Essay

When Is “DelIverIng the gooDs” not gooD enough?

how economic Disparities in latin American neighborhoods shape Citizen

trust in local government By ABBy CórDovA and MAttheW l. lAyton*

IntroductIon

the third wave of democratization that swept through the devel-oping world resulted in greater autonomy and a more important role for local governments in the provision and distribution of public goods and services, among other political transformations.1 scholars have since provided compelling empirical evidence that trust in local government is a vital component of the overall perceived legitimacy of a political system.2 As such, trust in local government is likely to bolster compliance with government regulations and taxation,3 thereby sustaining the rule of law and providing governments with crucial re- sources. extant research suggests that trust in local government can help pro- mote the stability of democracy, particularly in new democracies. even Matthew Cleary and susan stokes, who emphasize the merits of citi- zen skepticism toward democratically elected officials in latin Amer- ica, point to the importance of institutional trust for building stronger

* the data used in this project were collected as part of a larger study carried out by the latin American Public opinion Project (lapop), which evaluated the impact of the united states Agency for International Development’s (usaId) programs in Central America. We thank lapop for making the data available to us, and usaId for providing the funds to collect the data. We are also thankful to Mark Peffley and anonymous reviewers for their valuable feedback on previous drafts of this ar- ticle. An early version of this article was presented at the 2012 Midwest Political science Association conference, where we received excellent comments from sam handlin and other panel participants.

1 Bardhan and Mookherjee 2006; Campbell 2003; Faguet 2012. 2 Booth and seligson 2009; hiskey and Bowler 2005. 3 levi 1988; levi and stoker 2000; scholz and lubell 1998.

World Politics 68, no. 1 ( January 2016), 74–110 Copyright © 2015 trustees of Princeton university doi: 10.1017/s0043887115000441

d e l I v e r I n g t h e g o o d s 75

4 Cleary and stokes 2006; Cleary and stokes 2009. 5 espinal, hartlyn, and Kelly 2006; hetherington 1998; Mishler and rose 2001; rahn and ru-

dolph 2005. 6 hibbing and theiss-Morse 1995; hibbing and theiss-Morse 2002; levi 1998; van ryzin 2011. 7 Anderson et al. 2005; Kam 2005; slothuus and de vreese 2010; taber and lodge 2006. 8 huckfeldt’s classic study of neighborhood dynamics in the united states conceptualized the

“neighborhood” as a “shared geographic locale of a residential grouping” (huckfeldt 1986, 2). We refine this conceptualization slightly for application to the latin American context and define “neigh- borhoods” as the smallest geographic units within a local government jurisdiction that have been given a distinct proper name. In the latin American context, neighborhoods are typically called barrios, vecindarios, or colonias. the Portuguese equivalent is bairros.

9 van den Bos, lind, and Wilke 1998; van den Bos, lind, and Wilke 2001.

democracies when that trust results from the capacity of public institu- tions to hold elected officials accountable.4 yet despite the importance of institutional trust and the crucial role that local governments play in sustaining economic development and democratic governance, theories about what factors foster or undermine citizen trust in local govern- ments remain underdeveloped.

Prior research has demonstrated the importance of three factors in shaping citizen attitudes toward their government: (1) citizens’ eval- uations of government performance in terms of the quality of policy outcomes or the goods and services provided;5 (2) the process of gov- ernance, including the perceived fairness of state actions;6 and (3) the partisan or ideological congruence between citizens and the current administration.7 the theoretical framework we develop in this article moves this literature forward in two ways.

First, our theory posits that independent of citizens’ partisan sym- pathies, their position of relative deprivation will ultimately determine whether concerns with performance or fairness exert a stronger influ- ence on their level of trust in local government. More specifically, we argue that for poor people who live in highly unequal neighborhoods, concerns with fairness will outweigh their assessments of performance in shaping their trust in local government. Conversely, more affluent individuals will put more emphasis on government performance when placing trust in their local government, regardless of contextual eco- nomic inequality. We argue that these two dimensions of evaluation— performance and fairness—have varying effects on trust in local gov- ernment, depending on the relative economic standing of individuals within their social context.

second, because the formal processes of governance often go un- noticed by the average citizen, we posit that people largely judge the local government’s fairness by taking into account the most salient and accessible indicator available: the inequality of their neighborhood.8 Drawing on fairness heuristic theory (fht),9 we argue that the eco-

76 w o r l d p o l I t I c s

10 elbers et al. 2004; elbers et al. 2005. 11 Baker, Ames, and renno 2006; Canache 1996.

nomic inequality of one’s neighborhood is an important heuristic for evaluating the extent to which poor residents in particular can trust their local government. Although economists have documented high levels of economic inequality within neighborhoods in African and latin American countries,10 and political scientists have shown the im- portance of neighborhood dynamics for democracy in the current era of government decentralization in the developing world,11 the importance of neighborhood inequality for shaping trust in local government has rarely been studied. As such, our research fills an important gap in the literature and confirms the political relevance of the neighborhood as a unit of analysis.

the fieldwork that serves as the basis for this project entailed the collection of census, public opinion, and systematic observation data within resident-identified neighborhood borders, as opposed to using predetermined and arbitrary administrative units, such as clusters of census tracts. the data were collected in seventy-one neighborhoods sampled from six municipalities in el salvador. our empirical analysis of these data relies on multilevel models that incorporate measures of both individual-level characteristics and neighborhood context to test our argument. As our theory predicts, we find that among the poor- est third of respondents, the perceived performance of the local gov- ernment in the provision of public services becomes unimportant as a predictor of trust in local government when the respondents reside in neighborhoods where access to services and goods is highly unequally distributed across households. these findings imply that trust in local government can be broadly fostered across individuals of different so- cioeconomic backgrounds only when a government signals distributive fairness for the poor in addition to high effectiveness in public service provision.

our findings carry with them important policy implications. they suggest that government decentralization is more likely to promote widespread political trust if local governments implement development strategies that tackle within-neighborhood inequality in addition to shortcomings in service quality. We provide evidence that at the high- est levels of neighborhood inequality, local governments may be un- able to broadly increase perceptions of government legitimacy through investments in the quality of public services alone. As a result, in spite of the fixed costs associated with establishing progressive policies to alleviate inequality in living conditions at the neighborhood level, the

d e l I v e r I n g t h e g o o d s 77

long-term payoff of doing so may prove highly beneficial to local gov- ernments and, by extension, the national political system.

determInants of polItIcal trust

Political trust is typically defined as “a basic evaluative orientation to- ward the government founded on how well the government is operat- ing according to people’s normative expectations.”12 Consistent with this definition, numerous studies have demonstrated that citizens’ percep- tions of government performance in the delivery of expected policy outcomes are critical for building trust in government in both devel- oped and developing countries.13 For example, scholars studying trust in local governments in the united states have concluded that subjec- tive evaluations of the quality of services provided by the local govern- ment is one of the main determinants of political trust.14 similarly, in the context of developing countries, Michael Bratton finds that citizens in Africa inform their views about local government “by instrumental attitudes about whether governments deliver the goods.”15 In the latin American context, individual-level studies on the determinants of trust in local government have found similar results.16 each of these studies echoes the argument that “political trust is politically endogenous. … It is a consequence, not a cause, of institutional performance.”17 Despite compelling evidence that perceptions of government perfor- mance lead to institutional trust, research in the field of public admin- istration indicates that this is not always the case.18 In some contexts or for some citizens, developing trust in local government requires more than a government’s demonstrated capacity to “deliver the goods.” In other words, political trust is not always determined by perceived gov- ernment performance, as typically assumed. this is precisely the puzzle that motivates our research question: under what conditions are per- ceptions of government performance insufficient to promote institu- tional trust? We argue that part of the disjunction between evaluations of local government performance and political trust can be explained by the fact that some citizens are more likely to prioritize the fairness

12 hetherington 1998, 791, emphasis added. 13 For example, espinal, hartlyn, and Kelly 2006; hetherington 2005; Mishler and rose 2001;

Wolak and Palus 2010. 14 rahn and rudolph 2002. 15 Bratton 2012, 517. 16 Montalvo 2010. 17 Mishler and rose 2001, 31. 18 van ryzin 2011; van de Walle and Bouckaert 2003; yang and holzer 2006.

78 w o r l d p o l I t I c s

of government over their perceptions of government performance. We therefore maintain that some citizens value fairness as a “normative ex- pectation” of government action in addition to their expectation for high-quality performance. A substantial body of research suggests that an important criterion for trusting government is the perception that governmental authori- ties make decisions and distribute public goods and services fairly.19 As Margaret levi puts it, “Individuals also need to have evidence that government is relatively fair … if they are to have confidence that the state will harmonize the interest of otherwise competitive parties.”20 In other words, many people want government to demonstrate a “lack of bias or favoritism” toward certain groups in society,21 especially if they believe that they would lose out from such favoritism.

While these theoretical insights are valuable for understanding the importance of evaluations of fairness in the formation of trust in gov- ernment, two questions remain understudied: (1) how do perceptions of fairness form? And (2) under what conditions are citizens likely to place a higher priority on a government’s fairness over its performance when evaluating trust in local government? fht provides a framework for answering the first question.22 the theory maintains that individu- als value fairness information when deciding whether to trust authori- ties. Individuals turn to fairness assessments because they often lack readily available objective information about whether authorities can be trusted. Making a wrong judgment can be quite costly, given that an overestimation of how much one can trust an entity with power may increase the chances for future exploitation or exclusion from valuable benefits. Accordingly, fht posits that individuals use perceived fair- ness of past outcomes or current conditions as a heuristic to help make judgments about the level of trust to cede. fht further maintains that citizens draw on the fairness information that is most easily accessible in their social environment.23

Moreover, according to fht, fairness judgments are based on judg- ments of both distributive justice and procedural justice. Distributive fairness judgments include perceptions of equity in the distribution of outcomes or the belief that authorities have allocated to a given in- dividual what she deserves. A core assumption underlying the notion

19 hibbing and theiss-Morse 1995; levi 1998; tyler 2001. 20 levi 1998, 90. 21 van ryzin 2011, 747. 22 van den Bos, lind, and Wilke 1998; van den Bos, lind, and Wilke 2001. 23 tost and lind 2010.

d e l I v e r I n g t h e g o o d s 79

of distributive fairness is that individuals evaluate whether they have received a fair share by comparing their allocation to that of others in a given reference group. Consequently, perceptions of distributive fairness are central to theoretical frameworks that consider relative deprivation as a trigger of social discontent.24 In contrast, procedural fairness judgments assess the procedures by which policy decisions are made or through which outcomes are delivered, independently of how outcomes are distributed. For example, as gregg van ryzin suggests, citizens who have witnessed corruption when dealing with their local government are unlikely to perceive procedures as fair.25 similarly, citi- zens are unlikely to perceive procedural fairness if they have not been given a chance to voice their policy preferences or have been excluded from the policy-making process.26 While we incorporate measures of both distributive and procedural fairness in our analyses, the role of distributive fairness judgments is especially important for our argument about the impact of economic inequality on trust in local government.

We theorize that citizens’ observations of economic inequality in their neighborhood provide them with heuristic information related to the local government’s distributive fairness, or equity, in the delivery of goods and services, independent of evaluations of government perfor- mance or the quality of policy outcomes. Moreover, we posit that this distributional evidence will provide the basis for their level of trust in the local government. yet not all individuals should be equally attentive to signals of distributive fairness. Because the poor bear a much greater cost from inequality given that they are the ones who face deprivation in a relative sense,27 residents with the lowest economic status should be more likely to use observations of neighborhood inequality to evaluate the trustworthiness of local government. In contrast, the relatively af- fluent should be less concerned with distributive fairness and neighbor- hood inequality, ceteris paribus.

Fairness heuristic theory also helps to highlight the conditions under which citizens are likely to place greater weight on distributive fairness over performance when assessing local government trust. Poor citi- zens who live in highly unequal neighborhoods are expected to attach greater weight to distributive fairness than to local government per- formance when determining their local government’s trustworthiness. For the poor, improvements in the quality of services, such as keeping

24 tyler and van der toorn 2013, 631. 25 van ryzin 2011. 26 see van den Bos, lind, and Wilke 1998. 27 graham and Felton 2005; gurr 1970; luttmer 2005.

80 w o r l d p o l I t I c s

public spaces clean or renovating a park or soccer field, are unlikely to be viewed as providing for their basic needs. the effects of perceptions of government performance on political trust will depend therefore on perceptions of distributive fairness based on observations of economic inequality in the neighborhood and the economic standing of the indi- vidual. the poor will place higher value on fairness, whereas the afflu- ent will value government performance regardless of economic dispari- ties in their neighborhood.28 In the section that follows, we present the theoretical basis for our proposition that the neighborhood in particular, rather than the municipality, province, or nation, is an important con- text for the formation of judgments about distributive fairness.

neIghborhood context and dIstrIbutIve faIrness Judgments

In his classic work, robert huckfeldt emphasizes the importance of neighborhood context for the formation of citizens’ political attitudes by highlighting the crucial role that residential proximity plays in de- termining individuals’ personal and impersonal social encounters.29 Al- though individuals can avoid certain parts of the city or municipality in which they live, they cannot avoid their neighborhood of residence and thus will inevitably encounter other neighborhood residents. given that social interactions serve as a crucial source of political information, huckfeldt posits that social experiences within one’s neighborhood in- fluence citizens’ political attitudes and behaviors. In particular, he points to neighborhood composition as an important factor that conditions residents’ political orientations. heterogeneous neighborhoods are more likely to trigger dissonant political reactions among residents who do not identify with the elite or dominant group. this applies equally to economic heterogeneity. social-class inequal- ity may spur neighborhood-based political conflict because it may lead economically disadvantaged residents to perceive that a given political institution or actor takes sides with a relatively economically privileged group.30

28 the bulk of the literature that examines the attitudinal impacts of economic inequality has fo- cused on variables such as social trust (rothstein and uslaner 2005; uslaner 2002), or more generally examines how contextual factors condition the effects of social trust ( Jamal and nooruddin 2010). In this article we examine the impacts of contextual inequality on trust toward local government and explore the consequences of economic inequality at a level of analysis rarely considered: the neighbor- hood.

29 huckfeldt 1986. 30 huckfeldt 1986, 18–19.

d e l I v e r I n g t h e g o o d s 81

thus, to explicitly make the connection between fht and huckfeldt’s insights, we theorize that the level of economic inequality within the neighborhood will shape low-economic-status residents’ evaluations of a local government’s distributive fairness and thereby impact trust in that government. More specifically, the poor form their perceptions of distributive fairness as they share information about other neighbors with those in their immediate social circle within the neighborhood or simply through observation of their neighbors. evidence based on research in the united states is consistent with this account. For ex- ample, erzo luttmer finds that inequality in personal earnings among neighbors leads to a sense of personal deprivation for the less affluent.31 Although one rarely has direct knowledge of the salary of one’s neigh- bors, observable signals of that salary or more generalized economic success, such as the appearance of a new car in the driveway or building an addition to a home, may provide readily observable ways to monitor the economic situation of one’s neighbors even without direct informa- tion sharing. one way or another, people will tend to know where they stand economically in relation to their neighbors, arguably much more so than in relation to residents of other parts of the country, province, or even municipality, given the factor of residential proximity.

relevance of the neIghborhood for local governance In latIn amerIca

Although our theory is general, we argue that the neighborhood is par- ticularly important for shaping perceptions of local government fairness and legitimacy in the current era of government decentralization in latin America. While we do not disregard the possibility that neigh- borhood context might affect citizen attitudes toward other political institutions, such as the national government, we see an even stron- ger connection between neighborhood context and evaluations of local government in the context of latin America. As we explain in this section, this is in no small part because decentralization reforms in this region substantially increased the salience of the neighborhood as a so- cial space for citizens’ political participation and involvement with the policy-making processes of local governments. under decentralization, central governments in latin America de- volved power to local governments with the objective of “bringing the state closer to the people.”32 the expectation for this process was that

31 luttmer 2005. 32 selee 2004.

82 w o r l d p o l I t I c s

localized information and citizen participation in deliberative gover- nance at the local level would result in improved governance in terms of both performance and equity.33 As the process of decentralization gave distributive and budgetary responsibilities to local governments in an effort to ensure a more equitable distribution of resources and better- quality services, local governments turned to submunicipal organiza- tions to help aggregate political interests and organize the allocation of public resources.34 tim Campbell notes that “with more power and money at their disposal, local officials began to see the importance of listening to voices from ‘below,’ that is, tapping into the sentiments of voters, citizen groups, and neighborhood organizations as a part of making plans and budget tradeoffs.”35 In many municipalities in the region, citizens now participate in the local policy-making process pri- marily through their neighborhood.36 In fact, neighborhood organiza- tions often give residents a voice in local governmental budgetary deci- sions regarding the allocation and distribution of resources.37

the recent history of el salvador, the focal country of this article, reflects the regional trends in decentralization and the increasing im- portance of neighborhood politics in the region. the Código Municipal (municipal code), which regulates the establishment and functions of salvadoran municipalities, actively encourages the legal constitution of neighborhood associations. specifically, it mandates periodic meetings between members of neighborhood associations and their local gov- ernment to assess and solve neighborhood problems. salvadorans have the opportunity to participate in public policy decisions at the sub- municipal level through the Consulta Vecinal Sectorial (neighborhood sector consultations) and participatory budgeting meetings.38 survey data from the 2010 AmericasBarometer confirms the importance of the neighborhood as a space for political deliberation in el salvador: about 27 percent of the adult population in el salvador reports partici- pation in meetings of a neighborhood improvement committee at least once a year. this rate is statistically indistinguishable from the level of participation in neighborhood associations of citizens in the twenty other Central American and south American countries represented in the 2010 AmericasBarometer survey (that is, 27.8 percent, according to our own analysis of these data).

33 For example, Bardhan 1996; Faguet 2012. 34 eaton 2012. 35 Campbell 2003, 79. 36 Canel 2010. 37 Wampler 2007. 38 torres and lópez 2008.

d e l I v e r I n g t h e g o o d s 83

notably, in the salvadoran context, the rhetoric of ensuring both quality outcomes and equitable provision of services at the neighbor- hood level and the importance of neighborhood associations to fulfill that goal is also one of the central talking points used by political par- ties during and after political campaigns. the following excerpt from the 2009 inaugural speech of the mayor-elect of el salvador’s capital city, san salvador, illustrates this point:

today I want to remind the salvadoran voters what they voted for, what our vision is for the city, and tell them what our plan is. … over the last year during our campaign we visited thousands of homes in every district, from every social class; we met with neighborhood associations seeking an authentic participa- tory citizenship to nurture our plan of government. … you will see me walk- ing through various parts of the city aware that the mayor is the jurisdictional, administrative official who is closest to the citizens but above all that this is the way to best understand reality to then confront it. you will see me in siberias, in Chacra, in san Jacinto, in Colonia satelite … and in every community … re- solving problems with neighbors, and I will stay especially close to the neediest families with special programs to support their needs and quality of life. 39

given the substantial changes in state-society relationships wrought by decentralization in this and other countries in the region and with the ascent of the neighborhood as an important space for making deci- sions about the allocation and distribution of resources, we argue that the heuristic information provided by neighborhood inequality will be particularly relevant for the poor to evaluate whether their local gov- ernment has acted fairly and implemented equitable public policies. latin America provides a propitious setting to test the overall theory we propose, given that the high levels of economic inequality observed across the region are reflected at the local level in high inequality within neighborhoods.40

hypotheses

Following our theoretical discussion, we posit that high levels of in- equality in living conditions within neighborhoods will undermine the level of trust in local government among the relatively poor because they are the ones who are more likely to perceive that their preferences and needs have not been taken into account in the policy-making pro- cess. We also argue that high levels of neighborhood inequality will

39 translation by the authors. the full speech in spanish can be accessed at http://www.youtube .com/watch?v=9mkhyxnAoe4.

40 elbers et al. 2004; elbers et al. 2005; lópez-Calva and lustig 2010.

84 w o r l d p o l I t I c s

reduce political trust among the relatively poor, even if they perceive that the local government has provided good quality services. More concretely, we propose the following two complementary hypotheses:

—h1. high economic inequality within neighborhoods will have a negative effect on trust in local government, particularly among residents with the lowest relative socioeconomic status.

—h2. high economic inequality within neighborhoods will attenu- ate the positive effect of perceptions of high government performance on trust in local government, particularly among residents with the lowest relative socioeconomic status.

We illustrate these two hypotheses in Figure 1. Among low-eco- nomic-status neighborhood residents we expect to find a negative re- lationship between neighborhood inequality and trust in local govern- ment. Among better-off residents, trust in local government will not be dependent on neighborhood inequality, as depicted by the flat lines. In addition, we expect perceptions of local government performance to matter in the development of trust in local government on average, but this is conditional on contextual inequality and individual socioeco- nomic status.

As shown with the lines representing the poor in Figure 1, at high levels of inequality, we hypothesize that perceptions of high govern- ment performance will not be associated with greater trust in local government relative to perceptions of low government performance, but at low levels of inequality, perceptions of high performance will be associated with a higher level of trust compared to perceptions of low performance. In contrast, for the relatively wealthy, perceptions of high government performance will correlate with a higher level of trust in local government relative to perceptions of low government perfor- mance at all levels of contextual inequality. In our empirical analysis we are looking for evidence of these patterns to assess the validity of our hypotheses.

research desIgn

to conduct our empirical tests we rely on an original data set collected in 2010 and 2011 that includes information for seventy-one neighbor- hoods selected across six municipalities (local government jurisdictions) in el salvador: Chalchuapa, san Juan opico, santa Ana, santa tecla, Zaragoza, and san salvador. As in previous neighborhood-level stud- ies for the united states,41 the selection of municipalities, neighbor-

41 For example, huckfeldt and sprague, 1995.

d e l I v e r I n g t h e g o o d s 85

fIgure 1 IllustratIve hypotheses

relatively wealthy and perceive relatively wealthy and perceive high performance low performance

relatively poor and perceive relatively poor and perceive high performance low performance

low neighborhood Inequality

high

t ru

st in

l oc

al g

ov er

nm en

t

hoods, and individuals within neighborhoods was designed to increase the variance of socioeconomic characteristics across different units of analysis in our sample. Although the six municipalities selected are lo- cated in the same region of the country, they have very different levels of socioeconomic development. the municipalities of san salvador, santa Ana, and santa tecla are among the most urbanized and devel- oped in the country. Chalchuapa, san Juan opico, and Zaragoza are much poorer and more rural. Within each municipality there are also very marked differences in the extent of poverty across and within neighborhoods. to ensure that the sample would contain both poor and relatively affluent neighbor- hoods, the universe of neighborhoods in each municipality was clas- sified into two groups based on a poverty threshold.42 randomized neighborhood selections were then taken within each group. the re- sulting sample includes neighborhoods with very different socioeco- nomic characteristics and is large enough to allow us to explore indi- vidual- and neighborhood-level effects with a high level of statistical precision.

three different forms of data were collected across the seventy-one selected neighborhoods using original questionnaires designed to study neighborhood patterns in el salvador: a census, a public opinion sur- vey, and a systematic observation of each neighborhood. the census

42 the selection of neighborhoods was based on data from the 2007 census provided by the census bureau in el salvador.

86 w o r l d p o l I t I c s

was used to enumerate all households, to obtain information on the number of blocks in each selected neighborhood, and to collect infor- mation on household characteristics within blocks. Although the 2007 national census data made available by the census bureau was useful for the selection of neighborhoods within each municipality, it provided observations only for clustered census tracts, which did not necessar- ily coincide with actual neighborhood borders. In contrast, the cen- sus carried out for this study provided the opportunity to ensure that neighborhood indicators based on the census data, as well as the survey and the systematic observation, were gathered within the borders of resident-defined neighborhoods.

As dwellings were enumerated for the collection of the census data, maps were drawn in consultation with neighborhood leaders to identify neighborhood borders. As Cara Wong and her coauthors argue, “Pic- tures in our heads do not resemble governmental administrative units in shape or content,”43 and so can compromise the validity of studies that rely solely on census bureau–defined geographic units to examine contextual effects on citizen attitudes. By involving residents in the de- marcation of their neighborhood’s borders during the collection of our own census data, the design of our study overcomes that constraint, and therefore ensures that our ecological analysis effectively captures social dynamics in geographic spaces that are meaningful to citizens.

the census data collected across the seventy-one neighborhoods re- sulted in observations for a total of 30,791 household members corre- sponding to 8,516 households. As described below, we use this census data to estimate our main independent variable at the neighborhood level, inequality within neighborhoods, and to compute a control variable for the overall poverty level, absolute living conditions in neighborhoods. this census data also served as the sampling frame for the design of the public opinion survey. the neighborhood census data reveal that neighborhoods in our sample vary substantially in size, ranging from 48 to 748 dwellings. however, even in the smallest neighborhoods we find significant variation in the socioeconomic characteristics of residents, which allows us to estimate an aggregate indicator of inequality based on the census data and to draw a sample for the public opinion survey in each neighborhood with substantial variance in the socioeconomic background of individuals.44

43 Wong et al. 2012, 53. 44 table A1 in Córdova and layton 2015 presents the total number of dwellings in each neigh-

borhood reflected in the census data we collected. In addition, we present descriptive statistics of the socioeconomic characteristics of residents in each neighborhood based on the census and the public opinion data we gathered.

d e l I v e r I n g t h e g o o d s 87

the public opinion survey was carried out face-to-face among the voting-age population (individuals who were eighteen years old or older) in each neighborhood. to account for population characteristics from the census data gathered for this study, survey respondents were randomly selected within neighborhoods based on a preestablished gender and age quota for each block. only one household member was interviewed in each home. the total sample of the public opinion survey comprises 4,096 individual interviews across the seventy-one neighborhoods. the questionnaire of the public opinion survey con- tains items on a wide array of topics, including the key questions we use to measure attitudes toward political institutions and the local govern- ment, as well as questions on political participation, social capital, and residents’ experiences in their neighborhood.

Finally, following previous work in sociology,45 a systematic obser- vation questionnaire was administered with the purpose of collecting information on the physical condition of public spaces in each block of the seventy-one neighborhoods. trained observers visited each neigh- borhood and noted the presence or absence of “physical disorder” in public spaces or signs of deterioration in neighborhood conditions at the block level, including the presence of garbage mounds, sewage or waste in the streets, or a lack of public lighting, among others.46 A total of 2,012 blocks were observed across the seventy-one neighborhoods. In this article, we use the systematic observation data to assess the ro- bustness of our results to the inclusion of a “physical disorder” index, which objectively measures a neighborhood’s access to good-quality public services.

measurement and methods

dependent varIable to measure trust in the local government, respondents were asked, “to what extent do you trust the municipal government?”47 responses range

45 For example, sampson and raudenbush 1999. 46 the physical-disorder index is calculated by first taking the count of whether observers made

note of any of the following five items in the block-level observation: (1) garbage dumps or mounds of trash scattered outdoors throughout the area; (2) garbage or broken glass in the streets or on sidewalks; (3) empty lots with overgrown grass; (4) sewage or waste in the streets; and (5) lack of public electricity. the final index score is calculated at the neighborhood level by averaging the count scores for all the blocks within each neighborhood.

47 Questionnaires were pretested in el salvador and customized to the country’s lexicon. the wording in spanish of the survey item on trust in local government reads, “¿hasta qué punto tiene usted confianza en su municipalidad?” the word municipalidad makes direct reference to the local government in spanish-speaking countries in latin America, including el salvador. the same sur-

88 w o r l d p o l I t I c s

from 1 (not at all) to 7 (a lot). Based on the pooled results from the 4,096 individuals interviewed across the seventy-one neighborhoods, the average level of trust in the municipal government is 4.53, which is not statistically different from the national average of 4.5 from the same survey item used in the 2010 AmericasBarometer survey for el salvador (n = 1,550), carried out by lapop.48 In addition, we find very small differences between our neighborhood-based sample and the lapop survey on key demographic characteristics, such as education, gender, and age. As a result, we are confident that our neighborhood- based sample as a whole reflects the general contours of el salvador’s population.49 Comparing levels of trust in local government across the seventy- one neighborhoods, however, we find considerable variation, with mean values of trust ranging from a low of 3.6 to a high of 5.3 on the seven-point scale, with a standard deviation of 0.32. the results of a one-way analysis of variance test (anova) confirm that the varia- tion in trust across neighborhoods is statistically significant (F = 2.53; p<0.001), which suggests that neighborhood contextual factors may ac- count for the heterogeneity.

Independent varIables at the neIghborhood level At the neighborhood level, we examine the impact of inequality within neighborhoods (citizens’ relative living conditions) on trust in local gov- ernment, controlling for the neighborhood’s overall poverty level (ab- solute living conditions in neighborhoods) and its vulnerability to crime and violence (neighborhood crime). our neighborhood inequality mea- sure was computed with the census data collected for this study using the methodology developed by David McKenzie,50 which is based on household assets and principal component analysis (pca). McKenzie’s methodology has been widely used to measure inequality in living stan- dards in the absence of income data.51 As demonstrated by McKenzie, the index yields reliable measures of inequality in standards of living at the subnational level and has been found to be strongly associated

vey wording is used in the lapop national surveys. For more details on the measurement, descriptive statistics, and wording of the variables we employ in this article see tables A2–A4 in Córdova and layton 2015.

48 latin America Public opinion Project 2015. 49 table A5 in Córdova and layton 2015 compares the demographic characteristics of respon-

dents in our neighborhood-based sample to those in the 2010 nationally representative sample col- lected by lapop.

50 McKenzie 2005. 51 For example, labonne and Chase 2009.

d e l I v e r I n g t h e g o o d s 89

with traditional income-based measures of income inequality, such as the gini coefficient. Formally, McKenzie’s index is defined as follows:

I sj , (1) j = √ l

where for a given list of household assets, s j is the sample standard

deviation of the first principal component factor, y i , across households

in neighborhood j, and l is the largest eigenvalue, or variance, of y i

over the whole sample, typically the national sample. the first principal component factor across households, y

i , for household assets x, is the

linear combination

y i = α

1 ( x

1 – x−

1 ) + α 2 (

x 2 – x−

2 ) + ... + α k ( x

k – x−

k ), (2) s 1

s 2

s k

where x−

k and s

k are the mean and standard deviation of asset x

k and α

represents the weight for each asset x k . the first principal component y

i

yields a wealth index that assigns a larger weight to assets that vary the most across households so that an asset found in all households is given a weight of zero.52 thus, a luxury asset such as a vehicle or a computer is weighted more heavily to reflect a higher individual socioeconomic status.

In this article, scores based on the McKenzie index, I j , greater than,

less than, or equal to 1 indicate that a given neighborhood shows a higher, a lower, or the same level of inequality as the national average, respectively.53 our measure of neighborhood inequality was computed based on thirteen household assets.54 Across the seventy-one neighbor- hoods included in the sample, the observed level of inequality ranges from about 0.60 to 1.40 points, indicating that neighborhoods with the lowest and highest levels of inequality, respectively, show about 40 per- cent lower and 40 percent higher inequality than the nation as a whole. Consequently, our study further confirms the high levels of inequality

52 McKenzie 2005. 53 As a reference value, we use the overall level of inequality in the nation, which is based on house-

hold assets data from the 2010 national AmericasBarometer survey for el salvador. latin America Public opinion Project 2015. the census questionnaire administered at the neighborhood level con- tains the same battery of questions on household assets included in the national AmericasBarometer survey, making it possible to compare neighborhood- and national-level estimates using the first prin- cipal component methodology.

54 McKenzie’s index was computed on the basis of each respondent’s report of whether or not they possess the following household assets: cellular telephone, computer, flat panel tv, hot-water shower, indoor bathroom, Internet, landline telephone, microwave oven, motorcycle, refrigerator, television, vehicle, and washing machine.

90 w o r l d p o l I t I c s

within neighborhoods in the latin American context found in previ- ous studies, with the qualification that there is significant variation in those levels.

We also include a measure of absolute living conditions in the neigh- borhood to ensure that the effect of inequality within neighborhoods on trust in local government is independent of the overall poverty level. to measure absolute neighborhood conditions, we compute an index based on the census data gathered for this study using the widely em- ployed unsatisfied basic needs (ubn) methodology as described by Flo- ribel Méndez and Juan trejos.55 the index measures the percentage of households with at least one unsatisfied basic need across four dimen- sions: consumption capacity, education, health, and housing. unlike McKenzie’s measure of inequality, the ubn index does not reflect the social distance across households in a given neighborhood, only the extent of that neighborhood’s socioeconomic disadvantage. thus, each index (the McKenzie index of asset inequality and the ubn index of poverty) taps into theoretically distinct dimensions of neighborhood development. Across the seventy-one neighborhoods included in our sample, the percentage of households with at least one unsatisfied basic need ranges from 2 percent to 100 percent.

We also control for neighborhood vulnerability to crime and vio- lence because neighborhood insecurity is likely to be related to our core neighborhood variables of inequality and poverty, as well as to trust in local government. our measure of neighborhood crime is based on responses to seven items in the opinion survey that asked respondents whether they had knowledge of the following seven acts occurring in the past twelve months in their neighborhood: damage to private property, extortion, kidnapping, murder, robbery, sale of illegal drugs, and sexual violence. the index measures the average number of crimes (from the list of seven crimes) that respondents reported in each neigh- borhood and varies between zero and seven, meaning that there are neighborhoods in which all of the respondents in that neighborhood reported having knowledge of occurrences of all seven of the possible crimes on the list during the previous twelve months and neighbor- hoods in which no respondent had knowledge of any occurrence of those crimes. our data indicate that although el salvador as a whole has one of the highest crime rates in the world,56 violence is highly concentrated in certain neighborhoods, signifying that neighborhood

55 Méndez and trejos 2004. 56 united nations office on Drugs and Crime 2013.

d e l I v e r I n g t h e g o o d s 91

crime is possibly an important competing contextual determinant of trust in government.

Independent varIables at the IndIvIdual level At the individual level, our two core independent variables are person- al socioeconomic status and perceptions of local government perfor- mance. We measure these two variables using data from the neighbor- hood public opinion survey. Items on household-asset ownership were used to assess personal socioeconomic status by arraying respondents into high, medium, and low terciles of wealth, based on the wealth index, y

i , described above, as computed using the pca method. our theoreti-

cal focus is to examine how economic inequality within neighborhoods affects levels of trust in local government across terciles of wealth, and the effect of neighborhood inequality across terciles of wealth as per- ceptions of local government performance vary. Perception of local government performance in service provision was assessed by asking respondents, “Would you say that the services the municipal government is providing to the people are very good, good, neither good nor bad, bad, or very bad.”57 the scale is recoded on a scale of 1 (very bad) to 5 (very good). We find a high degree of cor- respondence between perceptions of low service quality at the neigh- borhood level and greater physical disorder in the neighborhood as observed by our trained observers (a correlation of −0.50, p< 0.001), which indicates that residents take into account the objective quality of public services in their neighborhood when they answer this subjec- tive survey question. Moreover, the data show that equal portions—31 percent—of the poorest individuals (the first tercile of wealth) and the wealthiest individuals (the third tercile) in our sample perceive that the local government is providing very good quality services. the question then becomes, do poor individuals who think their local government is providing quality services also show high levels of trust toward the government? We hypothesize that poor individuals living in neighbor- hoods with high levels of inequality are unlikely to do so.

We also include a series of control variables at the individual level. Because an individual’s trust in local government is also likely to be based on their support for the incumbent party, partisan predilections must be included as an important control variable in the analysis be- low. Indeed, a growing body of research in political psychology shows that partisan sympathies bias citizens’ political judgments in numerous

57 In spanish, this survey item reads, “¿Diría usted que los servicios que la municipalidad está dando a la gente son: Muy buenos, Buenos, ni buenos ni malos, Malos, o Muy malos.”

92 w o r l d p o l I t I c s

ways. Partisan sympathizers are more amenable to the cues and issue frames of party elites, and partisan identities promote strong affective and cognitive biases, such as motivated reasoning, in the way partisans attend to and weigh evidence that reflects on the reputation of the po- litical party they favor.58 Partisan biases tend to be particularly impor- tant across the divide between electoral winners and losers.59 We should then expect that survey respondents’ level of trust in their local govern- ment will be a product not only of the perceived performance of the government but also a reflection of their preexisting partisan identity. In the empirical models that follow, we control for such partisan con- gruence bias using a measure of whether the respondent reports having voted for the incumbent party in the previous municipal election; this is the only variable available in the survey that taps into political party preferences at the local level.

We also include three measures of potential sources of distrust in lo- cal government in our empirical analyses: respondents’ personal experi- ences with crime committed in the neighborhood, their fear of being a victim of a crime in the neighborhood, and their personal experiences with corruption involving local government officials. Finally, we in- clude controls for various other behavioral and attitudinal variables as- sociated with political trust. Because the literature suggests that citizen participation increases confidence in government,60 we include control variables on participation in neighborhood associations and participa- tion in meetings convened by the local government, with the expecta- tion that participation in local affairs will enhance trust in local govern- ment. As noted previously, the variables on experience with corruption and participation in local government meetings are likely to capture the effect of citizens’ assessment of government procedural fairness on their level of political trust.

regarding citizen attitudes, the social capital literature suggests that individuals who are sympathetic to their fellow citizens and live in plac- es characterized by solidarity and cooperation will have a higher level of trust in government because those communities will be more participa- tory and thus promote better performing governments.61 to account for residents’ ratings of the extent of solidarity and cooperation in their neighborhood, we control for an index of social cohesion.62 the models

58 Kam 2005; slothuus and de vreese 2010; taber and lodge 2006. 59 Anderson et al. 2005. 60 Brehm and rahn 1997. 61 Putnam 1993. 62 see sampson, raudenbush, and earls 1997. the social cohesion index is measured based on

responses indicating agreement with the following statements: “When there is a problem in the neigh-

d e l I v e r I n g t h e g o o d s 93

also control for standard sociodemographic characteristics: age, gender, and the respondent’s years of schooling.

methods We rely on multilevel modeling techniques to take into account the nested structure of our data in the estimation of standard errors.63 spe- cifically, we estimate a two-level multilevel model to account for the fact that an individual i in our sample resides in a given neighbor- hood j. thus, the intercept of the model is allowed to vary randomly across neighborhoods. Moreover, our multilevel analysis controls for likely differences across municipalities by incorporating dummy vari- ables (fixed effects) for the municipalities included in our sample, thus minimizing the possibility that our results are driven by local dynamics present in a given municipality.64 given that our dependent variable is ordinal, we estimate a two-level ordered logistic multilevel model.65

Moreover, because the expectation is that the level of trust in local government among residents with the lowest socioeconomic status will depend on the extent of neighborhood inequality, in the model that tests hypothesis 1, we allow the slope of the lowest tercile of wealth to vary randomly across neighborhoods. the multilevel model that tests hypothesis 2 also allows the slope of the variable on perceptions of lo- cal government performance to vary randomly across neighborhoods given that we theorize that the effect of this variable on trust depends not only on individuals’ socioeconomic status but also on neighborhood characteristics (the level of inequality).66 Accordingly, the model speci- fication for testing hypothesis 1 is as follows:

TrustLocalGov ij = b

0 + α

1 Inequality

1j + b

1 Tercile 1

1i j + (3)

b 2 Tercile2

2ij + g

1 Inequality

1j × Tercile 1

1ij + g

2 Inequality

1j × Tercile2

2ij +

b 3 Perc.LocalGov.Performance

3ij + ...

+ α

n X

nj + ...

b

n X

nij + ErrorTerm,

where α n X

nj are neighborhood-level control variables, and b

n X

nij are

individual-level control variables.

borhood, the neighbors usually organize themselves to try to fix it,” “this is a unified neighborhood,” “People around here are willing to help their neighbors,” “People in this neighborhood generally get along with each other,” and “People in this neighborhood share the same values.” these items form a unidimensional index with a Cronbach’s alpha of 0.85.

63 gelman and hill 2007. 64 If we do not include dummy variables for each municipality in our models, our results remain

substantively unchanged. 65 the multilevel analysis presented here was carried out in stata 13.1 using the meologit command. 66 see Figure 1 for a graphical representation of this expectation.

94 w o r l d p o l I t I c s

notice that this model, associated with hypothesis 1, includes two cross-level interactions: g

1 Inequality

1j × Tercile 1

1ij and g

2 Inequality

1j ×

Tercile 2 2ij

. our expectation is to find a significant negative effect for the former (that is, g

1 < 0), indicating that higher levels of neighborhood

inequality decrease trust in the local government, particularly among individuals with the lowest socioeconomic status. We include both in- teraction terms to be able to test this hypothesis and thus examine if, as we argue, neighborhood inequality mainly affects the poorest individu- als.

the model specification for testing hypothesis 2 is as follows:

TrustLocalGov ij = b

0 + α

1 Inequality

1j + b

1 Tercile 1

1i j + (4)

b 2 Perc.LocalGov.Performance

2ij + g

1 Inequality

1j × Tercile 1

1i j + g

2 Inequality

1j ×

Perc.LocalGov.Performance 2ij

+ g 3 Tercile 1

1i j × Perc.LocalGov.Performance

2ij +

g 4 Inequality

1j × Perc.LocalGov.Performance

2ij ×Tercile 1

1i j +

... + α

n X

nj + ...

b

n X

nij + ErrorTerm.

this model, associated with hypothesis 2, includes a three-way cross-level interaction between neighborhood inequality, perceptions of local government performance, and the lowest tercile of wealth. the coefficient associated with the triple interaction is expected to be nega- tive (g

4 < 0), meaning that the interaction we identify in hypothesis

1 also varies across levels of perceived local government performance. to simplify the model slightly, because our theoretical expectation is to find that neighborhood inequality will primarily affect the poorest individuals or those in the lowest (first) tercile of wealth, the model testing hypothesis 2 compares individuals in the first tercile to those in the second and third as a single reference category. elsewhere we show that the results remain substantively unchanged if we specify our three- way interaction model to compare the first to the third tercile and the second to the third tercile, as we do in the first model.67

results to show baseline effects, we first present a model without the inclusion of our hypothesized interactions (see the baseline model [model 1] in table 1). As can be seen, inequality within neighborhoods stands out as a significant contextual predictor of trust in local government. on

67 Córdova and layton 2015. As can be observed in table A9 in the supplementary material to this article, this model specification entails the inclusion of two three-way interaction terms in the model: inequality × perceptions of local government performance × tercile 1 and inequality × perceptions of local government performance × tercile 2.

d e l I v e r I n g t h e g o o d s 95

average, high neighborhood inequality is associated with lower levels of trust in local government, yet overall poverty (absolute living conditions in the neighborhood ) does not exert a significant effect on trust. Conse- quently, all else being equal, relative, but not absolute, living conditions in the neighborhood influence citizens’ trust in local government. We also find that individuals living in neighborhoods with a high incidence of crime and violence (neighborhood crime) are significantly less likely to trust their local government. At the individual level, we find that individuals with the lowest socioeconomic status show a significantly lower level of trust in lo- cal government than individuals with the highest socioeconomic status (the coefficient associated with the first tercile of wealth is statistically significant at p < 0 .05). the coefficient comparing individuals in the second and third tercile of wealth is also negative, but only statistically significant at p < 0.10. In addition, we also find that perceptions of local government performance in service provision are strongly associated with trust in local government. In contrast, political partisanship has a weak effect. Individuals who reported voting for the winning party in the most recent mayoral election showed higher levels of trust than those voting for the losing party, but this effect faded once we included the variable perception of local government performance in service provision in the model.68 In addition, neighborhood residents who have been vic- tims of crime in their neighborhood have lower levels of trust in local government. As expected, neighborhood cohesion is positively related to trust, but participation in meetings by neighborhood improvement associations has no independent impact on trust in local government. regarding the effect of our proxy variables that are likely to capture the impact of assessments of procedural fairness (that is, experience with corruption involving a local government official and participation in meetings convened by the local government), we find that only the indicator of local government corruption has a statistically significant effect. Individuals who have been asked for a bribe show lower levels of trust in local government. Moving to the direct tests of our hypotheses, we find support for hypothesis 1 in the two-way interaction model (table 1, model 2), in which the coefficient for the cross-level effect of tercile of wealth 1 × neighborhood inequality is statistically significant and negative. In oth- er words, as expected, trust in local government is significantly lower among the lowest economic status residents who live in contexts of

68 We report this result in table A6 in Córdova and layton 2015.

table 1 determInants of trust In local government a

Two-Way Baseline Model Interaction Model (Model 1) (Model 2)

Variables Coeff. Std. Error Coeff. Std. Error

Inequality within neighborhoods −0.777* (0.368) −0.222 (0.526) tercile of wealth 1 (1); tercile 3 (0) −0.146* (0.073) 1.043+ (0.584) tercile of wealth 2 (1); tercile 3 (0) −0.130+ (0.075) 0.090 (0.532) tercile of wealth 1 × Inequality within −1.341* (0.661) neighborhoods tercile of wealth 2 × Inequality within −0.243 (0.600) neighborhoods Perception of local government performance in 0.769*** (0.041) 0.773*** (0.041) service provision

Neighborhood-Level Control Variables

Absolute living conditions in neighborhoods 0.127 (0.215) 0.078 (0.211) (overall poverty) neighborhood crime −0.066* (0.026) −0.062* (0.026)

Individual-Level Control Variables

voted for incumbent political party (1); voted 0.096 (0.082) 0.088 (0.082) for opposition (0) Did not vote (1); voted for opposition (0) −0.072 (0.079) −0.079 (0.080) Did not reveal voting behavior (1); voted for −0.151+ (0.085) −0.154+ (0.085) opposition (0) Asked for a bribe (1); no bribe (0) −0.447* (0.194) −0.438* (0.195) no contact with municipality (1); no bribe (0) 0.005 (0.084) 0.013 (0.085) Attended local government meeting (1); no (0) 0.224 (0.153) 0.234 (0.154) Attended community meetings (1); no (0) −0.054 (0.101) −0.057 (0.101) Perceived social cohesion in neighborhood 0.006*** (0.002) 0.006*** (0.002) victimized by crime in the neighborhood (1); −0.335** (0.120) −0.351** (0.120) no (0) Fear of being victimized by crime in the −0.002+ (0.001) −0.002+ (0.001) neighborhood years of schooling −0.001 (0.008) −0.002 (0.008) gender (female: 1; male: 0) 0.053 (0.058) 0.057 (0.058) Age cohort 1 (1: 18–25; 0: 46 or more) −0.346*** (0.085) −0.350*** (0.085) Age cohort 2 (1: 26–35; 0: 46 or more) −0.293*** (0.082) −0.291*** (0.082) Age cohort 3 (1: 36–45; 0: 46 or more) −0.190* (0.080) −0.198* (0.080) san Juan opico (1); Chalchuapa (0) 0.009 (0.157) 0.039 (0.155) santa Ana (1); Chalchuapa (0) −0.049 (0.132) −0.036 (0.129) Zaragoza (1); Chalchuapa (0) −0.094 (0.168) −0.074 (0.165) santa tecla (1); Chalchuapa (0) 0.110 (0.199) 0.098 (0.194) san salvador (1); Chalchuapa (0) 0.240 (0.173) 0.282 (0.172) number neighborhoods [number observations] 71 [3,955] 71 [3,955]

+p<0.1; *p<0.05; **p<0.01, ***p<0.001 a two-level ordered logistic multilevel models. the two-way interaction model allows intercepts

and slope of tercile of wealth 1 to vary randomly across neighborhoods.

d e l I v e r I n g t h e g o o d s 97

high neighborhood inequality. Indeed, when we carried out further tests and kept the second and third terciles of wealth as the reference category (instead of only the third tercile as in table 1, model 2), the coefficient for the interaction term between the first tercile and neigh- borhood inequality is still statistically significant, indicating that at high levels of inequality, the largest difference in levels of trust in local government surfaces between neighborhood residents with the lowest level of wealth and the rest of the population.69

Figure 2 illustrates the results of the significant two-way interaction shown in model 2. At the highest level of neighborhood inequality, the predicted probability of trusting the local government “a lot” (a score of 7 on our dependent variable) is 5.3 percentage points among individu- als in the first tercile of wealth, while at the lowest level of neighbor- hood inequality it is 15.1 percentage points.70 In this first analysis, we find that the differences in the mean predicted probability of trusting the local government “a lot” between the poorest and richest residents are statistically significant (p< 0.05) when neighborhood inequality is equal to or greater than the national average (the equivalent of a score of 1.00 on the McKenzie index).71 Furthermore, trust in local government among higher socioeconomic status individuals is statisti- cally unaffected by neighborhood inequality. this result is confirmed when we split the sample and run separate models for individuals in the first, second, and third terciles of wealth. our split-sample analysis shows that within-neighborhood inequality has a negative effect on trust in local government only among individuals in the first tercile of wealth.72 this tendency is consistent with our expectations illustrated in Figure 1.

taking into account these first findings, we now proceed to explore the conditional effect of perceptions of local government performance across economic groups while still accounting for neighborhood in- equality. table 2, model 3, presents the results of the model when we include a three-way interaction term between our variables for neigh- borhood inequality, individual perceptions of local government per-

69 see table A7 in Córdova and layton 2015. 70 Predicted probabilities are estimated using the margins command in stata 13.1 based on the

results of the two-way interaction model presented in table 1. our conclusions remain substantively unchanged regardless of the cut-off point on the dependent variable used to estimate predicted prob- abilities. results remain similar if we estimate the predicted probability of providing an answer equal to 5, 6, or 7, on the 1–7 scale of the trust in local government survey question, or estimate the predicted probability of providing an answer equal to 6 or 7 on this survey item. these additional analyses are presented graphically in Figures A3–A6 in Córdova and layton 2015.

71 Figure A1 in Córdova and layton 2015 shows this result graphically. the statistical significance of differences in mean predicted probabilities was determined based on the Delta method.

72 see table A8 in Córdova and layton 2015.

98 w o r l d p o l I t I c s

formance, and the first tercile of wealth. As noted above, the refer- ence category in this analysis corresponds to individuals in the second and third terciles of wealth.73 As observed in table 2, the coefficient of the triple interaction term is negative and statistically significant (p< 0.01).

Figure 3 illustrates the results from model 3 more clearly. the pre- dicted values we observe are consistent with hypothesis 2; the predicted probability of trusting the local government “a lot” increases among individuals in the first tercile of wealth as perceptions of government performance improve when neighborhood inequality is relatively low, see Figure 3(a). In contrast, in the most unequal neighborhoods, fa- vorable perceptions of government performance do not increase trust among the poor. Figure 3(b), shows a very different pattern. For resi- dents at medium and high terciles of wealth, the predicted probability of trusting the local government “a lot” increases as perceptions of lo- cal government performance improve across all levels of neighborhood inequality. In other words, among the relatively affluent, trust in local government is much less dependent on neighborhood inequality and much more a function of the perception of local government perfor-

73 notably, as mentioned above, our conclusions remain unchanged when we test hypothesis 2 and include tercile 1 and 2 as distinct categories and compare each of them with tercile 3. only the three- way interaction associated with tercile of wealth 1 is statistically significant and negative, as expected. see table A9 in Córdova and layton 2015.

.60 .80 1.00 1.20 1.40 Within neighborhood Inequality (McKenzie Index)

M ea

n P

re di

ct ed

P ro

ba bi

lit y

t

ru st

in g

th e

l oc

al g

ov er

nm en

t “a

l ot

” tercile of wealth 3

tercile of wealth 2

tercile of wealth 1

.16

.14

.12

.1

.08

.06

.04

fIgure 2 predIcted probabIlItIes of trustIng local government “a lot” by

tercIle of wealth a

a Based on results from model 2.

table 2 determInants of trust In local government a

Three-Way Interaction Model (Model 3)

Variables Coeff. Std. Error

Inequality within neighborhoods 0.603 (1.611) tercile of wealth 1 (1); tercile of wealth 2 and 3 (0) −3.724* (1.848) tercile of wealth 1 × Inequality within neighborhoods 4.702* (2.155) Perception of local government performance in service provision 1.067* (0.428) tercile of wealth 1 × Perception of local government performance 1.467** (0.550) Perception of government performance × neighborhood inequality −0.307 (0.491) tercile of wealth 1 × Perception of local government −1.833** (0.640) performance × Inequality within neighborhoods

Neighborhood-Level Control Variables

Absolute living conditions in neighborhoods (overall poverty) 0.029 (0.220) neighborhood crime −0.053* (0.026)

Individual-Level Control Variables

voted for incumbent political party (1); voted for opposition (0) 0.084 (0.083) Did not vote (1); voted for opposition (0) −0.087 (0.080) Did not reveal voting behavior (1); voted for opposition (0) −0.162+ (0.085) Asked for a bribe (1); no bribe (0) −0.400* (0.197) no contact with municipality (1); no bribe (0) 0.022 (0.085) Attended local government meeting (1); no (0) 0.187 (0.156) Attended community meetings (1); no (0) −0.046 (0.102) social cohesion in neighborhood 0.006*** (0.002) victimized by crime in the neighborhood (1); no (0) −0.330** (0.121) Fear of being victimized by crime in the neighborhood −0.002+ (0.001) years of schooling −0.001 (0.008) gender (female: 1; male: 0) 0.054 (0.058) Age cohort 1 (1: 18–25; 0: 46 or more) −0.351*** (0.086) Age cohort 2 (1: 26–35; 0: 46 or more) −0.287*** (0.083) Age cohort 3 (1: 36–45; 0: 46 or more) −0.200* (0.081) san Juan opico (1); Chalchuapa (0) 0.045 (0.156) santa Ana (1); Chalchuapa (0) −0.021 (0.130) Zaragoza (1); Chalchuapa (0) −0.049 (0.167) santa tecla (1); Chalchuapa (0) 0.135 (0.199) san salvador (1); Chalchuapa (0) 0.307+ (0.173) number of neighborhoods [number of observations] 71 [3,955]

+p<0.1; *p<0.05; **p<0.01, ***p<0.001 a two-level ordered logistic multilevel model with random intercepts at the neighborhood level.

the slopes associated with the variables on tercile of wealth 1 and perceptions of government perfor- mance in the provision of services are allowed to vary randomly across neighborhoods.

100 w o r l d p o l I t I c s

mance in service provision. these patterns are also consistent with the expectations illustrated in Figure 1.

the statistical significance of this contrast between residents at dif- ferent levels of personal wealth is more clearly illustrated in Figure 4, where the analysis is restricted to individuals who perceive that the local government is providing very good services. As shown in Fig- ure 4(a), the contrast in the reactions of poor (first tercile) and afflu- ent (second and third terciles) individuals who rate local government

fIgure 3 predIcted probabIlItIes of trustIng local government “a lot” by

tercIle of wealth and perceptIon of local government performance In servIce provIsIon a

a based on results from model 3.

Individuals in the First tercile of Wealth (a)

.60 .80 1.00 1.20 1.40 Within neighborhood Inequality (McKenzie Index)

.8

.7

.6

.5

.4

.3

.2

.1

0

M ea

n P

re di

ct ed

P ro

ba bi

lit y

t ru

st in

g th

e l

oc al

g ov

er nm

en t

“a l

ot ”

very bad Bad neither good nor bad

good very good

Would you say that the services the municipal government is providing to the people are:

very bad Bad neither good nor bad

good very good

Would you say that the services the municipal government is providing to the people are:

Individuals in the second and third terciles of Wealth (b)

.60 .80 1.00 1.20 1.40 Within neighborhood Inequality (McKenzie Index)

.8

.7

.6

.5

.4

.3

.2

.1

0

M ea

n P

re di

ct ed

P ro

ba bi

lit y

t ru

st in

g th

e l

oc al

g ov

er nm

en t

“a l

ot ”

d e l I v e r I n g t h e g o o d s 101

performance as very good is dramatic as neighborhood inequality in- creases. As expected, the average level of trust in local government is the highest when citizens perceive good performance and live in neigh- borhoods with the lowest level of neighborhood inequality. Among the poor, those high rates of trust quickly evaporate as inequality increases. At the highest level of neighborhood inequality, the difference in the predicted probability of trusting the local government “a lot” between residents in the lowest and highest terciles of wealth is about 20 per- centage points, see Figure 4(b). Indeed, we find that the difference in the mean predicted probability of trust in local government between the two groups is statistically significant when the Mckenzie index is equal to or greater than one (the number representing the national av- erage).

Interestingly, as we look at the differences between the poor and the affluent in neighborhoods with low inequality in Figure 4, we also find that poor respondents who perceive that government performance is very good are significantly more likely to trust the local government “a lot” than affluent respondents. We find that this is the case in the five neighborhoods with the lowest level of inequality in our sample.74 taken together, these results reinforce our arguments that the poor are particularly attentive to neighborhood inequality and that the cost of inequality is substantial for the local government in terms of eroding citizen trust among the poor.

the weak association between perceptions of performance and trust in local government among the poor at high levels of neighborhood inequality demonstrates that these variables tap distinct concepts and thus are not tautological. this result also suggests that when percep- tions of performance and trust in government are correlated (that is, at low levels of inequality), the causal arrow runs from perceptions of performance to trust. the fact that at high levels of neighborhood in- equality trust in local government remains low among the poor despite perceptions of high performance suggests that inequality depresses the causal effect of perceptions of high performance on trust and not the other way around. Indeed, the opposite causal story makes no sense in this case: high inequality cannot possibly attenuate the impact of low trust on perceptions of high performance. taken together, these results give us confidence that our findings are not an artifact of an endogene-

74 this result is therefore not driven by the neighborhood with the lowest level of inequality in our sample; there are five neighborhoods with low inequality that show a similar trend (see Figure A2 in Córdova and layton 2015).

fIgure 4 predIcted probabIlItIes of trustIng local government “a lot” by tercIle of wealth for IndIvIduals perceIvIng “very good” local

government servIce provIsIona

a based on results from model 3.

Individuals Perceiving a “very good” service Provision (a)

Individuals Perceiving a “very good” service Provision (Difference in mean predicted probability between tercile groups with

95% confidence interval) (b)

.60 .80 1.00 1.20 1.40 Within neighborhood Inequality (McKenzie Index)

.60 .80 1.00 1.20 1.40 Within neighborhood Inequality (McKenzie Index)

.8

.7

.6

.5

.4

.3

.2

.1

0

.8

.7

.6

.5

.4

.3

.2

.1

0

D if

fe re

nc e

in M

ea n

P re

di ct

ed P

ro ba

bi lit

y t

ru st

in g

th e

l oc

al g

ov er

nm en

t a

l ot

M ea

n P

re di

ct ed

P ro

ba bi

lit y

t ru

st in

g th

e l

oc al

g ov

er nm

en t

a l

ot

tercile of wealth 2 and 3

tercile of wealth 1

d e l I v e r I n g t h e g o o d s 103

ity problem between perceptions of government performance and trust in local government.

notably, the results we report in tables 1 and 2 are robust to the in- clusion of a half-dozen additional neighborhood-level characteristics: observed physical disorder, neighborhood size, participation rates in neighborhood associations, participation rates in meetings convened by the local government, the extent of social cohesiveness in the neighbor- hood, and the percentage of neighborhood residents who voted for the incumbent political party in the municipal election carried out prior to the study.75 the consistency of our results after the inclusion of rigorous controls gives us confidence that our findings are reliable.

dIscussIon and conclusIon

the results of our empirical analysis have important theoretical and policy implications. First, our research demonstrates that contextual factors associated with living conditions in a neighborhood substan- tively shape trust in local government. In particular, economic inequal- ity within neighborhoods erodes trust in local government among the poor. this finding is consistent with the fht prediction that individuals draw on information about distributive fairness within their social con- text as a heuristic to determine how much they can trust an authority. While indicators of procedural fairness, such as experience with cor- ruption, reduce trust in local government, perceptions of local govern- ment’s distributive fairness derived from neighborhood inequality exert a particularly strong negative effect on poor respondents’ level of trust in local government. second, our research circumscribes one of the most consistent findings of individual-level studies on the origins of political trust—that people are more likely to trust their government if they perceive it is performing well and thus “delivering the goods.” We find that for the poor who live in a context of high neighborhood in- equality, subjective evaluations of good local-government performance are insufficient to increase political trust, which leads us to conclude that the poor value both performance and equity from their local gov- ernment. taken together, our findings indicate that high economic inequality in the developing world poses a challenge for local governance when it

75 to avoid multicollinearity, we evaluated the effect of neighborhood inequality vis-à-vis each of these six variables in separate models while simultaneously controlling for absolute living standards and crime in the neighborhood. We report the results of this analysis in tables A10–A21 in Córdova and layton 2015. Because el salvador is a highly ethnically homogenous country, ethnic heterogene- ity is also held constant in our analysis.

104 w o r l d p o l I t I c s

is reflected at the neighborhood level. While decentralization reforms have increased the importance of the neighborhood as a social space for “bringing the state closer to the people” by encouraging citizen par- ticipation in neighborhood organizations,76 this form of direct democ- racy will not automatically lead to greater political trust if economic inequality in the region remains significantly high. given the theorized importance of political trust for citizens’ respect for the rule of law, widespread compliance with local governments’ laws and regulations will be difficult to achieve given the high levels of economic inequality in latin American countries.

the drastic shift from a statist to a neoliberal economic model through “shock therapy” (the rapid adoption of free-market policies) resulted in an increase in economic inequality toward the end of the 1980s.77 More recently, a decline in economic inequality has been reg- istered in many countries in the region, which is largely attributed to the more progressive nature of policies implemented by leftist gov- ernments.78 the current economic trends in the region, however, cast doubt on the sustainability of these redistributive policies, and even with the registered decline in economic inequality during the last few years, the region as a whole still holds the title as the most unequal in the world.79 our estimates of neighborhood inequality indicate that the high overall economic inequality reported at the national level are largely reflected by deep inequalities that surface at the submunicipal level within neighborhoods.

We find that neighborhood inequality undermines local governance in the current era of decentralization. the confluent trend of provid- ing opportunities for citizen participation through neighborhood as- sociations and promising equity in the distribution of local government resources have arguably made the poor particularly attentive to local government’s distributive fairness in their assessments of local govern- ment’s trustworthiness. our evidence from el salvador is consistent with the claim that neighborhood inequality produces a reservoir of distrust in the local government among the poor. the policy implica- tion of our findings is clear: fulfilling decentralization’s dual goal of improving equity and performance will result not only in better devel- opment outcomes for all, but also in greater trust toward local govern- ments.

76 selee 2004. 77 lustig 1995. 78 huber and stephens 2012. 79 Birdsall, lustig, and Mcleod 2011; lópez-Calva and lustig 2010.

d e l I v e r I n g t h e g o o d s 105

the available evidence, however, shows that decentralization out- comes related to performance and equity have been mixed in latin America. For example, qualitative evidence for Chile suggests that after that country’s transition to democracy, “efficiency and equity did not emerge in large scale” in the areas of primary health care and primary and secondary education, and even where citizen participation actu- ally had the potential to bring about improved services and equity, lo- cal governments often lacked the resources to respond to citizens’ de- mands.80 evidence for Mexico leads to a similar conclusion; the impact of decentralization on the provision of water did not lead to uniform effects on performance and equity.81 More recently, based on evidence from Brazil and its efforts to decentralize primary education, Fabiana Machado concludes that in municipalities with high economic inequal- ity (and presumably in the most unequal neighborhoods in those mu- nicipalities), the promises of decentralization in terms of performance and equity were more difficult to fulfill due to political capture by local elites.82

this evidence supports the initial fear held by critics of the decen- tralization movement who cautioned against overly idyllic visions of decentralized power: that certain societal contexts would constrain the realization of these expectations. the foremost concern was that eco- nomic inequality has the potential to lead to starker political inequi- ties because, in the context of economic inequality, “it may be easy for the local overlords to capture the local community institutions and the poor may be left grievously exposed to their mercies and their mal- feasance.”83 For these critics, rather than ameliorating inequities and concerns with service quality, decentralization had the potential to fur- ther entrench already powerful factions or traditional elites and conse- quently reinforce historical patterns of exclusion.

the question is how to enact policies that allow local governments to fulfill their promise of better and more equitable policy outcomes. Part of the solution may emerge from the experiences that some municipal governments in latin American countries have recently had with pro- gressive policies, such as conditional cash transfer (cct) programs.84 In general, these programs offer cash benefits to poor households with the objective of incentivizing household investment in nutrition and human-capital development; by design, the programs aim to provide

80 Kubal 2006. 81 Wilder and romero 2006. 82 Machado 2013. 83 Bardhan 1996, 1355. 84 Adato and hoddinott 2010.

106 w o r l d p o l I t I c s

both short- and long-term solutions to poverty and inequality. nev- ertheless, to foster perceptions of government legitimacy, our results suggest that cct programs will need to be complemented by parallel efforts to close the gap in standards of living not only across munici- palities, but also within submunicipal units, including neighborhoods. Conditional cash transfer programs have focused their household tar- geting criteria mainly on proxy-means tests that account for standard monetary measures of poverty, such as household wealth or per capita income, and thus their procedures for the identification of beneficiaries are not based on a “multidimensional targeting” strategy that recogniz- es the complex determinants of poverty and inequality.85 Conditional cash transfer programs alone will not maximize the relative well-being of the poor, and subsequently, their trust in local government, without complementary public investments that more directly address struc- tural factors that perpetuate economic inequality.

Without the political will to value and address structural aspects of deep socioeconomic disparities within municipalities, inequality will continue to plague latin American societies and deter political trust. Although it might be possible for local political elites to repeatedly achieve reelection without addressing questions of inequality, economic disparities will weigh on and hold back the developmental prospects for the entire society, in part because long-term political stability will always be in question as long as a significant group of disenchanted and economically marginalized citizens remains. In the end we can only re- iterate the refrain of our argument: it is not good enough to just deliver the goods. Without achieving greater equity through government poli- cy efforts, many citizens will be unlikely to gain sufficient confidence in their governing institutions to ensure democratic stability.

supplementary materIal

supplementary material for this article can be found at http://dx.doi.org.10.1017 /s0043887115000441.

references

Adato, Michelle, and John hoddinott. 2010. “Conditional Cash transfer Pro- grams: A ‘Magic Bullet’?” In Michelle Adato and John hoddinott, eds., Conditional Cash Transfers in Latin America. Baltimore, Md.: Johns hopkins university Press.

85 Azevedo and robles 2013.

d e l I v e r I n g t h e g o o d s 107

Anderson, Christopher J., André Blais, shaun Bowler, todd Donovan, and ola lishaug. 2005. Losers’ Consent: Elections and Democratic Legitimacy. new york, n.y.: oxford university Press.

Azevedo, viviane, and Marcos robles. 2013. “Multidimensional targeting: Iden- tifying Beneficiaries of Conditional Cash transfer Programs.” Social Indicators Research 112, no. 2: 447–75.

Baker, Andy, Barry Ames, and lucio r. renno. 2006. “social Context and Cam- paign volatility in new Democracies: networks and neighborhoods in Brazil’s 2002 elections.” American Journal of Political Science 50, no. 2: 382–99.

Bardhan, Pranab. 1996. “efficiency, equity, and Poverty Alleviation in less De- veloped Countries.” Economic Journal 106, no. 438: 1344–56.

Bardhan, Pranab K., and Dilip Mookherjee. 2006. Decentralization and Local Governance in Developing Countries: A Comparative Perspective. Cambridge, Mass.: MIt Press.

Birdsall, nancy, nora lustig, and Darryl Mcleod. 2011. “Declining Inequality in latin America: some economics, some Politics.” Working Paper no. 251. Washington, D.C.: Center for global Development.

Booth, John A., and Mitchell A. seligson. 2009. The Legitimacy Puzzle in Latin America: Democracy and Political Support in Eight Nations. new york, n.y.: Cambridge university Press.

Bratton, Michael. 2012. “Citizen Perceptions of local government responsive- ness in sub-saharan Africa.” World Development 40, no. 3: 516–27.

Brehm, John, and Wendy rahn. 1997. “Individual-level evidence for the Causes and Consequences of social Capital.” American Journal of Political Science 41, no. 3: 999–1023.

Campbell, tim. 2003. The Quiet Revolution: Decentralization and the Rise of Po- litical Participation in Latin American Cities. Pitt latin American series. Pitts- burgh, Pa.: university of Pittsburgh Press.

Canache, Damarys. 1996. “looking out My Back Door: the neighborhood Context and Perceptions of relative Deprivation.” Political Research Quarterly 49, no. 3: 547–71.

Canel, eduardo. 2010. Barrio Democracy in Latin America: Participatory Decentral- ization and Community Activism in Montevideo. university Park, Pa.: Pennsyl- vania state university Press.

Cleary, Matthew r., and susan Carol stokes. 2006. Democracy and the Culture of Skepticism: Political Trust in Argentina and Mexico. new york, n.y.: russell sage Foundation.

———. 2009. “trust and Democracy in Comparative Perspective.” In Karen s. Cook, Margaret levi, and russell hardin, eds., Whom Can We Trust? How Groups, Networks, and Institutions Make Trust Possible. new york, n.y.: russell sage Foundation.

Córdova, Abby, and Matthew l. layton. 2015. supplementary material. At http://dx.doi.org.10.1017/s0043887115000441.

eaton, Kent. 2012. “Decentralization and Federalism.” In Peter Kingstone and Deborah J. yashar, eds., Routledge Handbook of Latin American Politics. new york, n.y.: routledge.

elbers, Chris, Peter lanjouw, Johan Mistiaen, Berk Özler, and Kenneth r. sim- ler. 2004. “on the unequal Inequality of Poor Communities.” World Bank Eco- nomic Review 18, no. 3: 401–21.

108 w o r l d p o l I t I c s

———. 2005. “Are neighbours equal? estimating local Inequality in three Developing Countries.” In s. M. ravi Kanbur and Anthony venables, eds., Spatial Inequality and Development. oxford, uK: oxford university Press.

espinal, rosario, Jonathan hartlyn, and Jana Morgan Kelly. 2006. “Performance still Matters. explaining trust in government in the Dominican republic.” Comparative Political Studies 39, no. 2: 200–23.

Faguet, Jean-Paul. 2012. Decentralization and Popular Democracy: Governance from Below in Bolivia. Ann Arbor, Mich.: university of Michigan Press.

gelman, Andrew, and Jennifer hill. 2007. Data Analysis Using Regression and Multilevel/Hierarchical Models. Analytical Methods for social research. new york, n.y.: Cambridge university Press.

graham, Carol, and Andrew Felton. 2005. “Inequality and happiness: Insights from latin America.” Journal of Economic Inequality 4, no. 1: 107–22.

gurr, ted robert. 1970. Why Men Rebel. Princeton, n.J.: Princeton university Press. hetherington, Marc J. 1998. “the Political relevance of Political trust.” Ameri-

can Political Science Review 92, no. 4: 791–808. ———. 2005. Why Trust Matters: Declining Political Trust and the Demise of Amer-

ican Liberalism. Princeton, n.J.: Princeton university Press. hibbing, John r., and elizabeth theiss-Morse. 1995. Congress as Public Enemy:

Public Attitudes toward American Political Institutions. Cambridge studies in Political Psychology and Public opinion. new york, n.y.: Cambridge uni- versity Press.

———. 2002. Stealth Democracy: Americans’ Beliefs about How Government Should Work. new york, n.y.: Cambridge university Press.

hiskey, Jonathan t., and shaun Bowler. 2005. “local Context and Democratiza- tion in Mexico.” American Journal of Political Science 49, no. 1: 57–71.

huber, evelyne, and John D. stephens. 2012. Democracy and the Left: Social Policy and Inequality in Latin America. Chicago, Ill.: university of Chicago Press.

huckfeldt, r. robert. 1986. Politics in Context: Assimilation and Conflict in Urban Neighborhoods. new york, n.y.: Agathon Press.

huckfeldt, r. robert, and John D. sprague. 1995. Citizens, Politics, and Social Communication: Information and Influence in an Election Campaign. Cambridge studies in Political Psychology and Public opinion. new york, n.y.: Cam- bridge university Press.

Jamal, Amaney, and Irfan nooruddin. 2010. “the Democratic utility of trust: A Cross-national Analysis.” Journal of Politics 72, no. 1: 45–59.

Kam, Cindy. 2005. “Who toes the Party line? Cues, values, and Individual Dif- ferences.” Political Behavior 27, no. 2: 163–82.

Kubal, Mary rose. 2006. “Contradictions and Constraints in Chile’s health Care and education Decentralization.” Latin American Politics and Society 48, no. 4: 105–35.

labonne, Julien, and robert s. Chase. 2009. “Who Is at the Wheel When Com- munities Drive Development? evidence from the Philippines.” World Develop- ment 37, no. 1: 219–31.

latin America Public opinion Project. 2015. AmericasBarometer survey 2010. nashville, tenn.: vanderbilt university. At http://www.vanderbilt.edu/lapop/, accessed June 24, 2015.

levi, Margaret. 1988. Of Rule and Revenue. Berkeley, Calif.: university of Cali- fornia Press.

d e l I v e r I n g t h e g o o d s 109

———. 1998. “A state of trust.” In v. A. Braithwaite and Margaret levi, eds., Trust and Governance. new york, n.y.: russell sage Foundation.

levi, Margaret, and laura stoker. 2000. “Political trust and trustworthiness.” Annual Review of Political Science 3: 475–507.

lópez-Calva, luis Felipe, and nora lustig. 2010. Declining Inequality in Latin America: A Decade of Progress? Washington, D.C.: Brookings Institution Press.

lustig, nora. 1995. Coping with Austerity: Poverty and Inequality in Latin America. Washington, D.C.: Brookings Institution Press.

luttmer, erzo F. P. 2005. “neighbors as negatives: relative earnings and Well- Being.” Quarterly Journal of Economics 120, no. 3: 963–1002.

Machado, Fabiana. 2013. “Decentralization and Accountability: the Curse of local underdevelopment.” Working Paper no. 397. Washington, D.C.: Inter- American Development Bank.

McKenzie, David J. 2005. “Measuring Inequality with Asset Indicators.” Journal of Population Economics 18, no. 2: 229–60.

Méndez, Floribel, and Juan Diego trejos. 2004. “Costa rica: un Mapa De Carencias Críticas Para el Año 2000” [Costa rica: A map of critical needs for the year 2000]. In luis rosero Bixby, ed., Costa Rica a La Luz Del Censo Del 2000 [Costa rica in light of the 2000 census]. san José, Costa rica: Centro Centroaméricano de Población de la universidad de Costa rica.

Mishler, William, and richard rose. 2001. “What Are the origins of Political trust?” Comparative Political Studies 34, no. 1: 30–62.

Montalvo, Daniel. 2010. “understanding trust in Municipal governments.” AmericasBarometer Insights series, no. 35. nashville, tenn.: latin American Public opinion Project, vanderbilt university. At http://www.vanderbilt.edu /lapop/insights/I0835en.pdf, accessed March 31, 2015.

Putnam, robert D. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, n.J.: Princeton university Press.

rahn, Wendy M., and thomas J. rudolph. 2002. “trust in local governments.” In Barbara norrander and Clyde Wilcox, eds., Understanding Public Opinion. Washington, D.C.: CQ Press.

rahn, Wendy M., and thomas J. rudolph. 2005. “A tale of Political trust in American Cities.” Public Opinion Quarterly 69, no. 4: 530–60.

rothstein, Bo, and eric M. uslaner. 2005. “All for All: equality, Corruption, and social trust.” World Politics 58, no. 1: 41–72.

sampson, robert J., and stephen W. raudenbush. 1999. “systematic social observation of Public spaces: A new look at Disorder in urban neighbor- hoods.” American Journal of Sociology 105, no. 3: 603–51.

sampson, robert J., stephen W. raudenbush, and Felton earls. 1997. “neigh- borhoods and violent Crime: A Multilevel study of Collective efficacy.” Sci- ence 277, no. 5328: 918–24.

scholz, John t., and Mark lubell. 1998. “trust and taxpaying: testing the heu- ristic Approach to Collective Action.” American Journal of Political Science 42, no. 2: 398–417.

selee, Andrew. 2004. “exploring the link between Decentralization and Demo- cratic governance.” In Joseph s. tulchin and Andrew selee, eds., Decentraliza- tion and Democratic Governance in Latin America. Washington, D.C.: Woodrow Wilson Center.

110 w o r l d p o l I t I c s

slothuus, rune, and Claes h. de vreese. 2010. “Political Parties, Motivated rea- soning, and Issue Framing effects.” Journal of Politics 72, no. 3: 630–45.

taber, Charles s., and Milton lodge. 2006. “Motivated skepticism in the evalu- ation of Political Beliefs.” American Journal of Political Science 50, no. 3: 755–69.

torres, raúl, and oscar h. lópez. 2008. Mecanismos de participación ciudadana y transparencia en la gestión municipal [Mechanisms for citizen participation and transparency in municipal administration]. san salvador, el salvador: Fun- dación nacional para el Desarrollo.

tost, leigh Plunkett, and Allan lind. 2010. “sounding the Alarm: Moving from system Justification to system Condemnation in the Justice Judgment Pro- cess.” In elizabeth Mannix, Margaret neale, and elizabeth Mullen, eds., Fair- ness and Groups. vol. 13, Research on Managing Groups and Teams. Bingley, uK: emerald group Publishing.

tyler, tom. 2001. “the Psychology of Public Dissatisfaction with government.” In John r. hibbing and elizabeth theiss-Morse, eds., What Is It about Gov- ernment that Americans Dislike? Cambridge, uK: Cambridge university Press.

tyler, tom, and Jojanneke van der toorn. 2013. “social Justice.” In leonie huddy, David o. sears, and Jack s. levy, eds., The Oxford Handbook of Political Psychol- ogy, 2nd ed. oxford, uK: oxford university Press.

united nations office on Drugs and Crime. 2013. Global Study on Homicide. vi- enna, Austria: unoDC.

uslaner, eric. 2002. The Moral Foundations of Trust. Cambridge, uK: Cambridge university Press.

van ryzin, gregg. 2011. outcomes, Process, and trust in Civil servants. Journal of Public Administration Research and Theory 21, no. 4: 745–60.

van den Bos, Kees, e. Allan lind, henk A. M. Wilke. 1998. “When Do We need Procedural Fairness? the role of trust in Authority.” Journal of Personal- ity and Social Psychology 75, no. 6: 1449–58.

———. 2001. “the Psychology of Procedural and Distributive Justice viewed from the Perspective of Fairness heuristic theory.” In russell Cropanzano, ed., Justice in the Workplace: From Theory to Practice. Mahwah, n.J.: lawrence erlbaum.

van de Walle, steven, and geert Bouckaert. 2003. “Public service Performance and trust in government: the Problem of Causality.” International Journal of Public Administration 29, 8-9: 891–913.

Wampler, Brian. 2007. Participatory Budgeting in Brazil: Contestation, Cooperation, and Accountability. university Park, Pa.: Pennsylvania state university Press.

Wilder, Margaret, and Patricia romero. 2006. “Paradoxes of Decentralization: Water reform and social Implications in Mexico.” World Development 34, no. 11: 1977–95.

Wolak, Jennifer, and Christine Kelleher Palus. 2010. “the Dynamics of Public Confidence in u.s. state and local government.” State Politics & Policy Quar- terly 10, no. 4: 421–45.

Wong, Cara, Jake Bowers, tarah Williams, and Katherine Drake simmons. 2012. “Bringing the Person Back In: Boundaries, Perceptions, and the Measurement of racial Context.” Journal of Politics 74, no. 4: 1153–70.

yang, Kaifeng, and Marc holzer. 2006. “the Performance-trust link: Implica- tions for Performance Measurement.” Public Administration Review 66, no. 1: 114–26.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.