Public Trust in Latin America's Courts: Do Institutions Matter?

Abstract Despite the integral role of the judiciary to democracy, and the importance of judicial trust for judicial system performance, we know relatively little about the bases of public trust in this institution. How does institutional quality affect judicial trust? We explore this question in the context of Latin America, using a multilevel data set comprising survey data spanning 2001 to 2016 and country-level institutional and economic factors. We find that the effects of institutional quality on judicial trust are highly circumscribed. Factors like rule of law and corruption impact the judicial trust of only the best-educated survey respondents. Among the broader public, however, judicial trust is shaped more strongly by individuals' subjective economic and regime evaluations, as well as one's personal experiences with the judiciary.

[U]nlike other arms of State, courts rely solely on the trust and confidence of the people to carry out their constitutionally-mandated function.
-Acting Chief Justice Sisi Khampepe, in Constitutional Court of South Africa (2021) The judiciary is a fundamental institution of democracy.From the perspectives of citizens, courts are intended as sites for peaceful dispute resolution, the pursuit of their interests and the defence of constitutionally guaranteed rights.Everyday citizens and political elites rely on a relatively powerful judiciary for the enforcement of horizontal accountability and the separation of powers.Further, legislators and policymakers, when challenged, depend on courts to lend legitimacy to laws and policies by decreeing them legal and just (Easton 1965;Gibson 1989;Tyler 1990).From the enforcement of accountability to the guarantee of citizenship, then, democratic governance itself relies in one form or another upon a functional and legitimate judiciary.
Public confidence is a linchpin in the judiciary's legitimacy, which derives not only from the authority it is delegated by the state or constitution, but also from the citizens over whom it presides.Indeed, popular support has been deemed critical to courts' ability to serve as checks on the executive and legislative branches (e.g.Ginsburg 2003;Prillaman 2000).In contexts where courts suffer from a lack of legitimacy, compliance with policy directives is also expected to suffer (Tyler 1990).And courts themselves must at times rely on their legitimacy to enforce their decisions, as they cannot induce compliance through either 'the sword' or 'the purse' (Caldeira 1986).Courts that enjoy public support are better able to withstand or counteract political attacks (Epstein et al. 2001;Helmke and Ríos-Figueroa 2011) and to take stronger stands against defiant political actors (Caldeira and Gibson 1992;Helmke and Staton 2011;Staton 2004).In short, the ability of the court to serve its vital democratic functions is unthinkable without the confidence citizens confer upon it.
Despite the integral role of the judiciary to democracy, the rule of law and the provision of justice, few studies focus attention on the determinants of public trust in the courts.Even fewer analyse contexts beyond advanced democracies in the United States and Europe (Bühlmann and Kunz 2011;Çakır and Şekercioğlu 2016;Gibson et al. 1998), and studies that do focus on the developing world often rely on single-survey rounds that offer only a snapshot of the temporal and cross-sectional dynamics of judicial trust (Boateng and Adjorlolo 2019;Driscoll and Nelson 2018;Salzman and Ramsey 2013).These theoretical and empirical gaps are particularly striking as it pertains to Latin America, where courts and other judicial actors have been deemed central to building the rule of law in the region's third-wave democracies (Carothers 2006;Méndez et al. 1999;Prillaman 2000;World Bank 2012) and where they are playing increasingly active roles in contemporary democratic politics (Helmke and Ríos-Figueroa 2011;Kapiszewski and Taylor 2008).This expanded role for the judiciary in the region makes it all the more important to understand when and why courts are deemed trustworthy in the eyes of citizens, and are thus able to serve as bulwarks of democracy and justice.
This article aims to identify the determinants of public trust in Latin America's courts.Specifically, we test whether institutional factors impact the public's willingness to trust the court, above and beyond subjective micro-level factors.Drawing on the institutional trust and judicial politics literature, we test the hypotheses that citizens ought to be more trusting of courts where rule of law is stronger and where corruption is lower.Drawing on the corruption literature, we also test the hypothesis that the most politically sophisticated citizens (i.e. the best educated) are most responsive to institutional quality when evaluating the trustworthiness of the court.To test these hypotheses, we assemble a multilevel data set that combines 13 waves of Latinobarometer survey data from 18 countries with macro-level economic and institutional variables from multiple sources.We find that the effects of institutional quality on judicial trust are highly circumscribed, as the corruption literature predicts.Factors like rule of law and corruption impact the judicial trust of only the best-educated survey respondents.Among the broader public, however, judicial trust is shaped more strongly by individuals' subjective economic and regime evaluations, as well as one's personal experiences with the judiciary.In sum, we find that personal experiences shape the trustworthiness of the court and that education conditions the impact of institutional quality on public confidence in Latin America's courts.

What explains judicial trust?
Scholars interested in understanding confidence in political institutions traditionally turn to David Easton's (1965) analysis of political systems, namely the concepts of diffuse and specific support.While diffuse support is 'a reservoir of favorable attitudes or good will that helps members to accept or tolerate outputs to which they are opposed or the effects of which they see as damaging to their wants' (Easton 1965: 273), specific support is more akin to a positive evaluation of a particular decision.Eugene Miller (1971: 201) further clarifies that diffuse support 'is support for a political object for its own sake rather than for what the individual expects to derive from it'.Expected personal gains fall within the domain of specific support.While some debate the need for such conceptual distinctions as well as their meanings (see Citrin 1974;Mishler and Rose 1997), we subscribe to Pippa Norris's (1999) view that diffuse and specific support are two ends of a continuum of trust.Institutional trust falls closer to specific support, as the line between the office and the officeholders is often unclear from the perspective of survey respondents (Citrin 1974).
Relatedly, confidence in institutions should not be interpreted as synonymous with institutional legitimacy (Driscoll and Nelson 2018).We focus on trust rather than legitimacy for several reasons.First, as a practical matter 'trust' more closely corresponds to 'confidence', which matches the question wording found in many regional and worldwide omnibus surveys.Additionally, individuals who report little confidence in the functioning of the judiciary may not necessarily view the judiciary as illegitimate, per se, or believe that it should be dismantled.Nonetheless, this lack of confidence may still impact how that individual interacts with or views the state.Likewise, believing that the judiciary should exist (that it is a legitimate and necessary part of a democratic government) is not the same as believing that the judiciary is a viable forum for the provision of justice in the status quo.Because we are interested in understanding how institutional factors impact citizens' perceptions of justice, we focus on confidence, or trust, rather than the overall legitimacy of judicial institutions.
We test hypotheses related to two main aspects of institutional quality and design, what scholars of institutional trust refer to as 'process', and control for factors that scholars of institutional trust characterize as 'performance' (system-wide outputs, such as economic conditions).With respect to 'process', scholars in this literature argue that citizens evaluate institutions based on the extent to which the ideals of fairness and equality are realized by the rules, procedures, norms and structures that guide and constrain the functioning of political institutions and the behaviour of the state and political actors.Where citizens can rely on institutions to provide expected services without resorting to extralegal measures, citizens ought to have greater faith in their government and political institutions overall (Norris 2011;van der Meer and Hakhverdian 2017).
Application of 'process' to the judiciary is straightforward: citizens will deem well-functioning (or high-quality) judicial systems as more trustworthy.In the judicial politics literature, judicial system quality can be understood in terms of 'rule of law ' (Domingo and Sieder 2001) or the idea that all individuals and institutions are equally held accountable to the law.Rule of law is of course multidimensional, but generally speaking it can be understood to indicate or proxy for the overall level of judicial system quality in a given context.The idea is that well-functioning courts are more likely to activate diffuse and specific support, especially with regard to the judiciary (Boateng and Adjorlolo 2019;Bühlmann and Kunz 2011;Salzman and Ramsey 2013).As such, we hypothesize that: Hypothesis 1: Higher judicial system quality will increase public trust in the judiciary.
In our analysis, we operationalize judicial system quality with a multidimensional index of rule of law, which also enables the disaggregation of 'rule of law' into many component sub-indices, which include specific institutional factors such as access to justice and legal transparency (see Online Appendix, Table A3, for a comprehensive list).
A second approach to 'process' focuses on the presence and extent of corruption as an indicator of institutional quality.Indeed, among the more robust findings in studies conducted across the world is that greater levels of society-wide corruption negatively correlate with trust in political institutions (Boateng and Adjorlolo 2019;Chang and Chu 2006;Hakhverdian and Mayne 2012;Norris 1999;Salzman and Ramsey 2013;Tyler 1990;van der Meer and Hakhverdian 2017;van Erkel and van der Meer 2016).Simply, the expectation is that where corruption is greater citizens will be cognizant of that corruption and be less trusting of institutions, especially those that are the ostensible stewards of justice, fairness and the rule of law.
Yet corruption takes a variety of different forms, ranging from million-dollar corruption scandals to small bribes or kickbacks that grease the wheels of day-to-day bureaucracy.Several studies assess the effects of corruption on trust by assuming society-wide corruption will be perceived and punished (e.g.Hakhverdian and Mayne 2012;Salzman and Ramsey 2013).But since corruption is (often) an illegal act, the corrupt attempt to conceal their actions.The consequence is that the true or objective rate of society-wide corruption might not be easily observable or perceptible for the average citizen. 1 Furthermore, such a formulation assumes a minimum level of concern, awareness or interest on the part of all citizens.Failing that, it might be the case that corruption affects trust by operating through personal exposure to such acts, either by offering or being solicited for bribes, or perhaps having personal knowledge of a corrupt act.We thus disaggregate this hypothesis into institutional and microlevel sub-hypotheses: Hypothesis 2a: Greater society-wide corruption will have a negative effect on individuals' judicial trust.
Hypothesis 2b: An individual's personal exposure to corruption will have a negative effect on judicial trust.
Finally, recent studies of corruption and institutional trust have found the effects of various institutional or macro-level factors are not uniform across citizenries.
Specifically, several studies have identified how education conditions or moderates such effects.The idea is that better-educated individuals (sometimes described as 'politically sophisticated' citizens) are equipped to identify and credibly discern information regarding corruption when it emerges (e.g.Weitz-Shapiro and Winters 2017). 2  The better educated, moreover, are expected to more strongly disapprove of corruption (Hakhverdian and Mayne 2012) and therefore punish corrupt politicians at the polls or downgrade their trust in institutions. 3Though it is not clear that disapproval expressed in surveys translates into real-world punishment of corruption, Taylor Boas et al. (2019) find that it nonetheless signals commitment to anticorruption norms.As a result, better-educated individuals might be most likely to become disillusioned with political institutions in contexts of corruption.This may be especially relevant to judicial trust because scholars generally describe judicial institutions as complex, bureaucratic and confusing to ordinary citizens to begin with (e.g., Cowan 2004).Following the corruption literature, we hypothesize that individuals' level of educationwhich is thought to indicate the capacities to discern corruption and understand the workings and complexities of the judicial systemwill moderate the effects of these institutional variables.We test this hypothesis in two ways: Hypothesis 3a: Institutional quality (greater judicial system quality and low corruption) will increase trust among the better educated more than among the lesser educated.

Hypothesis 3b:
Where institutional quality is high, education will have a positive effect on judicial trust; conversely, where institutional quality is low, education will have a negative effect on judicial trust.

Data and methods
To test these hypotheses, we compile a multilevel data set comprised of Latinobarometer public opinion surveys and experts' assessments of country-level institutions collected by the V-Dem Institute, Transparency International and other organizations. 4The resulting data set allows us to estimate the hypothesized effects of macro-level institutional and micro-level factors, as well as the hypothesized cross-level interactions.The longstanding Latinobarometer survey not only spans more than three decades, but offers the greatest consistency in question wordings and availability, enabling cross-sectional and cross-temporal analysis.Because the independence and legitimacy of courts and judicial institutions are in part functions of (and critical to) democracy, we exclude non-democracies from our sample (countries with a Polity IV score of less than 6 in a given country-year). 5After accounting for missingness, the sample contains 211,470 observations nested in countries, nested in years.Due to survey instrument variations that result in missingness in key variables, our analysis includes survey waves from 2001 to 2016.

Dependent variable
We measure trust in the courts, our dependent variable, with a survey item asking respondents how much confidence they have in the judiciary.Respondents report their level of confidence on a four-point Likert scale, comprising a lot, some, little, or no confidence.We analyse these responses as a continuous variable and centre values at zero such that reports of little or no confidence register as negative and some or a lot of confidence as positive.Figure 1 shows that public confidence in the courts oscillates considerably around the over-time mean in almost all countries.Overall, confidence hovers in the territory of scepticism to mistrust.Uruguay and Costa Rica exhibit exceptionally high levels of trust over time, with average evaluations just shy of positive.Peru and Ecuador, by contrast, exhibit the lowest levels of trust over time, with decidedly negative evaluations.Those familiar with the region will not be surprised to see Uruguay and Costa Rica, two of the region's most stable and wealthy third-wave democracies, with the highest levels of trust in a democratic institution.But wealth and democracy are poor predictors of judicial trust beyond these exceptional cases.Indeed, Argentina and Chile, two other wealthy and stable third-wave democracies, exhibit disparate levels of trust in the courts.Correlation coefficients bear out these casual observations, indicating weak relationships between judicial trust and GDP per capita (ρ = 0.26) and Polity IV (ρ = 0.22) in a given country-year.
Judicial trust is measured in the Latinobarometer survey using a four-point Likert scale (no, a little, some, or a lot of confidence).We centre the scale at zero, with responses ranging from −1.5 to 1.5, such that negative responses take on negative values and positive responses positive values.In analysing these responses, we present and discuss estimated effect sizes in Likert units.We also discuss these substantive effects as percentages of the full range of the Likert scale (effectively dividing substantive effects by three).

Independent variables
We assess the impact of macro-level factors by drawing on a number of measures compiled from multiple sources.Concepts like rule of law, democracy and corruption have generated a significant degree of scholarly debate, and as a result do not always have obvious or straightforward measurements.
We assess our first hypothesis regarding judicial system quality with V-Dem's multidimensional rule of law index (v2x_rule).This composite measure is computed with factor analysis of more than a dozen sub-indices related to legal compliance, judicial independence and accountability (see Online Appendix Table A3 for more on V-Dem measures).As an eminently important concept, rule of law has of course given rise to multiple measurements.Systematic comparisons of such measures reveal high correlations between all measures (Versteeg and Ginsburg 2017), which should mitigate concerns over the contingency of findings on measurement.We opt for V-Dem's index for its multidimensional approach to measuring judicial system quality overall.Nonetheless, we assess the robustness of our findings by substituting alternative rule of law measures from Freedom House and the World Bank in the supplementary analyses (Online Appendix, Table A14).Our main analyses also disaggregate the index to assess whether specific dimensions of rule of lawspecifically, the accessibility of the justice system (v2xcl_acjst) and the impartial application of the law (henceforth 'legal transparency', v2cltrnslw)impact judicial trust most.We highlight these sub-indices in our analysis as aspects that are likely to be especially relevant to citizens' everyday perceptions.Further, we fully disaggregate the rule of law index in supplemental analyses (see Online Appendix, Tables A10 and A11).Supplemental analyses serve to further elucidate this bundled concept but also assess whether specific features of the judiciary can impact judicial trust.
To measure society-wide corruption (Hypothesis 2a), we rely on Transparency International's Corruption Perception Index (CPI), based on surveys of country experts.We flip and rescale this variable to range from 0 to 1, where 0 indicates low or no corruption and 1 indicates high corruption.In addition, we measure individuals' exposure to a corrupt act with a survey item asking respondents if they or their family members have known of a corrupt act (Hypothesis 2b).To assess Hypotheses 3a and 3b, we interact respondents' education level with the country-level institutional variables related to judicial system quality and corruption.As with rule of law, we similarly do not rely solely on Transparency International's measure of corruption in drawing conclusions.We select this measure for our main analysis because its wide use provides commensurability with other studies, but our supplemental analyses also test the robustness of the estimates computed using this measure by substituting alternative corruption measures from the World Bank and V-Dem, including measures specifically capturing judicial corruption (see Online Appendix, Table A13).
Incorporation of macro-level factors in the analysis of micro-level attitudes might raise concerns related to the ecological fallacy, or the assumption that individual behaviour can be inferred from aggregated data (e.g.King 1997).Readers should be reassured that our analyses and hypotheses do not require, and do not make, such assumptions.Rather than incorporating macro-institutional factors with the goal of inferring individual attitudes, our analyses test whether broader contextual factors impact individual attitudes on average, both of which are observed directly and independently.Cross-level interactions shift the analysis from comparisons between individuals and within countries to comparisons of individual averages between countries.

Controls
We control for a host of factors that likely influence institutional trust more generally.Beyond education, we include basic demographic controls for respondent age, sex, household wealth, socioeconomic status (SES) and religion. 6To control for political views that might influence judicial trust in particular, we also include a measure of respondent ideology based on self-placement on a left-right scale.To isolate trust in courts from related institutions tasked with maintaining law and order, we control for trust in the police.We include a measure of interpersonal trust to control for the effect of this personality trait.And to isolate any effect of regime satisfaction, we control for the respondent's overall life satisfaction.Due to missingness that severely curtails sample size, we do not include presidential approval or news consumption (a proxy for political interest) in our main analyses.We replicate our findings with these added controls in Online Appendix, Tables A4 and A5. 7  We also control for other 'performance' factors that have been shown to influence trust in a broader array of political institutions.Following the institutional trust literature, we control for subjective, sociotropic economic evaluations (Dalton 2004;Mishler andRose 1997, 2001;van der Meer 2017). 8We also include a measure of overall regime performance evaluation, operationalized as satisfaction with democracy (Chang and Chu 2006;Linde and Ekman 2003;Mishler andRose 1997, 2001;Norris 1999Norris , 2011)).To control for variation in state and institutional capacity and the overall level of democracy, we control for GDP per capita (measured in thousands) and the Polity IV score in all country-level estimations.Though we rely on many V-Dem measures in our analyses and there is no shortage of criticism of Polity IV's democracy measures (Lindberg et al. 2014), V-Dem's hallmark electoral democracy measure is continuous and therefore does not facilitate removing non-democracies from our sample (Lindberg 2016).The V-Dem data set does contain ordinal and categorical measures of regime type, but relying on these measures (instead of Polity IV) produces a nearly identical sample.The exception pertains to Ecuador after 2006, which cannot be considered unambiguously democratic, especially as it pertains to the judiciary (e.g.Conaghan 2016; de la Torre and Lemos 2016).Though no measure of democracy is without issue, we opt for the Polity IV measure for its clarity in ordinal coding.Nonetheless, we replicate our analyses by substituting V-Dem's electoral democracy score (Online Appendix, Table A12).
To control for factors suggested by the judicial politics literature, we rely on the 2002 Latinobarometer survey round, which includes a battery of survey items pertaining to experiences with and perceptions of the courts.Scholars have suggested that individuals form impressions of the courts based on whether or not they view their direct, personal experiencesas plaintiffs, defendants, witnesses or jurorsas having served justice (Benesh 2006;Benesh and Howell 2001;Tyler 1990; but see Gibson 1989;Mondak 1993).They further hold that marginalized social groups (i.e.those experiencing discrimination) are more likely to hold negative views of access to justice, the rule of law, the police and the courts in general (e.g.Caldeira and Gibson 1992;Méndez et al. 1999).We combine survey items to create a measure of court experience, coding respondents as (1) having no contact with the justice system, as (2) having positive contact or (3) negative contact.To measure the extent to which individuals' perceptions of fairness and equality before the law shape trust in the judiciary, we construct a legal discrimination index capturing beliefs that members of specific groupsthe rich, women, the indigenous, the poor and immigrantshave their rights respected (1) always, (2) almost always, (3) almost never or (4) never. 9Though these controls cannot be included in analysis of the full time period, they nonetheless provide an opportunity to assess the robustness of estimates to the inclusion of these controls.

Estimation and models
Given the structured nature of our data and theoretical interest in both individual and country-level characteristics and their interactions, we progress in our analysis from fixed effects to random intercept models.First, we estimate fixed effects models to control for temporal and cross-sectional variation and estimate the effects of micro-level factors on judicial trust.To assess macro-level factors and their interactions, we estimate three-level random intercept models, with individuals nested in countries, nested in years.

Analysis
Table 1 displays OLS estimates from models of judicial trust, including country and survey-year fixed effects.Model 1 estimates the effect of education and exposure to corruption, relevant to Hypotheses 2 and 3.The model estimates that one's personal exposure to corruption significantly decreases judicial trust by 0.08 Likert points, as predicted (Hypothesis 2b).Relative to respondents with less than primary education, respondents with primary education are 0.02 points less trusting of the court, and those with university education more trusting by 0.04 points.We do not hypothesize any specific direct effect of education, though this baseline model suggests that education correlates positively with judicial trust.Given the large number of micro-level observations, however, the estimated effects are substantively small, ranging from 3% of the full scale range (exposure to corruption) to less than 1% (primary education).
Model 2 adds demographic and political controls.The estimated effect of corruption exposure is similar in size and significance, though the estimated effects of education differ, with only primary and secondary education correlating negatively with judicial trust.In models 2 and 3 we see a clear dip in trust among the primary educated, followed by a steady increase among the high-school and university educated.Controls for age, gender and household wealth have significant, but substantively small, negative effects, whereas SES has a small and significant positive effect.Model 2 also estimates that, relative to non-ideologues, all ideologues (and especially conservatives) are more trusting of the court.Conservative ideology is associated with a 0.22-point (7%) increase in judicial trust, roughly twice as high as centrist or leftist ideology.Relative to Catholics, those with other or no religious beliefs are also estimated to be less trusting of the court, though these effects are substantively small.Model 3 adds additional controls from the institutional trust literature and for political interest and engagement.The estimated effect of exposure to corruption decreases in size in this model but remains significant. 10The estimated negative effects of primary and secondary education are robust to the inclusion of all controls in this model.All added controls are significant correlates of judicial trust.Of these, the most substantively significant predictors are trust in the police (0.3 points or 10% of the scale range) and one's satisfaction with democracy overall (0.13 points or 4% of the scale range).As predicted by the institutional trust literature, such 'performance' factors as subjective economic evaluations (along with democratic satisfaction) are significant predictors of trust.
Model 4 includes controls from the judicial politics literature and limits the sample to the 2002 survey round.We offer a word of caution in placing too much weight on these findings, as they come from a single survey round from two decades ago.Nonetheless, these data provide a significant (and otherwise nonexistent) opportunity to test these hypotheses from the judicial political literature on a wide scale in Latin America.As predicted by the judicial politics literature, experience with the courts and perceptions of fairness before the law significantly shape judicial trust.Relative to those with no court experience, those with an experience deemed unjust are 0.09 points (or 3%) less trusting, whereas those with a positive experience are indistinguishable from those without court experience.Similarly, those who believe marginal or vulnerable groups do not have their rights respected are 0.13 points (or 4% of the scale range) less trusting of the court.Above and beyond these controls, this model estimates a significant negative effect of exposure to corruption, but the estimated effect of education is indistinguishable from zero.
Overall, models of micro-level correlates of judicial trust provide modest support for Hypothesis 2b, that individuals who are personally exposed to corruption are less trusting of the judiciary.Education, by contrast, does not appear to have any consistent, clearly monotonic or significant independent effect on judicial trust.
We turn now to macro-level institutional factors.Figure 2 displays scatter plots of country-level correlations between the relevant institutional variables (societywide corruption and judicial quality) and the mean level of judicial trust in a given survey year.Overall, institutional factors correlate modestly with overall levels of judicial trust among the public.As predicted, society-wide corruption (as measured by the CPI) is negatively correlated with mean judicial trust (ρ = −0.35)and V-Dem's rule of law index is positively correlated (ρ = 0.32).
Table 2 presents random intercept models of judicial trust and incorporates country-level institutional factors to assess Hypotheses 1, 2a and 3. We include both the CPI and V-Dem's rule of law index in our analyses, as well as the access to justice and legal transparency sub-indices.In each model we control for GDP per capita (state capacity proxy) and the Polity IV score.Because of the small number of country observations, we add each institutional variable individually in models 5 through 8.The effect of GDP per capita is significant in three of four models, but this effect is negative and substantively small.A $1,000 increase in GDP per capita is estimated to decrease average levels of judicial trust by 0.02 points (or less than 1% of the scale range).Notably, the Polity IV democracy measure does not correlate with judicial trust, though this may not be surprising given that our sample excludes nondemocracies.Table A12 in the Online Appendix replicates these models substituting V-Dem's electoral democracy index.These estimates are also insignificant.
Once controlling for micro-level determinants, institutional factors associated with corruption or the rule of law are estimated to correlate significantly, but unevenly, with judicial trust on average.The exception in Table 2 is corruption, which is estimated to affect judicial trust negatively and substantially.Moving from the least to the highest levels of the CPI score is estimated to decrease average trust by 0.4 points (or 13% of the scale range).The significance of this effect, however, is dependent on operationalization.Table A13 in the Online Appendix replicates model 5 with alternative measures of corruption (including measures of judiciary-specific corruption), and only the CPI measure correlates significantly.Rule of law and the access to justice and legal transparency sub-indices similarly do not correlate significantly, on average.In our supplemental analyses, we disaggregate the rule of law index and estimate the effect of sub-indices on judicial trust and similarly find uneven effects.Sub-indices that do correlate positively and significantly include judicial accountability and compliance, as well as the absence of judicial purges and reforms that aim to weaken the judiciary.
Of course, the small number of country observations in the sample means that country-level factors are estimated with less efficiency than micro-level factors, which remain strong and consistent predictors of judicial trust.Despite this,  Government and Opposition random intercept models provide some support for Hypotheses 1 and 2a: institutional factors are estimated to have modestly direct, if uneven, effects on the average levels of judicial trust.Finally, we assess whether these average effects vary significantly by level of education, (Hypotheses 3a and 3b).To do so, we estimate models that interact institutional factors with respondents' education.Figure 3 presents point estimates of the average effect of institutions conditional on respondent education (full model estimates are available in Online Appendix Table A6).Interactive models replicate those in Table 2 with the addition of interaction terms.As in Table 2, interaction terms are added individually for each institutional variable.All models contain main effects for education and institutional variables.
Estimates presented in Figure 3 lend support to Hypothesis 3a, which predicted that institutional quality will increase judicial trust among the better educated.Respondent education has a positive and statistically significant effect on the relationship between institutional quality and judicial trust.Though there is a clear monotonic relationship between education level and the effect of institutions, any effect of institutions is consistently identified only among university-educated respondents.Indeed, only the university educated are less trusting of the judiciary in countries with greater society-wide corruption, and are more trusting of it where there is a greater degree of rule of law overall, the judiciary is more accessible, and the law is applied more transparently.
In Online Appendix Tables A11, A13 and A14, we replicate these interactive models substituting alternative rule of law and corruption measures, and by disaggregating V-Dem's rule of law measure into its sub-indices.Results indicate robust and consistent estimates across specifications: with nearly all measures, bettereducated respondents are more likely to trust the judiciary where judicial institutional quality and the rule of law are high, and are less likely to trust the judiciary where corruption is worse.The exceptions are found in the rule of law sub-indices.The estimated effects of only two out of thirteen sub-indices do not vary significantly with education, judicial review and efforts to undermine the judiciary with judicial reform.All measures of corruption and alternative rule of law measures, by contrast, significantly increase the judicial trust of the university educated.
To assess Hypothesis 3b more substantively, we plot in Figure 4 the estimated marginal effect of university education (relative to less than primary education) on judicial trust at various levels of institutional quality. 11As predicted, the effect of education is conditional on institutional quality.Indeed, institutional quality does not simply attenuate the size of education's effects on judicial trust, but is estimated to change the direction of this effect.Where institutional quality is high, the highly educated are more trusting; where it is low, they are less trusting.The estimated effects of between plus or minus 0.1 and 0.2 points translate to 3% and 6% of the full scale, respectively.The extended analyses in Online Appendix, Tables A13,  A15 and A16 indicate similar patterns regardless of the particular institutional feature analysed.In sum, our findings strongly support Hypotheses 3a and 3b from the corruption literature: education and institutional factors interact to shape citizens' perceptions and evaluations of the judiciary.

Conclusions
This article presents a systematic assessment of trustor mistrustin the judiciary in Latin America.Our multilevel analysis of public opinion surveys spanning 13 years and 18 countries leveraged individual and country-level variation to assess the microfoundations of confidence in the courts.We find little evidence that institutional quality in the form of overall rule of law or the absence of corruptionat least as determined by expert surveysinfluences the mass public's evaluations on average.This finding ought to be especially relevant to policymakers, activists and  advocates, considering that 'rule of law reform' projects seeking to strengthen institutions and stabilize democracy have received considerable attention and an outpouring of financial aid from domestic and international organizations (see World Bank 2012).Instead, our analysis extends findings from the corruption literature, which suggests that only the best-educated citizens will be responsive to institutional quality in general.University-educated respondents are more trusting of the judiciary in less corrupt societies and where the rule of law is strongest.Insofar as rule of law is multidimensional, our findings suggest that better-educated respondents respond to myriad aspects of judicial quality.In our analysis, we highlighted that better-educated respondents are most trusting of legally transparent and accessible judiciariesthose where citizens can expect to bring cases before the court, see their rights defended and receive what they view as impartial justice.Among the broader public, however, traditional performance factors (subjective economic evaluations, democratic satisfaction and presidential approval) are stronger determinants of judicial trust.
Our findings on trust in Latin American courts complicate studies of judicial trust in the US.Studies that focus on the US imply that courts can work to increase citizen trust through at least two mechanisms: by publicizing their work (increasing knowledge about the courts), or by improving performance.We do not find evidence in support of 'positivity theory', the notion that knowledge of the courts is associated with esteem for the courts (Gibson and Caldeira 2009;Gibson et al. 1998Gibson et al. , 2014)).From this perspective, increased knowledge about the courts will lead citizens to view the courts as more trustworthy and legitimate.While this finding has been most robustly demonstrated in the context of the US Supreme Court, a 1995 survey by James Gibson (2007) found support for positivity in Bulgaria, Hungary, Poland, Russia, France, Spain and the US. 12 Our examination of this phenomenon outside of Europe and North America complicates this picture.At least in the context of Latin America, those who are more politically sophisticated or knowledgeable do not necessarily become more supportive of the courts.Whether or not courts are capable of inspiring confidence or esteem among the knowledgeable is contingent on institutional quality.
Relatedly, in contrast to most models of institutional trust, we do not find support for the notion that institutional performance is associated with trust.We assess the performance of the judiciary using several different measures: the rule of law, access to justice and legal transparency indices compiled by V-Dem, as well as the CPI, which may offer a more diffuse sense of how law functions (or does not function) in a particular country.Few of the direct relationships between these variables and judicial trust are robust or statistically significant.These findings indicate that apex courts globally may be more limited in their ability to impact public perceptions than scholarship on the US Supreme Court suggests.Widening the geographic scope of studies of judicial trust also demonstrates the limitations, and potential pitfalls, of adopting findings in one context as baseline assumptions.The variation we identify between countries within Latin America, and indeed between Latin America and other world regions, suggests there is more to unpack in the relationship between institutional performance and citizen trust.
Our analysis represents a first step towards understanding the dynamic relationship between citizens and courts, and has sought to integrate literatures on judicial politics, corruption and institutional trust.To the best of our knowledge, our analysis offers the greatest temporal coverage in the study of judicial trust in a developing world region and incorporates a greater breadth of institutional factors.But our study also faces limitations.First, while our observational survey data analysed here were sufficient for identifying correlations robust to temporal and crosssectional variability, such data do not allow us to make inferences on the direction of causality.Whether subjective evaluations of institutional performance are indeed causes of institutional trust, or whether institutional trust affects individuals' subjective evaluations of institutional performance, remains unclear and is worthy of further empirical scrutiny.The same can be said of the provision of justice and whether it shapes, or is shaped by, institutional trust.
A second caveat is in order regarding the extent to which omnibus survey data can provide insight into the meaning of institutional trust from the perspective of citizens.In some ways, this concern about what beliefs survey items tap into is reminiscent of past debates regarding diffuse and specific support (Easton 1965), or whether evaluations of trust reflect considerations of institutions or their officeholders (Citrin 1974).These questions remain relevant, but justice is a multidimensional and complex concept that, perhaps more so than other institutions, can be difficult to reduce to a single question included in a battery of survey items.Data limitations and a scarcity of qualitative research on how individuals perceive and understand judicial institutions means that we must be careful in attributing too much substance to reports of institutional trust in such survey data without a deeper understanding of how survey respondents interpret justice itself and the questions posed to them.Further research in this vein could provide welcome additions to this literature.
Despite these limitations, our findings nonetheless raise doubts about common expectations regarding the function of judicial institutions and state legitimacy.Scholars of institutional trust have ceased sounding the alarm regarding low public trust in political institutions under the logic that such deficits have not necessarily translated into preferences for non-democratic regimes.More sanguine observers point to the 'democratic deficits' observed in many advanced and stable democracies as evidence (Norris 1999(Norris , 2011)).These deficits may not be problematic for institutions whose purview is the business of politics itself.But, unlike executives, political parties or legislatures, robust and impartial judiciaries are supposed to be insulated from politics, adhering to and applying the law free from the pressures or influences of political elites and publics.Yet our findings suggest that in many ways the judiciary is viewed by most citizens no differently to other institutions occupied by overtly political actors, with institutional qualitywhen it affects trust at allbeing overshadowed by factors unrelated to institutional performance.This raises questions regarding the judiciary's ability to serve its vital functions and potentially protect democratic institutions if the forces that erode democratic satisfaction simultaneously erode confidence in the court as a bulwark of democracy.
Acknowledgements.For helpful comments and feedback, the authors would like to thank Taylor Boas and the participants of the 2020 meeting of the Southern Political Science Association, as well as workshop participants at the Center for Inter-American Policy and Research at Tulane University.

Figure 1 .
Figure 1.Proportion of Respondents Reporting Judicial Trust, 1995-2017Notes: Confidence is measured on a four-point Likert scale (no, little, some, or a lot of confidence), centred at 0 and ranging from −1.5 to 1.5.The straight line is the country's surveyweighted over-time mean.

Figure 2 .
Figure 2. Scatter Plots of Corruption and Rule of Law Indices with Judicial Trust Note: Y-axis indicates country-year means of judicial trust.CPI = Corruption Perception Index.

Figure 3 .
Figure 3. Effects of Institutional Factors Conditional on Education Note: Figure displays 95% confidence intervals.CPI = Corruption Perception Index.

Figure 4 .
Figure 4. Average Partial Effect of University Education by Institutional Quality Note: Figure displays 95% confidence intervals.CPI = Corruption Perception Index.

Table 1 .
OLS Models of Judicial Trust Notes: Standard errors in parentheses.* p < 0.05, + p < 0.1.All models include country and survey-year fixed effects.Model 4 analyses 2002 data only and includes country fixed effects.