessary services, and at times to ameliorate intercity conflict. Thus the citizen'due south expectations of local government are inherently more connected to daily life than are his or her expectations of other governmental bodies. And every bit such, how people view their local leadership and why they vote for them will likely reflect the proximity of their interests.

Ane implication is that constraints on local government power should influence the dimensions of regime functioning voters focus on when forming their retrospective evaluations. If municipal government influence is express to a small set of basic housekeeping services, information technology makes little sense for voters in local elections to hold city officials accountable for outcomes that are largely exterior of their straight control.

Surprisingly, however, much of the empirical literature on retrospective voting in the local context focuses on just such issues. For instance, Berry and Howell (2007) analyze how student achievement affects the reelection prospects of school board members,³ although more than 80% of the variation in academic achievement is explained by student-level factors, such as socioeconomic condition (Chingos, Whitehurst, and Gallaher 2015; Kogan, Lavertu, and Peskowitz 2016b) rather than local educational policies. In addition, the largest remaining determinant of achievement that is plausibly under the control of local officeholders—teacher quality—varies far more within schools (and districts) than betwixt them (encounter Chetty, Friedman, and Rockoff 2014), providing few clear expectations for balloter beliefs.

In a recent study, Hopkins and Pettingill (2015) examine the electoral functioning of mayoral incumbents in big The states cities. Although their model includes a number of policies plausibly under local government control, none of these factors announced to correlate with incumbent vote share. The only variable that consistently produces substantively big furnishings is the local unemployment charge per unit. While mayors tin can certainly touch on the local economy, factors such as the spatial distribution of industries, occupations, and natural resource endowments likely explain much more than of the variation in subnational economic performance. Indeed, Miller (2013) finds that the win-loss records of local professional sports teams—a statistic over which government officials conspicuously have no control—frequently influence local mayoral elections. He interprets this consequence to mean that voters follow a "Prosperity Model" of voting, holding incumbents accountable for their overall well-existence, regardless of whether authorities policies played any actual role in determining their level of happiness.

The existing literature rarely acknowledges the disconnect betwixt what local governments actually do and where analysts search for electoral accountability in local elections. When scholars do note the apparent discrepancy, they nevertheless draw sanguine conclusions about the normative implications for local democracy. Miller's (2013, 16–17) discussion of the probable welfare implications of voters relying on sports records when deciding whether to reelect their incumbent mayors offers a case in point:

The flip side to how voters cull their leaders is how this incentivizes politicians in office looking for reelection. Under the Prosperity Model, incumbent politicians are motivated to maximize voter happiness at the time of the election. This is inappreciably the gravest of threats, even if motivating politicians to maximize well-being in the long run is preferable. …

It is fair to say that voters are imperfect and occasionally irrational. It does non follow that commonwealth is seriously questioned. In fact, voting for incumbents when well-being is loftier is a sensible rule of thumb for voters who cannot reliably connect the dots betwixt political actions and outcomes. Although irrelevant events may seep into this calculation, voting remains an constructive method for selecting and disciplining leaders.

We argue that such conclusions may not always be warranted and certainly do not follow logically from existing research. Consider another finding from Arnold and Carnes (2012), who use several decades of panel information on New York City mayoral approving to prove that voters appear to blame local elected executives when crime levels rising and credit them when crime declines. In reality, all the same, there is little evidence that local governments accept much control over crime rates. In a recent exhaustive analysis, Roeder, Eisen, and Bowling (2015) prove that much of the decline in crime that occurred between 1990 and 2013 was a result of broad national trends, such as falling alcohol consumption, rising income, an aging population, and the legalization of ballgame. While local policies such as the introduction of information-driven law enforcement approaches (i.e., CompStat) and the growing size of local police force forces played a pocket-sized role on the margin, forces over which local governments had no direct control dwarfed these effects. Notwithstanding voters—at to the lowest degree in New York Metropolis—appear to take credited local officials anyway.^iv

If "retrospective voting" is merely a reflection of irrelevant considerations, the implications for local democracy are potentially far more troubling and the incentives for local officials more perverse than Miller (2013) suggests.⁵ If mayors know that voters will hold them accountable for crime, simply also realize that there are few tools at their disposal to influence offense levels directly, they may be tempted to shape the few things that are under their command, namely, how crime rates are measured and reported. Indeed, recent scandals in both Chicago (Bernstein and Isackson 2014) and Los Angeles (Poston and Rubin 2014) provide reason to worry. In both cities, mayoral administrations faced growing political force per unit area to reverse rising violent law-breaking rates. In the absenteeism of policy tools to do so and with elections budgeted, mayors in both cities put pressure level on their police departments to misclassify serious crimes to make the statistics appear less dire.

Thus, in determining whether local commonwealth "works," nosotros argue it is bereft to mensurate the extent to which observed policy and functioning outcomes shape incumbent electoral prospects. It is also necessary to consider whether voters allocate credit and blame appropriately, correctly linking these outcomes to the government officials responsible. Our study moves in this direction past shifting the focus to precisely the kinds of services controlled by city government—in our instance, the quality of local roads.

Our 2d contention is that, in evaluating voter beliefs, it is of import to pay attention to within-city heterogeneity in service quality. Equally Trounstine (2016, 720) notes, "Because local government decisions often concern spatial allocation, neighborhoods are important municipal actors in local politics." Voters often evaluate the operation of local government on the basis of the quality of life in their neighborhood rather than that in the city as a whole. If voters indeed vote retrospectively, we await their behavior to reflect the quality of the services that they receive, which we can rarely infer from average citywide outcomes that are the typical focus of analysis in the existing literature.⁶ In the sections that follow, we describe how our report incorporates both of these primal theoretical insights.

Financial Crisis and Government Services in "Enron-past-the-Sea"

Equally is the case with about of the research on local politics, our study examines a single city, San Diego. One disadvantage is that doing so potentially limits the generalizability of the findings. Fortunately, this is much less of a concern in our case because the municipality we examine is similar in important respects to most other major and mid-size Usa cities (meet app. A). In detail, San Diego is ethnically and racially diverse, overwhelmingly Autonomous, utilizes a "strong mayor" form of regime,⁷ and elects city officials through nonpartisan ballots. Across these and many other dimensions, America'southward eighth-largest metropolis is a representative case, and so the political dynamics we document are likely to provide useful insights near urban center politics writ large.

The primal advantage of a unmarried-city case study, all the same, is that we can utilize detailed data on both service quality and electoral outcomes at a very low level of aggregation, an empirical strategy that is simply non practical to implement across multiple jurisdictions. Since our unit of analysis is the electoral precinct and nosotros focus on variation across neighborhoods, the effective N of our report is still significantly higher than in a typical analysis of local government election outcomes. Nigh important for our purposes, San Diego is a useful context in which to examine voting behavior because the city faces—and, in some ways, exemplifies—many of the competitive and legal constraints confronted by many other local governments.^eight

Much of San Diego's population growth occurred in the three decades after World War II, an era that coincided with emerging financial pressures. In 1978, California voters passed Proposition xiii, a constitutional initiative that permanently reduced and froze property taxes and limited the reassessment of property values. Although the measure produced a sizable negative revenue shock for many local governments in California, the effect was peculiarly large in San Diego, which had unusually depression belongings taxes prior to the change. Equally a result, Proposition thirteen permanently locked in the city's low revenue (Erie et al. 2011, chap. 3).

As San Diego continued to grow, local officials found artistic ways to supplement local tax revenues to balance the annual budget. Starting in the early 1980s, for example, the city began diverting "surplus" earnings from its municipal pension organisation to pay day-to-mean solar day operating expenses. When several ambitious projects further strained the budget in the mid-1990s, the city struck a hugger-mugger bargain with its contained alimony board to underfund the pension system. In commutation for their support to allow the urban center to pay less than its actuarially required pension contribution, metropolis employees received a significant increase in pension benefits. Urban center finances faced additional pressures after the 2001 recession, prompting the pension bargain to exist expanded, with the city standing to underfund the pension organization in substitution for nevertheless more benefit increases for city workers.

After journalists brought to light these underfunding schemes in the early 2000s, the city was left on the brink of municipal bankruptcy. Information technology significantly increased its alimony contributions to help shut a now multi-billion-dollar unfunded pension liability. Because San Diego authorities officials did not adequately disembalm the pension underfunding on official fiscal statements, the city's external auditor refused to sign off on its financial statements, which effectively banished San Diego from the municipal bond markets. With the city poised at a political and financial precipice, San Diego attracted growing national attention, with a New York Times headline describing the metropolis every bit "Enron-by-the-Bounding main." Unable to infringe and facing a always-increasing pension beak,⁹ the city made dramatic reductions to public services and allowed basic infrastructure repairs to get unaddressed. By 2010, the pecker for unfunded majuscule maintenance exceeded $ane billion, more than the urban center'south almanac day-to-day operating budget.

Road quality and citizen satisfaction

Every bit documented in Erie et al. (2011), San Diego's fiscal and pension crisis resulted in a substantial reduction in service levels and quality. The urban center cut its library and recreation center hours, downsized its work force, and reduced essential services. One of the most striking changes, yet, was the continued deterioration of urban center streets and sidewalks. With no bonding capacity and picayune spare operating revenues with which to acquit out even the most basic street maintenance, road weather condition deteriorated rapidly. Ane of the city'due south major news organizations created a website on the photograph-sharing service Tumblr, which it named "The Stumblr," for readers to submit pictures of the worst streets and sidewalks. Even in 2015, years later the urban center's fiscal crisis abated and revenues rebounded, a national cess institute that over half of San Diego'due south streets were in poor condition, ranking it the eighth worst in the state ("Bumpy Roads Ahead: America's Roughest Rides and Strategies to Make Our Roads Smoother" [TRIP 2015]).

When San Diego officials commissioned a representative survey of city residents in 2010, satisfaction with the quality of city streets was ranked second-to-final amidst the 25 services evaluated past respondents, just ahead of downtown parking availability. In a Priority Spending Index, which combined citizen satisfaction with existing services and their willingness to pay higher taxes or fees to fund maintenance or avert farther cuts, urban center streets trounce out all other priorities by a large margin (see fig. 1).

Effigy one.

Constituents' top spending priorities. Source: "City of San Diego Residents' Opinions on Metropolis Services" (Behavior Inquiry Heart 2010)

Empirical Strategy

Given the salience of street quality, as documented in the city's survey, San Diego provides an excellent exam case to examine whether voters concur local government officials accountable at the ballot box for observed functioning. If we find little relationship between local street conditions and voting behavior in San Diego, we can exist confident that retrospective voting is unlikely to operate for less salient dimensions of functioning or in other locales. On the other hand, if we observe that voters do hold incumbents accountable, the magnitude of these effects tin can speak to the likely incentives that local officials face up while in role nether conditions about likely to produce accountability pressures. In this section, nosotros provide an overview of our empirical strategy for carrying out just such an analysis. It describes the variables nosotros use and the statistical models we estimate to measure out the impact of road quality on voter support for local incumbents.

Independent variable: Denizen pothole complaints

To identify the effect of municipal service quality on incumbent political prospects, nosotros leverage geographic variation in road weather condition betwixt neighborhoods in San Diego. In doing and then, we assume that voters intendance near about the quality of regime services that they personally receive rather than the average service quality citywide. Although there are a number of potential justifications for this assumption, the most plausible i is that voters are probable to form their retrospective judgments based on casual observations from their own solar day-to-day experiences (Popkin 1994). It seems much more realistic that voters volition notice whether a pothole has adult on their urban center block than to presume that they calculate the average asphalt quality index for all city-maintained streets

Neighborhoods in San Diego differ in the quality of their roads and the number of potholes that develop on them for a diversity of reasons. The city repaved or resealed individual streets in dissimilar years, for example, which means that asphalt in some areas is only older than in others. There is also variation in the material used to construct the roadway and the level of daily traffic flows, both of which can affect the rate of deterioration. Directly or indirectly, virtually all of these factors reflect policy decisions fabricated by city officials.¹⁰

The key reward in focusing on within-city variation in road quality is that such differences betwixt neighborhoods are plausibly contained of incumbents' expected electoral functioning. This is particularly true in our example, because we explicitly status on each incumbent candidates' lagged level of back up from the previous ballot. Provisional on previous electoral performance, at that place should be no significant differences between neighborhoods where potholes develop (or where they are reported) and those where no such reports are made, allowing us to credibly identify the effect of service quality on voting behavior. As we discuss in the next section, controlling for lagged vote share significantly increases the internal validity of our design by allowing usa to rule out plausible omitted variables. As an boosted robustness check, all the same, we present results from a placebo-style specification that further strengthens our power to make causal claims about the human relationship between route quality complaints and voting behavior.

Our key independent variable is the number of citizen pothole complaints, aggregated at the level of the electoral precinct. The data, covering the years 2008–11, were first obtained past the Vox of San Diego, an online investigative journalism organization, through a public records request.¹¹ The database contains more than 52,000 complaints and service requests involving city-maintained streets recorded past the San Diego's street maintenance department during these years.¹²Appendix B describes the geocoding protocols we used to map the potholes to voting precincts. Focusing on citizen complaints, rather than amass measures of route condition, likewise ensures that nosotros mensurate service quality using the information that is conspicuously visible and salient to local residents.^thirteen

Table ane provides a breakdown of service calls recorded by the Streets Department. Overall, more than two-thirds of the recorded complaints focused on potholes, although the information include a number of other, more than serious route repair requests. For convenience, we refer to all of these service requests as pothole complaints in the text below.

Table 1.

Summary of Service Calls

Problem Code	Total Complaints (%)
Major cobblestone repair	22.7
Minor asphalt repair	8.9
Pothole	68.four

One indicator that citizen complaints reflect actual problems with service quality is that the frequency of reported potholes is positively correlated with a pavement "overall status alphabetize," an objective measure out of road quality (see app. C). Another indicator is the tape of the city's responses to the complaints. As nosotros testify in appendix D, crews dispatched to assess the roadway in response to the complaints ended upwardly carrying out some kind of repairs in almost 98% of cases. This does not hateful that each pothole is every bit likely to be reported to the city, but it does propose that the reports made reflect legitimate grievances.

Our empirical strategy accounts for the fact that such grievances may not e'er translate perfectly into votes. Both theoretical and empirical literature on voting decisions illustrate that voters' use of functioning data is discipline to behavioral biases, such as myopia. Recent performance, such every bit macroeconomic weather in the half-dozen months earlier an election, has a disproportionate result on balloter outcomes (e.g., Healy and Lenz 2014; Huber, Loma, and Lenz 2012, Kogan et al. 2016b). We expect voters to showroom similar myopia in our local context. Every bit a result, we summate a count of total complaints filed in each precinct in the vi-month window immediately preceding each ballot using the notification dates recorded in the city's database.¹⁴ As we discuss below, we also utilize these date stamps to construct a count of pothole complaints made in the six-month window later on the election. Because potholes reported afterward Election Day should logically have little impact on incumbent performance, this post-election mensurate provides us with a useful "placebo" test, which we will draw in more detail below.

Dependent variable: Incumbent electoral functioning

The dependent variable used in our assay is the share of votes won by incumbent candidates in each ballot. In particular, we focus on ii elected offices nigh responsible for making policy and setting funding levels for metropolis services, the mayor and the city council. During our period of study, San Diego held ii citywide elections, in 2008 and 2010. In each year, local candidates competed in a June principal. If no candidate received more than 50% of all votes bandage for each role, the meridian two vote-getters proceeded to a runoff in Nov.

Although San Diego had viii city council districts during this period, term limits mean that we observed incumbent council members running for reelection in simply 2 districts in these years. Thus, our analysis focuses on 3 role holders: Mayor Jerry Sanders and Councilmen Kevin Faulconer and Tony Young.^xv Sanders stood for reelection in 2008, while the two councilmen both ran in 2010. All three candidates won more than 50% of the vote in the June primaries, securing their reelection without a Nov runoff. It is worth emphasizing that our data set includes incumbents from both major political parties—both Sanders and Faulconer are Republicans, while Young is a Democrat—although elections are technically nonpartisan in San Diego, and party labels do non appear directly on the ballot. Appendix E describes how we handle changing precinct boundaries over time, to make the election information comparable across electoral cycles.

Statistical models

Our baseline specification of each incumbent's vote share in precinct i at the time of his reelection, $\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{one\; thousand}\mathrm{b}\mathrm{due\; east}\mathrm{north}{\mathrm{t}}_{\mathit{i}\mathit{t}}$, is

$$\mathrm{I}\mathrm{north}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{n}{\mathrm{t}}_{\mathit{i}\mathit{t}}=\alpha +\tau \mathrm{P}\mathrm{o}\mathrm{t}\mathrm{h}\mathrm{o}\mathrm{l}\mathrm{due\; east}{\mathrm{s}}_{\mathit{i}\mathit{t}}+\beta \mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{due\; north}{\mathrm{t}}_{\mathit{i}\mathit{t}-1}+{ϵ}_i\mathrm{.}$$

Nosotros include each candidates' lagged vote share from the previous ballot, $\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{grand}\mathrm{b}\mathrm{e}\mathrm{northward}{\mathrm{t}}_{\mathit{i}\mathit{t}-\mathrm{one}}$, to accost several important concerns about omitted variable bias and reverse causality. In particular, there are iii possible threats to internal validity in our cross-exclusive setting. First, San Diego exhibits pregnant racial residential segregation. Considering race and ethnicity are strongly correlated with partisan identity, the residential patterns affect the political geography of the city. The heavily Democratic areas are concentrated in the oldest, poorest, and nigh industrial parts of the city south of Interstate 8. This introduces the potential for a spurious correlation betwixt local road quality and partisan political dynamics, creating the advent that voters are interim in a retrospective manner when in fact they are not. Indeed, if nosotros only regress Mayor Sanders'due south lagged vote share from 2005 on pothole complaints three years later, from the beginning half of 2008, we find a significant correlation, providing clear show that potholes are not "as-if randomly" assigned between neighborhoods.

Second, information technology is also possible that that local officials strategically allocate city resources and capital investments to sure constituents based in part on their expected or previous level of support in each neighborhood (due east.g., Cox and McCubbins 1986). If policy makers direct road resurfacing funds to neighborhoods containing their strongest supporters or set potholes quicker in these neighborhoods earlier voters have a chance to observe them and complain to the city, we would run across a correlation betwixt voting beliefs and road quality, just the direction of the causal pointer would be reversed.

Finally, one additional complication is that our independent variable is based on pothole complaints rather than an objective measure of road conditions, although we show in appendix C that there is a strong empirical relationship between the two. Theoretically, this is an advantage, since nosotros can focus only on considerations that are clearly salient to at to the lowest degree some constituents. In that location is, however, also a possibility that voters who already dislike the incumbent are merely more vigilant most noticing or reporting street maintenance issues, inducing a correlation between pothole counts and precinct-level voting behavior.

Fortunately, conditioning on lagged vote share helps address all three of these potential problems. This point is worth emphasizing: although it is improbable that pothole complaints are "as-if randomly" assigned, controlling for each incumbent's previous vote share in the model goes a long mode toward ensuring that this treatment is conditionally independent of our outcome of interest. Indeed, the only remaining omitted variables would need to (i) exist correlated with pothole complaints, (2) be correlated with candidate vote share at time t, simply (three) not be similarly correlated with the same candidates' vote share in the same precinct at fourth dimension t − i. In addition, these relationships must exist nowadays in both the 2008 mayoral race and the 2010 city quango contests for both Autonomous and Republican incumbents. Although we cannot completely rule out the possibility that such time-varying variables exist, information technology is hard to think of whatsoever that satisfy all of these weather condition. Nosotros hash out two other interpretation strategies—the stock-still effects and first-divergence specifications—in appendix F and explain why our modeling strategy makes the most sense in our context.

In addition to including a lagged dependent variable in our model, we tin also use the post-election complaints every bit a useful placebo test. If there are omitted variables that are correlated with voting behavior and the propensity to file pothole complaints (even afterwards controlling for voting behavior in the previous election), these variables should bear on complaints filed in the six months subsequently the ballot just equally much as they outcome complaints in the vi months beforehand. Clearly, nonetheless, mail-ballot complaints are significantly less likely to affect the retrospective judgments of voters on Election 24-hour interval.¹⁶ In our setting, information technology is not possible to carry out a traditional placebo exam, in which nosotros replace the pre-ballot complaints variable with the post-election complaints because the number of potholes within neighborhoods is strongly correlated over time. This is true because potholes are near probable to be reported during both periods in areas with the worst-maintained asphalt, and the status of the pavement does not change significantly in the vi months afterwards the ballot compared to the six months prior.¹⁷ Instead, we include the postal service-election complaints count as an additional control in our model. Conditioning on mail-election complaints should account for any other omitted variable that is not already captured by our lagged dependent variable.

Spatial interdependence

The nature of the data presents another potential econometric complication. Since service levels are probable to be similar between nearby neighborhoods—for example, voters residing on two adjacent streets are likely to patronize the aforementioned local library branch or recreation heart—a uncomplicated ordinary least squares (OLS) specification does non account for the spatial relationships between precincts.¹⁸ To account for the fact that service quality is likely to be more similar among nearby precincts than between distant locations, resulting in spatial clustering in the mistake terms, we also judge models that add a spatial error component to our baseline specification:

$$\mathrm{I}\mathrm{north}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{n}{\mathrm{t}}_{\mathit{i}\mathit{t}}=\alpha +\tau \mathrm{P}\mathrm{o}\mathrm{t}\mathrm{h}\mathrm{o}\mathrm{l}\mathrm{eastward}{\mathrm{due\; south}}_{\mathit{i}\mathit{t}}+\beta \mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{east}\mathrm{northward}{\mathrm{t}}_{\mathit{i}\mathit{t}-1}+\lambda {\mathbf{\text{w}}}_i{\xi }_i+{ϵ}_i,$$

where w _i is a spatial weights matrix specifying the human relationship between precincts and ξ_i is a vector of error terms.

Equally a final robustness bank check, we also estimate a spatial lag model, which allows voting beliefs in one precinct to be affected past its proximate neighbors. The motivation for the spatial lag is that voters likely detect and care about road quality not just in their neighborhood but also the condition of the other streets they utilise regularly, perhaps during their daily commutes. This specification thus allows vote share in precinct i to be affected by potholes in the same precinct just also in other nearby areas. The precise specification is

$$\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{n}{\mathrm{t}}_{\mathit{i}\mathit{t}}=\alpha +\tau \mathrm{P}\mathrm{o}\mathrm{t}\mathrm{h}\mathrm{o}\mathrm{fifty}\mathrm{due\; east}{\mathrm{south}}_{\mathit{i}\mathit{t}}+\beta \mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{yard}\mathrm{b}\mathrm{e}\mathrm{n}{\mathrm{t}}_{\mathit{i}\mathit{t}-1}+\rho {\mathbf{\text{w}}}_i\mathrm{I}\mathrm{n}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{eastward}\mathrm{n}{\mathrm{t}}_{\mathit{i}\mathit{t}}+{ϵ}_i,$$

where due west _i is again the spatial weights matrix and

$\mathrm{I}\mathrm{northward}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{n}{\mathrm{t}}_{\mathit{i}\mathit{t}}$

is a vector of election outcomes.¹⁹

Theoretically, information technology is most plausible that the interrelationships betwixt precincts are inversely proportional to their distance from one another. To reverberate this, our preferred specifications utilise spatial weights that are an changed of the quadratic distance between precincts. This corresponds to spatial effects that decay exponentially as the distance between precincts grows.²⁰

For the urban center council elections, we pool the results from the ii districts, only nosotros alter the models above past adding the subscript d to the intercept term, α _d , representing the district-specific fixed effects. Considering we observe urban center council elections with both Democratic and Republican incumbents, this ensures that the effects we certificate are not the result of partisan-motivated attribution of responsibility (due east.g., Rudolph 2006).²¹

Results

To ease substantive interpretation and allow for straightforward comparison across models and variables, we present our results visually in a series of plots, with full regression output available in appendix H. Figure 2 presents the results of our initial specifications. The top part of the effigy corresponds to the baseline OLS model, the middle to the spatial error specification, and the bottom portion contains the spatial-lag model.

Effigy 2.

Balloter consequences of potholes in San Diego. The coefficients (with 95% conviction intervals) represent the issue of each additional pothole reported within six months of each election on the vote share of the incumbent candidate. Each specification is described in the text.

Across all specifications and both office types, we find that pothole complaints have a significant negative result on incumbents' vote share. When constituents submit more road work requests in the vi months before the election, incumbent officeholders suffer at the polls. The coefficients are also similar in size for both the mayoral and council elections, with each additional pothole complaint reducing incumbent vote share by roughly 0.two percentage points. Increasing the number of potholes from the 25th percentile in 2008 (zero potholes) to the 75th percentile (six potholes) is thus expected to reduce the incumbents' share of the vote by about ane.ii percentage points. Although this outcome is not particularly large, information technology is arguably meaningful in substantive terms. In a close election, local road quality could evidence pivotal to whether the incumbent wins another term or loses the ballot. Note that the coefficients on the spatial-lag specifications are somewhat smaller because they include simply the short-run rather than the full-equilibrium effects that business relationship for the dynamics of the spatial spillovers betwixt precincts, and thus they are not directly comparable to the other 2 estimates (encounter Ward and Gleditsch 2008, 44–49).

Nosotros present our placebo specifications, which include separate variables for both pre- and post-election pothole complaints, in figure 3. Including the complaints made after the election does non have any measurable touch on on our estimates. Across the models, the coefficients on the pre-election variable are really somewhat larger. Past contrast, the coefficients on the post-election variable are statistically indistinguishable from zero, and they have the wrong sign across the lath.²² These results significantly strengthen our power to provide a causal estimation to the documented effects of denizen pothole complaints made before the election on incumbent political support.

Figure 3.

Electoral consequences of potholes in San Diego (placebo specification). The coefficients (with 95% confidence intervals) represent the effect of each additional pothole complaint on the vote share of the incumbent candidate. The "placebo" variables capture potholes reported in the six months later on the election. Each specification is described in the text.

In addition to our primary specifications, nosotros carried out a number of robustness tests. In appendix I, we re-estimate all models later on dropping the more serious asphalt repair requests, limiting our assay merely to pothole complaints as classified by the city. The results are substantively identical, and indeed the coefficient on the pothole count variable becomes somewhat larger.²³ In appendix J, we as well estimate split up models for each of the two council districts. The coefficients are most duplicate from those in the pooled models, although they are less precisely estimated in the case of Councilman Tony Young'south Commune 4 due to the much lower sample size in the disaggregated analysis. We have run several boosted checks, including transforming the pothole count variable into a relative quartile ranking to accost concern about outliers, re-estimating the mayoral ballot analysis in the subset of precincts that are used in our city council election results, and including President Obama'due south 2008 vote share as an additional control. Overall, the results are highly consequent across unlike specifications, elections, and subsets of the data, increasing our confidence in their validity.

It is also important to annotation that, across all elections and specifications, we find a strong human relationship betwixt current electoral performance and lagged vote share. That should not be surprising—incumbents who perform well in a neighborhood in the previous ballot are likely to do well at that place once more next time they run for function. The coefficient in the mayoral election, however, is non particularly large, with Mayor Sanders gaining roughly 0.3 votes in 2008 for each vote he won during the 2005 special election that first brought him to office. This relatively modest relationship likely reflects important differences in dynamics between these 2 elections. In the 2005 mayoral runoff, Sanders's opponent was Democratic Metropolis Councilwoman Donna Frye, a self-styled progressive and ecology activist. In 2008, past contrast, local Democrats failed to recruit a viable opponent. Instead, Sanders'southward leading challenger was Republican businessman Steve Francis. Although Francis tried to recast himself as an independent populist in 2008, he was widely considered to be more than bourgeois than Sanders, a moderate Republican who publicly embraced aforementioned-sexual practice marriage long before most other elected officials—and certainly before nearly all Republican elites—did so. Given these important differences, it is not surprising that Sanders'south winning coalition would alter significantly between these two elections.

Mechanisms: Irresolute minds or changing voters?

Overall, we find consistent bear witness that more than potholes in individual neighborhoods result in fewer votes for incumbent San Diego officials. But the findings presented above practise not speak to the precise mechanisms producing these effects. Incumbents could receive fewer votes because voters alter their beliefs—after observing the quality of their local roads, voters may update their beliefs almost the competency of current officeholders and their qualifications vis-à-vis the challengers. Alternatively, incumbents could also endure if potholes change the limerick of the electorate. For case, poor-quality streets might mobilize their opponents, increasing turnout among voters predisposed to vote against the incumbent, or, alternatively, they may depress participation amongst their political base.

We can offering some insights well-nigh the mechanisms involved past examining the impact of pothole complaints on voter turnout at the precinct level. As we report in appendix One thousand, at that place appears to be little human relationship between complaints filed before the election and the rate of voter participation on Ballot Day. We should stress that these results are merely suggestive: it is possible that the decrease in participation among the incumbents' supporters is nearly perfectly offset by higher turnout amid their opponents, producing no net change in the number of voters casting ballots. Nevertheless, these results suggest that the accountability effects we notice are likely driven by voters irresolute their minds rather than by potholes altering who participates in the elections.

Substantive significance

It is also useful to put the effects we document in context past comparing them to the findings on economic voting in US presidential elections. In these contests, incumbent presidents gain roughly iv pct points of vote share for each percentage point of election-year income growth (Healy and Lenz 2014). Thus, the electoral bear upon of a single pothole on city contests is roughly one-twentieth the magnitude of a single percentage point of income growth in national presidential elections. Of course, the average ballot for metropolis role in San Diego is considerable less competitive than is the case for a typical presidential ballot, so the political importance of the consequence may be more minor in terms of its capacity to motivate officeholders and alter their incentives while in office.²⁴ Still, if the electoral consequences we document for road quality extend to voter satisfaction with other metropolis services, retrospective considerations may help shape the electoral fortunes of some city officials and thus potentially serve equally a mechanism of accountability.

Word and Conclusion

Overall, we detect stiff show that government functioning in the provision of a peculiarly salient public service has a meaningful, albeit modest, impact on city elections in San Diego. That the results are consistent across three different modeling approaches, each of which makes different statistical assumptions well-nigh voter beliefs, gives further credence to their validity. Compared to previous studies, our analysis offers a much closer fit between theory and empirics by focusing our enquiry on outcomes that voters report are salient to them and that are actually under the command of local government. In improver, our analysis relies on a more credible identification strategy by exploiting plausibly exogenous variation in service quality between neighborhoods rather than differences in boilerplate outcomes between cities.

Nevertheless, our analysis is limited to one dimension of performance during two elections in a single city, and so it is important to consider the generalizability of our findings. Although San Diego'due south municipal government structure and election rules are typical of other large and mid-size cities, suggesting that we should expect the electoral dynamics to play out similarly in other locales, there were several attributes unique to these elections worth highlighting. As we noted previously, service quality featured prominently in civic dialogues and political campaigns in San Diego during this menstruum due to the cutbacks that occurred in the wake of the city's pension crisis, and our analysis further focuses on the service that voters themselves said they cared about the about. If salience and magnitude of the failures (or successes) moderate the extent to which voters weigh government operation in service delivery when casting their ballots, we might wait to come across smaller effects for other dimensions of operation and in other contexts. Similarly, in that location may be an disproportion in voter response if they punish incumbents more for service reductions than they reward them for expansions. On the other hand, near of the potholes in our data set did ultimately get stock-still in response to citizen requests. This might accept helped blunt the balloter consequences of individual potholes—both because individuals who filed the complaints may have rewarded the city government for its responsiveness to their requests and because making repairs in a timely manner reduced the overall number of drivers who had a chance to notice them. In contexts where government officials are less responsive, we might expect similar complaints to exact a larger electoral toll. All of these possibilities deserve close attending in future inquiry on retrospective voting in local elections, which should further explore such potential heterogeneity.

While our findings are hopeful in normative terms, nosotros believe they are not sufficient to reach definitive conclusions about the quality of American democracy. Whether the type of retrospective voting we document improves autonomous accountability or perversely leads to less optimal policies ultimately depends on how government officials respond to the resulting electoral incentives, something nosotros practice non explore in our analysis. Two sets of responses are possible. First, knowing that their electoral fortunes depend in part on voter satisfaction with regime operation, incumbents may piece of work to maximize the quantity and quality of services their constituents value the almost. This would exist a desirable outcome and one envisioned by the standard sanctioning models (Ferejohn 1986). Alternatively, however, incumbents could respond strategically to the balloter incentives nosotros certificate in ways that may ultimately reduce social welfare. For case, elected officials might leverage the temporal biases in voter behavior and systematically cut investment on preventative street maintenance early in their terms to build savings that might later be used to ramp up spending on repairs only in the months leading up to each ballot (Nordhaus 1975), increasing their credit challenge opportunities but too potentially the long-term costs. Alternatively, strategic politicians might respond to voter demands for well-maintained streets by just redirecting resources from other essential only perhaps less visible or salient services, such as sewer and storm water systems.

We believe information technology is also critical to consider whether voters accurately aspect credit and blame for observed performance outcomes to the right individuals in regime. Indeed, all of the incumbents we examine in this study were elected to office in the aftermath of San Diego'southward pension crisis, and none of them were arguably to blame for the dire service declines the city has experienced over the past two decades. Information technology is unclear (at to the lowest degree to u.s.) how holding Mayor Sanders and Councilman Faulconer and Councilman Immature accountable for the sins of their predecessors contributes to improve autonomous functioning or responsiveness. Ultimately, the relationship between commonwealth and social welfare depends crucially on the types of retrospective evaluations voters bring to the ballot box and on how elected officials respond to the resulting political incentives. While our study provides a useful ready of findings that speak to the former question, we do non view it as the final word on the topic.

Acknowledgments

We give thanks Scott Lewis and Keegan Kyle at the Vocalisation of San Diego for providing us with the pothole complaint data and Shaun Fontanella, Morton O'Kelly, and Chen Xiang at Ohio State University's Centre for Urban and Regional Assay for their geocoding assistance. Dan Hopkins, Stéphane Lavertu, and Jessica Trounstine provided us with invaluable feedback and comments on earlier drafts. The authors' names are in alphabetical order.

Notes

Craig Burnett ([electronic mail protected]) is an assistant professor in the Department of Political Scientific discipline at Hofstra University, Hempstead, NY 11549. Vladimir Kogan ([email protected]) is an banana professor in the Section of Political Science at the Ohio State University, Columbus, OH 43210.

Information and supporting materials necessary to reproduce the numerical results in the paper are available in the JOP Dataverse (https://dataverse.harvard.edu/dataverse/jop). An appendix with supplementary cloth is available at http://dx.doi.org/10.1086/688736.

1. Consider Arceneaux (2006). In examining unemployment, education, and traffic congestion, the author finds that voters aspect causal responsibleness for operation to the level of authorities they consider accountable for each area. Worryingly, however, in that location is no consensus among the respondents in the study well-nigh which level of government is really responsible for each policy expanse (encounter Arceneaux 2006, 742, table 2).

ii. A growing empirical literature challenges this view, showing that city governments take the capacity to tailor policies to reply to the ideological preferences of their voters (Einstein and Kogan 2016; Tausanovitch and Warshaw 2014).

three. See also Kogan, Lavertu, and Peskowitz (2016a).

4. Such policies may explicate why crime declined faster in some cities than in others. Arnold and Carnes (2012), however, evidence that New York voters responded to the aggregate changes in criminal offence over time, not simply to the portion of that change attributable to local policy choices.

five. Arnold and Carnes (2012) take a similarly glass-half-total view, concluding, "The electoral connection provides powerful incentives for mayors to practice their best in tackling whatever problems occur on their watches, but equally it does for elected executives effectually the world" (962).

6. Drupe and Howell'southward (2007) finding that voters appear to place greater weight on the achievement levels in their neighborhood schools, rather than the average achievement in each school commune, is consistent with this argument.

vii. This is technically known as a mayor-council, as opposed to council-metropolis director, form of government

8. The political evolution of the city in the second half of the twentieth century is examined in detail by Erie, Kogan, and MacKenzie (2011), then nosotros summarize but the most relevant facts in this department.

9. The 2008 recession made things even more than difficult. As a consequence of the recession, city revenues declined and the pension system suffered investment losses equal to roughly twenty% of its asset portfolio. This further increased the pension payment fifty-fifty as revenues declined, squeezing basic city services.

x. Of course all of these factors vary between neighborhoods in a way that might be correlated with political preferences, and we accost this endogeneity problem explicitly in our assay.

11. Vocalism of San Diego generously agreed to share these data with the authors.

12. Constituents have a number of means to transmit these complaints to the city. Outset, they can contact their local city council member, whose staff pass the information on to the Streets Department. Alternatively, residents tin send the information directly to the department using an online form or a phone-based web application. Each pothole is logged merely once into the urban center's database, based on the earliest complaint fabricated, and we do not notice the full number of complaints received about whatsoever given pothole in the data. Later a complaint is registered in the metropolis's database, a crew of either city employees or contractors is dispatched to evaluate the problem and deport out whatever repairs that are necessary. Each coiffure is responsible for an assigned geographic area of the metropolis. Considering more than complaints may come in from some neighborhoods than others on whatever given 24-hour interval, the response time may vary as well, simply city officials assert that "complaints are treated every bit regardless of source or area or origin" (Kyle 2009, 2012).

thirteen. Like the proverbial tree in the forest, if no one noticed or reported a pothole to the city, there is trivial reason to expect information technology to affect electoral outcomes.

14. The selection of a six-month window also reflects, to a certain extent, data availability. The earliest complaints nosotros observe are from January 2008, and the first election during the period covered in our data took identify in early June 2008.

15. We exclude open city council races for ii reasons. First, information technology is less plausible that responsibility follows the incumbents' co-partisans in these races, because partisan labels do not appear on the ballot, making partisanship less salient. Second, most of these races feature multiple candidates from the same party, and there are few theoretical expectations most which co-partisans voters should hold responsible for the incumbents' operation.

xvi. Of grade some potholes may have existed before Ballot Day and individuals just did not report them until subsequently the ballot. These potholes could conceivably have an impact on the election. Fifty-fifty in this scenario, notwithstanding, it seems reasonable to presume that these potholes were far less salient in terms of their touch on on voting behavior on Election 24-hour interval. Empirically, this is precisely what nosotros find in our placebo specifications.

17. Indeed, we found that the number of pre-election pothole complaints is positively correlated with post-election complaints at r ≈ 0.8. Thus, just putting post-election complaints into the model by itself, every bit would be the example in a traditional placebo test, would notwithstanding uncover a significant relationship because the number of post-election complaints would be a fairly good proxy for the number of pre-election complaints.

xviii. For case, using the spatial weights we describe below, the Moran's I for the residuals from the OLS specification of the mayoral vote share model is .nineteen (p < .001). This indicates significant positive spatial autocorrelation, with the residuals in nearby precincts being more similar than those in precincts that are more than distant from one another. Moran's I can range from −1 to 1, so while the spatial correlation is significant in our case, it is relatively pocket-size in magnitude.

19. As an additional robustness cheque, we have besides estimated a spatial Durbin model. The results were identical to those beneath, then nosotros present them in appendix Thousand.

20. Our results are not sensitive to this choice. We detect the same substantive effects if nosotros instead use an changed of the linear distance or employ equal weights to all observations within a mile of each precinct.

21. This might occur if voters blame the mayor'south co-partisans for potholes, regardless of whether they are the incumbents or challengers in the race.

22. Since both the pre- and post-election pothole coefficients are estimated somewhat imprecisely, we do note that they are non always statistically meaning from one another.

23. One caption is that voters may simply blame incumbents less for serious or catastrophic problems—such as water chief breaks—because these are less predictable and thus more hard to anticipate through policy. Alternatively, it is also possible that the virtually serious issues may affect voters across many precincts rather than merely in the neighborhood where the repairs are fabricated. Our precinct-based interpretation strategy does not model such widespread, geographically diffuse issues particularly well. Equally a issue, they essentially fall out of the analysis.

24. Since 2000, the average margin of victory in all metropolis contests has been more than xx% (the median was fifteen.3%).

References

• Arceneaux, Kevin. 2006. "The Federal Face up of Voting: Are Elected Officials Held Accountable for the Functions Relevant to Their Office?" Political Psychology 27 (five): 731–54.

• Arnold, Douglas R., and Nicholas Carnes. 2012. "Holding Mayors Accountable: New York's Executives from Koch to Bloomberg." American Journal of Political Science 56 (4): 949–63.

• Barro, Robert J. 1973. "The Control of Politicians: An Economic Model." Public Option 14 (1): xix–42.

• Behavior Enquiry Center. 2010. "Urban center of San Diego Residents' Opinions on City Services." Technical report, April, Behavior Research Center, Phoenix, AZ.

• Bernstein, David, and Noah Isackson. 2014. "The Truth virtually Chicago'southward Crime Rate: Part 2." Chicago Magazine, Apr 7.

• Berry, Christopher R., and William G. Howell. 2007. "Accountability and Local Elections: Rethinking Retrospective Voting." Journal of Politics 69 (3): 844–58.

• Brown, Adam R. 2010. "Are Governors Responsible for the State Economy? Partisanship, Arraign, and Divided Federalism." Journal of Politics 72 (3): 605–xv.

• Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. New York: Wiley.

• Chetty, Raj, John N. Friedman, and Jonah Due east. Rockoff. 2014. "Measuring the Impact of Teachers I: Evaluating Bias in Teacher Value-Added Estimates." American Economical Review 104 (9): 2593–2632.

• Chingos, Matthew G., Grover J. Whitehurst, and Michael R. Gallaher. 2015. "School Districts and Student Achievement." Teaching Finance and Policy 10 (3): 378–98.

• Cox, Gary W., and Mathew D. McCubbins. 1986. "Electoral Politics every bit a Redistributive Game." Journal of Politics 48 (2): 370–89.

• Delli-Carpini, Michael 10., and Scott Keeter. 1996. What Americans Know well-nigh Politics and Why It Matters. New Oasis, CT: Yale University Press.

• Einstein, Katherine Levine, and Vladimir Kogan. 2016. "Pushing the City Limits: Policy Responsiveness in Municipal Regime." Urban Affairs Review 52 (one): iii–32.

• Erie, Steven P., Vladimir Kogan, and Scott A. MacKenzie. 2011. Paradise Plundered: Fiscal Crunch and Governance Failures in San Diego. Palo Alto, CA: Stanford University Press.

• Ferejohn, John. 1986. "Incumbent Performance and Balloter Control." Public Choice 50 (1–3): 5–25.

• Healy, Andrew, and Gabriel S. Lenz. 2014. "Substituting the Finish for the Whole: Why Voters Respond Primarily to the Ballot-Yr Economy." American Journal of Political Science 58 (ane): 31–47.

• Healy, Andrew, and Neil Malhotra. 2013. "Retrospective Voting Reconsidered." Annual Review of Political Science 16:285–306.

• Hopkins, Daniel J., and Lindsay M. Pettingill. 2015. "Economical Voting in Big-Urban center U.Southward. Mayoral Elections." https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2567294 (accessed November 7, 2016).

• Huber, Gregory A., Seth J. Colina, and Gabriel S. Lenz. 2012. "Sources of Bias in Retrospective Decision Making: Experimental Show on Voters' Limitations in Decision-making Incumbents." American Political Scientific discipline Review 106 (4): 720–41.

• Kaufmann, Karen K. 2004. The Urban Voter: Grouping Conflict and Mayoral Voting Beliefs in American Cities. Ann Arbor: University of Michigan Printing.

• Kogan, Vladimir, Stéphane Lavertu, and Zachary Peskowitz. 2016a. "Practice School Study Cards Produce Accountability through the Ballot Box?" Journal of Policy Analysis and Management 35 (3): 639–61.

• Kogan, Vladimir, Stéphane Lavertu, and Zachary Peskowitz. 2016b. "Functioning Federalism and Local Democracy: Theory and Evidence from Schoolhouse Levy Elections." American Journal of Political Science threescore (ii): 418–35.

• Kyle, Keegan. 2009. "The City's Faux Pothole Pledge." Voice of San Diego, November 29.

• Kyle, Keegan. 2012. "Complainted nigh a Pothole? Meliorate Exist Patient." Vocalisation of San Diego, June xx.

• Lenz, Gabriel Fifty. 2012. Follow the Leader? How Voters Respond to Politicians' Policies and Functioning. Chicago: University of Chicago Printing.

• Miller, Michael Thousand. 2013. "For the Win! The Upshot of Professional Sports Records on Mayoral Elections." Social Science Quarterly 94 (1): 59–78.

• Nordhaus, William D. 1975. "The Political Business Wheel." Review of Economic Studies 42 (2): 169–xc.

• Oliver, Eric J., and Shang E. Ha. 2007. "Vote Pick in Suburban Elections." American Political Science Review 101 (iii): 393–408.

• Popkin, Samuel L. 1994. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. Chicago: University of Chicago Press.

• Poston, Ben, and Joel Rubin. 2014. "LAPD Misclassified Nearly ane,200 Violent Crimes equally Pocket-size Offenses." Los Angeles Times, August nine.

• Roeder, Oliver, Lauren-Brooke Eisen, and Julia Bowling. 2015. What Caused the Crime Decline? New York: Brennan Middle for Justice, New York Academy Schoolhouse of Constabulary.

• Rogers, Steven Michael. 2013. "Accountability in a Federal System." PhD diss., Princeton Academy.

• Rudolph, Thomas J. 2003a. "Institutional Context and the Assignment of Political Responsibility." Periodical of Politics 65 (ane): 190–215.

• Rudolph, Thomas J. 2003b. "Who'due south Responsible for the Economic system? The Formation and Outcome of Responsibility Attributions." American Periodical of Political Science 47 (four): 698–713.

• Rudolph, Thomas J. 2006. "Triangulating Political Responsibility: The Motivated Formation of Responsibility Judgments." Political Psychology 27 (1): 99–122.

• Tausanovitch, Chris, and Christopher Warshaw. 2014. "Representation in Municipal Authorities." American Political Science Review 108 (3): 605–41.

• TRIP. 2015. "Bumpy Roads Ahead: America's Roughest Rides and Strategies to Brand Our Roads Smoother." Technical report, July, TRIP, Washington, DC.

• Trounstine, Jessica. 2016. "Segregation and Inequality in Public Goods." American Periodical of Political Scientific discipline 60 (3): 709–25.

• Ward, Michael D., and Kristian Skrede Gleditsch. 2008. Spatial Regression Models. G Oaks, CA: Sage.

osunauncin1937.blogspot.com

Source: https://www.journals.uchicago.edu/doi/full/10.1086/688736

Share this post