The World Bank Economic Review, 37(2), 2023, 221–234 https://doi.org10.1093/wber/lhac030 Article Cash Transfers, Trust, and Inter-household Transfers: Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 Experimental Evidence from Tanzania David K. Evans and Katrina Kosec Abstract Institutionalized conditional cash transfer (CCT) programs may affect pre-existing, informal safety nets such as inter-household transfers and trust among community members. This study reports on a randomized controlled trial used to test the impact of CCTs on various measures of trust and informal safety nets within communities in Tanzania. It provides evidence that the introduction of a CCT program increased program beneficiaries’ trust in other community members and their perceived ability to access support from other households (e.g., childcare). Although CCTs reduced the total size of transfers to beneficiary households in the community in the short run (after 1.75 years of transfers), that reduction had disappeared 2.75 years after transfers began. Taken together, this evidence suggests that formal CCT programs do not necessarily crowd out informal safety nets in the longer term, and they may in fact boost trust and support across households. JEL classification: H31, H55, I38, O12, O15 Keywords: conditional cash transfers, informal safety nets, service delivery, trust 1. Introduction The transfer of cash and other goods to protect the well-being of households experiencing poverty and adverse shocks extends far back in history, and has occurred largely through informal transfers from household to household. Today, more and more countries are implementing government-administered safety nets like conditional cash transfer (CCT) programs to protect households from extreme poverty.1 The introduction of CCT programs may reduce the need for households to rely on each other for monetary or non-monetary assistance. However, one concern with this diminishing need is that formal programs may crowd out informal transfers, such that—if formal systems are disrupted or the size of formal trans- fers proves inadequate to fully insure against shocks—households may be left with insufficient access to David K. Evans (corresponding author) is a Senior Fellow at the Center for Global Development (CGD), Washington DC, United States; his email address is devans@cgdev.org. Katrina Kosec is a Senior Research Fellow at the International Food Policy Research Institute (IFPRI), Washington DC, United States; her email address is k.kosec@cgiar.org. The authors thank operational experts at the World Bank (especially Samantha de Silva), the International Food Policy Research Institute (IFPRI), the Tanzania Social Action Fund (TASAF), specifically Director Ladislaus Mwamanga, former Executive Director Servacius Likwelile, and Amadeus Kamagenge and his team, for insightful comments and substantive and logistical support. The authors are grateful to Eric Edmonds, Selim Gulesci, and three anonymous referees for helpful comments and, for expert research assistance, to Brian Holtemeyer, Amina Mendez Acosta, and Lucia Carrillo. A supplementary online appendix is available with this article at The World Bank Economic Review website. 1 Some cash transfer programs come with conditions on household behaviors (CCTs) and others do not (unconditional cash transfers), but almost none specifically restrict how money is spent. © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com 222 Evans and Kosec consumption smoothing mechanisms and other forms of assistance. At the same time, CCTs may boost beneficiaries’ engagement with other community members—potentially increasing trust in other commu- nity members and access to both informal transfers and other forms of support. This paper documents the impact of the randomized introduction of a pilot CCT program in Tanzania in 2010 on trust and informal safety nets across households. It utilizes individual- and household-level data from two follow-up surveys to analyze the effects of the program after 1.75 years of transfers and Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 after 2.75 years. It includes measures of both formal and informal transfers into and out of households as well as more subjective measures of trust in and feelings of being supported by other community members, thus providing a relatively nuanced look at the impacts of the program. The results show that trust in community members is unaffected after 1.75 years of transfers, but that it increases after 2.75 years have passed. By that time, beneficiary households report increases in trust both in people in general and in community members, as well as trust in shopkeepers and teachers specifically. Likewise, after 2.75 years, beneficiary households report positive impacts on two non-monetary aspects of informal safety nets, including the perceived ability to access childcare from other households and a greater likelihood that someone from another household has turned to their household for assistance with a personal problem recently. The impact on total inter-household monetary transfers dips after 1.75 years but then recovers by the time 2.75 years have passed; in other words, by 2.75 years, beneficiary households report receiving a comparable amount in inter-household transfers as they did at baseline. (At the extensive margin, there is some evidence of an enduring reduction in the proportion of households that report receiving inter- household transfers.) Even in the shorter-term follow-up, the apparent crowd-out of total inter-household transfers is small, equal to less than 16 percent of the size of formal government transfers. These results contribute to the literature in several ways. To our knowledge, this is among the first studies in Sub-Saharan Africa to examine the relationship between CCTs and either trust in other indi- viduals or informal safety nets (including both transfers and other forms of support). On such outcomes, one might expect very different results across countries and regions due to a variety of factors, including variation in baseline levels of trust (Knack and Keefer 1997). One other study to examine the relationship between cash transfers and interpersonal trust in the region examines the impact of unconditional cash transfers (in Kenya) and finds no impact on trust (Haushofer and Shapiro 2016). Another, mixed methods study from the region on unconditional cash transfers in Ghana shows increased inter-household support, both emotional and material, as a result of the transfers (de Milliano et al. 2021).2 However, uncondi- tional and conditional cash transfers may vary in their impacts on relationships between beneficiaries and other members of the community: because of the imposition of conditions, beneficiaries may be dis- gruntled or resentful toward government and even toward other members of the community (if those community members play a role in targeting). On the other hand, conditions may lead community mem- bers to view beneficiaries as more deserving than they would otherwise. Further, the conditions themselves may require beneficiary households to increase engagement with service providers and other community members (e.g., near schools and clinics). That said, unconditional transfers may also practically boost en- gagement in education and health services by boosting income, which may similarly increase interaction with other community members.3 Both types of transfers may involve community meetings (as is the case with the program evaluated in Tanzania), both to manage the program and to provide information about it, which could also serve to increase community engagement. 2 Evidence from other regions, such as Latin America, suggests that unconditional cash transfers have mixed effects— sometimes positive, sometimes neutral—on social cohesion, as documented in Ecuador (Valli, Peterman, and Hidrobo 2019). 3 The evidence on unconditional cash transfers and educational participation is largely but not universally positive, whereas the evidence on health services is weaker (Haushofer and Shapiro 2016; Kilburn et al. 2017; Pega et al. 2022). The World Bank Economic Review 223 Other evidence on CCTs and something akin to trust in other community members comes from lab-in- the-field evidence in Colombia, where CCT communities showed higher levels of cooperation than non- CCT communities (Attanasio, Pellerano, and Reyes 2009). This study in Tanzania also provides novel estimates of impact on trust in particular community members, including teachers and shopkeepers.4 While there is a larger literature on the relationship between CCTs and inter-household transfers, it almost entirely stems from Latin America—e.g., Colombia (Garcia and Cuartas 2021), Mexico (Attanasio Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 and Ríos-Rull 2000; Albarran and Attanasio 2003), Honduras, and Nicaragua (Olinto and Nielsen 2006). This previous work examines the impact of CCTs on informal transfers at a single point in time, whereas this study shows an evolving relationship between CCTs and informal transfers over time. These findings also relate to larger literatures on how formal insurance can crowd out informal insurance—both lab- in-the-field evidence, such as Auriol et al. (2020) and Cecchi, Duschoslav, and Bulte (2016), and quasi- experimental evidence, such as Strupat and Klohn (2018)—and how the broader range of social safety nets (including pensions) affects informal transfers (Nikolov and Bonci 2020; Burchi et al. 2022). 2. Experiment The Tanzania Social Action Fund (TASAF) and World Bank officials conceived the Tanzania Community- Based Conditional Cash Transfer (CB-CCT) pilot program and outlined its design during November 2007–September 2008 (see the full program timeline in table S1.1. TASAF is a government agency tasked with promoting economic and social development, established in 2000. During September–November 2008, TASAF carried out program sensitization at the regional, district, ward, and community levels in preparation for planned enrollment of beneficiaries during the fall of 2009. The program began delivering transfers in January 2010. Its aims were to increase investments in health for young children (ages 0–5) and the elderly (over 60) and to increase educational investments for children aged 7–15. A total of 80 eligible communities in three districts (Chamwino, Bagamoyo, and Kibaha) were randomized into treatment and control groups of 40 communities each, stratified on district and community size.5 Randomization was carried out after identification of (potential) beneficiary households in all 80 villages. At village meetings held prior to randomization, TASAF communicated program logistics and benefit amounts to households identified as prospective beneficiaries, explained that half of the villages would receive the program by 2010, and noted that the remaining (control) villages would receive the program in late 2012, as well as that the program would continue in treatment villages.6 As planned, the program was implemented for 2.75 years in treatment communities only, and then extended to all 80 communities after the last round of data collection. Treatment households received transfers every two months. Transfer amounts were linked to household size and composition. At the time of the first (2011) follow-up survey, the most recent transfer payment received by the median beneficiary household (they received payments every other month) was 22,500 Tanzanian shillings (TSh).7 At the time of the second (2012) follow-up survey, the most recent transfer 4 Several studies examine the impact of CCTs on trust in government or other institutions: e.g., in Tanzania (Evans, Holte- meyer, and Kosec 2019b) and Peru (Camacho 2014). Kosec and Mo (2023) show increases in trust in government and political institutions in Pakistan as a result of unconditional cash transfers when (for exogenous reasons) beneficiaries feel relatively poor and deprived. 5 The three districts were not randomly selected, but rather selected to be in rural areas proximate to Dar es Salaam (a major port city and the seat of government) and Dodoma (the capital); in particular, Bagamoyo is 70 km from Dar es Salaam, Chamwino is 500 km from Dar es Salaam but 40 km from Dodoma, and Kibaha is 35 km from Dar es Salaam. 6 The authors of this study are not aware of how discussions of program longevity were handled, but they are not aware of a date for conclusion of the program being communicated in any villages prior to the endline survey. 7 Using the average exchange rate in August 2011 (1,641.63 TSh per 1 US dollar) (Exchange Rates UK 2011), this is USD 13.71. 224 Evans and Kosec payment received by the median beneficiary household was 30,000 TSh. Converting both of these amounts to constant 2009 TSh,8 they amount to 18,800 TSh and 21,610 TSh, respectively. At both midline and endline, these equalled about 13 percent of household expenditures during the same two-month period.9 The program involved a number of individual-level conditions. Healthcare conditions applied only to young children and the elderly. Children aged 0–5 had to visit a public health clinic (for either preventive or curative care) six or more times per year (for children aged 2–5, the condition relaxed to only two visits Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 per year beginning in 2012), while those age 60 or over had to visit at least once per year. Additionally, children aged 7–15 had to stay enrolled in school and maintain an 80 percent attendance record. While this study lacks precise data on the strictness of enforcement, few households (2–3 percent) reported having experienced transfer reductions due to program non-compliance during the midline and endline surveys. Elected community management committees (CMCs) in each village worked with TASAF to select prospective beneficiary households before randomization into treatment and control.10 All communities had, in the past, implemented community development projects with resources from TASAF, and the CMC was responsible for administering those resources. TASAF asked the CMC to survey the poorest half (roughly) of households, collecting data on eight household characteristics that would subsequently enter a proxy means test (PMT): roof material, light supply, water supply, type of toilet, ownership of four different assets (vehicle/motorcycle, radio, iron, poultry), number of windows on the house, household size, and number of meals eaten per day. TASAF carried out the PMT, producing a ranking of house- holds by poverty level, which it then submitted to the CMC and village leader for approval. Households generally reported high levels of satisfaction with their CMCs. CMCs existed in both treatment and control communities. In control villages, the CMC continued to exist, but did not play an active role in administering resources during the time of the program. CMC over- sight and validation was intended to promote community buy-in. Following beneficiary selection, CMCs in treatment villages continued to screen potential beneficiaries, communicate program conditions, and transfer funds. The CMCs also helped monitor conditions; every four months, they were responsible for collecting monitoring forms from health clinics and schools, updating TASAF records, delivering warn- ings to those not meeting conditions, making home visits when needed, and conducting regular awareness sessions. After 1.75 years of transfers, over 86 percent of beneficiary households reported a visit to their household by a CMC member since the program began, and only 1.5 percent reported a CMC member asking for part of their transfer. 3. Data and Empirical Strategy This study reports on the impacts of the CCT program using three waves of data collected on beneficiaries in treatment communities and would-be beneficiaries in control communities. In total, 1,764 households (883 in treatment communities and 881 in control communities), including a total of 6,918 individuals, 8 This study uses consumer price index averages for Tanzania for 2009, 2011, and 2012 from World Bank (2020) to convert these amounts into constant 2009 TSh. 9 Statistics cited here use information on the most recent TASAF CCT program transfer to reduce concerns with inaccurate recall (e.g., such as might emerge in asking individuals to total a full year’s worth of transfers) and isolate CCT program transfers from other types of government support. Note that the analyses in the paper (e.g., tables 1 and 5), however, report amounts of total government transfers (which may include the CCT program as well as other transfers from government) received over the previous 12 months—which permits side-by-side comparisons of government transfers with other forms of transfers (individuals and non-governmental organizations, or NGOs) over the same time frame (i.e., 12 months). 10 CMC elections were convened at village meetings; in program communities, between 10 and 14 members were elected via secret ballot. Those who received financial training and successfully managed a past TASAF-supported project were eligible to run. The World Bank Economic Review 225 participated in a baseline survey in early 2009 (a random subset of households identified as sufficiently poor to receive transfers). No data were collected from households not identified as sufficiently poor. Table S1.1 presents the chronology of the impact evaluation and the program. Payments began in early 2010, a midline survey was conducted during July–September 2011 (about 1.75 years after transfers be- gan), and an endline survey was conducted during August–October 2012 (about 2.75 years after transfers began).11 Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 There is no evidence of differential attrition of either individuals or households across the treatment and control groups in this study (Evans, Holtemeyer, and Kosec 2019a). This is true for both survey rounds and is robust to various specifications, including those interacting treatment with various individual- and household-level characteristics and checking for the joint significance of those interactions and treatment in predicting attrition. The rate of household attrition was 8.5 percent at midline, and 13.0 percent at endline.12 This study reports causal intent-to-treat estimates from the following empirical specification: Yit = β0 + β1 2011t + β2 2012t + δ1 Ti × 2011t + δ2 Ti × 2012t + αi + it , where i indexes individuals and t indexes the survey round (baseline in 2009, midline in 2011, or endline in 2012). The variable Yit is an outcome, α i are individual fixed effects, Ti is an indicator for one’s community being assigned to treatment, 2011t is an indicator for midline, and 2012t is an indicator for endline. For some outcomes, data were only gathered at endline, and thus the analysis of those outcomes utilizes a sample that is roughly one-third the size of that used for other outcomes; results reported include regressions of the outcome on Ti , district fixed effects, a control for logged community population, and a vector of household-level controls measured at baseline.13 This study reports estimates using ordinary least squares regression models, including linear probability models where the outcomes are indicator variables. Standard errors are clustered at the village level, which is the unit of randomization. In addition to estimating effects overall, this study reports effects according to the household’s baseline asset wealth level (defining less-poor households as those with above-median asset wealth, and the poorest households as those with below-median asset wealth). The variable Yit comprises several different outcome variables. First, this study reports several measures of communal trust, which are in all cases indicator variables. The study reports two variables present in all survey rounds: indicators for believing most people can be trusted, and for believing community members can be trusted. The study also considers six variables present only at endline. Two are indicators for believing that trust in people in general has improved over the last three years (i.e., since baseline) and for believing that trust in community members has improved over this period. Four are indicators for trusting to a great or very great extent shopkeepers, teachers, nurses and doctors, and strangers. Teachers as well as nurses and doctors would be encountered by beneficiaries owing to the conditions of the CCT program, while shopkeepers and strangers might be encountered as members of the local CMCs elected to manage the program. Second, the study reports on three indicator variables capturing access to informal safety nets. These include indicators for knowing someone outside the household willing to provide 15K Tanzanian shillings 11 In cases where households split into two, enumerators visited both households. Both households share the treatment status of the original household from which they split. 12 For households in treatment communities, these figures were 8.3 percent and 11.8 percent, respectively. For households in control communities, they were 8.6 percent and 14.1 percent, respectively. Of the 1,764 households visited at baseline, 4 declined to be interviewed during the baseline visit; the analysis of attrition by midline and by endline omits these 4 observations in order to consider attrition after baseline. 13 These include head age, age squared, gender, education level, and literacy status; dummies for the number of household members; dummies for an improved roof, improved floor, toilet, and piped water; and an asset index that is the first principal component from a principal components analysis using indicators for 13 durable assets. 226 Evans and Kosec (TSh) suddenly and willing to provide childcare suddenly, and an indicator for someone with a personal problem having turned to the household for assistance sometime in the past 12 months. Third, the study reports on indicator variables for having received various forms of assistance from different places, as well as continuous variables for the amount of assistance received from each source. These places include individuals, government or TASAF, and NGOs, while the sources include cash, food, and other in-kind transfers. Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 Finally, in addition to incoming transfers, the study reports the amount of transfers the household paid out to others—the overall amount, as well as the amounts in cash, food, and other in-kind transfers. The analysis compares treatment and control households at baseline to ensure comparability on ob- servable characteristics (table 1). Of 27 variables compared, there are only a handful of significant differ- ences. Treatment households are 6 percentage points less likely to have an improved floor and 6 percentage points less likely to state that people in the community can be trusted. They are also slightly less likely to receive assistance from other individuals, which is reflected in both the binary (whether or not they received any assistance) and the continuous (how much assistance they received) outcome variables. Be- cause the regressions include household fixed effects, these modest differences at baseline should not affect the estimates. 4. Results First, this study reports impacts of government-provided CCTs on trust. Trust can be viewed as a critical requirement for the proper functioning of informal insurance arrangements between households. The results show no evidence that receiving CCTs eroded beneficiaries’ trust in others after 1.75 years. There is a statistically insignificant decline of 4.8 percentage points in the likelihood of trusting most people, and a statistically insignificant, 1.3 percentage point increase in the likelihood of trusting community members, as shown in table 2. However, after an additional year (2.75 years of transfers), there is a statistically significant (at the 0.05 level) 7.2 percentage point increase in trust of community members (though there are null impacts on trust in most people). This time path of findings on trust is striking when considered relative to the existing literature on cash transfers and trust. Haushofer and Shapiro (2016) observe inter-household trust after just one year of receiving cash transfers and find no change. Attanasio, Pellerano, and Reyes (2009) compare communities that have received CCTs for more than two years to communities that have not and find higher levels of social capital in CCT communities. While this relationship may vary across settings for multiple reasons, the fact that there is no change initially but then a positive change later potentially reconciles previous results. If increases in trust among beneficiaries in part stem from greater engagement with members of the community, then that trust could take time (and repeated interactions and engagement) to develop after the initiation of cash transfers. More detailed data on perceived changes in trust and on trust of particular groups (which are only avail- able in the endline survey) show further potential changes in trust patterns. At endline, individuals were first asked whether their level of trust in people in general, and in community members, increased over the last three years (i.e., since baseline). By endline, beneficiary households report significant improvements in trust in both groups. Next, individuals were asked about their level of trust in shopkeepers, teachers, nurses and doctors, and strangers. The analysis includes a set of dummy variables for trusting each group to a great or very great extent. Considering the impacts of treatment at endline, point estimates are in all cases positive, and they are statistically significant at conventional levels for the cases of shopkeepers and teachers. Thus, there is a broad pattern of increased trust in others after the program has been in place for a significant period of time. These novel findings complement previous analysis from the same project in Tanzania showing that CCTs boost trust in local elected leaders (Evans, Holtemeyer, and Kosec 2019b). Shopkeepers and The World Bank Economic Review 227 Table 1. Baseline Balance Treatment (T) Control (C) Difference (T − C) Outcome Mean N Mean N Mean S.E. Panel A: Household and community characteristics Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 Household characteristics Dummy: Household has improved roof 0.33 880 0.37 878 −0.04 (0.06) Dummy: Household has improved floor 0.03 880 0.09 878 −0.06** (0.02) Dummy: Household has toilet facilities 0.69 880 0.76 879 −0.07 (0.04) Dummy: Household has piped water 0.30 880 0.32 879 −0.01 (0.08) Dummy: Head of household is male 0.63 879 0.59 878 0.04 (0.03) Share of households reporting a...exists in village Parent association 0.14 40 0.13 40 0.01 (0.02) Health committee 0.61 40 0.59 40 0.02 (0.04) Panel B: Outcomes at baseline Trust Dummy: Leaders can generally be trusted 0.81 878 0.80 873 0.01 (0.03) Dummy: Most people can be trusted 0.26 875 0.23 874 0.03 (0.03) Dummy: Community people can be trusted 0.59 876 0.53 873 0.06* (0.03) Dummy: Received...gifts/assistance from individuals in last 12 months Dummy: Cash 0.24 874 0.28 876 −0.04 (0.03) Dummy: Food 0.21 869 0.27 874 −0.06** (0.03) Dummy: Other in-kind 0.22 875 0.26 872 −0.04 (0.03) Dummy: Received...gifts/assistance from NGOs in last 12 months Dummy: Cash 0.01 879 0.01 879 −0.01 (0.01) Dummy: Food 0.01 880 0.03 879 −0.02 (0.02) Dummy: Other in-kind 0.03 880 0.05 878 −0.02 (0.02) Dummy: Received...gifts/assistance from government in last 12 months Dummy: Cash 0.01 880 0.00 879 0.01 (0.01) Dummy: Food 0.02 879 0.02 879 0.00 (0.01) Dummy: Other in-kind 0.03 880 0.02 879 0.01 (0.01) Amount (1000s constant 2009 TSh) of transfers received in last 12 months from... Anyone 18.38 880 25.46 879 −7.08** (3.44) Individuals 16.85 880 23.62 879 −6.77* (3.48) Government or TASAF 0.83 880 0.46 879 0.37 (0.22) NGOs 0.70 880 1.38 879 −0.68 (0.49) Amount (1000s constant 2009 TSh) of transfers in last 12 months given in... Any form 1.32 880 1.73 879 −0.41 (0.67) Cash 0.42 880 0.41 879 0.009 (0.22) Food 0.48 880 0.66 879 −0.17 (0.25) Other in-kind 0.41 880 0.66 879 −0.25 (0.50) Source: Authors’ calculations based on baseline (2009) household survey data. Note: Treatment indicates assignment to treatment. TASAF is the Tanzania Social Action Fund, the government agency responsible for implementing the cash transfer program. TSh = Tanzanian shillings. NGOs = Non-government organizations. Standard errors are clustered at the village level. Outcome variables only available at endline are not shown. Baseline transfer amounts reported (expressed in 1000s of constant 2009 TSh over the past 12 months) compare with average midline and endline transfers (expressed in 1000s of constant 2009 TSh) over the past 12 months from government or TASAF in treatment village of 86.52 and 63.22, respectively. ***indicates p < 0.01; **indicates p < 0.05; *indicates p < 0.10. teachers are groups that beneficiaries may have greater reason to engage with after receiving CCTs, just as beneficiaries may have more interaction or awareness of elected leaders. This is also consis- tent with evidence from Peru that CCTs boosted trust in local institutions for beneficiary households (Camacho 2014). 228 Evans and Kosec Table 2. Communal Trust Dummy: Trust in... has improved over last Dummy:... can be trusted 3 years Dummy: Trusts... to a great/very great extent Most Community People in Community Nurses & Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 people members general members Shopkeepers Teachers doctors Strangers (1) (2) (3) (4) (5) (6) (7) (8) Treatment × 2011 (midline) −0.048 0.013 — — — — — — (0.036) (0.027) Treatment × 2012 (endline) 0.034 0.072** 0.032** 0.039*** 0.098*** 0.053** 0.041 0.029 (0.041) (0.035) (0.015) (0.012) (0.033) (0.025) (0.028) (0.021) 2011 (midline) 0.427*** 0.664*** — — — — — — (0.034) (0.027) 2012 (endline) 0.046* 0.633*** — — — — — — (0.027) (0.029) R2 0.198 0.369 0.029 0.030 0.037 0.038 0.035 0.031 Baseline mean 0.242 0.556 — — — — — — Observations 4,920 4,924 1,570 1,570 1,570 1,570 1,570 1,570 Source: Authors’ calculations based on baseline (2009), midline (2011), and endline (2012) household survey data. Note: Columns 1–2 include household fixed effects (the number of observations differs slightly due to small numbers of people answering only one of these two questions). Columns 3–8 are outcomes only available at endline, and thus these specifications include only one round of data and do not use household fixed effects. Instead, they utilize controls including district fixed effects and household-level controls measured at baseline capturing head age, age squared, gender, ed- ucation level, and literacy status; dummies for the number of household members; dummies for an improved roof, improved floor, toilet, and piped water; and an asset index that is the first principal component from a principal components analysis using indicators for 13 durable assets. Treatment estimates are estimates of the effect of living in a treatment village (intent to treat). Standard errors are in parentheses and clustered at the village level. ***indicates p < 0.01; **indicates p < 0.05; *indicates p < 0.10. Table 3. Informal Safety Nets Dummy: Knows someone outside the household willing to provide... Dummy: Someone with a personal 15K TSh Childcare problem has turned to HH for suddenly suddenly assistance in the past 12 mos. (1) (2) (3) Treatment × 2012 (endline) 0.048 0.087*** 0.045** (0.030) (0.032) (0.022) R2 0.055 0.045 0.070 Observations 1,570 1,570 1,570 Source: Authors’ calculations based on endline (2012) household survey data. Note: TSh = Tanzanian shillings; HH = household; mos = months; K = 1,000. These outcomes were only available at endline, and thus specifications include only one round of data and do not use respondent fixed effects. All specifications use controls including district fixed effects and household-level controls measured at baseline capturing head age, age squared, gender, education level, and literacy status; dummies for the number of household members; dummies for an improved roof, improved floor, toilet, and piped water; and an asset index that is the first principal component from a principal components analysis using indicators for 13 durable assets. Treatment estimates are estimates of the effect of living in a treatment community (intent to treat). Standard errors are in parentheses and clustered at the community level. ***indicates p < 0.01; **indicates p < 0.05; *indicates p < 0.10. Second, the study reports on whether beneficiary households are more likely to report engagement in informal safety nets (table 3). These include measures of monetary and other support from community members. While households do not report any change in knowing someone from whom they believe they could borrow a significant sum of cash, they do report positive, significant impacts on two other measures of informal safety nets. On the receiving end, they are significantly more likely (8.7 percentage points) to report knowing someone who they believe could provide childcare suddenly. On the providing end, The World Bank Economic Review 229 Table 4. Assistance Received from Individuals, Government, and NGOs by Type Dummy - received gifts/assistance from... in the past 12 months Individuals Govt. or TASAF NGOs Cash Food Other in-kind Cash Food Other in-kind Cash Food Other in-kind Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 (1) (2) (3) (4) (5) (6) (7) (8) (9) Treatment × 2011 (midline) −0.088** −0.072** −0.036 0.878*** 0.000 −0.003 −0.002 −0.011 0.016 (0.035) (0.032) (0.036) (0.017) (0.009) (0.016) (0.007) (0.018) (0.013) Treatment × 2012 (endline) −0.060* −0.041 −0.042 0.899*** −0.002 −0.003 −0.003 0.005 0.014 (0.035) (0.035) (0.034) (0.019) (0.007) (0.010) (0.009) (0.018) (0.018) 2011 (midline) 0.348*** 0.409*** 0.218*** 0.009 0.010 0.012 0.006 0.005 −0.026* (0.035) (0.028) (0.030) (0.010) (0.007) (0.011) (0.006) (0.018) (0.014) 2012 (endline) 0.298*** 0.341*** 0.152*** 0.018 −0.003 0.005 0.009 0.004 −0.020 (0.031) (0.028) (0.031) (0.011) (0.007) (0.008) (0.008) (0.015) (0.017) R2 0.105 0.264 0.070 0.806 0.036 0.069 0.047 0.040 0.101 Baseline mean 0.257 0.243 0.242 0.009 0.017 0.023 0.010 0.020 0.039 N 4,953 4,955 4,951 4,953 4,954 4,955 4,954 4,955 4,954 Source: Authors’ calculations based on baseline (2009), midline (2011), and endline (2012) household survey data. Note: TASAF = Tanzania Social Action Fund; NGO = non-governmental organization; Govt. = government. All specifications include household fixed effects. Treatment estimates are estimates of the effect of living in a treatment village (intent to treat). The number of observations differs slightly across specifications due to small numbers of people with missing data for some and not other transfer types. TASAF is the government agency responsible for implementing the cash transfer program. Standard errors are in parentheses and clustered at the village level. ***indicates p < 0.01; **indicates p < 0.05; *indicates p < 0.10. they are 4.5 percentage points more likely to have had someone with a personal problem turn to their household for assistance. These findings on household perceptions of access to multiple safety nets are a complement to aforementioned work on social capital in the face of CCTs in Colombia (Attanasio, Pellerano, and Reyes 2009), which relied on lab-in-the-field games. In both cases, CCTs were associated with increased social capital. Third, this study reports on monetary and in-kind transfers across households and from other sources (specifically, from government and from NGOs). At baseline, roughly one in four households reports receiving cash, food, or other in-kind assistance from other individuals (table 4). In contrast, earlier work from Mexico found that fewer than 6.5 percent of households in control villages were receiving inter- household transfers (Attanasio and Ríos-Rull 2000). There is some evidence of negative impacts on inter-household transfers on the extensive margin—i.e., whether or not households receive transfers. Specifically, there is a significant decline of 8.8 percentage points in the likelihood of receiving cash from other individuals after 1.75 years of transfers (at midline), but the coefficient declines in magnitude, to 6.0 percentage points after 2.75 years (at endline), and is only marginally significant. Transfers of food from individuals also decline by 7.2 percentage points at midline, though this reduction in food transfers is smaller in magnitude and statistically insignificant by endline. The only other significant effect in any survey round is the finding that beneficiary households are much more likely to report receiving transfers from the government at both midline and endline, as expected (i.e., households are receiving the formal CCTs they are due to receive). There is no evidence of a change in transfers from NGOs. On the intensive margin, considering the amount of transfers received (here the regression sample includes those receiving transfers of zero), households report somewhat smaller transfers from other in- dividuals after 1.75 years of the program (measured in thousands of constant 2009 TSh, where the cash value of in-kind transfers is added to the value of any cash received). However, there is no such crowd-out 230 Evans and Kosec Table 5. Amount of Assistance Received from Individuals, Government, and NGOs Total transfers from... in the past 12 months Total Individuals Govt. or TASAF NGOs (1) (2) (3) (4) Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 Treatment × 2011 (midline) 71.72*** −13.79*** 86.28*** −0.77 (5.950) (4.230) (4.216) (0.587) Treatment × 2012 (endline) 58.68*** −3.82 63.15*** −0.47 (5.892) (5.121) (2.626) (0.803) 2011 (midline) 34.50*** 33.52*** −0.07 1.05** (4.155) (3.795) (1.423) (0.486) 2012 (endline) 28.06*** 27.28*** 0.17 0.77 (4.330) (4.217) (1.121) (0.658) R2 0.304 0.215 0.398 0.096 Baseline mean 21.915 20.229 0.648 1.038 Observations 4,955 4,955 4,953 4,955 Source: Authors’ calculations based on baseline (2009), midline (2011), and endline (2012) household survey data. Note: TASAF = Tanzania Social Action Fund; NGO = non-governmental organization; Govt. = government. Transfer amounts are measured in thousands of constant 2009 Tanzanian shillings (TSh). All specifications include household fixed effects. Treatment estimates are estimates of the effect of living in a treatment village (intent to treat). The number of observations differs slightly across specifications due to small numbers of people with missing data for some and not other transfer types. TASAF is the government agency responsible for implementing the cash transfer program. Standard errors are in parentheses and clustered at the village level. ***indicates p < 0.01; **indicates p < 0.05; *indicates p < 0.10. of transfers from individuals after 2.75 years (table 5).14 Even after 1.75 years, the crowd-out is modest: the reduction in transfers from individuals is under 16 percent of the value of the increase in formal gov- ernment transfers. Specifically, 13,799 constant 2009 TSh (the reduction in transfers from individuals) is just under 16 percent of the additional 86,280 constant 2009 TSh received from government or TASAF.15 At the same time, households are no more likely to provide assistance to other households; table 6 considers as outcomes the value (measured in thousands of constant 2009 TSh) of assistance given to others in cash, food, other in-kind, and the total of the three, revealing null impacts of treatment on all four outcomes for both follow-up survey rounds. This lack of an increase on transfers out of beneficiary households is distinct from recent findings in Bangladesh, where ultra-poor households that receive an unconditional asset transfer were more likely both to give and to receive transfers (Gulesci 2020). The results on both incoming and outgoing transfers hold if the top 1 percent of observations are winsorized to account for outliers in transfer amounts (see tables S1.2 and S1.3, respectively). That the CCT program had an enduring impact (at midline and endline) on the likelihood of receiving transfers from other individuals at all (extensive margin results of table 4), but had no impact on the 14 Consumer price index averages for Tanzania for 2009, 2011, and 2012 from World Bank (2020) are used to convert all transfer amounts into constant 2009 TSh. 15 Table 5 highlights a reduction in government transfers from the CCT program between midline and endline (i.e., liv- ing in a treatment village provides an additional 86,280 constant 2009 TSh in government transfers at midline, but a more modest, 63,150 constant 2009 TSh increase at endline.) There are several possible reasons for this. First, at midline, households receiving some government transfers had received such transfers more times than had households at endline—possibly indicating that the midline took place right after receipt of a transfer payment while the endline took place right before it. Second, when households split, enumerators tracked all such households and retained all in the sample—meaning that the share of households in treatment villages actually receiving transfers declined over time. Finally, given this outcome captures all types of government assistance, it is possible that by endline, CCT program transfers had crowded out other forms of government assistance to a greater extent than they did at midline (when the program was still relatively new). Regardless, level differences over time should not affect the treatment estimates, given the inclusion of year fixed effects. The World Bank Economic Review 231 Table 6. Amount of Assistance Given to Community Members by Type Assistance given in the past 12 months in... Total Cash Food Other in-kind (1) (2) (3) (4) Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 Treatment × 2011 (midline) 0.58 0.16 −0.18 0.60 (0.682) (0.230) (0.435) (0.425) Treatment × 2012 (endline) 0.12 −0.03 −0.17 0.21 (0.631) (0.284) (0.373) (0.295) 2011 (midline) −0.03 −0.07 0.88* −0.88 (0.860) (0.190) (0.466) (0.745) 2012 (endline) 0.59 0.30 1.00*** −0.22 (0.974) (0.242) (0.365) (0.825) R2 0.005 0.003 0.006 0.010 Baseline mean 1.523 0.411 0.572 0.536 Observations 4,955 4,955 4,955 4,955 Source: Authors’ calculations based on baseline (2009), midline (2011), and endline (2012) household survey data. Note: Transfer amounts are measured in thousands of constant 2009 Tanzanian shillings (TSh). All specifications include household fixed effects. Treatment estimates are estimates of the effect of living in a treatment village (intent to treat). Standard errors are in parentheses and clustered at the village level. ***indicates p < 0.01; **indicates p < 0.05; *indicates p < 0.10. amount of transfers received from other individuals (table 5) suggests nuanced impacts on individuals’ transfer-providing behavior. It suggests no enduring crowd-out (after all, by endline, program beneficiaries collectively receive the same amount of transfers from other individuals on average), but greater likeli- hoods both of individuals receiving zero and of receiving higher than pre-program individual transfer amounts. This could be consistent with better targeting of informal transfers by other individuals: for example, if the program shed greater light on who are the poorest among the poor during the process of door-to-door visits to complete a poverty census, perhaps individuals began concentrating their transfers among the poorest of the beneficiaries, and reducing to zero their transfers to less-poor program beneficia- ries. To consider this possibility, this study includes analysis of heterogeneous treatment effects according to baseline household asset wealth (table S1.4 for the extensive margin and table S1.5 for the intensive margin). The resulting evidence is not consistent with improved targeting of the poorest of the poor by individ- uals. Among beneficiaries, the reductions in the likelihood of receiving transfers from other individuals that decline in magnitude over time are actually driven by the poorest beneficiaries, defined as those with below-median baseline asset wealth. As table S1.4 reveals, there are sizeable, significant negative effects on whether or not the poorest received transfers of cash from other households and near-zero, insignif- icant effects among the less-poor beneficiaries. (Remember that all recipients were targeted by poverty status, so all are considered to be at least somewhat poor by their communities.) Among the poorest, the reduction in whether or not they received transfers of cash from other households endures through 2.75 years as it did for the overall sample, although the point estimate after 2.75 years is about half its size at 1.75 years. While the finding that the poorest experienced the greatest crowding out is surprising, it is also consistent with evidence from Burkina Faso, where private transfers were most sensitive to income gains (not necessarily from CCTs) among the poorest of the poor (Grimm et al. 2021). Another possibility is that these findings are due to the distinct social networks of the poorest of the poor compared to less-poor households. Specifically, poorer individuals may have similarly poor individuals within their social network, for whom the utility loss of providing an informal transfer to another household may be proportionately greater (e.g., in the case of a household which can barely af- ford a minimally sufficient diet). Such households may be precisely those that eliminate their transfers 232 Evans and Kosec to a household once they learn that household is receiving government support from a CCT program. Meanwhile, households with above-median baseline asset wealth may have slightly better-off households in their social network, which are less likely to react to their receipt of transfers from a CCT program by withdrawing this informal support (as providing informal transfers has a smaller negative impact on the giving household’s utility). While that could explain the shift in whether or not households received transfers, when one considers Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 the amount of transfers received (table S1.5), there are statistically significant declines in the amount of transfers from individuals after 1.75 years that are statistically indistinguishable between the poorest and the less poor. For both the poorest and the less poor, there are no declines in transfers from individuals after 2.75 years, suggesting that the crowd-out of inter-household transfers on the intensive margin fades out completely over time, regardless of the severity of a household’s poverty. By endline, the poorest households may be more likely to face a situation of zero transfers from individuals, but on average they receive about the same amount of transfers from individuals as they did before the CCT program (suggesting that at least some of them observe increases from their pre-program levels of transfers from individuals). Together, these results suggest at least some changes in transfer behavior that endure to endline. Specif- ically, for the poorest households, there is a movement toward more instances of zero transfers alongside more instances of increased transfer amounts, relative to pre-program amounts. However, the results are inconsistent with an enduring crowd-out of total transfers—either overall (table 5), for the poorest bene- ficiaries, or for less-poor beneficiaries (table S1.5). This lack of an overall crowd-out over time is distinct from previous work from Mexico suggesting that CCTs do crowd out private transfers (Albarran and Attanasio 2003; Attanasio and Ríos-Rull 2000). At the same time, evidence from Colombia shows that CCTs actually crowd in private transfers (Garcia and Cuartas 2021), potentially through enhanced social capital—leading households in the community to be more likely to help beneficiary households in the face of adverse shocks. The finding of increased trust may point toward a similar phenomenon. One previous piece of evidence may also help to explain the lack of crowd-out findings. Olinto and Nielsen (2006) com- pare the impact of CCTs on private transfers in Honduras and in Nicaragua. In Nicaragua, the transfers were more than five times as large (relative to household expenditures) as in Honduras; at the same time, the evidence suggests that CCTs crowded out private food transfers in Nicaragua but not in Honduras. In the program this study reports on in Tanzania, the CCTs are the equivalent of about 13 percent of household expenditures, which falls somewhere between the level in Honduras and in Nicaragua. The relatively low level of transfers in Tanzania combined with high poverty rates may indicate that the CCTs were not sufficient to eliminate informal transfers. Considering transfers paid out, there is no pattern of difference between the poorest and the less poor in whether they provide assistance to other community members; for both groups, there are no impacts of formal CCTs on transfers to other households of any kind (cash, food, or other in-kind), as shown in table S1.5. 5. Conclusion This study reports on a randomized controlled trial to test the impact of CCTs on trust, inter-household transfers, and beneficiary perceptions of access to other forms of social support in Tanzania. Impacts are observed at two periods of time, which is distinct from most other work in this area. There is no evidence of a change in trust levels after 1.75 years, but there are positive changes in trust after 2.75 years, both in community members generally and in specific members of the community, such as shopkeepers and teachers. Beneficiary households also report a higher likelihood of believing they can rely on someone in the community for childcare and a higher likelihood that someone in the community has turned to them for help with a personal problem. There is a short-term dip in total transfers received from other The World Bank Economic Review 233 individuals, but by endline, the total transfers have returned to levels that are statistically indistinguishable from those at baseline. Looking beyond CCTs, the wide variation in results from previous studies on the interactions between formal safety nets and informal transfers (Nikolov and Bonci 2020) suggests that crowding out will depend on the structure of the program—e.g., the size of transfers, the completeness of coverage, and program duration. A key limitation of this work is that it does not estimate impacts on non-beneficiaries. The baseline Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 and follow-up household surveys interviewed only beneficiaries in treatment villages and would-be bene- ficiaries (selected before randomization of villages into treatment and control) from control villages. The targeting process for the cash transfer program sought to benefit the most vulnerable households in the village, but as these were vulnerable areas in what was at the time of the study a low-income country, it is likely that there are vulnerable households among non-beneficiary households as well. How cash transfers affect informal transfers to and from both vulnerable and less vulnerable non-beneficiary households is a question for future research. Existing evidence on non-beneficiary households in the context of cash transfer programs suggests negative impacts on subjective well-being (Haushofer, Reisinger, and Shapiro 2019) and on trust in community institutions and governance (Camacho 2014; Kosec and Mo 2023). These findings highlight the value of measuring impacts across multiple time periods, as results may well not be stable over time. They also continue to point to the many impacts that CCTs may have beyond the commonly stated objectives of alleviating extreme poverty and boosting human capital investments. Ultimately, this study finds little evidence that CCTs erode existing safety nets. On the contrary, this evidence suggests that at least for beneficiaries, engagement with and trust in community members may be enhanced as a result of the program. Data Availability The replication data for this article are available at https://doi.org/10.7910/DVN/JNIXA4. A larger col- lection of datasets from the Tanzania conditional cash transfer pilot evaluated in this article is available at https://microdata.worldbank.org/index.php/catalog/2669. Funding The research for this article was financed by the CGIAR Trust Fund (https://www.cgiar.org/funders), the International Initiative for Impact Evaluation (3ie), the Strategic Impact Evaluation Fund (SIEF) at the World Bank, and the Trust Fund for Environmentally and Socially Sustainable Development (TFESSD) at the World Bank. References Albarran, P., and O. P. Attanasio. 2003. “Limited Commitment and Crowding Out of Private Transfers: Evidence from a Randomised Experiment.” Economic Journal 113(486): C77–C85. Attanasio, O., L. Pellerano, and S. P. Reyes. 2009. “Building Trust? Conditional Cash Transfer Programmes and Social Capital.” Fiscal Studies 30(2): 139–77. Attanasio, O., and J.-V. Ríos-Rull. 2000. “Consumption Smoothing in Island Economies: Can Public Insurance Reduce Welfare?” European Economic Review 44(7): 1225–58. Auriol, E., J. Lassebie, A. Panin, E. Raiber, and P. Seabright. 2020. “God Insures those Who Pay? Formal Insurance and Religious Offerings in Ghana.” Quarterly Journal of Economics 135(4): 1799–848. Burchi, F., M. Loewe, D. Malerba, and J. Leininger. 2022. “Disentangling the Relationship between Social Protection and Social Cohesion: Introduction to the Special Issue.” The European Journal of Development Research 34: 1195– 1215. https://doi.org/10.1057/s41287-022-00532-2 234 Evans and Kosec Camacho, L. A. 2014. “The Effects of Conditional Cash Transfers on Social Engagement and Trust in Institutions: Evidence from Peru’s Juntos Programme.” German Development Institute Discussion Paper 24/2014. Retrieved from German Institute of Development and Sustainability website: https://www.idos-research.de/en/discussion- paper/article/the-effects-of-conditional-cash-transfers-on-social-engagement-and-trust-in-institutions-evidence- from-perus-juntos-programme/. Cecchi, F., J. Duschoslav, and E. Bulte. 2016. “Formal Insurance and the Dynamics of Social Capital: Experimental Downloaded from https://academic.oup.com/wber/article/37/2/221/7000204 by Joint Bank-Fund library user on 04 September 2023 Evidence from Uganda.” Journal of African Economies 25(3): 418–38. de Milliano, M., C. Barrington, G. Angeles, and C. Gbedemah. 2021. “Crowding-Out or Crowding-In? Effects of Leap 1000 Unconditional Cash Transfer Program on Household and Community Support among Women in Rural Ghana.” World Development 143: 105466. Evans, D. K., B. Holtemeyer, and K. Kosec. 2019a. “Cash Transfers and Health: Evidence from Tanzania.” World Bank Economic Review 33(2): 394–412. ———. 2019b. “Cash Transfers Increase Trust in Local Government.” World Development 114: 138–55. Exchange Rates UK. 2011. “US Dollar to Tanzanian Shilling Spot Exchange Rates for 2011.” Retrieved from: https://www.exchangerates.org.uk/USD-TZS-spot-exchange-rates-history-2011.html. Garcia, S., and J. Cuartas. 2021. “Can Poverty Alleviation Programs Crowd-In Private Support? Short-and Middle- Run Effects of a Conditional Cash Transfer Program on Inter-household Transfers.” Journal of Social Policy 50(3): 511–32. Grimm, M., R. Hartwig, A.-K. Reitmann, and F. Y. Bocoum. 2021. “Inter-household Transfers: An Empirical In- vestigation of the Income-Transfer Relationship with Novel Data from Burkina Faso.” World Development 144: 105486. Gulesci, S. 2020. “Poverty Alleviation and Interhousehold Transfers: Evidence from BRAC’s Graduation Program in Bangladesh.” World Bank Economic Review 35(1): 921–949. Haushofer, J., J. Reisinger, and J. Shapiro. 2019. “Is Your Gain My Pain? Effects of relative Income and Inequal- ity on Psychological Well-Being.” Working Paper. Retrieved from: https://jeremypshapiro.appspot.com/papers/ Pain%20is%20Gain.pdf. Haushofer, J., and J. Shapiro. 2016. “The Short-Term Impact of Unconditional Cash Transfers to the Poor: Experi- mental Evidence from Kenya.” Quarterly Journal of Economics 131(4): 1973–2042. Kilburn, K., S. Handa, G. Angeles, P. Mvula, and M. Tsoka. 2017. “Short-Term Impacts of an Unconditional Cash Transfer Program on Child Schooling: Experimental Evidence from Malawi.” Economics of Education Review 59: 63–80. Knack, S., and P. Keefer. 1997. “Does Social Capital Have an Economic Payoff? A Cross-Country Investigation.” Quarterly Journal of Economics 112(4): 1251–88. Kosec, K., and C. H. Mo. 2023. “Does Relative Deprivation Condition the Effects of Social Protection Programs on Political Support? Experimental Evidence from Pakistan.” American Journal of Political Science, Forthcoming. Nikolov, P., and M. Bonci. 2020. “Do Public Program Benefits Crowd Out Private Transfers in Developing Countries? A Critical Review of Recent Evidence.” World Development 134: 104967. Olinto, P., and M. E. B. Nielsen. 2006. “Do Conditional Cash Transfers Crowd Out Private Transfers? Evidence from Randomized Trials in Honduras and Nicaragua.” Working Paper. Retrieved from: https://www.researchgate.net/ publication/228463863_Do_Conditional_Cash_Transfers_Crowd_Out_Private_Transfers_Evidence_from_ Randomized_Trials_in_Honduras_and_Nicaragua. Pega, F., R. Pabayo, C. Benny, E.-Y. Lee, S. K. Lhachimi, and S. Y. Liu. 2022. “Unconditional Cash Transfers for Reducing Poverty and Vulnerabilities: Effect on Use of Health Services and Health Outcomes in Low-and Middle- Income Countries.” Cochrane Database of Systematic Reviews (3). DOI: 10.1002/14651858.CD011135.pub3. Strupat, C., and F. Klohn. 2018. “Crowding Out of Solidarity? Public Health Insurance versus Informal Transfer Networks in Ghana.” World Development 104: 212–21. Valli, E., A. Peterman, and M. Hidrobo. 2019. “Economic Transfers and Social Cohesion in a Refugee-Hosting Setting.” Journal of Development Studies 55(sup1): 128–46. World Bank. 2020. “World Development Indicators: Consumer Price Index – Tanzania. Retrieved from: https://data. worldbank.org/indicator/fp.cpi.totl?end=2012&locations=tz&start=2009.”