Policy Research Working Paper 10314 Food Insecurity Erodes Trust Woubet Kassa Michael D. Smith Dennis Wesselbaum Africa Region Office of the Chief Economist February 2023 Policy Research Working Paper 10314 Abstract This study examines the relationship between food inse- The findings show that food insecurity is associated with curity and trust using the 2014–17 waves of the Gallup a decrease in both measures of trust. The study further World Poll and the Food and Agriculture Organization’s document heterogeneous effects of food insecurity across Food Insecurity Experience Scale. Trust improves public economic development rankings. The results suggest a need institutions, social capital, public health interventions, and for governments to increase food security to bolster public economic development. Vertical trust is represented as an trust, strengthen the social contract, and enhance the effec- index of trust in national institutions, while horizontal trust tiveness of development efforts. is represented as a measure of trust in friends and family. This paper is a product of the Office of the Chief Economist, Africa Region. 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 authors may be contacted at wkassa1@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 Food Insecurity Erodes Trust * Woubet Kassa † Michael D. Smith ‡ Dennis Wesselbaum § World Bank National Oceanic and Atmospheric University of Otago Administration Keywords: Food Insecurity, Experiential Measures of Food Insecurity, Social Contract, Trust, Governance JEL: D73, O12, Q18, Z13 * The findings and conclusions in the paper are those of the author(s) and do not necessarily represent the views of the World Bank Group or the National Marine Fisheries Service, NOAA. † Corresponding author. The World Bank, Washington, DC, USA. Email: wkassa1@worldbank.org. ‡ National Oceanic and Atmospheric Administration, Economics and Social Sciences Research, Alaska Fisheries Science Center, 7600 Sand Point Way NE, Seattle, WA 98115-6349, USA. Email: michael.d.smith@noaa.gov. § Department of Economics, University of Otago, Dunedin, New Zealand. Email: dennis.wesselbaum@otago.ac.nz. 1. Introduction Despite widespread consensus on the importance of trust for long-term social and economic development (e.g., Acemoglu et al., 2001), research has not examined the role of food insecurity— the inability to obtain adequate food in quantity and quality because of a lack of money or other resources—in determining perceptions of trust. Food insecurity threatens both the vertical and horizontal dimensions of trust. Horizontal trust involves trust in friends and family, while vertical trust represents trust in institutions. First, since governments administer programs that are meant to combat food insecurity, their inability to do so may have an impact on vertical trust. Food insecurity is a failure of the social contract. That is, if people obey the law and pay taxes and cooperate for the social good, they expect the state to provide certain goods and services, including food security (OECD, 2013; World Bank, 2019). Trust is critical if this process is to function. Second, food insecurity causes distress and may reduce cooperation, thereby undermining horizontal trust. Food insecurity can affect collective social functioning (Dean and Sharkey, 2011; Dean et al., 2011) through decreases in subjective well-being (Frongillo et al., 2017; 2019) and increases in feelings of deprivation and alienation, and adverse family and social interactions (Wunderlich and Norwood, 2006). On the other hand, food insecurity could motivate civic cooperation and risk-sharing in social networks to strengthen the social safety net (Townsend, 1994; Ambrus et al., 2014), thereby increasing horizontal trust among those who work together. Thus, the link between food insecurity and trust is ambiguous and warrants empirical investigation. While the prevalence of global food insecurity declined over the last several decades, it has risen every year between 2015 and 2019 (FAO, 2020a) and has almost certainly risen disastrously during the COVID-19 pandemic (FAO, 2020b; Laborde et al., 2020; Smith and Wesselbaum, 2020) and more recently the war in Ukraine. This would have important implications on societal trust – both vertical and horizontal dimensions of trust. In 2014, the Food and Agriculture Organization’s (FAO) Voices of the Hungry project developed an experiential measure of food insecurity called the Food Insecurity Experience Scale (FIES), and contracted Gallup, Inc. to use its Gallup World Poll (GWP) to collect data in over 150 countries (Cafiero et al., 2018). The rich individual-level data in the GWP combined with asking individuals directly about their experiences and behaviors related to their access to food in the FIES module, provides an unprecedented opportunity to examine the relationship between food insecurity and trust globally. 2 Figure 1. Vertical and horizontal trust and food insecurity by regions and economic development (GNI per capita). Figure Notes: Means calculated using sample-weighted individual-level data from the 2014-17 Gallup World Poll. Food insecurity and trust perceptions vary by geographical area and economic development (GNI per capita; Figure 1). The Sub-Saharan Africa region leads the world in experiencing food insecurity (62%). South Asia leads the world in trust in national institutions (69%), followed by North America (67%); trust is lowest in Latin America and the Caribbean (37%). For trust in friends and family, North America leads the world (94%) and it is lowest in South Asia (67%). Low-income countries also have the highest proportion of respondents who trust their national institutions (57%), as well as the lowest proportion of respondents who trust their friends and family (68%). 5 5 We measure vertical trust in national institutions (military, judicial system and courts, national government, and electoral honesty) and horizontal trust in friends and family. 3 Using a series of linear regressions with country fixed effects, and data from the 2014-17 waves of the GWP/FIES, we find that food insecurity is associated with a decrease in both trust in national institutions and trust in friends and family. These findings also apply in all four World Bank income classes of countries. The association with trust in friends and family is larger than any other explanatory variable. For trust in national institutions, the associations of food insecurity increase monotonically with income ranking. For trust in friends and family, the results reveal nonlinear effects of food insecurity as national incomes rise. There are also differences in the determinants of vertical and horizontal trust. For example, rural/urban location strongly determines trust in national institutions, whereas being an immigrant matters more to trust in friends and family. As a robustness check on the possible endogeneity of food insecurity, we use non- parametric entropy matching methods to isolate the relationship between food insecurity and perceptions of trust. While the results may not be considered causal, they remain robust under alternative model specifications. 2. Empirical Strategy 2.1. Estimation Framework The baseline model that estimates various forms of trust on our main variable of interest (food insecurity) and other control variables, is given by: , = + , +, + +,, , (1) where i and c are indices for individuals and countries; represents respondents’ food insecurity status, where represents the correlation of interest; consists of individual- and household-level socioeconomic characteristics; contains country and year fixed effects; and εi,c represents unobserved individual heterogeneity. The country and year fixed effects remove the cross- sectional variation related to unobserved heterogeneity associated with individuals in each country for each year. Possible endogeneity may still exist between food insecurity and trust. Food insecurity is not distributed randomly among the population. Observed and unobserved individual factors influence 4 food insecurity and trust-related behavior. Consequently, the estimates may be biased and inconsistent. Following recent research (Ruyssen and Salomone, 2018; Smith and Floro, 2020), we address potential endogeneity issues using a matched sample of food secure and food insecure respondents with identical variable distributions. Entropy matching methods enable us to compare individuals such that, after matching, the only difference between the two subsamples is their food insecurity status. Given the strong correlation between observable and unobservable characteristics, matching on observable characteristics implies at least some matching on unobservable characteristics (Stuart et al., 2010; Ferraro and Miranda, 2014; Ruyssen and Salomone, 2018). Matching produces an unbiased measure of the influence of food insecurity on trust if the entropy algorithm captures all relevant differences between individuals who are food insecure and those who are food secure (see Appendix A for more details). 2.2. Data The data for the study draws from the 2014-17 waves of the Gallup World Poll, including FAO’s FIES. The GWP collects information on individuals’ labor force participation, income, educational attainment, future aspirations, subjective well-being, demographic characteristics, and country- identifiers. In most countries, the GWP interviews 1,000 individuals and is nationally representative. Researchers use a random route procedure to select sample households within each country and select the respondent randomly within each household using a Kish grid method (Gallup, 2016). Observations for respondents without valid food insecurity responses or who failed to provide valid information on one or more questions used to construct the control variables were dropped from the sample. The final sample is 387,385 individuals aged 15 years and older in 134 countries. 2.3. Measures of Vertical and Horizontal Trust Trust is defined as holding a positive perception about the actions of an individual or an organization (OECD, 2013). Generally, trust has two components: 1) vertical (political) trust, citizens’ faith in government and its institutions; and 2) horizontal (social) trust, citizens’ interpersonal confidence in their social community (OECD, 2013). Trust is harder to quantify and model formally than traditional economic measures (Barrett, 1997). Despite the difficulty of 5 formal theoretical and empirical testing, the value of trust to social and economic progress is not diminished (Barrett, 1997). 6 As a measure of vertical trust, we use GWP’s Trust in National Institutions Index (NI index). The NI index is composed of four items measuring respondents’ trust in key national institutions (Gallup, 2016). Respondents were asked whether they had “confidence in each of the following, or not?” with the options of “(1) How about the military? (2) How about the judicial system and courts? (3) How about the national government? and (4) How about the honesty of elections?” Responses were combined into a single index score for each individual. The index is the mean of valid items multiplied by 100, and is only calculated if respondents had two or more out of four valid scores. As a sensitivity check on the NI index, we study each of the four questions independently. We also examine the effects of food insecurity on more local public institutions (as opposed to national institutions or interpersonal levels of trust), using binary measures of trust in the local police and trust in the financial system and local banks. Further, we also use a principal component approach to reduce potential measurement error (Gillen et al., 2019). For the horizontal dimension of trust, we use respondents’ trust in friends and family. This binary measure equals one if the respondent answers positively to the question: “If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them?” This measure allows us to test whether food insecurity influences interpersonal relationships and social capital rather than only with authority and formal institutions. People who trust friends or family are also more likely to exhibit altruistic behavior and reciprocity, boosting civic cooperation (Putnam, 2000; Uslaner, 2002; Corbacho et al., 2015). Both GWP trust measures are translated into different languages and standardized, facilitating worldwide comparability. However, the inherent difficulties in assessing a latent trait like trust persist (Barrett, 1997; OECD, 2013). For example, the GWP’s questions measuring trust in national institutions are ultimately subject to ambiguity around respondents’ interpretation of the 6 Research identifies trust as one of the most crucial pillars upon which the legitimacy and sustainability of social and economic systems are built (Knack and Keefer, 1997; Alesina and La Ferrara, 2002; Barrett, 2005; OECD, 2013, Algan and Cahuc, 2010). 6 terms “trust” and “confidence” and respondents’ definition of government (OECD, 2013). Cultural factors and societal norms influence how individuals define public institutions, making valid cross- national comparisons of trust difficult (OECD, 2013). These issues affect all trust-related, international surveys. However, this study takes advantage of the detailed information on individual-level food insecurity provided by the FIES. This information, combined with the globally standardized trust-related questions in the GWP, make it possible for the first time to examine the direct impact of food insecurity on both dimensions of trust around the world. 2.4. Measuring Food Insecurity FAO created the FIES to gather consistent and comprehensive information on the prevalence and severity of global food insecurity. The GWP/FAO FIES survey module contains eight questions designed to assess the adequacy of individuals’ access to food, adapted from the long-established United States Household Food Security Survey Module (US HFSSM) and the Latin American and Caribbean Food Security Scale (ELCSA). The FIES survey questions focus on respondents’ behaviors and experiences when they have encountered difficulties in meeting their basic food needs over the past 12 months. Similar to other food insecurity surveys, the FIES questions are ranked by severity. Each question specifies that food insecurity results from a lack of money or other resources to obtain adequate food (see Appendix, Table A1). In this analysis, we use a binary measure of individuals’ severity of food insecurity. The measure, food insecurity, is coded as one if an individual experienced moderate or severe food insecurity in the past 12 months; zero otherwise. This measure captures experiences ranging in severity from a reduction in food quality and variety to the reporting of physiological hunger (Nord et al., 2016; Cafiero et al., 2018). 2.5. Control Variables In line with the literature, we control for other determinants of trust that include individual and household characteristics. These include gender, age, and lifecycle stage (age squared), household composition, marital status, education, income, consumption, location (urban/rural), labor force status, whether or not the respondent is an immigrant, and the respondent’s religiosity. Lastly, we 7 control for perceptions of corruption in government, quality of infrastructure (satisfied with roads), and if the respondent has had money or property stolen in the past 12 months. 7 We estimate the model with a varying mix of controls to ensure robustness and avoid using closely related control variables. We have some priors on how these controls could affect trust. For example, education correlates with respondents’ expectations of government performance. If respondents’ expectations rise faster than the actual performance of governments, trust could decline (Dalton, 2005; OECD, 2013). Previous research has shown married people have higher levels of institutional trust than single people (Hudson, 2006). Household composition can dictate who can participate in civic activities that build interpersonal trust and determines the financial needs of the household. Studies have shown that working people trust national institutions more than the unemployed or those out of the workforce (Mishler and Rose, 2001; Hudson, 2006). In addition, household income is positively related with trust (Leblang et al., 2022). 8 We use GWP’s Per Capita Income Quintile variable, where the lowest income quintile indicates the respondent’s per capita household income lies among the poorest 20% of the country’s population, and the highest income quintile indicates it lies among the richest 20% of the country’s population. Citizens’ dissatisfaction with public infrastructure may be an important determinant of trust (Mischler and Rose, 2001; Corbacho et al., 2015). Thus, we include the dummy variable Satisfied with roads, which equals one if the respondent is satisfied with the roads and highways in their area. Government corruption can determine perceptions of justice and fairness in encounters with national institutions (Corbacho et al., 2015). To control for this, we include a dummy variable for Corruption in government which equals one if the respondent thinks that corruption is “widespread throughout the government in this country.” Similarly, since crime affects both vertical and horizontal dimensions of trust (Corbacho et al., 2015), we include the dummy variable Money or property stolen, which equals one if the respondent has had money or property stolen in the last 12 months. 7 Our results are robust to excluding these three variables over concerns that they could also measure some amount of trust. 8 Respondents who have difficulty answering the income question are presented a set of ranges in local currency and are asked which group they fall into (Gallup, 2016). This measure also relies on multiple imputation methodology to replace missing values. 8 Likewise, individuals may trust those who are more similar to themselves (i.e., the same social, racial, religious or ethnic group), and may trust others they have known longer (Alesina and La Ferrara, 2002; 2005). Hence, we control for religiosity and immigration status. The country and year fixed effects capture any national-level policies or shocks, such as storms, conflict, and economic downturns, and controls for any changes in the overall trend of trust in a country. 3. Results Table 1 presents the descriptive statistics for trust and food insecurity, as well as mean comparison tests examining differences across economic development rankings. 9 Roughly 53% of individuals trust their national institutions, and roughly 80% trust their friends and family. 10 About 30% of the world’s population experienced food insecurity in 2014-17. The descriptive statistics suggest there are systematic differences between trust and food insecurity depending on economic development. Respondents in low-income countries have higher likelihoods of trust in national institutions, but significantly lower trust in their friends and family. As expected, food insecurity is much more prevalent in low-income countries. Table 1. Descriptive statistics. Economic development rankings (GNI per capita) Lower-middle- Upper-middle- World sample Low-income income income High-income Dependent variables Trust in national institutions 0.530 0.566 0.564*** 0.439*** 0.557*** (0.367) (0.369) (0.371) (0.370) (0.347) Trust in friends and family 0.803 0.672 0.754*** 0.832*** 0.902*** (0.398) (0.469) (0.431) (0.374) (0.298) Food insecurity variables Food insecurity 0.303 0.623 0.348*** 0.277*** 0.088*** (0.460) (0.485) (0.476) (0.447) (0.283) Number of observations 387,385 68,925 106,018 100,397 112,045 9 Descriptive statistics by regions and economic development and for all control variables are shown in Appendix B. 10 Sample weights provided by Gallup are used to estimate all descriptive statistics and prevalence rates but are not used in the regression analyses. 9 Notes: Means calculated using sample-weighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses. Asterisks indicate whether the difference in means is statistically significantly different from low-income countries at the 0.01 level. Low-income countries are defined as those with a gross national income (GNI) per capita of $995 or less, lower-middle-income countries are those with a GNI per capita of more than $996 but less than $3,895, and upper-middle-income countries are those with a GNI per capita of more than $3,895 but less than $12,055, and high-income countries are those with a GNI per capita of more than $12,055. The remaining results are from the regression model discussed in section 2. Table 2 examines how food insecurity affects trust in national institutions (vertical trust, column 1) and trust in friends and family (horizontal trust, column 2). The estimates presented are the average marginal effects estimated by the linear model with country and year fixed effects. Table 2: The effect of food insecurity on vertical and horizontal trust. Trust Variables Trust in national institutions Trust in friends and family Food insecurity -0.027*** -0.117*** (0.004) (0.005) Female -0.010*** 0.013*** (0.003) (0.002) Age -0.004*** -0.008*** (0.000) (0.000) Age squared 0.000*** 0.000*** (0.000) (0.000) Number of adults 0.001 0.008*** (0.001) (0.001) Number of children 0.004*** -0.001* (0.001) (0.001) Married or Domestic partner 0.016*** -0.001 (0.002) (0.003) Separated Widowed or Divorced 0.001 -0.012*** (0.003) (0.003) Secondary education -0.046*** 0.056*** (0.004) (0.003) Post-secondary education -0.039*** 0.082*** 10 Trust Variables Trust in national institutions Trust in friends and family (0.005) (0.004) Employed Part Time -0.014*** -0.017*** (0.002) (0.004) Unemployed 0.001 -0.008* (0.003) (0.004) Out of workforce -0.013*** -0.016*** (0.003) (0.004) Second quintile household income 0.005* 0.026*** (0.003) (0.003) Middle quintile household income 0.007** 0.046*** (0.003) (0.004) Fourth quintile household income 0.007* 0.065*** (0.004) (0.004) Richest quintile household income 0.006 0.085*** (0.005) (0.005) Small town or suburb -0.022*** 0.000 (0.004) (0.003) Large city -0.047*** -0.005 (0.005) (0.004) Satisfied with roads 0.112*** 0.020*** (0.003) (0.003) Corruption in government -0.201*** 0.004 (0.008) (0.003) Money or property stolen -0.048*** -0.017*** (0.003) (0.003) Immigrant 0.000 -0.033*** (0.007) (0.005) Religiosity 0.046*** 0.024*** (0.005) (0.004) Country and year fixed effects yes yes Log likelihood -97827.642 -166327.437 R-squared 0.282 0.124 11 Trust Variables Trust in national institutions Trust in friends and family Number of Observations 388,602 388,602 Number of Countries 134 134 Note: Models estimated using unweighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses and clustered by country and robust to heteroskedasticity. References categories are single or never married, elementary education, employed full-time, poorest quintile, rural area or farm, and the year 2014. Each model includes country and years fixed effects. * Significance at the 0.10 level. ** Significance at the 0.05 level. *** Significance at the 0.01 level. Estimates show that food insecurity is strongly associated with both vertical and horizontal trust among the sample respondents. Trust is lower as the severity of food insecurity is higher. Food insecurity is associated with a decrease in trust in national institutions by 2.7 percentage points and a decrease in trust in friends and family by 11.7 percentage points. This is a sizable effect considering that the average probabilities of trusting national institutions and friends and family are roughly 53% and 80%, respectively. The effect of food insecurity on trust in friends and family is larger than for any other explanatory variable, including government corruption, poor infrastructure, or if the respondent has been robbed. Friends and family have more direct everyday interactions with people than the national government, or other national institutions such as the judiciary, which may explain the differences in magnitudes. These results imply that alleviating food insecurity would have a spillover effect of increasing both types of trust. Our control variables are well-behaved. For example, being an immigrant is a significant factor for trust in friends and family but is statistically insignificant for trust in institutions. Living in a city, compared to that of a rural area, is a significant factor to trust in institutions but is statistically insignificant to trust in friends and family. Column 2 shows that, controlling for other individual and household characteristics, immigrants are 3.3 percentage points less likely to trust friends and family. This suggests that, controlling for the level of food insecurity, immigrants may have access to less social capital to rely on in times of crisis, than native-born adults. Column 1 shows that, compared to living in a rural area, living in a large city is associated with a 4.7 percentage point decrease in trusting institutions. Putnam (2000) argues that rural areas have more social capital 12 than urban areas, and therefore those who live in small towns and rural areas have greater trust in acquaintances. Being in the richest quintile of household income, compared to being in the poorest quintile, is associated with an increase in trust in friends and family, but is insignificant to trust in institutions. This follows previous findings on the impact of inequality on trust (Uslaner, 2002; Elgar, 2010; Stephany, 2017). Women are more likely to have trust in friends and family than men, but less likely to trust in institutions. We find that an increase in the number of children in the household is associated with an increase in trust in institutions but a decrease in trust in friends and family (although the estimate is weakly significant). High levels of education are associated with increases in trust in friends and family but associated with decreases in trust in institutions. Research shows that income and education correlate with interpersonal trust, where a successful professional experience is likely to make individuals more trusting (Knack and Keefer, 1997; Alesina and La Ferrara, 2002; Li et al., 2005; Dincer, 2011). Conversely, if education leads expectations of government performance to rise faster than actual performance, education may diminish trust (Dalton, 2005; OECD, 2013). This illustrates the advantages of examining the two dimensions of trust independently. The likelihoods of both dimensions of trust decrease with age, but at a decreasing rate, and levels off with middle age (at roughly 40 for national institutions and 58 for friends and family). Table 2 also shows that working less than full-time and experiencing an episode of theft are both associated with decreases in both trust in institutions and trust in friends and family. While being a middle- income household, being satisfied with local infrastructure, and religiosity are all associated with increases in both trust in institutions and in friends and family. 3.1. Heterogeneity by Economic Development The relationship between food insecurity and trust varies across rankings of economic development (Table 3). Column 1 presents the results from the total sample, while Columns 2 through 5 represent the sample decomposed into low-income, lower-middle-income, upper- middle-income countries, and high-income (as defined earlier). 13 Table 3. The effect of food insecurity on vertical and horizontal trust by economic development. Economic development rankings (GNI per capita) Lower-middle- Upper-middle- World sample Low-income income income High-income Trust in national institutions Food insecurity -0.027*** -0.020*** -0.019*** -0.035*** -0.067*** (0.004) (0.006) (0.006) (0.007) (0.011) Trust in friends and family Food insecurity -0.117*** -0.120*** -0.109*** -0.111*** -0.129*** (0.005) (0.009) (0.008) (0.012) (0.009) Controls yes yes yes yes yes Number of observations 388,602 69,101 106,410 100,728 112,363 Notes: Models estimated using unweighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses and clustered by country and robust to heteroskedasticity. References categories are single or never married, elementary education, employed full-time, poorest quintile, rural area or farm, and the year 2014. Each model includes country and years fixed effects. Low-income countries are defined as those with a gross national income (GNI) per capita of $995 or less, lower-middle-income countries are those with a GNI per capita of more than $996 but less than $3,895, and upper-middle-income countries are those with a GNI per capita of more than $3,895 but less than $12,055, and high-income countries are those with a GNI per capita of more than $12,055. * Significance at the 0.10 level. ** Significance at the 0.05 level. *** Significance at the 0.01 level. We find that each category of food insecurity is associated with a decrease in both dimensions of trust across all levels of development. For trust in national institutions, the associations increase with income ranking, ranging from a 2 percentage-point decrease for low-income countries to a 6.7 percentage-point decrease for high-income countries. This implies that food insecurity has a larger effect on trust in richer countries. For trust in friends and family, the associations are nonlinear, starting at a 12 percentage-point decrease for low-income countries, rising to 11 percentage-point decrease for lower- and upper-middle countries, but then falling back to 12.9 percentage-point decrease in high income countries. 14 3.2. Sensitivity and Robustness Checks To address the endogeneity issues that may arise from the two-way relationship between food insecurity and trust, we perform a series of robustness and sensitivity checks to validate the results. First, we use entropy matching to create a matched sample of food-secure and food-insecure respondents (Table 4). Table C1 in Appendix C shows that entropy matching perfectly balances the distributions of all covariates. Table 4. The effect of food insecurity on vertical and horizontal trust: Entropy-matched sample. Economic development rankings (GNI per capita) Lower-middle- Upper-middle- World sample Low-income income income High-income Trust in national institutions Food insecurity -0.030*** -0.020*** -0.021*** -0.035*** -0.071*** (0.002) (0.003) (0.003) (0.003) (0.004) Trust in friends and family Food insecurity -0.119*** -0.128*** -0.112*** -0.113*** -0.127*** (0.002) (0.005) (0.004) (0.004) (0.005) Controls yes yes yes yes yes Number of observations 388,602 69,101 106,410 100,728 112,363 Notes: Models estimated using entropy-weighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses. References categories are single or never married, elementary education, employed full- time, poorest quintile, rural area or farm, and the year 2014. Each model includes country and years fixed effects. Low-income countries are defined as those with a gross national income (GNI) per capita of $995 or less, lower- middle-income countries are those with a GNI per capita of more than $996 but less than $3,895, and upper-middle- income countries are those with a GNI per capita of more than $3,895 but less than $12,055, and high-income countries are those with a GNI per capita of more than $12,055. * Significance at the 0.10 level. ** Significance at the 0.05 level. *** Significance at the 0.01 level. Column 1 of Table 4 presents the results from the total sample after matching, while Columns 2 through 5 represent the matched sample decomposed into economic development rankings. Overall, the results are very similar when compared to the non-matched sample (Table 3): food insecurity in each income category is associated with a decrease in both dimensions of trust across 15 all levels of development. The matching results offer robust evidence that food insecurity determines both vertical and horizontal dimensions of trust. However, while matching methods ensure findings are free of observable bias, it is still possible for unobserved factors to introduce hidden bias. Hidden bias may be caused by unobservable variables that cannot be controlled for in the matching process that simultaneously affect trust and the probability of being food insecure. To explore possible unobserved mechanisms, we examine whether measures of poor health and subject well-being mediate the results discussed above (Table C2, Appendix C). Health problems is a binary variable that equals one if the respondent has “any health problems that prevent you from doing any of the things people your age normally can do.” We also use GWP’s Life Evaluation Index, which is a categorical measure of respondents’ subjective well-being. The life evaluation index has two items where respondents were asked to place their perceived life status on two ladders that ranged from 0 to 10 corresponding to the worst and best possible life. The first ladder refers to the present and the second ladder refers to the future (i.e., about five years from the present time). The responses were categorized as Thriving if they scored 7 or higher on current life score and 8 or higher on future life score; Suffering if they scored 4 or lower on current and future life scores; and Struggling otherwise (Gallup, 2016). Overall, Table C2 shows that while experiencing a health problem and poor subjective wellbeing are significant negative factors determining both dimensions of trust, the effects of food insecurity are substantively unchanged from previous results. The coefficient for health problems is statistically insignificant in the total sample but is associated with an increase in trust in national institutions for low-income countries, and a decrease in trust in national institutions in high-income countries. Experiencing a health problem is also negatively associated with trust in friends in family for each income sample. Struggling and suffering, compared to thriving, are strongly negatively associated with both dimensions of trust, across all income classes. While poor health and subjective well-being are clearly important determinants of trust, their inclusion in the model 16 does not mediate the effect of food insecurity. Thus, there is no evidence that the relationship between food insecurity and trust operates through poor health or subjective well-being. 11 Next, to test the sensitivity of the institutions index we run the model for each of the four items that make up the index separately (trust in military, judicial and court systems, national government, and honesty of elections) and examine two additional measures of vertical trust: trust in local police and trust in the financial system and local banks (Table C3, Appendix). We find that the associations of food insecurity are stable, ranging from a decrease of 2.3 percentage points for trust in the honesty of elections to a 3.5 percentage-point decrease for trust in the military. It may also be the case that food insecurity impacts trust in local public institutions (at the meso- scale) more than national institutions.12 To test this hypothesis, we examine the relationship between food insecurity and trust in the local police department and trust in financial institutions and local banks (Columns 5 and 6). We find a significant negative effect of food insecurity on both local public institutions. Interestingly, we find that the magnitude of association at the meso-scale is somewhere between that of trust in friends and family (individual-level) and trust in national institutions (5.2 percentage-point decrease for local police and a 5.7 percentage-point decrease in trust in local banks). In sum, the sensitivity checks indicate the results are robust. The associations of food insecurity on trust in institutions and trust in friends and family remain relatively unchanged and statistically significant after using entropy matching methods. While health problems and subjective well- being are important determinants of trust, they do not mediate the effect of food insecurity. Additionally, the magnitude of this association is determined by the scale at which trust is measured. Food insecurity has a large significant effect on interpersonal levels of trust (like friends and family), somewhat less impact on local institutions (local police and local banks), and even less of an impact—but still strongly statistically significant—on national institutions (government, elections, judicial system, military). This reveals the importance of interpersonal social networks 11 To further account for potential omitted variable bias, we also conducted the Oster (2019) test and found that the delta coefficient is relatively small (<0.4) for all outcome variables, showing that food insecurity is virtually uncorrelated with unobservables. Full results are available upon request. 12 Bouckaert (2012) argues that trust in government can be analyzed at the macro-level (political institutions), meso- level (policy making), and at the micro-level (people’s daily lives). 17 to alleviating food insecurity (Smith et al., 2017a; 2017b) and the role of local decentralized governance structures. 4. Conclusion Trust in institutions is crucial during crises like the COVID-19 pandemic since a lack of trust can impede governments’ responses and slow emergency and recovery efforts (OECD, 2013; Bavel et al., 2020; Siegrist and Bearth, 2021). Interpersonal trust is a public good (Ostrom, 1997) based on people’s willingness to follow norms and follow institutional health guidelines to prevent pandemics (Johnson et al., 2020). Trust in friends and family is essential for risk-sharing and social networks that protect against shocks, especially in developing countries (Townsend, 1994; Ambrus et al., 2014; Barrett et al., 2022). Crime, corruption (Clausen et al., 2011), and inequality are well-studied trust factors, but food insecurity is not. We fill this gap in the literature using data from the 2014–2017 Gallup World Poll Survey. In a global sample of 387,385 adults from 134 countries, food insecurity is negatively associated with both vertical (trust in national institutions) and horizontal (trust in friends and family) measures of trust. The probability of trusting in national institutions is 2.7 percentage- points lower among the food insecure. The probability of trusting in friends and family is 11.7 percentage-points lower among the food insecure. The magnitude of the association of food insecurity with trust in friends and family is larger than any other explanatory variable. These relationships also vary significantly by GNI per capita country rankings. As a robustness check, we employ an entropy matching strategy to control for the possible endogeneity of food insecurity. Results remain robust and sensitivity analyses using other measures of trust support our results. Several limitations are worth mentioning. Since the data are cross-sectional, we are unable to control for unobserved individual heterogeneity and thus some hidden bias may remain. While a robustness check confirms that food insecurity is not mediated by poor health or subjective well- being, other unobservable factors may bias the results. Thus, the estimates are correlational and only reservedly causal under strong assumptions. Additionally, the survey questions on trust could be defined more precisely. GWP respondents are asked about the confidence they have in national government, without first defining “national government.” Some respondents might equate 18 government with political leadership while others may equate it with the bureaucracy (OECD, 2013). Food insecurity may act as a multiplier for the pandemic due to its negative economic and health effects (Smith and Wesselbaum, 2020). This paper suggests that food insecurity may exacerbate the pandemic by decreasing vertical and horizontal trust, delaying recovery. In Africa and South Asia, the ongoing conflict in Ukraine has worsened the already grave threat of food insecurity. It is crucial to address food insecurity not only for its immediate humanitarian benefits, but also for the deeper issues of trust, social contract, and stability that are necessary for overall development. 19 References Acemoglu, D., Johnson, S., and Robinson, J. A., 2001. The colonial origins of comparative development: an empirical investigation American Economic Review, 91(5): 1369-1401 Alesina, A. and La Ferrara, E., 2002. Who trusts others? Journal of Public Economics, 85: 207- 234. Alesina, A. and La Ferrara, E., 2005. Ethnic diversity and economic performance. Journal of Economic Literature, 43: 762-800. Ambrus, A., Mobius, M., and Szeidl, A., 2014. Consumption risk-sharing in social networks. American Economic Review, 104: 149-182. Algan, Y. and Cahuc, P., 2010. Inherited trust and growth. American Economic Review, 100: 2060- 2092. Barrett, C. B., 1997. Idea gaps, object gaps, and trust gaps in economic development. Journal of Developing Areas, 31: 553-568. Barrett, C. B., 2005. On the relevance of identities, communities, groups, and networks to the economics of poverty alleviation. The social economics of poverty. Routledge New York; 1-11. Barrett, C. B., Reardon, T., Swinnen, J., & Zilberman, D. (2022). Agri-food value chain revolutions in low-and middle-income countries. Journal of Economic Literature, 60(4), 1316-77. Bavel, J. J. V., Baicker, K., Boggio, P. S. et al., 2020. Using social and behavioural science to support COVID-19 pandemic response. Nature Human Behavior, 4: 460-471. Bouckaert, G., 2012. Trust and public administration. Administration, 60: 91-115. 20 Cafiero, C., Nor,d M., Viviani, S., et al., 2016. Methods for estimating comparable rates of food insecurity experienced by adults throughout the world. Technical report, Food and Agriculture Organization of the United Nations, Rome. Clausen, B., Kraay, A. and Nyiri, Z., 2011. Corruption and confidence in public institutions: Evidence from a global survey. The World Bank Economic Review, 25(2), pp.212-249. Corbacho, A., Philipp, J., and Ruiz-Vega, M., 2015. Crime and erosion of trust: Evidence for Latin America. World Development, 70: 400-415. Dalton R. J., 2005. The social transformation of trust in government. International Review of Sociology, 15: 133-154. Dean, W. R. and Sharkey, J. R., 2011. Food insecurity, social capital and perceived personal disparity in a predominantly rural region of Texas: an individual-level analysis. Social Science & Medicine, 72: 1454-1462. Dean, W. R., Sharkey, J. R., and Johnson, C. M., 2011. Food insecurity is associated with social capital, perceived personal disparity, and partnership status among older and senior adults in a largely rural area of central Texas. Journal of Nutrition in Gerontology and Geriatrics, 30: 169- 186. Elgar, F. J., 2010. Income inequality, trust, and population health in 33 countries. American Journal of Public Health, 100: 2311-2315. Dincer, O. C., 2011. Ethnic diversity and trust. Contemporary Economic Policy, 29: 284-293. FAO, UNICEF, WFP, WHO, IFAD, 2020a. The State of Food Security and Nutrition in the World 2020. Transforming food systems for affordable healthy diets. Rome, FAO. 21 FAO, 2020b. Urban food systems and COVID-19: the role of cities and local governments in responding to the emergency. Policy brief. FAO Rome. Ferraro, P. J. and Miranda, J. J., 2014. The performance of non-experimental designs in the evaluation of environmental programs: a design-replication study using a large-scale randomized experiment as a benchmark. Journal of Economic Behavior & Organization, 107: 344-365. Frongillo, E. A., Nguyen, H. T., Smith, M. D., and Coleman-Jensen, A., 2017. Food insecurity is associated with subjective well-being among individuals from 138 countries in the 2014 Gallup World Poll. Journal of Nutrition, 147: 680-687. Frongillo, E. A., Nguyen, H. T., Smith, M. D., and Coleman-Jensen, A., 2019. Food Insecurity Is More Strongly Associated with Poor Subjective Well-Being in More-Developed Countries than in Less-Developed Countries. Journal of Nutrition, 149: 330-335. Gallup, 2016. Worldwide research methodology and codebook. Gillen B., Snowberg E., Yariv L., 2019. Experimenting with measurement error: techniques with applications to the Caltech cohort study. Journal of Political Economy, 127(4): 1826-1863. Hainmueller, J., 2012. Entropy balancing for causal effects: A multivariate reweighting method to produce balanced samples in observational studies. Political Analysis, 25-46. Hellegers, P. (2022). Food security vulnerability due to trade dependencies on Russia and Ukraine. Food Security, 14(6), 1503-1510. Hudson, J., 2006. Institutional trust and subjective well‐being across the EU. Kyklos. 59: 43-62. Johnson, T., Dawes, C., Fowler, J., and Smirnov, O., 2020. Slowing COVID-19 transmission as a social dilemma: Lessons for government officials from interdisciplinary research on cooperation. Journal of Behavioral Public Administration, 3. 22 Knack, S. and Keefer, P., 1997. Does social capital have an economic payoff? A cross-country investigation. Quarterly Journal of Economics, 112: 1251-1288. Laborde, D., Martin, W., Swinnen, J., and Vos, R., 2020. COVID-19 risks to global food security. Science, 369: 500-502. Leblang, D., Smith, M. D., and, D., 2022. The effect of trust on economic performance and financial access. Economics Letters, 220: 110884. Mishler, W. and Rose, R., 2001. What are the origins of political trust? Testing institutional and cultural theories in post-communist societies. Comparative Political Studies, 34: 30-62. Li, Y., Pickles, A., and Savage, M., 2005. Social capital and social trust in Britain. European Sociological Review, 21: 109-123. Nord, M., Cafiero, C., and Viviani, S., 2016. Methods for estimating comparable prevalence rates of food insecurity experienced by adults in 147 countries and areas. In Journal of Physics: Conference Series, 772(1). OECD, 2013. Trust in government, policy effectiveness and the governance agenda. Government at a Glance 2013. OECD Publishing, Paris. Oster, E., 2019. Unobservable selection and coefficient stability: Theory and evidence. Journal of Business & Economic Statistics, 37(2), 187-204. Ostrom, E., 1998. A behavioral approach to the rational choice theory of collective action: Presidential address, American Political Science Association, 1997. American Political Science Review, 1-22. 23 Putnam, R. D., 2000. Bowling alone: The collapse and revival of American community. Simon and Schuster. Ruyssen, I. and Salomone, S., 2018. Female migration: A way out of discrimination? Journal of Development Economics, 130: 224-241. Siegrist, M. and Bearth, A., 2021. Worldviews, trust, and risk perceptions shape public acceptance of COVID-19 public health measures. Proceedings of the National Academy of Sciences, 118(24). Smith, M. D. and Wesselbaum, D., 2020. COVID-19, Food Insecurity, and Migration. The Journal of Nutrition, 150(110): 2855-2858. Smith, M. D. and Floro, M. S., 2020. Food insecurity, gender, and international migration in low- and middle-income countries. Food Policy, 91: 101837. Smith, M. D., Kassa, W., and Winters P., 2017a. Assessing food insecurity in Latin America and the Caribbean using FAO’s Food Insecurity Experience Scale. Food Policy, 71: 48-61. Smith, M. D., Rabbitt, M., and Coleman-Jensen, A., 2017b. Who are the world’s food insecure? New evidence from the Food and Agriculture Organization’s Food Insecurity Experience Scale. World Development, 93: 402-412. Stephany, F., 2017. Who are your joneses? socio-specific income inequality and trust. Social Indicators Research, 134: 877-898. Stuart, E. A., 2010. Matching methods for causal inference: A review and a look forward. Statistical Science, 25(1): 1-21. Townsend, R. M., 1994. Risk and insurance in village India. Econometrica, 62: 539-591. Uslaner, E. M., 2002. The moral foundations of trust. Cambridge University Press. 24 World Bank, 2019. Social Contracts and World Bank Country Engagements: Lessons from Emerging Practices. IEG Meso Evaluation. Independent Evaluation Group. World Bank, Washington, DC. Wunderlich, G. S. and Norwood, J. L., 2006. Food insecurity and hunger in the United States. National Academies Press. 25 Appendix A Measuring Food Insecurity In line with the US HFSSM and ELCSA, an individual’s food security status is determined by summing the affirmed FIES responses. This raw score classification is not comparable across countries, however, because the same number of affirmed responses would not necessarily correspond to the same level of severity in different countries. Differences across countries in, for example, languages, livelihood arrangements, and food-related cultural norms and expectations may affect how respondents understand the FIES questions, and, in turn, may affect their responses. The FAO therefore renders nation-specific scales comparable by creating food insecurity thresholds that partition the continuum of food insecurity into meaningful and comparable ranges of food insecurity, an FIES Global Standard Scale (FIES GSS). To ensure the measured severity of food insecurity is comparable across countries, FAO uses Item Response Theory (i.e., the Rasch model) to equate the food insecurity scales for each country to the FIES Global Standard Scale (FAO, 2016). The FAO equating procedure maintains cross-country comparability by creating two standard food insecurity thresholds: moderate food insecurity and severe food insecurity. The thresholds are adjusted to place each country’s scale on the same metric as the global standard. The resulting food security prevalence rates and severity measures are then equivalent and comparable across countries (for more details, see Cafiero et al., 2016). Individuals whose raw score equals zero are classified as food secure. Individuals with a raw score of at least one and less than the country specific FIES GSS threshold for moderate food insecurity are classified as experiencing mild food insecurity. Those who report a raw score at least equal to the FIES GSS threshold for moderate food insecurity but less than the FIES GSS threshold for severe food insecurity are deemed moderately food insecure. Those above the country specific FIES GSS threshold for severe food insecurity are classified as such. Following the SDG Target 2.1 indicator definition, the prevalence of moderate and severe food insecure individuals represents the sum of those two categories. 26 Entropy Matching Entropy balancing (EB) produces individual weights that are used to reweight the control group so that they have the same sample moments (i.e., mean and variance) as the treatment group (Hainmueller, 2012). EB gives each observation in the treatment group (food insecure) a weight equal to one and each observation in the control group (food secure) is assigned a balancing weight. The EB algorithm first imposes balance constraints on the sample moments of the control variables, and then selects a weighting scheme that minimizes a loss function, as measured by directed entropy divergence. We use the first three moments, namely the mean, variance, and skewness, to ensure the exact balancing of treatment and control groups. EB offers several advantages over more conventional propensity score methods. First, it is more effective at balancing the treatment and control groups and never produces a worse balance. Propensity score matching on the other hand can improve balance on some covariates by decreasing balance on other covariates. Second, EB is fully nonparametric and does not rely on the functional form assumptions necessary for propensity score models. Third, EB eliminates the need for manual balance checks by researchers since the covariate moments are automatically balanced by the algorithm. Table A1. Survey Questions: FAO Food Insecurity Experience Scale (FIES) Q1. During the last 12 MONTHS, was there a time when You were worried you would not have enough food to eat because of a lack of money or other resources? Q2. Still thinking about the last 12 MONTHS, was there a time when you were unable to eat healthy and nutritious food because of a lack of money or other resources? Q3. Was there a time when you ate only a few kinds of foods because of a lack of money or other resources? Q4. Was there a time when you had to skip a meal because there was not enough money or other resources to get food? Q5. Still thinking about the last 12 MONTHS, was there a time when you ate less than you thought you should because of a lack of money or other resources? Q6. Was there a time when your household ran out of food because of a lack of money or other resources? Q7. Was there a time when you were hungry but did not eat because there was not enough money or other resources for food? 27 Q8. During the last 12 MONTHS, was there a time when you went without eating for a whole day because of a lack of money or other resources? Source: United Nations, Food and Agriculture Organization, Voices of the Hungry project. 28 Appendix B Table B1. Correlation matrix of the trust and food insecurity variables. Trust in national Variables institutions Trust in friends and family Food insecurity Trust in national institutions 1.000 Trust in friends and family 0.030*** 1.000 Food insecurity -0.045*** -0.220*** 1.000 Note: Data from the 2014-17 Gallup World Poll. Asterisks indicate statistical significance of correlations at the 0.01 level. Table B1 shows the correlations between trust in national institutions, trust in friends and family, and food insecurity. There is a relatively strong negative correlation between both trust measures and food insecurity, but the magnitude of association is much larger for trust in friends and family than for trust in national institutions. 29 Table B2. Descriptive statistics of control variables. Economic development rankings (GNI per capita). Variables World sample Low-income Lower-middle-income Upper-middle-income High-income Female 0.502 0.494 0.497 0.506*** 0.508*** (0.500) (0.500) (0.500) (0.500) (0.500) Age 39.28 33.32 36.11*** 39.74*** 45.51*** (17.45) (15.27) (16.11) (17.24) (18.07) Number of adults 3.310 3.981 3.666*** 3.214*** 2.649*** (1.909) (2.376) (1.886) (1.591) (1.632) Number of children 1.461 3.159 1.718*** 1.077*** 0.524*** (2.115) (3.076) (1.975) (1.557) (0.988) Single 0.325 0.323 0.346*** 0.349*** 0.286*** (0.468) (0.468) (0.476) (0.477) (0.452) Married or Domestic partner 0.569 0.587 0.571*** 0.538*** 0.585** (0.495) (0.492) (0.495) (0.499) (0.493) Separated Widowed or Divorced 0.106 0.0904 0.0835*** 0.114*** 0.129*** (0.307) (0.287) (0.277) (0.317) (0.335) Elementary education 0.390 0.704 0.469*** 0.321*** 0.187*** (0.488) (0.457) (0.499) (0.467) (0.390) Secondary education 0.487 0.271 0.453*** 0.545*** 0.598*** (0.500) (0.444) (0.498) (0.498) (0.490) Post-secondary education 0.123 0.0258 0.078*** 0.134*** 0.215*** (0.328) (0.158) (0.268) (0.340) (0.411) Employed Full Time 0.406 0.356 0.372*** 0.405*** 0.468*** (0.491) (0.479) (0.483) (0.491) (0.499) 30 Economic development rankings (GNI per capita). Variables World sample Low-income Lower-middle-income Upper-middle-income High-income Employed Part Time 0.146 0.188 0.145*** 0.148*** 0.118*** (0.353) (0.391) (0.352) (0.355) (0.323) Unemployed 0.0840 0.144 0.0873*** 0.0724*** 0.0546*** (0.277) (0.351) (0.282) (0.259) (0.227) Out of workforce 0.365 0.312 0.396*** 0.374*** 0.359*** (0.481) (0.463) (0.489) (0.484) (0.480) Poorest quintile household income 0.193 0.192 0.194 0.195 0.191 (0.395) (0.394) (0.396) (0.396) (0.393) Second quintile household income 0.199 0.197 0.198 0.199 0.199 (0.399) (0.398) (0.399) (0.399) (0.399) Middle quintile household income 0.199 0.201 0.199 0.198 0.200 (0.399) (0.401) (0.399) (0.399) (0.400) Fourth quintile household income 0.203 0.204 0.202 0.202 0.203 (0.402) (0.403) (0.401) (0.402) (0.402) Richest quintile household income 0.206 0.206 0.207 0.205 0.207 (0.405) (0.405) (0.405) (0.404) (0.405) Rural area or farm 0.267 0.435 0.336*** 0.239*** 0.122*** (0.442) (0.496) (0.472) (0.427) (0.328) Small town or suburb 0.446 0.426 0.419*** 0.403*** 0.524*** (0.497) (0.495) (0.493) (0.490) (0.499) Large city 0.287 0.139 0.245*** 0.358*** 0.353*** (0.452) (0.346) (0.430) (0.479) (0.478) Satisfied with roads 0.523 0.350 0.536*** 0.486*** 0.651*** 31 Economic development rankings (GNI per capita). Variables World sample Low-income Lower-middle-income Upper-middle-income High-income (0.499) (0.477) (0.499) (0.500) (0.477) Corruption in government 0.755 0.778 0.798*** 0.826*** 0.636*** (0.430) (0.416) (0.402) (0.379) (0.481) Money or property stolen 0.166 0.248 0.173*** 0.167*** 0.108*** (0.372) (0.432) (0.379) (0.373) (0.311) Immigrant 0.0417 0.0236 0.0131*** 0.030*** 0.0905*** (0.200) (0.152) (0.114) (0.171) (0.287) Religiosity 0.729 0.937 0.881*** 0.742*** 0.445*** (0.445) (0.243) (0.324) (0.438) (0.497) Health problems 0.252 0.308 0.271*** 0.247*** 0.204*** (0.434) (0.462) (0.444) (0.431) (0.403) Thriving 0.262 0.108 0.180*** 0.278*** 0.414*** (0.440) (0.310) (0.385) (0.448) (0.493) Struggling 0.606 0.689 0.681*** 0.583*** 0.508*** (0.489) (0.463) (0.466) (0.493) (0.500) Suffering 0.132 0.204 0.139*** 0.140*** 0.0781*** (0.339) (0.403) (0.346) (0.347) (0.268) Number of observations 388,602 69,101 106,410 100,728 112,363 Notes: Means calculated using sample-weighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses. Asterisks indicate whether the difference in means is statistically significantly different from low-income countries. 32 Appendix C Table C1: Covariate balancing: differences in means before and after matching. Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference Female mean 0.495 0.518 0.023*** 0.538 0.538 0.000 sd 0.500 0.500 0.499 0.499 variance 0.250 0.250 0.249 0.249 skewness 0.019 -0.071 -0.154 -0.154 kurtosis 1.000 1.005 1.024 1.024 Age mean 40.319 36.887 -3.432*** 37.878 37.906 0.028 sd 17.846 16.235 16.410 16.346 variance 318.484 263.583 269.280 267.188 skewness 0.497 0.751 0.700 0.730 kurtosis 2.314 2.898 2.721 2.826 Number of adults mean 3.210 3.541 0.332*** 2.961 2.953 -0.007 sd 1.795 2.131 1.910 1.922 variance 3.220 4.541 3.650 3.693 skewness 7.004 3.920 3.181 3.109 kurtosis 289.656 99.949 61.453 60.259 Number of children mean 1.080 2.335 1.255*** 2.103 2.118 0.015 sd 1.721 2.617 2.486 2.474 variance 2.962 6.848 6.183 6.119 skewness 8.476 5.265 6.640 6.739 kurtosis 345.619 120.223 195.960 191.175 33 Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference Single mean 0.324 0.329 0.005*** 0.317 0.317 0.000 sd 0.468 0.470 0.465 0.465 variance 0.219 0.221 0.216 0.216 skewness 0.754 0.729 0.788 0.788 kurtosis 1.569 1.531 1.621 1.621 Married or domestic partner mean 0.575 0.556 -0.019*** 0.541 0.541 0.000 sd 0.494 0.497 0.498 0.498 variance 0.244 0.247 0.248 0.248 skewness -0.304 -0.226 -0.166 -0.166 kurtosis 1.092 1.051 1.028 1.028 Separated widowed or divorced mean 0.101 0.115 0.014*** 0.142 0.142 0.000 sd 0.302 0.319 0.349 0.349 variance 0.091 0.102 0.122 0.122 skewness 2.640 2.413 2.052 2.052 kurtosis 7.971 6.822 5.211 5.211 Elementary education mean 0.301 0.596 0.295*** 0.530 0.530 0.000 sd 0.459 0.491 0.499 0.499 variance 0.210 0.241 0.249 0.249 skewness 0.868 -0.392 -0.121 -0.121 kurtosis 1.753 1.154 1.015 1.015 Secondary education mean 0.539 0.367 -0.171*** 0.421 0.421 0.000 sd 0.499 0.482 0.494 0.494 variance 0.249 0.232 0.244 0.244 34 Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference skewness -0.155 0.550 0.322 0.322 kurtosis 1.024 1.303 1.104 1.104 Completed four years beyond hs mean 0.160 0.036 -0.124*** 0.049 0.049 0.000 sd 0.367 0.187 0.216 0.216 variance 0.135 0.035 0.047 0.047 skewness 1.851 4.949 4.169 4.170 kurtosis 4.426 25.495 18.384 18.385 Employed full time mean 0.432 0.345 -0.087*** 0.350 0.350 0.000 sd 0.495 0.475 0.477 0.477 variance 0.245 0.226 0.227 0.227 skewness 0.275 0.652 0.630 0.630 kurtosis 1.076 1.425 1.397 1.397 Employed part time mean 0.125 0.194 0.070*** 0.192 0.192 0.000 sd 0.330 0.396 0.394 0.394 variance 0.109 0.156 0.155 0.155 skewness 2.274 1.546 1.564 1.564 kurtosis 6.171 3.390 3.447 3.447 Unemployed mean 0.064 0.129 0.065*** 0.129 0.129 0.000 sd 0.245 0.335 0.335 0.335 variance 0.060 0.113 0.112 0.112 skewness 3.552 2.210 2.216 2.216 kurtosis 13.616 5.885 5.909 5.909 Out of workforce mean 0.379 0.331 -0.048*** 0.329 0.329 0.000 35 Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference sd 0.485 0.471 0.470 0.470 variance 0.235 0.222 0.221 0.221 skewness 0.498 0.716 0.726 0.726 kurtosis 1.248 1.513 1.527 1.527 Poorest quintile household income mean 0.155 0.281 0.126*** 0.244 0.244 0.000 sd 0.362 0.450 0.429 0.429 variance 0.131 0.202 0.184 0.184 skewness 1.907 0.974 1.194 1.194 kurtosis 4.636 1.949 2.426 2.426 Second quintile household income mean 0.182 0.236 0.054*** 0.217 0.217 0.000 sd 0.386 0.425 0.412 0.412 variance 0.149 0.180 0.170 0.170 skewness 1.646 1.243 1.376 1.376 kurtosis 3.711 2.545 2.892 2.892 Middle quintile household income mean 0.200 0.198 -0.002 0.198 0.198 0.000 sd 0.400 0.399 0.399 0.399 variance 0.160 0.159 0.159 0.159 skewness 1.502 1.516 1.514 1.514 kurtosis 3.255 3.297 3.293 3.293 Fourth quintile household income mean 0.219 0.165 -0.054*** 0.182 0.182 0.000 sd 0.414 0.371 0.386 0.386 variance 0.171 0.138 0.149 0.149 skewness 1.358 1.806 1.652 1.652 36 Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference kurtosis 2.844 4.262 3.728 3.728 Richest quintile household income mean 0.244 0.120 -0.124*** 0.160 0.160 0.000 sd 0.429 0.325 0.366 0.366 variance 0.184 0.106 0.134 0.134 skewness 1.193 2.340 1.856 1.856 kurtosis 2.423 6.476 4.446 4.446 Rural area or farm mean 0.231 0.347 0.116*** 0.335 0.335 0.000 sd 0.422 0.476 0.472 0.472 variance 0.178 0.227 0.223 0.223 skewness 1.273 0.642 0.697 0.697 kurtosis 2.622 1.413 1.486 1.486 Small town or suburb mean 0.450 0.438 -0.013*** 0.438 0.438 0.000 sd 0.498 0.496 0.496 0.496 variance 0.248 0.246 0.246 0.246 skewness 0.200 0.251 0.249 0.249 kurtosis 1.040 1.063 1.062 1.062 Large city mean 0.318 0.215 -0.103*** 0.226 0.226 0.000 sd 0.466 0.411 0.419 0.419 variance 0.217 0.169 0.175 0.175 skewness 0.780 1.386 1.307 1.307 kurtosis 1.609 2.922 2.709 2.709 Satisfied with roads mean 0.570 0.417 -0.153*** 0.419 0.419 0.000 sd 0.495 0.493 0.493 0.493 37 Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference variance 0.245 0.243 0.243 0.243 skewness -0.282 0.338 0.327 0.327 kurtosis 1.080 1.114 1.107 1.107 Corruption in government mean 0.732 0.808 0.077*** 0.813 0.813 0.000 sd 0.443 0.394 0.390 0.390 variance 0.196 0.155 0.152 0.152 skewness -1.047 -1.567 -1.602 -1.602 kurtosis 2.096 3.457 3.567 3.567 Money or property stolen mean 0.128 0.254 0.126*** 0.252 0.252 0.000 sd 0.334 0.435 0.434 0.434 variance 0.111 0.189 0.189 0.189 skewness 2.230 1.132 1.140 1.140 kurtosis 5.974 2.281 2.299 2.299 Immigrant mean 0.048 0.028 -0.020*** 0.028 0.028 0.000 sd 0.213 0.164 0.166 0.166 variance 0.045 0.027 0.028 0.028 skewness 4.240 5.747 5.686 5.686 kurtosis 18.979 34.027 33.335 33.335 Religiosity mean 0.667 0.869 0.202*** 0.875 0.875 0.000 sd 0.471 0.337 0.331 0.331 variance 0.222 0.114 0.109 0.109 skewness -0.710 -2.192 -2.268 -2.268 kurtosis 1.504 5.807 6.142 6.142 38 Before entropy matching After entropy matching Matching variable: Food insecurity Statistics Food security Food insecurity Mean difference Food security Food insecurity Mean difference Observations n 274,166 114,436 274,166 114,436 Note: Entropy matching for food insecurity. 39 Table C2: The effect of food insecurity on vertical and horizontal trust: Experiential well-being and health problems. Economic development rankings (GNI per capita) Lower-middle- Upper-middle- World sample Low-income income income High-income Trust in national institutions Food insecurity -0.020*** -0.021*** -0.015** -0.024*** -0.053*** (0.003) (0.006) (0.006) (0.006) (0.009) Health problems -0.001 0.019*** 0.000 -0.001 -0.017*** (0.002) (0.006) (0.004) (0.004) (0.003) Struggling -0.046*** -0.030*** -0.035*** -0.048*** -0.044*** (0.003) (0.008) (0.006) (0.006) (0.004) Suffering -0.093*** -0.050*** -0.079*** -0.118*** -0.102*** (0.006) (0.010) (0.011) (0.010) (0.012) Trust in friends and family Food insecurity -0.097*** -0.105*** -0.093*** -0.083*** -0.100*** (0.005) (0.009) (0.008) (0.010) (0.008) Health problems -0.041*** -0.062*** -0.033** -0.043*** -0.030*** (0.005) (0.009) (0.012) (0.007) (0.005) Struggling -0.040*** -0.037*** -0.049*** -0.041*** -0.037*** (0.003) (0.008) (0.007) (0.004) (0.004) Suffering -0.163*** -0.142*** -0.161*** -0.182*** -0.171*** (0.007) (0.010) (0.010) (0.013) (0.022) Controls yes yes yes yes yes Number of observations 359,330 64,228 96,609 91,115 107,378 40 Notes: Models estimated using entropy-weighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses. References categories are single or never married, elementary education, employed full-time, poorest quintile, rural area or farm, and the year 2014. Each model includes country and years fixed effects. Low-income countries are defined as those with a gross national income (GNI) per capita of $995 or less, lower-middle-income countries are those with a GNI per capita of more than $996 but less than $3,895, and upper-middle-income countries are those with a GNI per capita of more than $3,895 but less than $12,055, and high-income countries are those with a GNI per capita of more than $12,055. * Significance at the 0.10 level. ** Significance at the 0.05 level. *** Significance at the 0.01 level. Table C3: The effect of food insecurity on vertical and horizontal trust: Robustness checks Robustness checks Trust in national institutions Trust in financial Trust in Trust in judicial Trust in national Trust in honesty of Trust in local institutions and local Variables military system and courts government elections police banks Food insecurity -0.035*** -0.030*** -0.031*** -0.023*** -0.052*** -0.057*** (0.004) (0.005) (0.004) (0.005) (0.004) (0.005) Controls yes yes yes yes yes yes Country fixed effects yes yes yes yes yes yes Log likelihood -190971.89 -227545.45 -221680.61 -220973.53 -221942.74 -226810.97 R-squared 0.166 0.192 0.231 0.217 0.153 0.167 Number of Observations 361,535 367,510 372,799 366,142 373,915 367,258 Note: Models estimated using unweighted individual-level data from the 2014-17 Gallup World Poll. Standard errors are in parentheses and clustered by country and robust to heteroskedasticity. References categories are single or never married, elementary education, employed full-time, poorest quintile, rural area or farm, and the year 2014. Each model includes country and years fixed effects. * Significance at the 0.10 level. ** Significance at the 0.05 level. *** Significance at the 0.01 level. 41