Policy Research Working Paper 10372 Institutional Trust, Perceptions of Distributive Unfairness, and Income across Salvadoran Municipalities Emilio Depetris-Chauvin Poverty and Equity Global Practice March 2023 Policy Research Working Paper 10372 Abstract Using multiple waves of two public opinion surveys and a trust in the Catholic Church. The relationship between two-way fixed effect model, this paper analyzes how people’s income and trust toward the president and municipalities perceptions and attitudes towards public institutions shifted masks a relevant heterogeneity from a rural-urban divide as with the business cycle in El Salvador during 2004–2018. well as from differences in municipal state capacity. Further, It finds that individuals’ levels of trust toward both the views of income distribution fairness as well as preferences president and the municipal government are positively asso- for democracy are positively shaped by municipality-spe- ciated with higher levels of income at the municipality level. cific business cycles. In contrast, neither generalized trust Income is also a strong predictor of trust in mass media, nor satisfaction with democracy is empirically associated confidence in the judicial system and, to a lesser extent, with income at the municipality level. trust in the national legislature but income does not affect This paper is a product of the Poverty and Equity Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The author may be contacted at mrobayo@worldbank.org and brude@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Institutional Trust, Perceptions of Distributive Unfairness, and Income across Salvadoran Municipalities* Emilio Depetris-Chauvin† JEL Classification: E32, H11, Z13 Keywords: trust, institutions, income, El Salvador. * Paper prepared for the World Bank Group as part of an Institutional and Political Economy Analysis for El Salvador Systematic Background Country Diagnostic. TTL: Monica Robayo † ıa, Pontificia Universidad Cat´ Associate Professor, Instituto de Econom´ olica de Chile. I thank Monica Robayo, Francesca Recanatini, and Rafael Chelles Barroso for suggestions that greatly improved the paper. Leonor Castro provided outstanding research assistance. 1 Introduction Trust and confidence in institutions are necessary for policy implementation and are particularly key ingredients to the sustainability and legitimacy of long-needed social and economic reforms. Indeed, it has long been argued that a high level of trust is a virtue for well-governed societies and a building block for prosperity (Papaioannou, 2013). While a burgeoning literature in economics has shown that differences in levels of trust across societies are deeply rooted and can be traced back to historical events or linked to pre-modern times’ societal characteristics, it is also known that trust can change substantially in a shorter period of time (Putnam, 1995).1 Previous work has shown that levels of interpersonal trust in general, and trust in public institutions in particular, can be linked to economic shocks in middle and high-income countries (see, e.g., Stevenson and Wolfers (2011), Ananyev and Guriev (2018), and Papaioannou (2013)).2 Yet little is known about how the business cycle can shape trust and confidence in public institutions in the developing world. In this paper, I aim at filling this gap in the literature by studying how individuals’ attitudes towards institutions and perceptions of distributive unfairness vary with municipality-specific economic shocks in El Salvador during the 2004-2018 period. El Salvador offers an ideal testing ground to study the link between the business cycle and atti- tudes towards public institutions for several reasons. First, levels of generalized trust, satisfaction with democracy, perceptions of fairness in income distribution, and government approval in El Salvador are among the lowest in Latin America (Corporaci´ on Latinobar´ometro, 2018). Second, coinciding with mild economic growth in recent times, and as is shown below, El Salvador has witnessed a steady decline in trust towards state institutions and an erosion in satisfaction with the democracy, especially after the 2008 global financial crisis. Third, despite being one of the coun- tries with the lowest income inequality in Latin America, a vast majority of Salvadorans thinks the income distribution is unfair. Fourth, El Salvador is currently experiencing an unprecedented political alignment between the executive and legislative branches, which represents a unique pos- sibility for the discussion and approval of reforms. If trust and confidence are indeed malleable by the business cycle, policymakers could build positive attitudes towards reforms by choosing pro- growth policies. For instance, counter-cyclical fiscal policy during a time of economic downturns may prevent the erosion of trust and may even boost institutional legitimacy for reforms. 1 The literature on the deep-rooted determinants of trust and attitudes towards institutions has studied, among others, the importance of the slave-trade during pre-colonial time in Africa(Nunn and Wantchekon, 2011), climatic variability (Buggle and Durante, 2021), the impact of communism in Eastern Europe (Alesina and Fuchs-Sch¨ undeln, 2007), and pre-modern political institutions (Guiso et al., 2016) 2 Stevenson and Wolfers (2011) show that state-specific economic shocks lead to lower trust in institutions in the United States. Papaioannou (2013) shows that trust in institutions declined substantially more in the European countries more severely hit by the 2008 global financial crisis. Ananyev and Guriev (2018) show how income impacted trust in the Russian Federation after the 2008 global shock. 2 How could the business cycle shape attitudes (particularly trust) towards public institutions? First, a positive relationship between income and trust can operate through changes in loss aver- sion. Individuals experiencing a positive (negative) shock in their standard of living may feel more (less) secure and then more (less) willing to trust others in general and institutions in general. Sec- ond, trust and positive attitudes can be understood as an assessment of the political system and its institutions. An individual observing the business cycle may perceive shocks as sufficient statis- tics of institutions’ performances. For instance, a good economic situation could indicate good institutional performance (regardless of whether the attribution and assessment are correct), thus boosting trust in (and satisfaction with) state and related institutions. Third, and connected to the previous point, income shocks can directly affect the resources needed for well-functioning insti- tutions. Therefore, changes in attitudes towards institutions may reflect changes in institutional equilibrium due to initial changes on the supply side of institutions. In this paper, I use multiple waves of two public opinion surveys (i.e., LAPOP and Latino- bar´ometro) to analyze how people’s perceptions and attitudes towards public institutions respond to the business cycle in El Salvador. To do so, I estimate several econometric models explain- ing respondents’ attitudes towards different institutions (e.g., trust in the president or satisfaction with democracy) as a function of municipality-specific income levels. Nonetheless, the uncovered robust statistical associations do not necessarily imply causality. Indeed, economic development within a country is not a random process in which income has been randomly assigned across municipalities over time. The process of economic development and the formation of trust and confidence towards institutions (as well as its evolution), and the factors underlying the emergence and persistence of these attitudes are complex phenomena. Furthermore, some of those factors are unobserved, which hinders the identification of the causal effect of income on the aforemen- tioned attitudes. Nevertheless, I argue that it is unlikely that my OLS results are fully driven by omitted factors or a reverse causality from trust to income. Importantly, my econometric models include both year fixed effects, which partial out the aggregate business cycle for El Salvador, and municipality fixed effects to control for persistent cross-municipality differences. Thus, my speci- fications explore the relationship between attitudes and deviations of the municipality income from the country average. The lack of subnational-level statistics for GDP or measures of household income is a com- mon problem in developing countries, and El Salvador is not the exception where a limited set of municipalities has income measures. To expand the set of municipalities under analysis and cover periods and regions for which income data is not available, I use a proxy of income commonly exploited in the growth and development literature: density of nighttime lights (Henderson et al., 2012). I show below that this proxy indeed works well predicting household income (and even cross-sectional variation in poverty rates) in El Salvador over time when such data exist. 3 My econometric analysis reveals that individuals’ trust towards both the president and the mu- nicipal government is positively associated with a higher level of income at the municipality level. I also document that municipality-specific business cycle is a strong predictor of trust in mass me- dia and, to a lesser extent, of trust in the national legislature and confidence in the judicial system but does not affect trust in the Catholic Church. By performing a heterogeneity analysis, I also show that the relationship between income and trust towards the president and municipalities is affected by a rural urban divide and differences in municipal state capacity. That is, most of the positive association between income and trust in the Salvadoran president is explained by individ- uals living in rural areas and municipalities with low levels of state capacity (as proxied by the share of households with access to electricity and water services), whereas most of the positive association between income and trust in the municipal government comes from individuals living in urban areas. Additionally, views of fairness in income distribution, as well as preferences for democracy, are positively shaped by municipality-specific business cycles, while neither generalized trust nor satisfaction with democracy is empirically associated with variations in income across Salvadoran municipalities. The rest of this paper is organized as follows. In Section 2, I discuss the data used in this paper, including a validation of nighttime light density as proxy for income at the municipality level, as well as recent trends in attitudes toward public institutions in El Salvador. Section 3 presents the main empirical strategy to study the link between attitudes and municipality-specific business cycle. Section 4 presents and discusses the main empirical results whereas section 5 concludes. 2 Data and Recent Evolution in El Salvador In this section I describe the data used and provide a brief description of the recent evolution in attitudes toward a set of public institutions in El Salvador. I also discuss the relevance of Nighttime Light Density as proxy for income across Salvadoran municipalities. 2.1 Attitudinal Data from Opinion Surveys In order to study Salvadorans’ attitudes towards public institutions over the 2004-2018 period, I use several waves of two opinion surveys, namely, LAPOP (9 waves) and Latinobar´ ometro (13 waves). Both surveys are implemented in El Salvador since the 1990s and focus primarily on citizen’s perceptions and attitudes toward different institutional actors. LAPOP is conducted ev- ery two years and interviews around 1,500 individuals in each wave whereas Latinobar´ ometro interviews around 1,000 Salvadorans every year. Both surveys provide information on a small set 4 of respondents’ demographic characteristics such as gender, years of schooling, and age. While LAPOP allows the identification of respondents living in rural and urban areas, Latinobar´ ometro is exclusively conducted in urban areas (albeit it allows to distinguish between small and large urban- ization). Importantly, both survey provide information regarding the municipality of residence for each respondent which allows to link income data discussed below as well as other time varying municipality-level characteristics (e.g., crime rates or national government transfers). I focus my analysis in 10 outcome variables. From LAPOP I focus on 1) trust in the president, 2) trust in the municipal government, 3) trust in the national legislature, 4) confidence in the judicial system, 5) trust in mass media, and 6) trust in the Catholic church. All trust questions asked about trustworthiness on the aforementioned institutions. Specifically, LAPOP asks “To what extent do you trust the President/Municipality/National Legislature/Mass Media/Church”. Respondents provide ordinal answers from 1 (not at all) to 7 (a lot). I construct dichotomic variables taking value of 1 if the answer is 5 or more, 0 otherwise. For the case of confidence in judicial system the question asked is slightly different: “If you were a victim of a robbery or assault how much faith do you have that the judicial system would punish the guilty?” The answers go from 1 (a lot) to 4 (little). I coded confidence in the judicial system as 1 if the answer is “a lot” or “some”, 0 otherwise. I dichotomize ordinal dependent variables for three main practical reasons: to ease the interpretation of the results, to employ more reliable measure of trust/confidence, and to facilitate the computation of my econometrics models.3 Nonetheless, it is important to note that none of the main results depends on this dichotomization approach (see Table A.2 in the appendix). From Latinobar´ ometro I focus on the following 4 variables: 1) perception of income distribu- tion in the country being unfair, 2) generalized trust, 3) preference for democracy, and 4) satisfac- tion with democracy. The first question asked is “How fair do you think income distribution is in El Salvador?” The possible answers are“very fair”, “fair”, “unfair” or “very unfair”. Following Reyes and Gasparini (2021) I coded this variable as 1 if individuals answer “unfair” or “very unfair”, 0 otherwise. Generalized trust is based on the question“Generally speaking, would you say that you can trust most people, or that you can never be too careful when dealing with others” I coded gen- eralized trust equals 1 if answer is “you can trust most people”, 0 otherwise. Finally, the variable preference for democracy takes the value 1 if the respondent states “Democracy is preferable to any other kind of government”, 0 otherwise. The variable satisfied with democracy takes the value 1 if the respondent answers “very satisfied” or“fairly satisfied” to the question “In general, would you say that you are very satisfied, fairly satisfied, not very satisfied or not at all satisfied with the 3 Arguably, ordered responses from 1 to 7 are not highly reliable in terms of providing precise information about respondent differences in trust. In this sense, dichotomizing the highest values of response may better indicate whether a given respondent has a high level of trust. On the other hand and as discussed below, my main econometric model includes fixed effects at both the municipality and year levels. This makes the computation of a non-linear model difficult as no default approach for fitting fixed-effects models for ordered responses exists. 5 way democracy works in El Salvador?” Again the main results do not depend on dichotomization decisions. Table 1 shows some basic summary statistics for the aforementioned dependent variables as well as for some set of individual-level characteristics. Panel A focuses on LAPOP data while Panel B on Latinobar´ ometro data. My samples of analysis come from pooling different waves available for the time period 2004-2018. Note that the total number of observations for each variable vary because some of the questions were not asked every wave and, in some cases, when asked some respondents choose to not answer. Roughly half of the respondents are men (48%) and the average age is 39 in both samples. In the LAPOP sample, 38% are individuals living in ometro sample lives in small urban areas. rural areas while 18% of the individuals in the Latinobar´ Average years of schooling is above 8 years while, depending on the sample, between 7% and 9% of the individuals are unemployed (i.e,individuals who are not working and look for a job). The LAPOP sample reveals substantial heterogeneity in the level of trust depending on which institutions one looks at. According to the respondents, the Catholic Church is the most trusted institution (60% of Salvadoran strongly trust in the Church) followed by mass media and municipal government (both with 56%). On the other hand, the national legislature (with 39%) and the president (with 45%) are the less trusted over the time period of analysis. Finally, only 38% of Salvadorans think that the judicial system would punish the guilty if they were a victim of a robbery or assault. The Latinobar´ometro sample shows that 52% of respondents thinks that income distribution in El Salvador is “unfair” or “very unfair”. Few Salvadorans are optimistic about current economic conditions (only 16% answered that the economic conditions of the country were better than a year before). A large share of respondents (i.e., 70%) thinks that democracy is the preferable to other kinds of governments, but only half of them (i.e., 35%) were satisfied with the democracy. Finally, ometro reveals a very high level of mistrust among Salvadorans since only 22% of the Latinobar´ respondent said that one could trust most people (as opposed to the alternative “you can never be too careful when dealing with others”). 2.1.1 Recent Trends in Attitudes Towards Institutions Figure 1 depicts the temporal evolution of the aforementioned attitudinal measures over the 2004- 2018 period. In Panel A I plot the evolution of trust measures from LAPOP. Two main patterns emerge from this figure. First, all measures but trust towards municipal government have experi- enced a steady reduction over the last years. Second, trust in the president seems to be the most volatile measure. It reached its highest value after the 2008 global financial crisis (almost 70% of respondents trusted the president in 2010) and then experienced a rapid drop reaching 27% in 2018 6 Table 1: Summary Statistics Variable Obs Mean Std. Dev. Min Max Panel A: LAPOP (2004-2018) Trust in President 8983 .454 .498 0 1 Trust in Municipality 12283 .558 .497 0 1 Trust National Legislature 12076 .393 .488 0 1 Confidence in Judicial System 12107 .382 .486 0 1 Trust in Media 10721 .558 .497 0 1 Trust in Church 12201 .605 .489 0 1 Male 12392 .481 .5 0 1 Age 12354 39.063 16.432 17 99 Age Squared 12354 1796 1493 289 9801 Years of Schooling 12227 8.287 4.977 0 18 Unemployed 12392 .07 .256 0 1 Living in Rural Area 12392 .379 .485 0 1 ometro (2004-2018) Panel B: Latinobar´ Income Distribution Perceived as Unfair 12580 .515 .5 0 1 Generalized Trust 12128 .215 .411 0 1 Economy is Improving 11433 .158 .365 0 1 Democracy is Better 11542 .705 .456 0 1 Satisfied with Democracy 11817 .347 .476 0 1 Male 12580 .475 .499 0 1 Age 12580 39.011 16.02 18 94 Age Squared 12580 1778.486 1436.62 324 8836 Living in Small City 12580 .184 .388 0 1 when the president is described as the lowest trusted institution among all the institutions studied here. In Panel B I plot attitudinal measures from Latinobar´ ometro. Several patterns also emerge from this figure. First, the 2008 crisis seemingly triggered a positive shock on people’s views about democracy. The fraction of respondents saying that democracy was the best kind govern- ment and were satisfied with democracy reached their highest value in 2009. Second, starting in 2010, satisfaction with democracy dramatically plummeted from 60% to less than 20% in 2018. People’s preference for democracy followed a similar downward trend. Third, while generalized trust remains very low over the whole period, it started to experience a steady decline since 2008. Figure 2 reveals another interesting pattern: for all trust measures in LAPOP, respondents living in rural areas show higher levels of trust relative to those living in urban areas. The same patterns emerge in Figure 3 when distinguishing between individuals living in municipalities with weak and strong state capacity. Those living in municipalities with low levels of public goods provision (proxied by the share of households with access to electricity and water services) tend to trust relatively more. Interestingly, this pattern is less clear when one looks at attitudinal measures from Latinobar´ ometro. Indeed, it can be seen in Figure 4 that views of unfairness of income distribution are not systematically stronger in any particular type of municipality. Likewise generalized trust 7 Figure 1: Overall Trends in Attitudes 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 2004 2006 2008 2010 2012 2014 2016 2018 0 2004 2006 2008 2010 2012 2014 2016 2018 year year Income Distribution Perceived as Unfair President Municipal Gov. Generalized Trust National Leg. Judicial Syst. Democracy is Better Media Church Satisfied with Democracy (a) LAPOP ometro (b) Latinobar´ Figure 2: Urban vs Rural 1 1 1 .8 .8 .8 .6 .6 .6 .4 .4 .4 .2 .2 .2 0 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year year Rural Urban Rural Urban Rural Urban (a) Trust in President (b) Trust in Municipal Gov. (c) Trust in Nat. Legislature 1 1 1 .8 .8 .8 .6 .6 .6 .4 .4 .4 .2 .2 .2 0 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year year Rural Urban Rural Urban Rural Urban (d) Confidence in Jud. Syst. (e) Trust in Media (f) Trust in Church 8 Figure 3: Weak vs Strong State Capacity (LAPOP) 1 1 1 .8 .8 .8 .6 .6 .6 .4 .4 .4 .2 .2 .2 0 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year year Strong State Weak State Strong State Weak State Strong State Weak State (a) Trust in President (b) Trust in Municipal Gov. (c) Trust in Nati. Legislature 1 1 1 .8 .8 .8 .6 .6 .6 .4 .4 .4 .2 .2 .2 0 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year year Strong State Weak State Strong State Weak State Strong State Weak State (d) Confidence in Jud. Syst. (e) Trust in Media (f) Trust in Church and the two measures of attitudes towards democracy do not follow a weak-strong state capacity divide.4 2.2 Nighttime Light Density (Proxy for Income) The lack of subnational-level statistics for GDP or measures of household income is a common problem in developing countries and El Salvador is not the exception. Indeed, no municipality- level measures of income, let alone other measures of economic development, are consistently available over time for the universe of Salvadoran municipalities. While the Multipurpose House- hold Survey (EHPM) does provide income data for some municipalities, its coverage is limited to 50 municipalities. Nonetheless, even for these so-called self-represented municipalities some years are missing for the period under analysis. That is, for the period 2001-2018 the EHPM al- lows the construction of an unbalanced panel with 772 municipality-year observations, excluding 78 municipality-year observations due to the lack of data. In response to the lack of GDP figures on a consistent basis at the subnational level, the seminal work by Henderson et al. (2012) proposes the use of a proxy that cover periods or regions for which GDP data are not available at all or not available in a timely fashion: density of nighttime lights (NTL density from now on). Henderson et al. (2012) show that NTL density indeed works well as a proxy for economic development at the 4 Figures A.1 and A.2 depict the evolution of all the attitudinal measures by each of the four Salvadoran macro regions. 9 Figure 4: Weak vs Strong State Capacity (Latinobar´ ometro) 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Strong State Weak State Strong State Weak State (a) Unfair Distribution (b) Generalized Trust 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Strong State Weak State Strong State Weak State (c) Democracy is Better (d) Satisfied with Democracy 10 subnational level, particularly in the developing world. I thus use NTL data to expand the sample of Salvadoran municipalities in my analysis. NTL density data come from the Defense Meteorological Satellite Program (DMS) - Opera- tional Linescan System (OLS). Every night from 8 PM to 9:30 PM the DMSP captures night light intensity from space. The high-frequency measure of NTL is a six bits digital number calculated for each location (i.e., pixel) of approximately 0.86 square kilometers in the Equator (its annual measure is an average across all nights for a given year). A digital number can go from 0 (a non-lit area) to a maximum of 63. Panel (a) in Figure 5 shows the distribution of digital numbers across pixels in El Salvador and its neighboring countries for the year 2004. The darker the pixel, the higher the density of NTL. As can be seen in the map, the intensity of NTL is higher in the most developed parts of El Salvador, such as San Salvador, Santa Ana, and San Miguel. It is important to note that the satellites’ technologies changed in 2013, which makes the direct comparison between the 2000-2012 and 2013-2018 periods difficult. To overcome this problem Li et al. (2020) created a harmonized global nighttime light dataset for the entire period for which NTL data is available (i.e., 1992–2018). I exploit this granular data to compute density of NTL at the municipality level by averaging across all pixels within the boundaries of each Salvadoran municipality. Panel (b) in Figure 5 depicts the average NTL density by municipality in 2004. A closer look at the data reveals that even within regions of El Salvador substantial heterogeneity in development exists across municipalities.5 2.2.1 Validation of the Proxy for Income at the Municipality Level To validate the use of NTL data as a proxy for income in El Salvador I first show NTL density is a strong predictor of income at the municipality level. For the unbalanced panel of municipal- ities for which income data is available, Table 2 shows that, indeed, NTL density does a great job predicting per capita household income, average household income, and per capita household labor income.6 Even after controlling for municipality and year fixed effects, NTL density and the different measures of income are strongly and positively associated. All these associations are statistically significant under the standard levels of confidence. The implied elasticity from the different specifications in Table 2 ranges between 0.2 and 0.32. Moreover, the results in Table 2 show that using the average of NTL for two consecutive years does a better job than just using 5 Li et al. (2020) recommends removing non-lit areas with low values of digital numbers. I removed from the analysis three municipalities with an average NTL density below 1.5 for the period of analysis. These excluded an, and Nuevo Ed´ municipalities are Cacaopera and Joateca in the department of Moraz´ en de San Juan in the department of San Miguel. 6 All income measures come from SEDLAC’s harmonized versions of Encuesta de Hogares de Prop´ ositos ultiples (EHPM). All waves from 2000 to 2018 were available through the World Bank’s DATALIBWEB. M´ 11 Figure 5: NTL Density in El Salvador (a) NTL Density in 2004 (pixels) (b) Average NTL Density by Municipality in 2004 Table 2: NTL density and Income Per Capita Household Average Household Per Capita Household Income (in logs) Income (in logs) Labor Income (in logs) (1) (2) (3) (4) (5) (6) NTL Density at Municipality (in logs) 0.229*** 0.290*** 0.192** 0.244*** 0.250*** 0.319*** [0.068] [0.075] [0.074] [0.082] [0.083] [0.092] years used to compute NTL same 2-yrs average same 2-yrs average same 2-yrs average Observations 772 772 772 772 772 772 R-squared 0.736 0.738 0.687 0.689 0.718 0.720 Notes: Sample is based on unbalanced panel of 50 self-represented municipalities in Multipurpose Household Survey (EHPM) from 2001 to 2018. All regressions include municipality and year fixed effects. Robust standard errors clustered at the municipality level in brackets. Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. the contemporary value consistently with the idea that averaging two consecutive years effectively smooths out measurement error. Reassuringly, Table A.1 shows that for the cross section of the universe of Salvadoran municipalities in 2004, NTL density is also a very strong (negative) predic- tor of (household) poverty rates. Figure A.3 shows different scatter plots for NTL density, income, and poverty rates in cross-sections for different years. 2.3 Other Data Sources In addition to survey, household income, and NTL density data I exploit two additional sources of data. In order to account for the potential time-varying effect of crime on individuals’ attitudes I exploit homicide data from the National Civil Police of El Salvador.7 To do so I compute homicide rates using total counts per 1,000 inhabitants by municipality-year (using 1992 population data). 7 Data available through the World Bank 12 Using respondent’s municipality of residence I match homicide rates by municipality-year. For the LAPOP sample, homicide rates have a mean value of 0.75 and standard deviation of 0.52 (with a maximum value of 3.1 homicides per 1,000 inhabitants). Crime statistics merged to the ometro sample are fairly similar: mean value of 0.8 and standard deviation of 0.57 (and Latinobar´ maximum of 4.2). I also take into account differences across municipalities and years in central government un- conditional transfers (FODES, by its acronym in Spanish) per capita received the year before the implementation of the survey.8 Data comes from the Instituto Salvadore˜ no de Desarrollo Munic- ipal.9 FODES represent the main financial resource for municipalities. The central government allocates 6% of the national budget to municipal governments, 80% of these funds are intended to fund investment and 20% expenses. Importantly, the allocation mechanism for these funds is based on population (50% ), equity (25%), poverty (20%), and territorial size (5%) of each mu- nicipality. For the LAPOP sample, FODES Transfer per 1,000 inhabitants (in 1992) has a mean value of USD 45, a standard deviation of USD 39, and maximum of USD 206. On the other hand, Latinobar´ ometro sample shows a mean value of USD 46, a standard deviation of USD 46, and maximum of USD 478. 3 Empirical Strategy ometero) from Exploiting several waves of two public opinion surveys (i.e., LAPOP and Latinobar´ 2004 to 2018, I estimate the following equation: yimt = βEconmt + γ Xi + θ Mmt + φm + ϕt + mt (1) where yimt is one of each of the reported attitudes by individual i, living in municipality m, and year t. The variable Econmt is the proxy of economic activity (NTL density) or per capita household income for a restricted sample. Both vary at the municipality level and over time. X is a vector of individual-level controls such as gender, age, years of schooling, rural and employment status.10 This set of individual-level controls is constrained by data availability and are included to improve the precision of the estimates. The vector Mmt includes two time-varying measures at the munici- pality level: annual homicide rates (per 1,000 inhabitants) and Central Government unconditional 8 I used 1992 population counts to normalize the amount of government transfers. 9 Retrieved at https://www.transparencia.gob.sv/institutions/isdem/documents/otra-informacion-de-interes 10 Since Latinobar´ometero does not survey individuals in rural areas, I use a “small town” indicator when perform- ing the analysis with this public opinion survey. 13 transfers (FODES) per capita received in t -111 The econometrics model includes municipality and year fixed effects. Thus, the empirical analysis explores the relationship between individuals’ at- titudes and deviations of the municipality income from the national average. Since the treatment comes at the municipality*year level the error term, ( m, t), is allowed to be correlated within a municipality*year. I also perform a heterogeneity analysis by restricting the same across different groups based on rural-urban status and weak vs strong state capacity municipalities.12 4 Results I organize results in two parts. First, I analyze data from LAPOP for the time period 2004-2018. Second, and for the same time period, I analyze data from Latinobar´ ometro. As discussed above the main difference between the two surveys is that the latter is exclusively urban while the former interview individuals in both urban and rural areas. In spite of that, the two samples do not appear to substantially differ in terms of other observable demographic characteristics. I use LAPOP to analyze the link between income at the municipality-year level and individu- als’ attitudes towards the president of El Salvador and the municipal government. I also study how attitudes towards other institutions such as national legislature, judicial system, mass media, and the Catholic Church vary with the business cycle. Finally, I explore whether these empirical asso- ciations mask any heterogeneity based on location characteristics (i.e., rural vs urban and strong vs weak state capacity of the municipality). ometro to explore the link between income (proxied by NTL density) Finally, I use Latinobar´ and perceptions of fairness in income distribution, generalized trust, views on country’s economic situation, and attitudes towards democracy. I also perform a heterogeneity analysis based on the strength of municipal state capacity. 4.1 LAPOP Analysis (2004-2018) Table 3 presents the main results on the relationship between income and trust in the president and the municipal government. Regardless of whether I use income data (for a smaller sample of municipalities in columns 1 to 3) or NTL density as proxy for income in an extended sample (in column 4), individuals’ trust towards the president positively correlates with higher level of development at the municipality level. Specifically, the empirical association between income and trust in the president is virtually unaffected by the inclusion of individual levels controls (column 11 I used 1992 population counts to normalize both the number of homicides and the amount of government transfers 12 A municipality is defined as weak (strong) state capacity if the average share of household with access to elec- tricity and water is below (above) 80 percent. 14 1 vs column 2) and statistically significant under the standard level of confidence ( p-value<0.05). To give a sense of the magnitude of the point estimate, an individual living in a municipality experiencing a one-standard deviation increase in the log of average per capita family income is 5 percentage points more likely to trust in the president. This increase represents 11% of the mean value of trust in the analyzed sample. Further, controlling for the level of crime and the size of federal funds received by the municipality of residence a year before the interview slightly increase the size of the point estimate (column 3). Four additional patterns emerge from specifications in columns 1 to 3 in Table 3. First, individual-level covariates such as gender, unemployment status, and age do not predict trust in the president.13 Second, individuals living in rural areas are roughly 3 percentage points more likely to trust in the president. Third, as years schooling increase, trust in government decreases. Fourth, higher homicide rates appear to erode trust in the president. Fourth, FODES transfer does not predict trust in the president. When I extend the sample of municipalities under study and use NTL density as proxy for in- come, I find very similar results. Noteworthy, the precision of the estimate for the main coefficient of interest increases ( p-value<0.01). After standardizing the implied effects, the magnitude of the coefficient in column 4 is slightly larger than in column 3: a one-standard deviation increase in NTL density is associated with a 0.17-standard deviation increase in trust towards the president. Put differently, individuals living in municipality having a one-standard deviation increase in NTL density are 8.2 percentage points more likely to trust in the Salvadoran president. In columns 5 and 6 of Table 3 I focus on trust in the municipal government. Average per capita family income at the municipality level is a strong predictor of trust in local government ( p-value<0.01). Individuals in municipalities experiencing a one-standard deviation increase in the log of average per capita family income are 3.5 percentage points more likely to trust in this institution. When extending the sample and using NTL density as main explanatory variable I find again a slightly larger standardized effect (albeit statistically weaker): a one-standard deviation increase in NTL density is associated with a 4.5 percentage points increase in the likelihood of trusting the municipal government. The results in columns 5 and 6 of Table 3 also shed light on other determinants of institutional trust: individuals living in rural areas tend to trust more whereas men, older and more educated individuals tend to trust less in local government. (Generally, more educated individuals tend to trust less regardless of the institutions.) Interestingly, the size of central government transfers (FODES) positively correlates with trust in municipal government. The magnitude of the latter effect is economically significant: a one-standard deviation increase in per capita central government transfer (i.e., USD 38.2 per inhabitant) is associated with a 7 percentage points increase in the likelihood of trusting the municipal government. 13 Using data from the World Value Surveys for 79 countries Algan and Cahuc (2013) show that age positively correlates with generalized trust across while gender does not predict trust. 15 Table 3: Income and Trust Towards President and Municipal Government Trust President Trust Municipal Government (1) (2) (3) (4) (5) (6) Average Per Capita Family Income (in logs) 0.145** 0.157** 0.206*** 0.083*** [0.072] [0.070] [0.074] [0.026] NTL Density at Municipality (in logs) 0.080*** 0.037* [0.025] [0.022] Male -0.005 -0.005 -0.004 -0.021* -0.016* [0.013] [0.013] [0.011] [0.012] [0.009] Age 0.003 0.003 0.002 -0.006*** -0.005*** [0.002] [0.002] [0.002] [0.002] [0.001] Age Squared -0.000 -0.000 0.000 0.000*** 0.000*** [0.000] [0.000] [0.000] [0.000] [0.000] Years of Schooling -0.006*** -0.006*** -0.005*** -0.009*** -0.007*** [0.002] [0.002] [0.001] [0.002] [0.001] Unemployed 0.011 0.010 0.007 -0.032 -0.019 [0.020] [0.020] [0.017] [0.022] [0.019] Living in Rural Area 0.047** 0.047** 0.039*** 0.030* 0.022* [0.019] [0.018] [0.014] [0.016] [0.012] Homicide Rates -0.036* -0.022* -0.025 -0.008 [0.021] [0.014] [0.021] [0.013] FODES transfer per inhab. (previous year) -0.002 -0.001 0.004*** 0.002*** [0.001] [0.001] [0.001] [0.001] Observations 6,056 5,930 5,930 8,751 7,438 12,031 Number of Municipalities 42 42 42 88 45 114 Number of Waves 6 6 6 6 8 8 R-squared 0.113 0.125 0.126 0.122 0.045 0.045 Notes: All regressions include municipality and year fixed effects. Robust standard errors clustered at the municipality- year level in brackets. FODES refers to Development Municipal Fund Economic and Social transfers. Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. 16 Table 4: Income and Trust in Other Institutions Trust in National Legislature Judicial System Media Church (1) (2) (3) (4) NTL Density at Municipality (in logs) 0.042* 0.137** 0.094*** 0.000 [0.022] [0.060] [0.020] [0.027] Homicide Rates -0.032** -0.095*** -0.042*** -0.008 [0.013] [0.028] [0.015] [0.013] FODES transfer per inhab. (previous year) 0.000 -0.002 0.002*** 0.002*** [0.001] [0.001] [0.001] [0.001] Observations 11,826 11,864 10,471 11,952 Number of Municipalities 114 114 114 114 Number of Waves 8 8 7 8 R-squared 0.068 0.092 0.068 0.057 Notes: All regressions include municipality and year fixed effects as well as individual controls (e.g., gender, age, age squared, and rural dummy). Robust standard errors clustered at the municipality-year level in brackets. FODES refers to Development Municipal Fund Economic and Social transfers. Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. In Table 4 I look at trust towards other institutions. Specifically, I study the empirical link between income (proxied by NTL density) and attitudes towards the national legislature, judicial system, mas media, and the Catholic Church. Income appears to be a strong predictor of confidence in the judicial system (column 2), trust in mass media (column 3), and to a lesser extent of trust in national legislature (column 1). These results are consistent with the one in Stevenson and Wolfers (2011) showing that higher unemployment rates in the US predict lower levels of trust in public institutions such as the congress and journalists. Interestingly, income does not predict trust in the Catholic Church (column 4). Another interesting pattern emerges from Table 4: homicide rates are negatively associated to trust in all the institutions but the Catholic Church. As an example, moving an individual from a municipality in the 10% percentile to another in the 90% percentile distribution of crime rates would be associated with an approximately 10 percentage point drop in the confidence that the judicial system would punish the guilty in case respondent were victim of a crime. Further, the size of central government transfers is positively and statistically associated to higher levels of trust in mass media and the Catholic Church wheres it is negatively associated to confidence in the judicial system. Finally, Table A.2 shows that all previous results hold when using the original ordered dependent variables and ordered probit models are estimated instead. Next, I study potential heterogeneous effects underlying the aforementioned empirical associ- ations. Results in Table 5 shows that most of the positive association between income and trust in the Salvadoran president is explained by individuals living in rural areas (column 1 in Panel A) and municipalities with low levels of state capacity (column 1 in Panel C) -as proxied by the share of households with access to electricity and water services. A similar pattern emerges when looking 17 Table 5: Income and Institutional Trust (heterogeneity) Trust in President Municipal Gov. National Leg. Judicial Syst. Media Church (1) (2) (3) (4) (5) (6) Panel A: Individuals Living in Rural Areas NTL Density at Municipality (in logs) 0.092*** 0.001 0.051 0.265*** 0.109*** -0.000 [0.035] [0.029] [0.032] [0.070] [0.023] [0.037] Observations 3,187 4,508 4,362 4,425 3,911 4,495 Number of Municipalities/Waves 75/6 101/8 101/8 101/8 101/7 101/7 R-squared 0.114 0.045 0.042 0.065 0.055 0.113 Panel B: Individuals Living in Urban Areas NTL Density at Municipality (in logs) 0.031 0.097** 0.030 -0.037 0.057 -0.049 [0.040] [0.047] [0.032] [0.080] [0.037] [0.042] Observations 5,564 7,523 7,464 7,439 6,560 7,457 Number of Municipalities/Waves 85/6 112/8 112/8 112/8 112/7 112/7 R-squared 0.135 0.051 0.077 0.086 0.078 0.051 Panel C: Individuals Living in Municipalities with Weak State Capacity NTL Density at Municipality (in logs) 0.094*** 0.031 0.036 0.187** 0.098*** 0.014 [0.029] [0.024] [0.028] [0.073] [0.021] [0.030] Observations 3,082 4,323 4,209 4,237 3,756 4,302 Number of Municipalities/Waves 47/6 64/8 64/8 64/8 64/7 64/8 R-squared 0.104 0.036 0.055 0.074 0.055 0.074 Panel D: Individuals Living in Municipalities with Strong State Capacity NTL Density at Municipality (in logs) 0.000 0.005 0.000 0.133 0.089** -0.063 [0.037] [0.051] [0.041] [0.095] [0.040] [0.045] Observations 5,669 7,708 7,617 7,627 6,715 7,650 Number of Municipalities/Waves 41/6 50/8 50/8 50/8 50 /7 50/8 R-squared 0.126 0.044 0.064 0.086 0.074 0.046 Notes: All regressions include individual-level controls (gender, age, age squared, years of schooling, unemployment status, and small town dummy), homicide rates, FODES transfers in previous periods, municipality and year fixed effects. Robust standard errors clustered at the municipality-year level in brackets. FODES refers to Development Municipal Fund Economic and Social transfers. A municipality is defined as weak (strong) state capacity if the average share of household with access to electricity and water services is below (above) 80 percent.Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. at confidence in the judicial system (see columns 4 in Panel A and C). Further, most of the income impact on trust in mass media is explained by individuals living in rural areas (see columns 5 in Panel A). On the contrary, the positive association between income and trust in the municipal gov- ernment comes from individuals living in urban areas (column 2 in Panel B). There does not seem to exist an important heterogeneity in the association of income and trust in the Catholic Church. 4.2 ´ LATINOBAROMETERO Analysis (2004-2018) Table 6 presents the main results on the relationship between income and a set of attitudinal vari- ables in Latinobar´ometro. Specifically, I look at perceptions of fairness in income distribution 18 Table 6: Income and Other Perceptions Income Distribution Generalized Economy is Democracy is Satisfied with Perceived as Unfair Trust Improving Better Democracy (1) (2) (3) (4) (5) Panel A: Full Sample NTL Density at Municipality (in logs) -0.066* -0.050 0.064** 0.067* 0.008 [0.035] [0.036] [0.027] [0.040] [0.034] Observations 12,580 12,128 11,433 11,542 11,817 Number of Municipalities/Waves 166/13 166/13 163/13 166/12 166/13 R-squared 0.481 0.054 0.046 0.064 0.100 Panel B: Individuals Living in Municipalities with Weak State Capacity NTL Density at Municipality (in logs) -0.017 -0.018 0.019 0.005 0.016 [0.047] [0.055] [0.039] [0.047] [0.050] Observations 4,470 4,288 4,116 4,004 4,130 Number of Municipalities/Waves 100/13 100/13 99/12 100/13 100/13 R-squared 0.445 0.068 0.070 0.077 0.123 Panel C: Individuals Living in Municipalities with Strong State Capacity NTL Density at Municipality (in logs) -0.068 -0.119** 0.073* 0.096 -0.053 [0.056] [0.052] [0.043] [0.075] [0.055] Observations 8,110 7,840 7,317 7,538 7,687 Number of Municipalities/Waves 66/13 66/13 64/13 100/13 100/13 R-squared 0.502 0.052 0.036 0.061 0.093 Notes: All regressions include individual-level controls (gender, age, age squared, unemployment status, and small town dummy), homicide rates, FODES transfers in previous periods, municipality and year fixed effects. Robust standard errors clustered at the municipality-year level in brackets. FODES refers to Development Municipal Fund Economic and Social transfers. A municipality is defined as weak (strong) state capacity if the average share of house- hold with access to electricity and water services is below (above) 80 percent.Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. (column 1), generalized trust (column 2), views on country’s economic situation (column 3), and attitudes towards democracy (columns 4 and 5). Panel A shows the results for the whole sam- ple whereas panels B and C show the results for weak and strong state capacity municipalities, respectively. Result from column 1 in Panel A of Table 6 suggests that individuals living in municipality experiencing improving economic conditions in the current year tend to be less likely to report that income distribution in the country is unfair. This effect is remarkably absent in locations with weak state capacity (column 1 of Panel B) and exclusively explained by individuals living in location with strong state capacity (Panel C). The association is not economically large though. To convey the main magnitude better: one-standard deviation increase in our proxy for economic development at the municipality level is associated with 6 percentage points (0.12 standard deviation of the outcome variable) decrease in the probability of assessing income distribution as unfair. Unlike findings for the Russian Federation documented by Ananyev and Guriev (2018), I do not find that income is positively associated with generalized trust (column 2 in Panel A). If any- 19 thing, the association is negative for the case of individuals living in municipalities with stronger provision of basic public goods (column 2 in Panel C). Perhaps unsurprisingly, individuals living in municipalities experiencing improving economic conditions do consider that the economic sit- uation of the country at the moment of the interview was better than a year before (column 3). This association is exclusively explained by responses from individuals living in municipalities characterized as “strong state capacity” (column 3 in Panel C). I next analyze how individuals’ attitudes toward democracy respond to variation in income at the municipality level. Higher income levels in the municipality are associated with preferences for democracy over other types of government (column 4). This effect is essentially explained by individuals living strong state capacity (albeit standard errors for the implied effect are large, see column 4 in Panel C). Further, income does not appear to shape people’s satisfaction with democracy (column 5). 5 Conclusion Trust and confidence in most Salvadoran institutions have declined over time. This decline started in the aftermath of the 2008 global financial crisis and has particularly accelerated over recent years, coinciding with a path of low economic growth. In this paper, I study whether individuals’ attitudes toward Salvadoran institutions are shaped by respondents’ municipality-specific business cycles. Exploiting multiple waves of two opinion surveys and a two-way fixed effect model, I find that individuals’ trust towards both the president and the municipal government are positively associated with a higher level of income at the municipality level. These documented empirical relationships mask relevant heterogeneity from a rural-urban divide and differences in municipal state capacity. Most of the positive association between income and trust in the Salvadoran pres- ident is explained by individuals living in rural areas and municipalities with low levels of state capacity, whereas most of the positive association between income and trust in the municipal gov- ernment comes from individuals living in urban areas. I also find that municipality-specific business cycles strongly predicts trust in mass media and, to a lesser extent, trust in the national legislature and confidence in the judicial system. Interest- ingly, income does not affect trust in the Catholic Church. Moreover, views of income distribution fairness and preferences for democracy over other types of government are positively shaped by improving economic conditions at the municipality level. On the contrary, neither generalized trust nor satisfaction with democracy is empirically associated with the business cycle. Except for general attitudes towards the government, the implied associations between income at the municipality level and attitudes towards institutions do not seem to emanate from respon- dents’ own economic situation. That is, respondent’s unemployment status does not predict trust 20 in most of the institutions analyzed in this paper. There is a strong complementary between well-functioning public institutions and individuals’ trust (Papaioannou, 2013). Further, trust in and support for public institutions are crucial for the implementation of reforms. While an important caveat of the present analysis is the lack of ex- perimental variation necessary to draw causal statements, the foregoing results suggest important implications for policymakers. First, Salvadoran policy makers should closely monitor attitudes towards institutions as they do get affected by economic shocks. Second, while policies aiming at reinforcing or mitigating the impact of short-run business-cycles on trust are essential, Salvadoran policymakers should consider implementing specific policies that can foster persistent and higher levels of generalized trust and confidence in institutions in particular. This is particularly important since generalized trust levels are at their historical lows in El Salvador, which would likely impose a constraint on the sustainability of structural reforms. In this regard, education policies aiming at increasing average schooling may constitute great vehicles to improve institutional trust and civic engagement in the long run (Papaioannou, 2013). To conclude, the empirical associations docu- mented in this paper should be regarded as exploratory since further empirical analysis is required to identify causal relationships between aggregate income and individuals’ level of trust within El Salvador. 21 References Alesina, A. and Fuchs-Sch¨undeln, N. (2007). Good-bye lenin (or not?): The effect of communism on people’s preferences. American Economic Review, 97(4):1507–1528. Algan, Y. and Cahuc, P. (2013). Trust and growth. Annual Review of Economics, 5(1):521–549. Ananyev, M. and Guriev, S. (2018). Effect of Income on Trust: Evidence from the 2009 Economic Crisis in Russia. The Economic Journal, 129(619):1082–1118. Buggle, J. C. and Durante, R. (2021). Climate Risk, Cooperation and the Co-Evolution of Culture and Institutions. The Economic Journal, 131(637):1947–1987. Consejo Centroamericano de Procuradores de los Derechos Humanos (2008). Pol´ ublicas ıticas p´ on de la pobreza en Centroam´ regionales sobre la reducci´ erica y su incidencia en el pleno dis- frute de los derechos humanos: Informe nacional El Salvador. Inter-American Institute of Human Rights. on Latinobar´ Corporaci´ ometro (2018). Informe 2018. Guiso, L., Sapienza, P., and Zingales, L. (2016). Long-Term Persistence. Journal of the European Economic Association, 14(6):1401–1436. Henderson, J. V., Storeygard, A., and Weil, D. N. (2012). Measuring economic growth from outer space. American Economic Review, 102(2):994–1028. Li, X., Zhou, Y., Zhao, M., and Zhao, X. (2020). A harmonized global nighttime light dataset 1992–2018. Sci Data, 7(1):168. Nunn, N. and Wantchekon, L. (2011). The slave trade and the origins of mistrust in africa. Ameri- can Economic Review, 101(7):3221–3252. Papaioannou, E. (2013). Trust(ing) in europe? how increased social capital can contribute to economic development. European View, 12(2):323–323. Putnam, R. D. (1995). Bowling alone: America’s declining social capital. Journal of Democracy, 6(1):65–78. Reyes, G. and Gasparini, L. (2021). Are Fairness Perceptions Shaped by Income Inequality? Evidence from Latin America. Stevenson, B. and Wolfers, J. (2011). Trust in public institutions over the business cycle. American Economic Review, 101(3):281–87. []article 22 APPENDIX Institutional Trust, Perceptions of Distributive Unfairness, and Income Across Salvadoran Municipalities Emilio Depetris-Chauvin i Figure A.1: Regional Variation (LAPOP) 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Central I Central II Central I Central II Occidental Oriental Occidental Oriental (a) Trust in President (b) Trust in Municipal Gov. 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Central I Central II Central I Central II Occidental Oriental Occidental Oriental (c) Trust in National Legislature (d) Confidence in Judicial Syst. 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Central I Central II Central I Central II Occidental Oriental Occidental Oriental (e) Trust in Media (f) Trust in Church ii Figure A.2: Regional Variation(Latinobarometer) 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Central I Central II Central I Central II Occidental Oriental Occidental Oriental (a) Unfair Distribution (b) Generalized Trust 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 2004 2006 2008 2010 2012 2014 2016 2018 2004 2006 2008 2010 2012 2014 2016 2018 year year Central I Central II Central I Central II Occidental Oriental Occidental Oriental (c) Democracy is Better (d) Satisfied with Democracy iii Figure A.3: NTL Density as Proxy for Development 6 6 Per Capita Household Income in 2001 (in logs) Per Capita Household Income in 2010 (in logs) 5.5 5.5 4.5 5 5 3.5 4 4.5 0 1 2 3 4 0 1 2 3 4 Night Time Light Density in 2001 (in logs) Night Time Light Density in 2010 (in logs) (a) NTL Density and Income in 2001 (b) NTL Density and Income in 2010 100 6.5 Per Capita Household Income in 2018 (in logs) 80 6 Poverty Rate in 2004 60 5.5 40 5 20 4.5 0 2 2.5 3 3.5 4 -4 -2 0 2 4 Night Time Light Density in 2018 (in logs) Night Time Light Density in 2004 (in logs) (c) NTL Density and Income in 2018 (d) NTL Density and Poverty in 2004 Notes: Income measures per capita household income from EHPM (Harmonized by SEDLAC). Size of plots denotes population counts in 2007. iv Table A.1: NTL density and Poverty Household Poverty Rate in 2004 (1) (2) (3) (4) NTL Density at Municipality (in logs) -9.547*** -8.560*** -9.630*** -8.620*** [0.827] [0.738] [0.852] [0.758] years used to compute NTL 2004 2004-2003 2004 2004-2003 Sample All Excluding Very Low Lit Observations 255 255 254 254 R-squared 0.43 0.439 0.428 0.437 Notes: Robust standard errors in brackets. Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. Poverty rates comes from Consejo Centroamericano de Procuradores de los Derechos Humanos (2008) Table A.2: Income and Institutional Trust (Ordered Probit) Trust in President Municipal Gov. National Leg. Judicial Syst. Media Church (1) (2) (3) (4) (5) (6) NTL Density at Municipality (in logs) 0.138*** 0.113** 0.119** 0.138** 0.177*** -0.003 [0.053] [0.055] [0.049] [0.061] [0.043] [0.080] Observations 8,751 12,031 11,826 11,864 10,471 11,952 Number of Municipalities/Waves 75/6 101/8 101/8 101/8 101/7 101/7 Notes: Ordered Probit estimates. All regressions include individual-level controls (gender, age, age squared, years of schooling, unemployment status, and small town dummy), homicide rates, FODES transfers in previous periods, municipality and year fixed effects. Robust standard errors clustered at the municipality-year level in brackets. FODES refers to Development Municipal Fund Economic and Social transfers. Statistical significance: *** p <0.01, ** p <0.05, * p <0.1. v