Disaster and political trust: A natural experiment from the 2017 Mexico City earthquake

Frost, Margaret; Kim, Sangeun; Scartascini, Carlos G.; Zamora, Paula; Zechmeister, Elizabeth J. Working Paper Disaster and political trust: A natural experiment from the 2017 Mexico City earthquake IDB Working Paper Series, No. IDB-WP-1192 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Frost, Margaret; Kim, Sangeun; Scartascini, Carlos G.; Zamora, Paula; Zechmeister, Elizabeth J. (2024) : Disaster and political trust: A natural experiment from the 2017 Mexico City earthquake, IDB Working Paper Series, No. IDB-WP-1192, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0005511 This Version is available at: https://hdl.handle.net/10419/289909 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode Disaster and Political Trust: A Natural Experiment from the 2017 Mexico City Earthquake Margaret Frost Sangeun Kim Carlos Scartascini Paula Zamora Elizabeth J. Zechmeister IDB WORKING PAPER SERIES Nº IDB-WP-1192 January 2024 Department of Research and Chief Economist Inter-American Development Bank January 2024 Disaster and Political Trust: A Natural Experiment from the 2017 Mexico City Earthquake Margaret Frost* Sangeun Kim** Carlos Scartascini*** Paula Zamora**** Elizabeth J. Zechmeister***** * University of Rhode Island ** Aspen Institute *** Inter-American Development Bank ‡ Texas A&M University ‡‡ Vanderbilt University Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Disaster and political trust: a natural experiment from the 2017 Mexico City earthquake / Margaret Frost, Sangeun Kim, Carlos Scartascini, Paula Zamora, Elizabeth J. Zechmeister. p. cm. — (IDB Working Paper Series ; 1192) Includes bibliographic references. 1. Earthquake relief-Political aspects-Mexico-Econometric models. 2. Natural disasters-Political aspects-Mexico-Econometric models. 3. Trust-Political aspectsMexico-Econometric models. I. Frost, Margaret. II. Kim, Sangeun. III. Scartascini, Carlos G., 1971IV. Zamora, Paula. V. Zechmeister, Elizabeth J., 1972VI. InterA merican Development Bank. Department of Research and Chief Economist. VII. Series. IDB-WP-1192 Copyright © Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. http://www.iadb.org 2024 Abstract Political trust is foundational to democratic legitimacy, representative governance, and the provision of effective public policy. Various shocks can influence this trust, steering countries onto positive or negative trajectories. This study examines whether natural disasters can impact general political trust and if disaster relief efforts can mitigate these effects. We investigate the relationships between disaster, trust, and aid using novel survey data collected before and after a 7.1-magnitude earthquake struck Mexico City in September 2017. Our findings reveal that the disaster resulted in an 11% decrease in general political trust. Additionally, we demonstrate that geographical proximity to disaster relief efforts may counterbalance this decline in trust. This study contributes to the scholarship on the politics of disasters and offers policy implications, highlighting the role of disaster assistance in potentially restoring general political trust after a disaster. JEL classifications: H84, Q54, D72, Z13 Keywords: Political trust, Natural disaster, Natural experiment, Aid relief, Development Data collection was IRB approved by the Behavioral Sciences Committee of Vanderbilt University’s Institutional Review Board. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Neither the IDB nor any other party had any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The information and opinions presented herein are entirely those of the authors, and no endorsement by the Inter-American Development Bank, its Board of Executive Directors, or the countries they represent is expressed or implied. 1 Introduction Political trust is essential to the well-functioning of democracy. Representative democracy rests on delegation, which requires trust (Urbinati and Warren, 2008). Trust increases public demand for welfare-enhancing policies and facilitates social cohesion, which in turn enhances trust (Keefer and Scartascini, 2022), and trust is crucial in times of disaster when privatepublic collaboration is necessary for effective recovery (Aldrich, 2017). It matters, then, if disasters undermine trust in public officials precisely when it is most needed. Scholarship on post-disaster trust levels has most often focused on incumbents and/or specific offices (Reinhardt, 2015,2019), governments (Han et al., 2011; Nicholls and Picou, 2013), and social trust (Bai and Li, 2021; Becchetti et al., 2017; Carlin et al., 2014b; Stephane, 2021). We add to this body of research by addressing the effect of disaster on general political trust, and the conditioning role of disaster relief, in a comparatively less developed context. Our core contention is that large-scale disasters lower general political trust—meaning, individuals’ confidence in the reliability and integrity of public officials. We argue trust falls post-disaster via two paths: stressing state capacity and increasing avenues for corruption. First, disasters signal competence and may overwhelm state capacity. Subsequent negative experiences ought to affect assessments of government benevolence, competence, and the ability to deliver to citizens (Olson and Gawronski, 2010). Additionally, the chaotic aftermath of a disaster and the need for fast disbursements create opportunities for corruption (Nikolova and Marinov, 2017; Yamamura, 2014) at a time when the public is highly sensitive to malfeasance (Gawronski and Olson, 2000). Therefore, especially where weak institutions prevail, general political trust would decline in the aftermath of a major natural disaster. But can this fall in trust be prevented? A promising avenue for theorizing on the relationship between disaster and general political trust comes from research suggesting that the provision of welfare mitigates against negative turns in political evaluations (Lazarev et al., 2014) and social or interpersonal trust (Carlin et al., 2014b; De Juan et al., 2020; Petrova and Rosvold, 2024). In theory, aid carries the potential for the state to demonstrate capacity and beneficence. Some scholars have shown that aid can generate incumbent support (Healy and Malhotra, 2009; Bechtel and Hainmueller, 2011; Gallego, 2018; Lazarev et al., 2014). Others, however, find that aid is insufficient to counter increases in negative political evaluations, even when distributed fairly (Cole et al., 2012; Heersink et al., 2017). Moreover, aid inflows can be mishandled (Leeson and Sobel, 2008; Yamamura, 2014) or ineffectively distributed (Eichenauer et al., 2020; Francken et al., 2012; Sobel and Leeson, 2006), leading to declines in political evaluations. In short, whether aid can mitigate against currents that 2 diminish trust is an open question in need of empirical testing. We test relationships among disaster, trust, and aid using data gathered immediately before and after a 7.1-magnitude earthquake struck Mexico City on September 19, 2017. The earthquake killed 369 people and injured approximately 2,000 (Andone et al.,2017;United States Geological Survey,2017;Villegas and Ahmed,2017). To assess the consequences of disaster for general political trust, we employ a novel design based on two rounds of surveys: one conducted immediately before (569 observations) and another two months after the earthquake (1,164 observations) in the greater Mexico City metropolitan region. These 1,164 individuals were surveyed to match the pre-earthquake sample, with each preearthquake individual matched to at least two comparable (on location, gender, and age) post-earthquake individuals. Our first objective is to assess the effect of the earthquake on political trust, measured via a composite of four questions on confidence in politicians and civil servants. Although running a randomized controlled trial of exposure to an earthquake is not possible, the matched (by design) sample supports a causal interpretation of the analyses under a set of assumptions: that the approach creates two otherwise homogeneous groups and earthquake damage is independent of pre-disaster levels of trust.1We further test the robustness of our initial results via statistical matching. We conclude that the disaster event negatively affected general political trust: the earthquake caused a substantial drop of 11% in general political trust. We then consider the question of disaster relief: average results may mask substantial heterogeneity if those who receive disaster relief have their general political trust restored by this demonstration of state effectiveness and benevolence. We run analyses with two measures of access to aid: subjective awareness and objective location data. We find a positive, significant relationship between trust and perceived proximity to distribution centers for food, water, and other essential items. Precisely, those who reported having access to such assistance reported relatively higher levels of trust. Although not as significant, we also find a positive relationship for analyses of actual proximity to a distribution center. Both measures have limitations. Although we control for political variables at the municipality level before the earthquake, subjective reports may be endogenous to factors we are not considering. While we scoured the web to find aid center data, the list of actual locations is likely incomplete. We conclude that the results provide suggestive evidence that aid can counter the adverse effects of a disaster on general political trust. 1We also assume no relevant compositional changes in the population in surveyed areas. 3 The paper makes four contributions. First, unlike previous studies focusing on particular political figures or offices, we examine the connection between disaster and general political trust. This scope is relevant because, once low, general political trust is more difficult to restore than trust in specific individuals’/offices’ handling of a disaster (Levi and Stoker, 2000). Second, we add strong evidence that disaster decreases trust using a unique dataset with advantageous timing and survey design in which we matched post-disaster respondents to those in a pre-disaster survey. Third, we extend scholarship on the consequences of postdisaster welfare provision beyond their implications for incumbent support and social trust to the domain of general political trust. We show that, even in a context in which efficient and graft-free aid distribution is a challenge, the establishment of distribution centers—often through public-private or NGO partnerships—contains the potential to counter declines in general political trust. Fourth, we extend the scope of studies on disasters and trust from the United States and Western Europe to a comparatively less developed context. In this context, baseline general trust is low, and corruption, slow delivery, and the potential coopting of aid delivery by third parties often complicate aid distribution. Concentrating on a developing country is important because most prior studies have focused on contexts where general trust is moderate to high, as is the state’s capacity—albeit perhaps not its willingness or organization—to deliver aid quickly and efficiently. 2 Analytical Framework Most studies of post-disaster political trust that use systematically-collected empirical data concentrate on public confidence in incumbents in specific government offices (Reinhardt, 2015,2019), government at different levels (Han et al., 2011; Nicholls and Picou, 2013), and society (Carlin et al., 2014b). Table A.1 in the Online Appendix Asummarizes most of these studies despite not being an exhaustive list. Of these, Albrecht (2017) is closest to our study, as it examines trust in general politicians with multiple rounds of cross-country surveys and, interestingly, finds no evidence of a connection between disaster and political trust. Nevertheless, this cross-national study includes comparatively minor disasters (e.g., extreme temperatures) that may have more marginal implications for trust. According to Hardin (1993, p.506), trust is defined as a three-part relationship where “A” trusts “B” to perform “X.” Our research expands upon existing literature by considering a broader scope for “B” and “X.” Specifically, we investigate the public’s general trust in politicians and civil servants (B) concerning their commitment to keeping promises— 4 reliability—and compliance with the law—integrity (X). In essence, our focus centers on the public’s general political trust, where they place their trust in the general political community (public officials) in matters of core responsibilities that include but also extend beyond disaster relief. We fuse several lines of scholarship into a unified framework on the connections between disaster and general political trust. We detail a micro-logic that supports the expectation that general political trust will tend to fall after a major disaster. Then, we consider whether providing disaster aid from any source may mitigate that drop. In this case, we state an open expectation. On the one hand, aid availability should restore perceptions of state competence. The literature has shown that incumbents are rewarded for aid provision. However, because aid can be corrupted and ineffective, particularly in developing countries, the connection between disaster aid and general trust is far from guaranteed. Our core thesis is that disasters place stress on political trust due to two non-rival dynamics. The first relates to the dimension of trust concerning competence (Levi and Stoker, 2000). Natural disasters overload systems, placing a significant burden on state capacity. Disasters disrupt economies, disturb infrastructure, and disorganize bureaucracies, while they multiply societal demands (Drury and Olson, 1998; Olson, 2000; Schneider, 1992). With limited time horizons and a motivation to cater to public opinion, governments tend to under-invest in preparation (Healy and Malhotra, 2009). The combination of pre-disaster under-preparedness and post-disaster disruptions and demands risks leaving states vulnerable to appearing incompetent following natural disasters (Olson and Gawronski, 2010). The second dynamic relates to the state’s vulnerability to corruption and accusations of corruption following disaster. Perceptions of pervasive corruption matter because they signal a lack of beneficence, which may shape political trust (Chang and Chu, 2006). Widespread corruption also undermines political trust by hampering the state’s capacity to deliver resources to the people in need (Lavallée et al., 2008). Trustworthy agents do not act opportunistically even when it is beneficial for them (Keefer et al., 2020). Nevertheless, natural disasters can potentially increase opportunity and demand for corruption, especially in developing country contexts (Nikolova and Marinov, 2017; Yamamura, 2014). Research reveals that individuals are highly sensitive to corruption in bad times, including following disasters (Gawronski and Olson, 2000; Olson and Gawronski, 2010; Zechmeister and ZizumboColunga, 2013). In brief, to the extent that corruption seeps into post-disaster dynamics and perceptions, there is reason to suspect that political trust will decline after a disaster. Declines in trust’s two core dimensions—increased burden and vulnerability to corruption— 5 who had no confidence in politicians and civil servants before and after the earthquake. Less than 25% of respondents thought that politicians and civil servants fulfilled their promises or complied with the law before the earthquake. This fact not only offers a glimpse into the low levels of trust in the country but may also generate a ceiling effect in our estimations for reduced trust. For our core analysis, we conducted Principal Component Analysis (PCA) using the four survey items: i) fulfillment of politicians’ promises, ii) fulfillment of civil servants’ promises, iii) law compliance of politicians; and 4) law compliance of civil servants. All four variables exhibit a high degree of positive correlation on the first component.11 Subsequently, we use it as a latent variable for “general political trust.” Figure 2: PCA: Trust before and after the Earthquake 0 .5 1 1.5 2 2.5 Density 0 .2 .4 .6 .8 1 “Pre” “Post” Figure 2shows the distribution of the dependent variable before and after the earthquake once the first component of the PCA is taken as the general political trust variable and is normalized. The latent variable ranges from low (0) to high (1) trust levels. The high concentration of post-earthquake respondents’ answers on the left-hand side of the distribution indicates that more people reported having “very low" general political trust after the earthquake than before the event. and after the earthquake (Wilcoxon rank-sum test: p-value = 0.020). 11See Section D.2 in the Online Appendix for details regarding the Principal Component Analysis (PCA). 12 Table 1: Effect of the Earthquake on General Political Trust General political trust (1) (2) (3) (4) (5) (6) Earthquake -0.029*** -0.029*** -0.029*** -0.029*** -0.029*** -0.029*** (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Male -0.000 0.001 -0.000 -0.000 -0.000 0.001 (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Age -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Last year of education approved -0.002 -0.002 -0.002 -0.002 -0.002 -0.002* (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Number of adults in the household -0.001 -0.001 -0.001 -0.001 -0.001 -0.002 (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Number of children in the household 0.002 0.002 0.002 0.002 0.002 0.002 (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) PRI: pre 0.031 0.069 (0.056) (0.073) PRD: pre 0.146*** 0.167*** (0.053) (0.057) Wealth: pre 0.013 0.087 (0.081) (0.118) Risk Aversion: pre 0.003 -0.002 (0.004) (0.003) Patience: pre 0.001 0.000 (0.003) (0.003) Social Trust: pre -0.023 -0.017 (0.036) (0.042) Constant 0.340*** 0.314*** 0.333*** 0.238** 0.405*** 0.358* (0.028) (0.031) (0.049) (0.120) (0.103) (0.196) Observations 1,651 1,651 1,651 1,651 1,651 1,651 R-squared 0.039 0.043 0.039 0.039 0.039 0.044 Notes: Clustered standard errors are in parentheses. ∗p < 0.10,∗∗p < 0.05,∗∗∗p < 0.01. The average levels of political affiliation, risk aversion, patience, and social trust are calculated at the municipality level before the earthquake. We estimate the effect of the earthquake on the general political trust variable using ordinary least squares. Table 1presents the results and supports the conclusion that trust substantially decreased after the earthquake, independently of the set of controls included in the regressions. The first model controls for variables unlikely to be affected by the earthquake (i.e., gender, age, education, and number of adults and children per household) to improve efficiency. Models (2) - (5) include political affiliation, wealth, risk aversion, patience, and social trust. Given that the earthquake may affect these variables, we use pretreatment averages at the municipality level.12 For political affiliation, we focus on the two 12We considered averaging by clusters, but each cluster included only about 10 respondents, while mu13 main parties in this subregion: PRI (Institutional Revolutionary Party) and PRD (Party of the Democratic Revolution). For example, suppose 60% of the respondents in a given municipality supported PRI before the earthquake. In that case, we assign this percentage to respondents who resided in the municipality during the post-earthquake survey. All models control for regional indicator variables. Therefore, the results support H1: the earthquake reduced the level of general political trust.13 The estimated effect of the earthquake on general trust remains the same: -0.029, which is equivalent to a drop of 11% (0.029/0.26). Interestingly, the table also shows a positive and significant correlation between pre-earthquake support for the party of the head of the Federal District/Mexico City at the time of the earthquake (PRD: pre) and the effect on general political trust. What could be driving the drop in trust? We mentioned two paths, one involving state capacity and another one involving corruption. We do not have micro-level data to test these mechanisms in detail. However, we do have data on blame attribution. Perceptions of government credibility and integrity ought to be affected if citizens blame public actors for the disaster’s occurrence. The post-earthquake survey shows that 62.2% of respondents believed the earthquake damage might have been averted if building regulations had been enforced. 6 Disaster Relief’s Role in Rebuilding Political Trust Thus far, the evidence supports the hypothesis that severe disasters undermine political trust. However, the results presented for the 2017 Mexico earthquake may mask whether disaster relief in the aftermath of the earthquake, particularly aid distribution centers, positively affected political trust. If that were the case, there are two important implications. First, disaster relief would countervail the impact of the earthquake. Second, the results from the test of H1 would be a lower bound of the actual effect; without aid, the drop in trust would have been higher. Thus, this section focuses on one of the most common forms of postearthquake aid: distribution centers (tables and tents) set up in affected neighborhoods for individuals to pick up essential goods (e.g., water and food). We investigate the relevance of distribution centers via how close they were to respondents’ households in two ways: i) nicipalities averaged about 60 survey respondents. 13Additional analyses show that the result holds when controlling for the actual values of political support, wealth, patience, risk aversion, and social trust without averaging at the municipality level. 14 respondents’ self-reports about the presence of distribution centers in their neighborhood, and ii) gathering the location data of distribution centers. We note that both measures have strengths and weaknesses. Subjective assessments are helpful because they measure awareness of the resource. If a distribution center exists, but an individual is not aware of it, in theory, it should not affect attitudes. However, subjective assessments may be endogenous to factors that shape political trust or subject to expectations to receive additional aid. We control for some potential confounders but cannot eliminate this threat. Objective indicators avoid subjective biases but have two challenges: individuals may not be aware of them, and it is impossible to develop a complete list of all aid distribution locations. The latter is because, in the chaotic aftermath of a severe disaster like that in Mexico, aid stations vary in size and duration, they are set up by many different actors, and there is no single authoritative list of the placement and timing of all centers. We assess the evidence offered by these measures with these caveats in mind. 15 6.1 Self-reported Proximity to Distribution Centers Table 2: Association between Self-reported Distribution Centers and General Political Trust General political trust (1) (2) (3) (4) (5) (6) Self-reported DC 0.032** 0.030** 0.032** 0.032** 0.032** 0.030** (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Male 0.003 0.004 0.003 0.003 0.003 0.004 (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) Age -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Last year of education approved -0.001 -0.001 -0.001 -0.001 -0.001 -0.002 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Number of adults in the household 0.001 0.001 0.001 0.001 0.001 0.001 (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Number of children in the household 0.003 0.003 0.002 0.003 0.003 0.003 (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) PRI:pre 0.034 0.054 (0.065) (0.087) PRD:pre 0.148** 0.192*** (0.065) (0.071) Wealth: pre -0.057 0.014 (0.087) (0.143) Risk Aversion: pre -0.002 -0.006 (0.005) (0.005) Patience: pre 0.001 0.000 (0.003) (0.003) Social Trust: pre 0.012 -0.008 (0.037) (0.047) Constant 0.284*** 0.260*** 0.316*** 0.318* 0.252** 0.440* (0.035) (0.037) (0.056) (0.163) (0.106) (0.254) Observations 1,069 1,069 1,069 1,069 1,069 1,069 R-squared 0.056 0.061 0.057 0.057 0.056 0.063 Notes: Clustered standard errors are in parentheses. ∗p < 0.10,∗∗p < 0.05,∗∗∗p < 0.01. The average levels of political affiliation, risk aversion, patience, and social trust are calculated at the municipality level before the earthquake. Table 2reports the results of self-reported distribution centers on general political trust. This table includes the same control variables and regional fixed effects in the previous section. We conduct this analysis only for post-earthquake respondents (proximity to distribution center information is irrelevant before the earthquake). We also note that the “treatment” (i.e., distribution center) is no longer assumed to have been assigned randomly. The table shows that respondents who reported having (at least) one distribution center in their neighborhood 16 reported higher general political trust in politicians and civil servants, and its effect size is sizeable (ranging from 0.030 to 0.032, it represents an increase of about 12% with respect to the pre-earthquake mean). As the average treatment effect of the earthquake was -0.029, those who reported observing a distribution center in their neighborhood would appear to have had their general political trust restored to the pre-earthquake level. 6.2 Proximity to the Objective Location of Distribution Centers While the self-reports on distribution centers capture the awareness of the availability of aid, they may be confounded with respondents’ prior beliefs about the politicians and civil servants (reverse causality: those who trust more report having seen a distribution center.) There might also be self-selection—those who needed them more took the extra steps to locate and use them. To address these concerns, we test the robustness of the results using an objective measure. To locate distribution centers, we gathered and cross-referenced information from government, newspaper, and social media sources reporting on distribution centers established in Mexico City in the wake of the earthquake (see section E.1 for detailed sources). These sources provided partial identifying information for each distribution center, such as the name, provider, and, in some cases, latitude and longitude of a distribution center. Some information was lacking, such as the street name, neighborhood name, municipality name, and zone number. We filled in the missing information for each distribution center using publicly available maps and government documents on postal information and zoning in Mexico City. We checked each observation twice, using two different maps. We took most of the primary source materials about distribution centers from the Mexico City government website and the Collaborative Map “Rescue Mexico,” which then were corroborated with news articles from El Universal,Animal Político, and Mexican public institutions’ Twitter feeds.14 We could not locate similar data on distribution centers for the State of Mexico, so our analysis here is limited to Mexico City and, thus, to fewer survey respondents. Figure 3 shows the location of the distribution centers in Mexico City identified through our scouring of available information. 14According to most sources, this collaborative map was the most used and accurate source of information after the earthquake. 17 Figure 3: Location of Distribution Centers in Mexico City Source: Authors’ compilation. Notes: The gray lines indicate the boundaries of the colonias (“neighborhoods” in Mexico City). The diamonds represent the distribution centers found. We determined whether a distribution center was in or near a respondent’s neighborhood if the difference in zip codes was smaller than 30 and in the same municipality (for example, 77810 was the zip code of the place where the distribution was located and 77830 was the code of the respondent’s residence).15 Respondents were assigned a value of 1 for having a distribution center in their neighborhood and 0 otherwise. We arrived at 30 for the cut-off via an inductive approach, where we estimated the distance between distribution centers and survey respondents’ houses in kilometers and walking time in minutes using their geographical coordinates.16 The summary statistics in Table 3demonstrate that respondents 15We use the neighborhood and not the municipality because we want to match the question asked in the survey. See Online Appendix C. 16We calculated the distance between each household and its closest distribution centers. After obtaining the information from Google Maps, we checked the correspondence of the addresses and the ZIP codes using the official tables of ZIP codes and human settlements provided by the Mexican government, which can be found at https://datos.gob.mx/busca/dataset/codigos-postales-coordenadas-y-colonias/ resource/7675f085-6a8f-4b20-8091-ff5117fe964e. 18 coded to be residing near a distribution center live near the centers by the distance measures. Table 3: Summary Time and Distance between Distribution Center and Households Mean S.D. Min p25 Median p75 Max Time (minutes) 21.21 15.10 1.10 10.70 15.13 32.30 50.30 Distance (Km) 1.66 1.21 0.10 0.83 1.12 2.56 4.01 Table 4reports the results of objective proximity to distribution centers, as opposed to perceived closeness to the centers, on general political trust. The regressions remain the same as before, except that the variable of interest now captures proximity to a verified location of distribution centers. The table shows that having a distribution center nearby positively correlates with respondents’ general political trust. Notably, additional analyses demonstrate that whether the government provides the distribution center or not does not affect the conclusion. The coefficient is larger but less precisely estimated than in Table 2. Since the sample size has reduced substantially, it lowers the statistical power to detect a significant and consistent effect. In the most parsimonious model shown in column 1, the coefficient for the distribution center is significant at the 10% level. However, the significance varies depending on the controls we include.17 Considering these results and the prior ones that use the self-reported existence of distribution centers, we conclude that there is suggestive evidence supporting H2a. 17We also ran the regressions using different coding schemes for the Objective DC variable by relaxing the zip code differences to be 50 and 100. The statistical significance vanishes with the alternative variables, although the positive correlations remain. However, the two alternative coding schemes increased the distance between survey respondents’ residences and distribution centers to about 30 and 40 minutes, meaning that the distribution centers were potentially not precisely in the neighborhood. 19 Table 4: Association between Objective Distribution Centers and General Political Trust General political trust (1) (2) (3) (4) (5) (6) Objective DC 0.045* 0.042* 0.045* 0.039 0.045* 0.026 (0.025) (0.024) (0.024) (0.024) (0.025) (0.025) Male -0.002 -0.002 -0.002 -0.002 -0.002 -0.002 (0.016) (0.016) (0.016) (0.016) (0.016) (0.016) Age -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Last year of education approved -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Number of adults in the household -0.001 -0.001 -0.001 -0.002 -0.001 -0.002 (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) Number of children in the household -0.003 -0.004 -0.003 -0.003 -0.003 -0.003 (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) PRI:pre -0.163 -0.408 (0.196) (0.282) PRD:pre 0.040 0.094 (0.124) (0.179) Wealth: pre 0.039 0.505 (0.263) (0.501) Risk Aversion: pre 0.008 0.007 (0.009) (0.009) Patience: pre 0.001 0.005 (0.005) (0.006) Social Trust: pre 0.005 0.065 (0.062) (0.106) Constant 0.275*** 0.288*** 0.254 0.059 0.260 -0.398 (0.043) (0.055) (0.156) (0.287) (0.170) (0.582) Observations 515 515 515 515 515 515 R-squared 0.038 0.039 0.038 0.039 0.038 0.043 Notes: Clustered standard errors are in parentheses. ∗p < 0.10,∗∗p < 0.05,∗∗∗p < 0.01. The average levels of political affiliation, risk aversion, patience, and social trust are calculated at the municipality level before the earthquake. 6.3 Correspondence between Subjective and Objective Measures The somewhat varying results from self-reported vs. objective locations of distribution centers raise the question: How much overlap exists between the two variables in Mexico City? We found that the variables based on self-reports and actual locations match for 62.64% of respondents. How were the distribution centers targeted? Analyses of predictors of the self-report and 20 objective location measures are informative, though they also diverge in some ways. With the self-report variable, we find that those who experienced water service interruptions, had difficulties finding food and emergency products for their household, or supported PRD (i.e., the party of the head of the Federal District/Mexico City at the time of the earthquake) are more likely to report that there was a distribution center in their neighborhood compared to those who did not experience those difficulties (see Table E.1). Analyses of the actual location-based variable yield no statistically significant correlations between distribution centers in respondents’ neighborhood and experiencing water service disruptions or having difficulties locating food and other necessities. Instead, only the political variables show statistically significant correlations with the actual distribution center location variable (See Table E.2). The probability of having a distribution center nearby is positively correlated with support for the incumbent party. Altogether, the evidence indicates some misalignment between actual and perceived realities, although we caution that the potential incompleteness of the objective center list may be partially responsible for these differences. 6.4 Robustness Checks and Placebo Analysis While the two cross-sectional data sets are useful for assessing the earthquake’s impact on trust, we want to ensure there was no selection on the sample surveyed in the second wave.18 We use weighting and different matching techniques to account for the fact that people’s baseline conditions, like education, number of children, number of adults in the household, and region, might affect the probability of facing a harder situation after the earthquake, having a distribution center in the neighborhood, and experiencing lower levels of trust.19 To guarantee that the matching estimators consistently estimate the effects of interest, we assume that surveying a person in round one or two was independent of the outcomes, conditional on the covariates and that the probability of being surveyed in the second wave is bounded away from zero and one. Using weighting and matching techniques, we confirm that the results are robust and that the two samples are comparable. The results in F.1 in the Online Appendix show that general political trust decreased significantly after the earthquake across models controlling for different control variables and matching methods. Additionally, as we have shown in Section 6, the actual distribution center variable may include some noise. To address this concern and ensure that the reported statistical sig188%(93 individuals) of the post-quake sample report that they were surveyed in both waves. 19Reweighting is competitive with the most effective matching estimators when the overlap is good (Busso et al., 2014), as in our case. 21 Nikolova, Elena, and Nikolay Marinov. 2017. “Do Public Fund Windfalls Increase Corruption? Evidence from a Natural Disaster.” Comparative Political Studies 50, no. 11 (September): 1455–88. O’Donnell, Guillermo. 2004. “The Quality of Democracy: Why the Rule of Law Matters.” Journal of Democracy 15 (4): 32–46. Accessed September 29, 2021. Olson, Richard Stuart. 2000. “Toward a Politics of Disaster: Losses, Values, Agendas, and Blame.” Crisis Management 18 (2): 154. Olson, Richard Stuart, and Vincent T. Gawronski. 2010. “From Disaster Event to Political Crisis: A “5C +A” Framework for Analysis.” International Studies Perspectives 11 (3): 205–21. Petrova, Kristina, and Elisabeth L. Rosvold. 2024. “Mitigating the legacy of violence: Can flood relief improve people’s trust in government in conflict-affected areas? Evidence from Pakistan.” World Development 173:106372. https://www.sciencedirect.com/scien ce/article/pii/S0305750X23001900. Reinhardt, Gina Yannitell. 2015. “First-Hand Experience and Second-Hand Information: Changing Trust across Three Levels of Government.” Review of Policy Research 32 (3): 345–64. . 2019. “The Intersectionality of Disasters’ Effects on Trust in Public Officials.” Social Science Quarterly 100 (7): 2567–80. Accessed August 3, 2021. Rudolph, Thomas J. 2003. “Institutional Context and the Assignment of Political Responsibility.” The Journal of Politics 65, no. 1 (February): 190–215. Scartascini, Carlos, and Joanna Valle L. 2020. “The Elusive Quest for Growth in Latin American and the Caribbean: The Role of Trust.” Washington, DC. Schneider, Saundra K. 1992. “Governmental Response to Disasters: The Conflict between Bureaucratic Procedures and Emergent Norms.” Public Administration Review 52, no. 2 (March): 135. Sobel, Russell S., and Peter T. Leeson. 2006. “Government’s Response to Hurricane Katrina: A Public Choice Analysis.” Public Choice 127 (1-2): 55–73. Stephane, Victor. 2021. “Hiding behind the veil of ashes: Social capital in the wake of natural disasters.” World Development 145:105518. https://www.sciencedirect.com/science/ article/pii/S0305750X21001303. 28 Townshend, Ivan, Olu Awosoga, Judith Kulig, et al. 2015. “Social Cohesion and Resilience Across Communities That Have Experienced a Disaster.” Natural Hazards 76:913–38. https://doi.org/10.1007/s11069-014-1526-4. United States Geological Survey. 2017. Magnitude 7.1 Earthquake in Mexico. USGS News, September. https://www.usgs.gov/news/featured-story/magnitude-71-earthquakemexico. Urbinati, Nadia, and Mark E Warren. 2008. “The Concept of Representation in Contemporary Democratic Theory.” Annual Review of Political Science 11:387–412. Villegas, Paulina, and Azam Ahmed. 2017. Powerful Earthquake Strikes Mexico, Killing Dozens. The New York Times, September. https://www.nytimes.com/2017/09/19/ world/americas/mexico-earthquake.html. Wong, Mathew Y. H., and Ying-ho Kwong. 2021. “Warning signal: Political Trust, Typhoons and the Myth of the ‘Li’s Field’ in Hong Kong.” Asia Pacific Viewpoint 62 (2): 206–22. Wortman, Camille B. 1976. “Causal Attributions and Personal control.” In New Directions in Attribution Research, edited by John H. Harvey, William J. Ickes, and Robert F. Kidd, 23–52. Hillsdale, NJ: Erlbaum. Yamamura, Eiji. 2014. “Impact of Natural Disaster on Public Sector Corruption.” Public Choice 161 (3-4): 385–405. You, Yu, Yifan Huang, and Yuyi Zhuang. 2020. “Natural Disaster and Political Trust: A Natural Experiment Study of the Impact of the Wenchuan Earthquake.” Chinese Journal of Sociology 6, no. 1 (January): 140–65. Zechmeister, Elizabeth J., and Daniel Zizumbo-Colunga. 2013. “The Varying Political Toll of Concerns about Corruption in Good versus Bad Economic Times.” Comparative Political Studies 46 (10): 1190–218. 29 Online Appendix A Literature review Table A.1: Selected Recent Empirical Investigations of Political Trust After A Natural Disaster Paper Case Data Explanatory Trustee (DV) variable Albrecht 10 disasters Survey Disaster Politicians 2017 in Europe (cross-national, multiple rounds) Aldrich Earthquake Observations, Disaster Central & 2017 & Tsunami interviews, local govt in Japan & news articles Carlin et al. Earthquakes in Survey Direct vs. indirect People in the 2014b El Salvador, (cross-national, disaster experience community Haiti & Chile post-disaster) Han et al. & Earthquake Observations, Disaster Local govt 2011 in China interviews, & focus group Mathew & Typhoon News articles Trust in politicians Bureaucrat Kwong 2021 in Hong Kong after disaster Nicholls & Hurricane Survey Direct vs. indirect Local, state, Picou 2013 in the U.S. disaster experience federal govt Reinhardt Hurricane Survey Direct vs. indirect president, governor, 2015 in the U.S. disaster experience Mayor, FEMA, state/local EMA, Reinhardt Hurricane Survey Gender Police & 2019 in the U.S. ambulance service Scott et al. Coal waste Survey Time Expert, general 2016 rupture ppl, public in the U.S. officials, local govt Scott et al. Coal waste Interviews Disaster Corporations, 2005 rupture govt, & regulatory in the U.S. authorities Strömbäck Tsunami Survey, Disaster Politicians, & Nord 2006 in Indonesia focus group govt, media You et al. Earthquake Survey Disaster Local, provincial, 2020 in China (2 rounds) central govt official Notes: DV listed include only outcomes related to trust. Studies about incumbent approval ratings and electoral outcomes are related but not mentioned here. B Survey Design and Implementation The pre-earthquake survey (August 31 - September 19, 2017) and the post-earthquake survey (November 18, 2017 - January 18, 2018) had IRB approval and conformed to APSA’s Principles and Guidance for Human Subjects Research. Recruited individuals were read a study information sheet that informed them that they were being invited to participate in a 1 study about public opinion in Mexico, conducted by LAPOP and administered by DATAOPM. They were then allowed to consent or decline to participate. There was no deception in the study. Those who consented to the study could decline any question and terminate the study at any time. As is common for standard public opinions in Mexico, and elsewhere, participants were not compensated. The study did not intervene with real-world events and only functioned as an opinion survey.1 We designed the original survey to be representative of non-institutionalized adults (18 or above) living in the greater Mexico City metropolitan area. The original target was 900 interviews in the pre-earthquake survey. Each region of the metropolitan area was designated a proportion of the total number of interviews based on population data. A sample of people was then chosen in four stages, as shown in Table B.1. We used geographical stratification since this increases precision by yielding smaller random sample errors than those obtained with a simple random sampling of the same sample size (Gerber and Green, 2012). A stratified sample also tends to be more representative and dispersed since it guarantees the inclusion of municipalities in the entire metropolitan area. Table B.1: Stratified Multi-stage Clustered Sampling Design Sampling Unit Unit selected Sampling Selection Process Stratification Strata Regions Stratified Sampling Multistage sample from each stratum First stage Secondary Sampling Unit (SSU)* Electoral sections‡PPS sampling§ Second stage Tertiary Sampling Unit (TSU) Blocks of dwellings PPS sampling Third stage Quaternary Sampling Unit (QSU) Households Systematic sampling Fourth stage Final Sampling Units (FSU) Person in household Quota by sex and age Notes: *‡An electoral section is the basic geographical unit into which the national territory is divided for electoral purposes. §PPS denotes Probability Proportional to Size. We used a probability proportional to the adult population sampling (PPS) method to select the electoral sections (Secondary Sampling Units, SSUs) within each region.2Since six interviews were to be conducted per electoral section, the number of interviews per region directly determined the number of electoral sections to be randomly drawn from each region. 1In addition to survey responses, the devices were programmed to capture location, audio segments, a picture of the enumerator from the front-facing camera, and timing data; the local team audited 100% of surveys for quality control on each of these dimensions, and our team conducted a second audit of just over 20%. A small proportion of initially-recorded interviews were canceled for quality control issues (about 6-7% of completed interviews); the computer-assisted quality control procedures meant that those poor quality interviews were detected and replaced with valid high-quality interviews from the correct blocks while fieldwork was still in progress. 2There are federal electoral districts and electoral sections. We use electoral sections, the basic geographical unit into which the national territory is divided for electoral purposes. (INE, FEPADE, UNAM, Tribunal Electoral del Poder Judicial de la Federación,2016). 2 Table B.2 shows each region’s participation in the survey, the number of municipalities and electoral sections, the number of interviews that had to be carried out in each region, and the number of interviews that were actually carried out before and after the earthquake. Table B.2: Number of Interviews per Region (target and actual sample) Sample Actual sample (Target) Pre - Earthquake Post - Earthquake Region Number of Number of Number of Proportion Number of Number of Number of Proportion Number of Number of Proportion Municipalities SSUs Interviews Municipalities SSUs Interviews SSUs Interviews Center 6 28 168 18.67%4 19 109 19.16%32 229 19.67% North 14 27 162 18.00%9 19 111 19.51%24 229 19.67% South 10 25 150 16.67%6 13 78 13.71%26 156 13.40% East 9 47 282 31.33%6 36 213 37.43%62 431 37.03% West 6 23 138 15.33%4 10 58 10.19%38 119 10.22% Total 45 150 900 100%29 97 569 100%162 1,164 100% Once we had randomly selected the electoral sections, we used a new PPS sampling method to select the blocks of dwellings in each SSU. The number of households in it defined the size of each block. (see National Housing Inventory 2016). Next, enumerators systematically chose households, skipping two housing units after each completed interview within a block.3Finally, a person in each household was chosen according to gender and age quotas, which were estimated based on the distribution of the population registered in the electoral sections.4Three different quota forms (A, B, and C), shown in Table B.3, were used to approximate the reference parameters for each SSU. Table B.3: Household Forms Form A Form B Form C Gender/Age 1829 3050 <50 Total 1829 3050 <50 Total 1829 3050 <50 Total Men 1 2 0 3 1 1 1 3 1 1 1 3 Women 1 1 1 3 1 2 0 3 1 1 1 3 Total 2 3 1 6 2 3 1 6 2 2 2 6 The pre-earthquake survey was about two-thirds complete (569 interviews) when the earthquake struck. The post-quake survey was conducted two months after the earthquake. The post-earthquake survey was designed to draw 900 individuals from the same SSUs as in the original design. In addition, we layered in an oversample of individuals who fit the profile of the 569 individuals who were interviewed before the earthquake. We only keep the data from respondents in the same SSUs as respondents in the first wave of the survey. 3The enumerators were instructed to locate the block’s north-eastern point to start the interviews and continue walking clockwise. In case of rejection, vacant housing, or people’s absence, the enumerator selected the adjacent housing. When survey personnel reached the end of a block without completing the quota of 3 interviews per block, they continued to the next block, following the same routine as in the previous block. 4Interviewees had to reside permanently in the household, not be a domestic worker or visitor. If more than two people in a house were in the same age group and gender, we recruited a person with the closest birthday. 3 C Survey Questions, Variables, and Balance Check Table C.1: Questions Questions for general political trust variable Variable name Question Response options Note General political trust 1./2. [Politicians in general/ public officials], do you think it is very common, somewhat common, not very common, or not common at all that when they promise something they fulfill it? 3./4. [Politicians in general/ public officials], do you think it is very common, somewhat common, not very common, or not common at all that they comply with the laws and regulations of the country? 1. Not common at all 2. Not very common 3. Somewhat common 4. Very common One composite variable is constructed from the four survey questions using PCA Question for self-reported distribution center variable Code Question Response options Note Self-reported DC After the earthquake, were there distribution centers for food, water and other essential items in your neighborhood? 1. Yes 2. No Recoded as: 1. Yes 0. No 4 Table C.2: Questions Questions for control variables Code Question Response options Note Male Gender 1. Male 2. Female Recoded as: 1. Male 0. Female Age How old are you? Numeric value Education How many years of schooling have you completed? Ranges from 0 (none) to 24 (doctorate degree) N. Adults Including you, how many adults live in your home? Numeric value N. Children How many children under the age of 18 live in your home? Numeric value PRI/PRD Of these parties, which are you most willing to support? 1. PRI 2. PAN 3. PRD 4. MORENA 5. Green Party 6. New Alliance 7. Citizen Movement 8. Labor Party First, coded one if a respondent supported PRI/PRD, 0 otherwise. Then, estimated the proportion of party supporters before the earthquake and assigned the average to residents of a corresponding municipality Wealth Could you tell me if you have the following in your house: (1) Bathroom inside the house (2) Salaried employee(s)/Domestic worker/Domestic service (3) Automobile( s)/car(s) (4) Microcomputer(s), laptops, tablets, ipads and netbooks (5) Dishwasher (6) Refrigerator (7) Freeze(s) (8) Washing machine (9) Microwave (10) Motorcycle (11) Clothes dryer (12) Television (13) Cable television, satellite television, Netflix (14) Landline telephone (15) Cellular telephone (16) Drinking water inside the house (17) Internet service inside the house 1. Yes 2. No Recoded each survey question as an indicator variable assigning a value of 1 if owned, then summed up the number of items that a respondent possessed, and estimated the municipality average of it before the earthquake and assigned the average to residents of a corresponding municipality Patience Imagine you have won 1,800 Mexican pesos in a lottery and have the option to receive the reward in a payment today of 1,800 pesos or a higher payment that you would receive in 12 months. I am going to ask you a series of questions about your preferences about receiving 1800 today or receiving a higher payment in a year. The sequence of questions has 32 possible ordered outcomes, such that we can derive a measure of patience ranging from 1 to 32. Example: Between receiving a prize of 1,800 pesos today or 2,770 pesos in 12 months, what would you prefer? 1. To receive 1,800 pesos today. 2. To receive 2,770 pesos in a year. We use the variable containing values from 1 to 32, where 1 indicates the lowest level of patience and 32 the highest level. We calculated the average before the earthquake and assigned it to residents of the corresponding municipality after the earthquake. Risk aversion Imagine you can choose between receiving a sure prize in pesos, or receiving a lottery ticket with which you could win 5,400 pesos, but you could also win nothing. Below I am going to name different prize alternatives so you can tell me which ones you would choose. The sequence of questions has 32 possible ordered outcomes, such that we can derive a measure of risk aversion ranging from 1 to 32. Example: Between a lottery ticket, in which you have an equal chance of winning 5,400 pesos or not winning anything, and receiving a sure prize of 360 pesos, which option would you choose? 1. The lottery ticket. 2. The sure prize of 360 pesos. We use the variable containing values from 1 to 32, where 1 indicates the lowest level of risk aversion (also called risk-loving person) and 32 the highest level of risk aversion (risk-averse person). We calculated the average before the earthquake and assigned it to residents of the corresponding municipality after the earthquake. Social trust In general, would you say that the majority of people are very trustworthy, somewhat trustworthy, not very trustworthy, or untrustworthy? 1. Very trustworthy 2. Somewhat trustworthy 3. Not very trustworthy 4. Untrustworthy Estimated the municipality average before the earthquake and assigned the average to residents of a corresponding municipality 5 Table C.3: Descriptive Statistics Pre-earthquake Post-earthquake Mean s.d. Mean s.d. Difference p-value Npre Npost Male 0.511 (0.500) 0.508 (0.500) 0.004 0.885 569 1,164 Age 39.306 (16.023) 38.942 (15.563) 0.364 0.651 568 1,164 Education 10.743 (4.292) 10.683 (3.979) 0.060 0.774 569 1,149 N. Adults 3.497 (1.830) 3.331 (2.004) 0.166 0.097 563 1,148 N. Children 1.528 (1.612) 1.431 (1.565) 0.098 0.229 562 1,145 PRI 0.148 (0.355) 0.141 (0.348) 0.007 0.707 569 1,164 PRD 0.130 (0.337) 0.083 (0.277) 0.047 0.002 569 1,164 Wealth 9.396 (3.026) 9.259 (3.187) 0.137 0.397 560 1,116 Social trust 2.692 (0.782) 2.704 (0.805) -0.012 0.768 569 1,164 Center 0.192 (0.394) 0.197 (0.398) -0.005 0.799 569 1,164 North 0.195 (0.397) 0.197 (0.398) -0.002 0.935 569 1,164 South 0.137 (0.344) 0.134 (0.341) 0.003 0.861 569 1,164 East 0.374 (0.484) 0.370 (0.483) 0.004 0.869 569 1,164 West 0.102 (0.303) 0.102 (0.303) -0.000 0.985 569 1,164 Patience 11.380 (11.715) 11.989 (11.868) -0.609 0.343 560 850 Risk Aversion 28.073 (7.212) 26.011 (8.853) 2.062 0.000 560 816 General Trust 0.260 0.200 0.193 0.232 0.027 0.007 560 1,134 Notes: Standard deviation in parentheses. There is 1 missing value in age, 15 in education, 22 in the number of adults, 26 in the number of children, 357 in risk aversion, 323 in time-preferences (patience), and 39 in general trust . D General Political Trust D.1 Distribution of Dependent Variables Table D.1: Distribution of Dependent Variables Panel 1 - Promises Politicians Civil servants Pre Post Total Pre Post Total N. Col % N. Col % N. Col % N. Col % N. Col % N. Col % (1) Not common at all 240 42.25% 575 49.83% 815 47.33% 254 45.12% 600 52.04% 854 49.77% (2) Not very common 228 40.14% 453 39.25% 681 39.55% 209 37.12% 391 33.91% 600 34.97% (3) Somewhat common 77 13.56% 94 8.15% 171 9.93% 89 15.81% 128 11.10% 217 12.65% (4) Very common 23 4.05% 32 2.77% 55 3.19% 11 1.95% 34 2.95% 45 2.62% Total 568 100% 1,154 100% 1722 100% 563 100% 1,153 100% 1,716 100% Pearson chi2(3) = 17.852 p-value < 0.001 Pearson chi2(3) = 12.813 p-value = 0.005 6 Panel 2 - Compliance Politicians Civil servants Pre Post Total Pre Post Total N. Col % N. Col % N. Col % N. Col % N. Col % N. Col % (1) Not common at all 225 39.68% 477 41.41% 702 40.84% 227 39.96% 513 44.49% 740 43.00% (2) Not very common 255 44.97% 491 42.62% 746 43.40% 223 39.26% 456 39.55% 679 39.45% (3) Somewhat common 78 13.76% 157 13.63% 235 13.67% 108 19.01% 162 14.05% 270 15.69% (4) Very common 9 1.59% 27 2.34% 36 2.09% 10 1.76% 22 1.91% 32 1.86% Total 567 100% 1,152 100% 1,719 100% 568 100% 1,153 100% 1,721 100% Pearson chi2(3) = 1.804 p-value = 0.614 Pearson chi2(3) = 7.843 p-value = 0.049 D.2 Principal Component Analysis The principal component analysis on four survey items shows that the variations in the data can be explained by four components, indicated as in Components 1-4 in Table D.2. However, the first latent variable alone explains about 60% of the variation and only its eigenvalue exceeds 1, meaning that it is the only underlying latent variable worth exploring as others do not have significant covariance. Table D.2: Components from PCA Component Eigenvalue Difference Proportion Cumulative Component 1 2.295 1.480 0.573 0.573 Component 2 0.814 0.327 0.204 0.777 Component 3 0.487 0.084 0.122 0.899 Component 4 0.403 0.000 0.1007 1.000 Notes: Number of observations = 1,694. Table D.3 shows that the loadings of all four survey items are high on the first component. Therefore, there are sufficient variations that correlate to one another and can be interpreted as tapping into one latent variable. Table D.3: Loadings Variable Component 1 Promises - Politicians 0.483 Promises - Civil servants 0.500 Compliance - Politicians 0.506 Compliance - Civil servants 0.510 7 E Distribution Centers E.1 Observed Distribution Centers Data Collection We analyzed post-earthquake social media posts in three phases. First, we searched Google Scholar for literature on the Mexico City government earthquake response and the role of social media. Second, we created a list of hashtags for Twitter searches. Third, we collected photos of distribution centers that provided aid in the aftermath of the earthquake. As a first step, we found articles that examined the government response via—and in comparison to—civil society responses to the earthquake. These articles recount that members of civil society organizations in the hours after the September 19 earthquake assembled a large social network to exchange real-time information about building collapses, missing persons, and large-scale damage on WhatsApp (Campos Rivera, 2018; Mora et al., 2018). The information being transmitted via WhatsApp about building collapses and damage became a collaborative map called “Mapeo colaborativo RescateMX,” which civil society and governmental actors used to find survivors of the earthquake, assess where damage had occurred, and locate shelters and distribution centers. After we conducted the Google Scholar searches, we created a list of potentially popular post-earthquake Twitter hashtags. Our goal in these searches was to find information about who ran distribution centers, where they were located, and how this information was disseminated. We found that individuals primarily communicated publicly via Twitter (following Mora et al. (2018)). We then used the first set of tweets to create a longer list of the most popular post-earthquake hashtags: #19S, #Verificado19S, #CentrodeAcopio, #sismoCDMX, #19sAcopio, #19svoluntarios, #AyudaSismo, #AyudaCDMX, #reconstucciónCDMX, #6meses19S, #FuerzaMéxico, #AmplificaMexico, and #Edoméx19S. In addition, we searched for hashtags that mentioned the Mexican Army (SEDENA), Navy (SEMAR), and Plan DN-III-E—a post-disaster assistance plan, in which the military is deployed to assist in disaster recovery. Third, we conducted Google searches for distribution center images. We downloaded 40 photos in which the image metadata associates the center with the September 19 earthquake. Most individuals in the photos are not wearing uniforms; however, in five photos, individuals are wearing a brightly colored vest (neon green or purple) or a sticker that says “CDMX.” In two of the photos, volunteers are wearing vests that clearly say “Cruz Roja” (Red Cross). Distribution centers were staffed by individuals from diverse organizations, including civil 8 Table G.2: Marginal Effects of Observed Distribution Center on General Trust General political trust (1) (2) (3) (4) (5) (6) Objective DC -0.024 -0.017 -0.027 -0.024 -0.023 -0.030 (0.022) (0.022) (0.021) (0.020) (0.022) (0.024) Male -0.018 -0.018 -0.017 -0.018 -0.018 -0.017 (0.018) (0.018) (0.018) (0.018) (0.018) (0.019) Age -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Last year of education approved -0.006* -0.006** -0.006** -0.006** -0.006** -0.006** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Number of adults in the household -0.013* -0.013* -0.013* -0.014* -0.013* -0.013* (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) Number of children in the household -0.004 -0.004 -0.004 -0.004 -0.004 -0.003 (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) PRI:pre 0.256 -0.036 (0.233) (0.288) PRD:pre -0.110 -0.098 (0.151) (0.127) Wealth: pre 0.387 0.568 (0.244) (0.479) Risk Aversion: pre -0.012** -0.015** (0.006) (0.006) Patience: pre 0.001 0.002 (0.005) (0.006) Social Trust: pre -0.052 0.108 (0.095) (0.132) Constant 0.407*** 0.396*** 0.195 0.739*** 0.553** 0.220 (0.051) (0.059) (0.151) (0.196) (0.259) (0.540) Observations 271 271 271 271 271 Clustered standard errors are in parentheses. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 15 References Abadie, Alberto, David Drukker, Jane Leber Herr, and Guido W. Imbens. 2004. “Implementing Matching Estimators for Average Treatment Effects in Stata.” Stata Journal 4 (3): 290–311. Abadie, Alberto, and Guido W. Imbens. 2006. “Large Sample Properties of Matching Estimators for Average Treatment Effects.” Econometrica 74 (1): 235–67. . 2011. “Bias-corrected Matching Estimators for Average Treatment Effects.” Journal of Business & Economic Statistics 29:1–11. . 2016. “Matching on the Estimated Propensity Score.” Econometrica 84 (2): 781–807. Busso, Matias, John DiNardo, and Justin McCrary. 2014. “New Evidence on the Finite Sample Properties of Propensity Score Reweighting and Matching Estimators.” The Review of Economics and Statistics 96 (5): 885–97. Campos Rivera, Héctor. 2018. “# Verificado19s: la Fortaleza de las Redes Sociales ante un Terremoto.” Garrido, Melissa, Amy Kelley, Julia Paris, Katherine Roza, Diane Meier, R Sean Morrison, and Melissa Aldridge. 2011. “Methods for Constructing and Assessing Propensity Scores.” Health Services Research 49 (5). Gerber, A.S., and D.P. Green. 2012. Field Experiments: Design, Analysis, and Interpretation. W. W. Norton. https://books.google.com.co/books?id=yxEGywAACAAJ. INE, FEPADE, UNAM, Tribunal Electoral del Poder Judicial de la Federación. 2016. Compendio Legislación Nacional Electoral. https://portalanterior.ine.mx/archivos3/portal/ historico/recursos/IFEv2/DS/DSVarios/docs/2016/CompendioLegislacionNal/ Compendio-TomoII.pdf. Leuven, Edwin, and Barbara Sianesi. 2003. “PSMATCH2: Stata Module to Perform Full Mahalanobis and Propensity Score Matching, Common Support Graphing, and Covariate Imbalance Testing.” Accessed on June 11, 2020. Available at https://econpapers. repec.org/software/bocbocode/s432001.htm. Mora, Mariana, María Paula Saffon, and Pablo Gómez. 2018. “Investigación-acción durante Desastres: Uso de Redes y Derechos.” Revista Mexicana de Sociología 80 (SPE): 95–119. 16