Policy Research Working Paper 10103 Distributional Policies and Social Cohesion in a High-Unemployment Setting Jorge M. Agüero Eniola Fasola Social Sustainability and Inclusion Global Practice June 2022 Policy Research Working Paper 10103 Abstract This paper studies the impact of distributional policies on accounting for multiple hypothesis testing. Results show a social cohesion. The focus is on South Africa, a country limited impact of the transfer on social cohesion. Transfer with the highest unemployment rate worldwide and a major increases life satisfaction and views favorable towards racial destination hub for the forcibly displaced. The paper uses diversity. However, it has only a marginal effect on inter- a regression discontinuity design based on the eligibility personal trust and a very small effect on attitudes towards rule of an unconditional cash transfer program (Old Age immigration. These findings are consistent with theoretical Pension) together with multiple rounds of the country’s models where anti-immigrant behaviors are not the result Social Attitudes Survey and estimates the impact of the of low-income but rather due to non-wage factors such as cash transfer to the local population on over 100 variables ethnic background or language barriers. capturing different dimensions of social cohesion, while This paper is a product of the Social Sustainability and Inclusion Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at jorge.aguero@uconn.edu. 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 Distributional Policies and Social Cohesion in a High-Unemployment Setting∗ Jorge M. Ag¨uero Eniola Fasola University of Connecticut University of Connecticut Keywords : Social cohesion, anti-immigration attitudes, cash transfers, South Africa. JEL-codes : J15, H53, H55, O5. ∗ Contact author: Jorge M. Ag¨ uero, University of Connecticut, Department of Economics, 365 Fairfield Way, Storrs, CT 06269-1063, USA. E-mail: jorge.aguero@uconn.edu. Online appendix is available here: https://www.dropbox.com/s/qmsxcohmv0ot9zy/appendix_social_cohesion.pdf?dl=0. This paper was commissioned by the World Bank Social Sustainability and Inclusion Global Practice as part of the activity “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts.” The activity is task managed by Audrey Sacks and Susan Wong with assistance from Stephen Winkler. This work is part of the program “Building the Evidence on Protracted Forced Displacement: A Multi-Stakeholder Partnership”. The program is funded by UK aid from the United Kingdom’s Foreign, Commonwealth and Development O ce (FCDO), it is managed by the World Bank Group (WBG) and was established in partnership with the United Nations High Commissioner for Refugees (UNHCR). The scope of the program is to expand the global knowledge on forced displacement by funding quality research and disseminating results for the use of practitioners and policy makers. This work does not necessarily reflect the views of FCDO, the WBG or UNHCR. 1 Introduction South Africa faces enormous challenges integrating international migrants and refugees. This is exemplified by the multiple xenophobic-motivated acts of violence post-apartheid, which have increased in recent years. At least two factors exacerbate this tension. First, the county is a major destination hub for the forcibly displaced. From 2008 until 2012, the country received the largest number of new asylum applications worldwide and it remains a major destination country for asylum seekers according to the United Nations High Commissioner for Refugees (2020). Second, South Africa has the highest unemployment rate of any country in the world, quadrupling the unemployment in Sub-Saharan Africa (World Bank, 2020). The COVID-19 pandemic is already creating sharp increases in unemployment in many developing countries, where 80% of the world’s refugees are hosted. This could expand the tension between immigrants and refugees with the local population observed in South Africa before 2020. Thus, it is of extreme policy relevance to understand what tools could be used to reduce anti-immigration attitudes overall, prevent social conflict and promote social cohesion in refugee-hosting settings with high unemployment levels such as South Africa.1 Our research question explores whether distributional policies could promote social co- hesion. This is clearly not the only factor that could explain it. Indeed, as discussed below, the are salient historical and structural triggers in the context of South Africa. However, our goal is to be able to causally identify the impact of a distributional policy alone. In particular, we evaluate the role of an unconditional cash transfer program, South Africa’s Old Age Pension (OAP). Focusing on the OAP provides several important contributions to the literature. First, the OAP is a countrywide, means-tested, non-contributory cash grant targeting South African citizens or permanent residents aged 60 or older. As explained later, this criteria permits an identification strategy to estimate the causal impact of the cash transfer, which expands and complements previous work on the economic drivers of anti-immigration attitudes and social cohesion (Mayda, 2006; Facchini et al., 2013). Second, the OAP is a large transfer. It is between 1.6 and 2 times the monthly median per capita household income of non-eligible individuals and currently provides monthly payments of R1890 (around US$137) until death. These features allow us to expand recent work that, for example, focused on a smaller (US$40) short-term (six months) transfer in localized parts of a refugee-hosting country (Valli et al., 2019). Third, the data source described below allow us to explore possible mechanisms by testing for heterogenous e↵ects, for example, by the 1 Following the literature, in this paper forcibly displaced individuals refers to those who have been forced to flee their home. They could cross an international border or not. Immigrants refers to individuals moving to a country di↵erent from their country of birth. 2 flow of asylum seekers.2 Our identification strategy takes advantage of the eligibility rules for the OAP. In partic- ular, we exploit the discontinuous jump in eligibility at the age of 60.3 This discontinuity has been well-document in previous work (e.g., Case and Deaton, 1998; Duflo, 2003; Edmonds, 2006; Bertrand et al., 2003; Hamoudi and Thomas, 2014; Ambler, 2016; Ag¨ uero, 2019). The use of a valid identification strategy help us eliminate the possibility of an e↵ect in the op- posite direction: from conflict, displacement or low social cohesion to poverty and income. For examples of recent work focusing on this opposite direction see Foltz and Shibuya (2021) on Mali and Sedova et al. on Nigeria. In that regard, we see our work as complementing these studies. As in the papers about the OAP, we will focus on the sample of Black and Coloured (mixed race) South Africans and employ a fuzzy regression discontinuity design (RDD) because the probability of receiving the OAP jumps discontinuously at age 60. Yet, we are able to expand on these papers in three ways. First, our time horizon ranges from 2008 to 2017. As such, our time frame covers the changes in the age eligibility criteria that have remained unexplored in prior work. Second, we study the role of the OAP on social cohesion, an outcome never considered before as all prior work has limited the analysis to outcomes a↵ecting the beneficiaries themselves or their immediate families and has ignored broader societal implications of the pension. Third, we take advantage of recent develop- ments in RDD for the case of high mass points in the running variable (age in years) given its discrete nature (Cattaneo et al., 2019; Koles´ ar and Rothe, 2018). Another key reason to focus on the OAP is the role that the population under study (those aged 50-70 at the time of the surveys) plays on voting. For example, while this age group represents close to 30% of the registered voters, they are the voters more likely to participate and decide elections (e.g., Nhlapo et al., 2017; Schulz-Herzenberg, 2019) Thus, their opinions and views are important for social cohesion and policy decisions, even when most of the anti-immigration acts tend to be conducted by younger groups. Also, Hamoudi and Thomas (2014) show that the OAP plays an important role on household composition as it attracts younger family members to live with grant recipients. Thus, as long as heads of households influence the views of younger co-resident family members, studying the impact of a large cash transfer for the elderly is an important policy question. Furthermore, in South Africa, anti-immigration attitudes do not vary with age, as shown by Facchini et al. (2013). Thus, focusing on an older group of the population allows us to use a strong identification strategy without sacrificing “external validity” given the flat age-gradient on anti-immigration views. 2 See Ferguson et al. (2021) for examples of alternative programs to promote social cohesion in developing countries hosting forcibly displaced populations. 3 Until 2008, women were eligible at 60 and men at 65. By 2010, the age criteria was standardized to 60. Our analysis will take into account these changes. See section 2 for details. 3 The main data source come from multiple rounds of the South African Social Attitudes Survey (SASAS). The SASAS is a large nationally representative, cross-sectional survey that has been conducted annually by the Human Sciences Research Council (HSRC) and is avail- able until 2017. The survey allows us to consider over 100 variables that we grouped into seven indices (accounting for multiple hypothesis testing) that capture various dimensions of social cohesion previously used in the literature Valli et al. (2019): (i) attitudes on immi- grants, (ii) interpersonal trust (iii) trust in institutions, (iv) satisfaction with the government, (v) life satisfaction, (vi) preferences for racial diversity and (vi) social participation.4 Our findings suggest that South Africa’s Old Age Pension has no major impact on overall social cohesion. We then show that this result comes from two competing e↵ects. On one hand, we found strong evidence that the pension increases the recipients’ life satisfaction and their attitudes towards racial diversity. We argue that all these e↵ects are consistent with a net gain in income and wellbeing. On the other hand, this increase in wellbeing does not translate into more interpersonal trust and neither on trust on institutions and satisfaction with the government. We do find, however, that the transfer reduces social participation. Most importantly, the Old Age Pension does not change attitudes toward immigrants. Based on the theories discussed below, this evidence is consistent with models where anti-immigrant behaviors are less likely to be the result of low-income levels and instead, respond to non- wage factors such as ethnic background or language barriers. This also suggest a critical role of historical triggers for anti-immigrant attitudes. The rest of the paper is composed of five additional sections. Section 2 describes the South African context and presents recent trends in the arrival of the forcibly displaced as well as the distributional policies implemented in the country. The theoretical considerations behind the impact of the Old Age Pension on social cohesion are discussed in section 3. The datasets and the identification strategy are presented in section 4. The validation of our empirical strategy and main results and shown in section 5. Section 6 summarizes the findings and the paper’s implications for policy.5 4 These questions mimic and expand those included in well-known but infrequent surveys such as the Afrobarometer (Nunn and Wantchekon, 2011), the World Values Survey (Mayda, 2006) and the International Social Surveys Programme (ISSP). 5 Online appendix is available here: https://www.dropbox.com/s/qmsxcohmv0ot9zy/appendix_ social_cohesion.pdf?dl=0. 4 2 Context 2.1 Immigration in South Africa International migration from neighboring countries has been a long feature in South Africa, dating back to the mid 19th century when foreign workers were brought mainly for the booming mining sector (Crush, 2000). This long history has not prevented the issue to become a contested topic today and it is often “entangled in political discourse that blames them [migrants and refugees] for ‘stealing’ local jobs” (World Bank, 2018, p. 3). For instance, the Apartheid regime severely limited foreign workers as they considered them a source of political threat (Facchini et al., 2013). The post-apartheid era, South Africa became once again a destination for foreign workers from the region.6 . Yet as discussed in section 3 there are historical and structural triggers remain in place. Compared to its neighbors and to other countries in the Sub-Saharan region, South Africa is clearly at a higher level of economic development (see Table 1). For example, the country’s GDP per capita is more than 4.6 times larger than the average country in the region and is substantially richer that its neighbors in Southern Africa. This status, together with its post-Apartheid policies, makes South Africa a common destination of immigrants from neighboring countries fleeing economic, humanitarian and climate crises among other reasons. This is observed in Figure 1. Both, in terms of refugees and asylum seekers the country has seen a massive increase in the past 20 years. The rise was so considerable that from 2008 until 2012, the country received the largest number of new refugees and asylum applications worldwide (United Nations High Commissioner for Refugees, 2020). The massive exodus from Syria and Venezuela in recent years have outpaced the numbers of refugees arriving to South Africa, yet it remains a major destination country for the forced displaced. Using data from the World Values Survey, Facchini et al. (2013) show that the o cial discourse from the Immigration Act is at odds with the view of South African voters. The authors find that the share of the population in favor of migration has declined by ten points from 2001 to 2007 and it is supported only by 23 percent of the people. This decline is further confirmed by our own analysis. In Figure 2, we plot our index measuring favorable attitudes towards immigration from 2008 to 2017 created using data from the South African Social Attitudes Survey (see section 4.1 for details on the data and the index creation). This evidence of a less favorable attitudes towards immigrants correlates with the substantial increase in xenophobic attacks in South Africa since 2008 (see Steenkamp (2009); Everatt (2011); Friebel et al. (2013); Mamabolo (2015) for a detailed discussion of these attacks and 6 A major change came with the Immigration Act in 2002 (amended in 2004) 5 possible explanations).7 The large influx of people to South Africa occurs in a context of high levels of unem- ployment. As shown in Figure 3, the country has the largest unemployment rate in the world. This high rate is an outlier even when compared to countries with a similar GDP per capita but also relative to countries in a similar stage of economic development (see Figure B1 for additional comparisons). The high unemployment rate precedes the large influx of immigrants of the last two decades and scholars have discussed the structural nature of its origin (Kingdon and Knight, 2004, 2007; Banerjee et al., 2008).8 Our paper seeks to evaluate whether such views towards immigrants in particular, and towards social cohesion overall, can be a↵ected by distributional policies. As discussed below in section 3, there are theoretical arguments to expect that given South Africa’s high unemployment levels, distributional policies could be e↵ective in promoting social cohesion and more favorable attitudes towards immigrants. In this paper we focus on such a policy: South Africa’s Old Age Pension. 2.2 Distributional policies The Old Age Pension (currently called Old Age Grant) is a monthly cash transfer for older adults.9 It is non-contributory, so payments are drawn from the central government’s general revenue and not from a person’s savings, labor history or payroll deductions. In that sense, it is a clear distributional policies from the general funds to older individuals.10 The current eligibility rules are based on age (60 and older) and applicants must be South African citizens or permanent residents.11 People receiving any other social grant for themselves and those cared for in a state institution are not eligible. The Old Age Pension (OAP) is also means- tested: pensioners cannot earn more than R86,280 if single or R172,560 if married. Also, they cannot have assets worth more than 1,227,600 (single) or R2,455,200 if married. The pension provides a monthly transfer of R1,890 (around US$136.5). Grantees can have their payments suspended when their circumstances change or after their case is reviewed negatively. Other causes for suspension of the grant include failing to cooperate when their grant is reviewed, 7 According to the The African Centre for Migration & Society at the University of the Witwatersrand, 5.3% of the current working age people (15-64) were born outside South Africa (2020 ACMS Fact Sheet). 8 See Biavaschi et al. (2018) as well as World Bank (2018) for a discussion of the impact of immigrants on the South African labor market. 9 See Lund (1993) and Case and Deaton (1998) for a detailed discussion of the grant including its origin and how it changed towards the end of the Apartheid regime. 10 For a discussion of other types of cash transfer programs worldwide see Fiszbein and Schady (2009) and Hanlon et al. (2012). 11 In recent years it has been extended to refugees living in South Africa. However, refugees are underrep- resented in surveys preventing us from study the impact of the pension for this population. 6 fraud or misrepresentation or if there was a mistake when their grant was approved. The grant ends when the person dies, is admitted to a state institution, leaves the country, or does not claim it for three consecutive months.12 The OAP provides a large cash transfer. Ambler (2016) estimates that the amount is two times the “monthly median per-capita household income of noneligible older women and 1.6 times the monthly median per-capita income of noneligible older men” (p. 904). Until 2007, women were eligible at 60 but men only at 65. Starting in 2008, men have become eligible sooner: 63 in 2008, 61 in 20019. Since 2010, the age criteria was standardized to 60 for all. As explained below, our analysis covers the years 2008 to 2017. Given the changes in age eligibility cuto↵ during this period, we re-center the age criteria in each year to zero. Due to its wide coverage (with over 3 million grantees), its large amount and eligibility rules, there is an ample literature evaluating its impacts. Many studies have focused on the labor market impacts (Bertrand et al., 2003; Posel et al., 2006; Lam et al., 2006; Ranchhod, 2006; Ardington et al., 2009; Abel, 2019), while others have studied the e↵ect on children (Duflo, 2003; Edmonds, 2006), household composition (Edmonds et al., 2005; Hamoudi and Thomas, 2014; Ambler, 2016), private transfers within family members (Jensen, 2004), men- tal health (Ag¨uero, 2019) as well as the pension take up patterns and changes in the allocation of income to food, schooling, transfers, and savings (Case and Deaton, 1998). Our focus on social cohesion and attitudes towards immigrants sets us apart from prior work on the im- pacts of the Old Age Pension as our paper is the first to explore outcomes that matter for people outside the beneficiaries themselves or their families. The Old Age Pension is not the only welfare grant provided by the South African govern- ment. The Child Support and the Care Dependency grants focus on children (from birth to age 17) and currently provide R460 (per month per child) and R1,890, respectively.13 The Foster Child grant covers children under 18 who have been placed under custody by a court as a result of being orphaned, abandoned, at risk or abused.14 This grant provides R,1050 (per month per child). The Grant-in-aid consists of a monthly payment of R460 for those living on a social grant but cannot look after themselves. This additional grant is to pay the person taking care on the recipient.15 None of these grants have an eligibility criteria that relates to the OAP’s. However, there are two grants whose age eligibility does coincide with 12 For additional rules and details please visit https://www.gov.za/services/ social-benefits-retirement-and-old-age/old-age-pension. Accessed on May 15, 2021. 13 For more information on the Child Support, a need-based grant, please see https://www.gov.za/ services/child-care-social-benefits/child-support-grant. The Care Dependency grant is to take care of a child who has a severe disability and is in need of full-time and special care. For details visit https: //www.gov.za/services/services-residents/parenting/child-care/care-dependency-grant. 14 https://www.gov.za/services/child-care-social-benefits/foster-child-grant. 15 See also https://www.gov.za/services/social-benefits/grant-aid. Accessed on May 15, 2021. 7 the Old Age Pension. First, there is the War Veterans grant. This is provided to those aged 60 or more who fought in World World II or the Korean War.16 This is not a binding grant for the Black and Coloured populations under study. Finally, the Disability Grant focuses on people aged 18 and 59 and for those who have a physical or mental disability that impedes them to work for a period of longer than six months.17 The grant provides a month payment of R,1890 (same as the OAP). However, as shown below, this grant covers a small fraction of the population under study. See section 5.1 for more details on how does this a↵ect the interpretation of our findings. 3 Theoretical motivation 3.1 Historical and structural triggers for xenophobic violence While the focus of this paper is on the role of cash transfers as way to increase social cohesion and reduce anti-immigration attitudes, there is important scholarship that argues for historical and structural causes for these attitudes that we want to briefly acknowledge.18 For instance, the Apartheid laws heavily controlled the movement of the black majority in South Africa and they were confined to live in townships right outside urban areas. There is an important strand of the literature that argues that current foreign migrants but also internal migrants are seen as actors without a “legitimate” claim to urban spaces (e.g., Landau, 2012). Others have argued that the lack of trust towards public institutions creates non-state authorities claiming de facto control of parts of the territory including urban areas (Monson and Arian, 2012; Kihato, 2011). Furthermore, Misago (2009) discusses how in urban impoverished contexts, violence is the socially acceptable way of manifesting and dealing with grievances. These can manifest against the government or against other people. Structural causes have also been proposed. For example, Crush and Ramachandran (2010) point out that government failures to address endemic poverty, lack of jobs, shelter, and basic services “led to the scapegoating of foreign migrants by frustrated citizens” (p. 16). Recent quantitative work using correlation analysis suggest that poverty, relative depri- vation as well as frustration with the government are some of the triggers for anti-immigrant attitudes (e.g., Fauville and Segatti, 2011; Claassen, 2017). Additionally, Pillay et al. (2008) concludes that “poverty is clearly exacerbating tensions” (p. 12). Our paper seeks to con- 16 For details and other requirements see https://www.gov.za/services/ social-benefits-retirement-and-old-age/war-veterans-grant. 17 https://www.gov.za/services/social-benefits/disability-grant. 18 We want to thank a reviewer for suggesting we add this discussion and for sharing a private document from which this sub-section draws heavily. 8 tribute to this literature by providing a causal analysis of the role of cash transfers on social cohesion and anti-immigration attitudes. The next subsection summarizes the economic theories behind these possible e↵ects.19 3.2 The role of cash transfers Theoretically, cash transfers to poor citizens of the hosting nation could reduce negative attitudes towards immigration and contribute to conflict reduction as well as increase so- cial cohesion. For example, consistent with work on structural triggers summarized above, Mamabolo (2015) and others, have argued that the xenophobic attacks of 2008 in South Africa were driven by –but not limited to– poverty and unemployment. Under such view, distributional policies that reduce poverty via income transfers, as in the case of the Old Age Pension, should lead to more favorable views towards immigrants. In economics, the positive role of cash transfers to natives is derived from models where individuals form their opinion based on the impact that immigration has on their utility (e.g. Mayda, 2006). However, in these models, the impact is uneven across the population and depends on the skill composition of immigrants relative to natives in the destination country. When migrants are on average less skilled than natives, they will hurt unskilled natives and benefit skilled ones through an e↵ect on wages. Thus, unskilled (and poor) natives benefiting from a cash transfer would have a more pro-immigration attitude than a counterfactual situation where they do not have access to such transfer. Instead, when migrants are on average more skilled than natives, as in the South African case (Facchini et al., 2013), the skilled natives would be the ones with more negative views about migrants, leaving the e↵ect of cash transfer to the poor with an unspecified sign, including zero. There is another theoretical argument that could reduce, if not mute, the e↵ect of large a government transfer on attitudes towards immigrants even if when poverty is the main driver. Barro (1974) and Becker (1974) show that public support programs could displace or “crowd out” private support. In their models, the reduction in private support could completely undo any gain from the public transfer. This crowding out e↵ect is highly relevant to the South African context where there is an extensive tradition of within-family transfers (e.g., Posel, 2001; Bowles and Posel, 2005; Hall and Posel, 2019). In fact, Jensen (2004) finds evidence of crowding out in the OAP. He estimates that for each rand from the pension to the elderly there is a 0.25–0.30 rand reduction in private transfers from children living away from home. This behavioral change could limit the impact of the OAP on social cohesion. 19 For additional discussions about these two triggers as well as for additional analysis of see the work by The African Centre for Migration & Society at the University of the Witwatersrand. For a list of their publications see http://www.migration.org.za . This sections cites many of those papers. 9 A similar null e↵ect from cash transfers could also be obtained if the main driver of the anti-immigrant sentiment is based on non-wage factors such as sharing a common eth- nic background (Epstein and Gang, 2010), language barriers (Bauer et al., 2005) or crime (Mayda, 2006). This would also be the case based on the historical triggers discussed earlier. In all these cases, cash transfers would have a limited e↵ect on preventing social conflict and promoting social cohesion, even if the transfers increase income and reduce poverty. All this theoretical ambiguity about the impact of distributional policies, such as cash transfers, demands an empirical evaluation. In the next section we describe the data and the identi- fication strategy that would allows to estimate the causal e↵ect of the Old Age Pension on social cohesion. 4 Research design 4.1 Data sources We use two main data sources to evaluate the impact of the Old Age Pension on social cohesion. The first dataset is composed by multiple rounds of the General Household Survey (GHS). The GHS is an annual (cross-sectional) household survey designed to measure the living circumstances of South African households.20 As such, the GHS collects data on ed- ucation, health and social development, housing, household access to services and facilities, food security, and agriculture. Relevant to our project, the GSH has a dedicated module measuring access to all social programs including the Old Age Pension. We use this in- formation to validate our identification strategy in order to estimate the causal impact of distributional policies on social cohesion. To match the time frame of the dataset used to measure social cohesion, we use GHS from 2008 to 2017. The second source of data is the South African Social Attitudes Survey (SASAS). This is a nationally representative, repeated cross-sectional survey conducted annually by the Human Sciences Research Council (HSRC) since 2003.21 The survey is designed to serve as a “time series” so it maintains a consistent sample design and questionnaire to track “the speed and direction of change in underlying public values and the social fabric of modern South Africa” according to SASAS website. Each round of SASAS interviews individuals 16 and older, regardless of nationality or citizenship, in all nine provinces. Relevant to the calculation of the standard errors for our analysis, the sample has been drawn from the HSRC’s Master Sample, which consists of a sampling frame from 1,000 Population Census 20 The data is available from Data First’s website: https://www.datafirst.uct.ac.za/dataportal/ index.php/catalog 21 The surveys are available from HSRC’s website http://www.hsrc.ac.za/en/departments/sasas 10 enumeration areas (EAs). Every year, a sub-sample of 500 EAs are randomly drawn from the Master Sample, stratified by province, geographical sub-type and majority population group. For the purpose of our paper, the SASAS questionnaire contains a standard set of demo- graphic characteristics including age at the time of the survey as well as over 100 variables to measure a variety of social, economic and political values over time.22 In this regard, SASAS is the South African equivalent to the International Social Survey Programme (ISSP) used by scholars to understand social cohesion and attitudes towards immigrants. As such, SASAS o↵ers a very large set of variables to measure social cohesion. To estimate the impact of OAP on a these variables, we employ two ways to account for the multiple number of outcomes considered. First, we create indices for each family of outcomes resembling the grouping proposed by Valli et al. (2019).23 Let s refers to one of our seven indices so that s = 1, . . . , 7. Specifically, these indices are (i) attitudes toward immigration; (ii) interpersonal trust; (iii) confidence in institutions; (iv) satisfaction in government; (v) life satisfaction; (vi) attitudes towards racial diversity and (vii) social participation. These indices are further aggregated to create an overall measure of social cohesion, by taking the average of all these indices. For each index s there are Js variables. In Table B1 we describe the list of 108 variables considered (by index) and their summary statistics. To construct the indices, we define each outcome j in index s so that higher values correspond with better outcomes (i.e., more social cohesion) following the methodology Kling et al. (2007). We standardize each outcome (yijs ) into a Z-score by subtracting the mean (µc j ) and dividing it c by the standard deviation ( j ). Both µc c j and j are computed over the control group (those not yet elegible for the OAP based on age and depending on the prevalent age cuto↵ in each survey year). The average of these Z-scores is the index. Formally, for each s, the index SCIijs is created based on Equation (1) below: 1 X yijs µc js SCIijs = c j = 1, . . . , Js ; s = 1, . . . , 7 (1) Js j js Second, we correct for the potential issue of simultaneous inference using multiple hy- pothesis testing. Based on Benjamini and Hochberg (1995), we apply the concept of a false discovery rate (FDR) to allow inference when conducting many tests. Intuitively, FDR al- lows a researcher to tolerate a certain number of tests to be incorrectly discovered. An FDR adjusted q-value of 0.05 implies that 5 percent of significant tests result in false positives 22 In addition to the core module, each year a specific theme is included in response to current events. 23 As discussed by these authors, there “there has not been a unified understanding of the specific compo- nents or of the measurement of the concept” (p. 129) of social cohesion. Thus, the inclusion of a vast set of indicators provides the most comprehensive way to measure it. 11 compared with an unadjusted p-value of 0.05 that implies 5 percent of all tests result in false positives. In the regression tables we show standard errors based on unadjusted p-values as well as FDR adjusted q-values that address the multiple hypotheses being tested in a given family of outcomes. 4.2 Identification strategy Evaluating the impact of distributional policies such as the Old Age Pension on social cohe- sion represents a challenge due to the possible presence of unobserved variables. For example, if people who applied and received the pension tend to be more (or less) in favor of social cohesion, then it would not be possible to separate the impact of the cash transfer from their beliefs. What is needed is an exogenous variation that facilitates (or limits) access to the pension in a way that is independent of social cohesion beliefs. Such variation is obtained by the eligibility rules of the Old Age Pension. As discussed in section 2, a critical condition to receive the pension is based on age. For example, since 2010, those aged 60 or more are eligible. Those younger than 60 are not. Of course, as explained before, age is not the only condition for the pension, but at the cuto↵, it alters the probability of receiving the grant in a discontinuos way. Our identification strategy uses this sudden probability jump in access to the pension in a fuzzy Regression Discontinuity Design (RDD). Thus, those age 59 serve as the counterfactual for pension recipients aged 60. For all other ages away from 60, we expect the probability to vary smoothly with age. Since we re-center the data based on the cuto↵ age in each survey year, those aged -1 years younger than the cuto↵ serve as the counterfactual for pensioner just eligible. This is the same identification strategy employed in previous work using the Old Age Pension (e.g., Edmonds, 2006; Ambler, 2016). The fuzzy RDD allows us to predict access to the pension and we use that prediction to estimate the impact on social cohesion by Two-Stage Least Squares (2SLS). Formally, our identification strategy is represented by these two equations: E [OAPit |Ageit ] = 1(Ageit ¯t ) + f (Ageit ) + ✓1 xit a (2) SCIit = E [OAPit |Ageit ] + g (Ageit ) + ✓2 xit + eit (3) The first stage is given by Equation (2), where the probability of receiving the Old Age Pension for person i observed in survey year t is measured by a binary variable (OAPit ). Based on the eligibility rules, this probability depends on age and this is captured by a flexible polynomial on the running variable: f (Ageit ). To account for the discontinuos change in the probability at the cuto↵, which varies over time (¯ at ), we include the indicator variable 12 1(Ageit ¯t ) that becomes a one when the statement inside the parenthesis is true and a zero otherwise. This jump is captured by , which is expected to be positive. The equation includes a few controls (vector xit ) such as fixed e↵ects by province of residence and by survey year as well as controls for gender (male or female) and race (Black or Coloured). Given the nature of the coverage of the Old Age Pension, in both datasets we restrict our working sample to these two races and to those aged ±10 years around the cuto↵ according to the survey year and eligibility rules. The second stage is shown in Equation (3). This equation will be estimated for each measure of social cohesion (SCI ) as captured by the aggregate index, the seven (sub) indices as well as their individual components. As before, the model includes polynomials on age, g (Ageit ), and the previous controls (xit ). The parameter of interest is and represents the magnitude and sign of the impact of Old Age Pension on our measures of social cohesion. In both equations we cluster the standard errors by the surveys’ primary sample unit following the discussion by Koles´ ar and Rothe (2018) for inference when the assignment variable (age) is discrete. In the next section we present evidence in favor of our identification strategy as well as the results of applying it to the datasets discussed earlier. 5 Results 5.1 First stage: access to the Old Age Pension Several assumptions are needed to validate our identification strategy. To do so, we use the General Household Surveys as they contained very rich information regarding access to all the distributional policies in South Africa. The first assumption is that the probability of access to the Old Age Pension must discontinuously jump at the cuto↵ age (a ¯t ). This is clearly shown in the top left panel of Figure 4. To account for changing values of a ¯t over time, we show the running variable as deviations (in years) from the year-specific cuto↵. We can see that individuals younger than the mandated cuto↵ age do not receive the pension. As expected, the probability jumps from zero to 60 percent at the cuto↵, the first age of eligibility. This sudden jump helps validate our identification strategy visually. Note that the probability does not reach one because there are other criteria that matter as explained in section 2. The regression counterpart of this graphical analysis is shown in Table 2. In column (1), we consider a bandwidth of 10 years to each side of the cuto↵ together with the simplest spline for f (Ageit ) as a linear function (but di↵erent for each side of the cuto↵ ) and without the xit controls. The estimated increase in the probability of receiving the Old Age Pension 13 is around 67 percentage points and statistically significant at the one percent. Adding all the controls, as in column (2), does not change our findings. Another assumption to validate our identification strategy requieres all other “baseline” variables to be smooth around the cuto↵. If that is not the case, one cannot isolate the e↵ect of the pension from those characteristics. This is shown graphically in the reminder panels of Figure 4 as well as in Figure 5. In the former, we plot the probability of receiving all other welfare grants available in South Africa. As discussed in section 2, eligibility for all other welfare programs focused on younger individuals and therefore there is no jump at the OAP cuto↵. The exception is the Disability Grant that ends at 59. However, as shown in Figure 4, take up is low (under 17 percent). This means, as in all previous papers analyzing the impact of the pension, that the e↵ect of the Old Age Pension is net of the di↵erences with respect to the Disability Grant. Figure 5 also helps validate the assumption showing the smoothness around the cuto↵ for all the control variables: gender, province of residence, survey year and race. The regression counterparts are shown in Table B2. Out of 16 tests only two report statistically significant results (but at only at 10% and 5%). Yet, none are significant when adjusting the analysis to account for the discrete nature of the running variable as shown by the local randomization estimates at the bottom of the table. We conducted multiple additional robustness tests to further validate our identification strategy. See Appendix A for details. The extensive set of robustness checks accounting for the discrete nature of the assignment variable shows that for the case of the Old Age Pension such additional considerations do not di↵er from the estimates treating it as a continuos at least within the bandwidth used in previous work (e.g., Ambler, 2016; Ag¨ uero, 2019). 5.2 Distributional policies and social cohesion We now focus on the e↵ect of the pension on social cohesion. We consider a reduced-form approach using data from SASAS between 2008 to 2017. That is, we are interested in whether we observe a discontinuity in the outcomes of interest around the cuto↵ of the running variable. To obtain the 2SLS one would have to divide our reduced-form estimates by the 0.60, the jump in the probability of receiving the Old Age Pension at the cuto↵.24 We first analyze the impact on the social cohesion overall index. This is done by combin- ing all 108 variables into seven indices and using the latter to create an overall index. We start with a visual inspection as shown in the top left panel Figure 6. There is almost no evidence of a jump in the social cohesion index at the cuto↵. In Figure 7 we report the esti- 24 This approach is equivalent to a two-sample instrumental variable. See Inoue and Solon (2010) for details. 14 mated coe cient (with the 95% confidence interval) and Table 3 complements it (column 1). In both cases, we see a small positive e↵ect. At the cuto↵ the social cohesion index increases by 1.7 percent of a standard deviation (0.017 ) but it is not statistically significant. The same can be said about the impact on attitudes towards immigrants. Figure 6 also shows no major jump at the cuto↵ and the point estimate reports a positive but even smaller e↵ect on the index. As shown in Figure 7 and Table 3, the estimated e↵ect on this index is around 0.006 and not statistically di↵erent from zero. When exploring the 21 variables used to create this index we see very small e↵ects (Figure 8 and Table B5). Two exceptions are worth mentioning. First, we found that the OAP increases the view that immigrants bring skills (column 11 of Table B5). This statistically significant at the 10 percent level. However, when considering the q-value adjustment for multiple hypothesis testing, this positive e↵ect cannot be separate from mere chance given the large number of outcomes. Something di↵erent is observed for the impact about the view that the government should welcome immigrants. We find an increase of 6.1 percent at the cuto↵ relative the mean of the control group (=0.125/2.034). This is significant at the one percent using traditional p-values but also after accounting for multiple hypothesis testing. Thus, this positive e↵ect is not driven by chance. The implied 2SLS would be an increase of 10.2 percent due to the OAP. Nonetheless, the overall small or even null e↵ect on social cohesion and on attitudes to- wards immigrants would be consistent with the three theories discussed in section 3. For in- stance, following Mayda (2006), the poor would be less likely to have a more pro-immigration attitude with an income transfer given the higher skills set of immigrants to South Africa. Also, this small impact would be consistent with the crowding-out e↵ect of the OAP re- ported by Jensen (2004). Third, if anti-immigration attitudes are more rooted on non-wage factors such as ethnic background (Epstein and Gang, 2010), language barriers (Bauer et al., 2005) or crime (Mayda, 2006) or on historical triggers (see section 3), cash transfers would not a↵ect social cohesion much. By exploring the other components of the social cohesion index we would rule out the first two explanations in favor of the third: the role of non-wage factors. Our findings are in line with recent work in other developing countries using credible iden- tification strategies. For example, Valli et al. (2019) use a clustered randomized controlled trial and find that cash transfers to locals in a refugee-hosting setting such as Ecuador has no impact on social cohesion. However, for papers using alternative strategies, our results point to a much smaller e↵ect (e.g., Pavanello et al., 2016). Yet, drawing a strong comparison is always limited by the di↵erences in the historical, cultural and structural issues across 15 settings.25 We now focus on the other components of the social cohesion index. For interpersonal trust, the impact is positive (0.043 ) but imprecise given the large standard errors. The same is observed when looking at the five variables that create the index (see Figure 9 and Table B6). We arrive to the same conclusion when considering the impact on either trust in government institutions –both in the index and its 15 components (Figure 10 and Table B7)– or satisfaction with government policies (Figure 11 and Table B8) and its 11 variables. When considering life satisfaction the e↵ects are di↵erent (Figure 12 and Table B9). The increase in the index at the cuto↵ is a notable 0.086 and statistically di↵erent from zero. The implied 2SLS suggest that the OAP increases this index by 0.14 . Half of the 14 variables of this index report a significance robust to the adjustment for multiple hypothesis testing. This means that beneficiaries do perceive an improvement in their wellbeing as a result of the grant. This is consistent with prior work focusing on other outcomes. For example, Duflo (2003) finds that anthropometric measures of girls increase when their grandmother receives the pension. Ag¨ uero (2019) shows that the pension improves the mental health status of the recipients. Ambler (2016) reports an increase in women’s income (and less on men due to labor force withdrawal). Edmonds (2006) finds results consistent with the pension lifting liquidity constraints by studying the rise in school enrollment of children of beneficiary families. All this evidence suggests that there is a clear monetary gain from the pension despite the crowding out e↵ect found by Jensen (2004). This is further reinforced by our findings on racial diversity. This index jumps by 0.028 at the age cuto↵, significant at the 10 percent, with an implied 2SLS estimate of 0.047 . Exploring the 35 variables that composed this index we find positive impacts on trusting Black South African and considering Coloured as friendly that remain significant even after accounting for multiple hypothesis testing (Figure 13 and Table B10). This suggests that cash transfers can alter beliefs about other people, but this is limited to people “closer” to them as our sample centers only on Black and Coloured South Africans. Our last index of social cohesion focuses on social participation (see Figure 14 and Table B11). We observe a statistically significant decline (at the 10 percent) at the age cuto↵ of about 0.04 . We found that recipients are less likely to contact the media, the government or participate in a protest. Again, this is consistent with the increase in overall wellbeing discussed above. These findings are robust to changes in the bandwidth as expected given the analysis of the first stage. For instance, in Figure 15 we limit the analysis to observations within 25 See also Lehmann and Masterson (2015) for an evaluation of transfers to refugees instead of the hosting population. 16 three years of the cuto↵ age. As observed there, the e↵ects are in line with the main results, except for the much larger standard errors due to the smaller sample size. Finally, we also explore whether the e↵ects vary by the intensity of the flow on immigrants. To do so, we merge the data from United Nations High Commission of Refugees (used in Figure 1) and interacted each of them with the splines (polynomials on Ageit ) as well as the indicator variable for the cuto↵ age: 1(Ageit ¯t ). In Figures 16 and 17, we respectively a consider number of refugees and number of asylum seekers. The interaction with these variables does not have an impact on social cohesion, supporting our main findings. 6 Policy and program implications Whether distributional policies can lead to more social cohesion is a critical question given the large amount of people forcibly displaced worldwide who are facing severe anti-immigrant reactions from the local population. We answer this question in the context of South Africa, a country that is a major hub of refugees and asylum seekers and that has seen an increase in xenophobic behaviors since 2008. The country has the largest unemployment rate worldwide and a widespread and important distributional policy that permits a rigorous causal estima- tion of its impact. Exploring whether cash transfers policies could provide an e↵ective tool to reduce anti-immigration attitudes and promote social cohesion in refugee-hosting settings with high unemployment rates is of even more relevance given the economic decline and increase in refugees expected in many developing countries due to the COVID-19 pandemic. Our findings show that the South Africa’s Old Age Pension has a limited impact on social cohesion. However, this hides two competing forces. On the one hand, we found an increase in the recipients’ life satisfaction and in their attitudes towards racial diversity, including in-group trust. These impacts are consistent with a net gain in income and wellbeing. This is despite the crowding out e↵ect reducing private transfers within family members to elder individuals documented elsewhere. On the other hand, this increase in wellbeing does not translate into more interpersonal trust and neither on trust or satisfaction with the government. We identify a decline in social participation. Overall, there is no change in attitudes toward immigrants, except for a higher desire that the government welcomes more foreigners. Based on our results and the previous literature on South Africa’s Old Age Pension, this unconditional cash transfer program has improved welfare and life satisfaction but did not improve social cohesion. Whether similar results would be observed for cash transfers programs conducted in di↵erent contexts remains an open question and should be explored in future research. This suggest that the anti-immigrant behaviors in South Africa are less likely to be related 17 to income and more about non-wage factors such as ethnic background or language barriers or crime but also to historical triggers given the legacy of Apartheid. It is a pending question whether cash transfers could improve social cohesion in other contexts or where inequality and unemployment are less salient. Other strategies should be explored too. For example, King et al. (2010) review the e↵ectiveness of community-driven development and curriculum interventions on improving social cohesion in sub-Saharan Africa and find mixed results. Koehler (2021) proposes “access to employment and income programmes, and equitable access to social services and social infrastructure” (p. 643) echoing suggestions made by Schuettler and Caron (2020) on employment channels. Future work should address these alternative strategies in more detail while finding credible identification strategies to isolate those e↵ects from possible confounders. References Abel, M. (2019). Unintended labor supply e↵ects of cash transfer programs: New evidence from South Africa’s pension. Journal of African Economies, 28(5):558–581. uero, J. M. (2019). Sit back and relax? the e↵ect of income on mental health in South Ag¨ Africa. Unpublished working paper, University of Connecticut. Ambler, K. (2016). Bargaining with grandma: The impact of the South African pension on household decision-making. Journal of Human Resources, 51(4):900–932. Ardington, C., Case, A., and Hosegood, V. (2009). Labor supply responses to large social transfers: Longitudinal evidence from South Africa. American Economic Journal: Applied Economics, 1(1):22–48. Banerjee, A., Galiani, S., Levinsohn, J., McLaren, Z., and Woolard, I. (2008). Why has unemployment risen in the new South Africa? Economics of Transition, 16(4):715–740. Barro, R. J. (1974). Are government bonds net wealth? Journal of Political Economy, 82(6):1095–1117. Bauer, T., Epstein, G. S., and Gang, I. N. (2005). Enclaves, language, and the location choice of migrants. Journal of Population Economics, 18(4):649–662. Becker, G. S. (1974). A theory of social interactions. Journal of Political Economy, 82(6):1063–1093. 18 Benjamini, Y. and Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57(1):289–300. Bertrand, M., Mullainathan, S., and Miller, D. (2003). Public policy and extended families: Evidence from pensions in South Africa. The World Bank Economic Review, 17(1):27–50. Biavaschi, C., Facchini, G., Mayda, A. M., and Mendola, M. (2018). South–south migration and the labor market: evidence from South Africa. Journal of Economic Geography, 18(4):823–853. Bowles, S. and Posel, D. (2005). Genetic relatedness predicts South African migrant workers’ remittances to their families. Nature, 434(7031):380–383. Case, A. and Deaton, A. (1998). Large cash transfers to the elderly in South Africa. The Economic Journal, 108(450):1330–1361. Cattaneo, M. D., Idrobo, N., and Titiunik, R. (2019). A practical introduction to regression discontinuity designs: Extensions. Cambridge University Press. Claassen, C. (2017). Explaining south african xenophobia. Afrobarometer Working Paper, (173). Crush, J. (2000). Migrations past: An historical overview of cross-border movement in Southern Africa. In McDonald, D., editor, On borders: Perspectives on international migration in Southern Africa, pages 12–24. St. Martin’s Press, New York. Crush, J. and Ramachandran, S. (2010). Xenophobia, international migration and develop- ment. Journal of Human Development and Capabilities, 11(2):209–228. Duflo, E. (2003). Grandmothers and granddaughters: old-age pensions and intrahousehold allocation in South Africa. The World Bank Economic Review, 17(1):1–25. Edmonds, E. V. (2006). Child labor and schooling responses to anticipated income in South Africa. Journal of Development Economics, 81(2):386–414. Edmonds, E. V., Mammen, K., and Miller, D. L. (2005). Rearranging the family? income support and elderly living arrangements in a low-income country. Journal of Human Resources, 40(1):186–207. Epstein, G. S. and Gang, I. N. (2010). Migration and culture. Emerald Group Publishing Limited. 19 Everatt, D. (2011). Xenophobia, state and society in South Africa, 2008–2010. Politikon, 38(1):7–36. Facchini, G., Mayda, A. M., and Mendola, M. (2013). What drives individual attitudes towards immigration in South Africa? Review of International Economics, 21(2):326–341. Fauville, C. and Segatti, A. (2011). People, space and politics: An exploration of factors explaining the 2008 anti-foreigner violence in south africa. Exorcising the demons within: Xenophobia, violence and statecraft in contemporary South Africa. Ferguson, N. T. N., Wolfeb, R. J., Aminec, L., Ramadic, E., and Shahind, L. (2021). Build- ing stability between host and refugee communities: Evidence from a TVET program in Jordan and Lebanon. Unpublished Working Paper. Commissioned as part of the “Pre- venting Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series, World Bank Group, Washington, DC. Fiszbein, A. and Schady, N. R. (2009). Conditional cash transfers: reducing present and future poverty. World Bank Publications. Foltz, J. and Shibuya, S. (2021). The e↵ects of internally displaced peoples on income and inequality in Mali. Unpublished Working Paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series, World Bank Group, Washington, DC. Friebel, G., Gallego, J. M., and Mendola, M. (2013). Xenophobic attacks, migration inten- tions, and networks: evidence from the South of Africa. Journal of Population Economics, 26(2):555–591. Hall, K. and Posel, D. (2019). Fragmenting the family? the complexity of household migra- tion strategies in post-apartheid South Africa. IZA Journal of Development and Migration, 10(2):1–20. Hamoudi, A. and Thomas, D. (2014). Endogenous coresidence and program incidence: South Africa’s Old Age Pension. Journal of Development Economics, 109:30. Hanlon, J., Barrientos, A., and Hulme, D. (2012). Just give money to the poor: The devel- opment revolution from the global South. Kumarian Press. Inoue, A. and Solon, G. (2010). Two-sample instrumental variables estimators. The Review of Economics and Statistics, 92(3):557–561. 20 Jensen, R. T. (2004). Do private transfers ‘displace’ the benefits of public transfers? evidence from South Africa. Journal of Public Economics, 88(1-2):89–112. Kihato, C. W. (2011). The city from its margins: rethinking urban governance through the everyday lives of migrant women in johannesburg. Social Dynamics, 37(3):349–362. King, E., Samii, C., and Snilstveit, B. (2010). Interventions to promote social cohesion in sub-saharan africa. Journal of development e↵ectiveness, 2(3):336–370. Kingdon, G. and Knight, J. (2007). Unemployment in South Africa, 1995–2003: Causes, problems and policies. Journal of African Economies, 16(5):813–848. Kingdon, G. G. and Knight, J. (2004). Unemployment in South Africa: The nature of the beast. World Development, 32(3):391–408. Kling, J. R., Liebman, J. B., and Katz, L. F. (2007). Experimental analysis of neighborhood e↵ects. Econometrica, 75(1):83–119. Koehler, G. (2021). E↵ects of social protection on social inclusion, social cohesion and nation building. In Handbook on Social Protection Systems. Edward Elgar Publishing. ar, M. and Rothe, C. (2018). Inference in regression discontinuity designs with a discrete Koles´ running variable. American Economic Review, 108(8):2277–2304. Lam, M., Leibbrandt, G., and Ranchhod, V. (2006). Labor force withdrawal of the elderly in South Africa. In Aging in Sub-Saharan Africa: Recommendation for Furthering Research. National Academies Press (US). Landau, L. B. (2012). Exorcising the demons within: Xenophobia, violence and statecraft in contemporary South Africa. Wits University Press. Lehmann, C. and Masterson, D. T. (2015). Impact evaluation of a cash-transfer programme for syrian refugees in lebanon. Field Exchange 48, page 56. Lund, F. (1993). State social benefits in South Africa. International Social Security Review, 46(1):5–25. Mamabolo, M. (2015). Drivers of community xenophobic attacks in South Africa: poverty and unemployment. The Journal for Transdisciplinary Research in Southern Africa, 11(4):143–150. Mayda, A. M. (2006). Who is against immigration? a cross-country investigation of individ- ual attitudes toward immigrants. The Review of Economics and Statistics, 88(3):510–530. 21 Misago, J. P. (2009). Xenophobic violence in south africa: reflections on causal factors and implications. Synopisis: Centre for policy studies. Policy studies bulletin of CPS, 3(3):3–28. Monson, T. and Arian, R. (2012). Media memory: A critical reconstruction of the may 2008 violence. United Nations University Press with Wits University Press. Nhlapo, M. S., Anderson, B. A., and Wentzel, M. (2017). Trends in voting in south africa 2003-2014. University of Michigan Population Studies Center Research Report, (17-881). Nunn, N. and Wantchekon, L. (2011). The slave trade and the origins of mistrust in Africa. American Economic Review, 101(7):3221–52. Pavanello, S., Watson, C., Onyango-Ouma, W., and Bukuluki, P. (2016). E↵ects of cash transfers on community interactions: emerging evidence. The Journal of Development Studies, 52(8):1147–1161. Pillay, S., Barolsky, V., Naidoo, V., Mohlakoana, N., and Hadland, A. (2008). Citizenship, violence and xenophobia in south africa: Perceptions from south african communities. Posel, D., Fairburn, J. A., and Lund, F. (2006). Labour migration and households: A reconsideration of the e↵ects of the social pension on labour supply in South Africa. Economic Modelling, 23(5):836–853. Posel, D. R. (2001). Intra-family transfers and income-pooling: a study of remittances in KwaZulu-Natal. South African Journal of Economics, 69(3):501–528. Ranchhod, V. (2006). The e↵ect of the South African old age pension on labour supply of the elderly. South African Journal of Economics, 74(4):725–744. Schuettler, K. and Caron, L. (2020). Jobs interventions for refugees and internally displaced persons. World Bank Jobs Working Paper, (47). Schulz-Herzenberg, C. (2019). To vote or not? testing micro-level theories of voter turnout in south africa’s 2014 general elections. Politikon, 46(2):139–156. Sedova, B., Ludolph, L., and Talevi, M. Inequality and security in the aftermath of internal population displacement shocks: evidence from Nigeria. Unpublished Working Paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series, World Bank Group, Washington, DC. 22 Steenkamp, C. (2009). Xenophobia in South Africa: What does it say about trust? The Round Table, 98(403):439–447. United Nations High Commissioner for Refugees (2020). UNHCR Global Trends Forced Displacement in 2019. UNHCR. Valli, E., Peterman, A., and Hidrobo, M. (2019). Economic transfers and social cohesion in a refugee-hosting setting. The Journal of Development Studies, 55(sup1):128–146. World Bank (2018). Mixed migration, forced displacement and job outcomes in South Africa. World Bank. World Bank (2020). World Development Indicators. The World Bank. 23 Figure 1: Refugee and asylum-seekers in South Africa: 2000-2020 Refugee population Asylum-seekers 120000 1000000 100000 80000 500000 60000 40000 20000 0 2000 2005 2010 2015 2020 2000 2005 2010 2015 2020 Year Year Note: Authors’ calculation based on data from the United Nations High Commission of Refugees. Obtained from https://www.unhcr.org/refugee-statistics/download/?url=fkIpe3. 24 Figure 2: Trends in attitudes toward immigrants .1 Attitudes towards immigrants (index) -.05 0-.1 .05 2008 2009 2010 2011 2012 2013 2015 2016 2017 Note: See text for details on the construction of the index. Higher values represent more positive attitudes towards immigrants. Sample is restricted to Blacks and Coloured within ±10 years of the age cuto↵ for the Old Age Pension in each year. Data source: South African Social Attitudes Survey 2008-2017. 25 Figure 3: Unemployment rate and Income: 2018 30 South Africa Total unemployment rate (%) 20 10 0 6 8 10 12 GDP per capita (constant 2010 US$, log) Note: Each symbol represents a country. Total unemployment rate is based on ILO estimates. Data source: World Bank’s World Development Indicators. 26 Figure 4: Access to Old Age Pension and age Old age pension Disability grant Child support grant Care dependency grant 1 1 1 1 .8 .8 .8 .8 .6 .6 .6 .6 .4 .4 .4 .4 .2 .2 .2 .2 0 0 0 0 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 Foster child grant War veterans grant Grant in aid Social relief 27 1 1 1 1 Proportion .8 .8 .8 .8 .6 .6 .6 .6 .4 .4 .4 .4 .2 .2 .2 .2 0 0 0 0 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 Age in years (deviation from threshold) Note: Each circle represents the average participation for the welfare programs by age measured as deviations from the cuto↵ age in each survey year. The red vertical lines show the standardized cuto↵ age. Data source: General Household Survey 2014-2017. Figure 5: Smoothness test: covariates 28 Note: Each circle represents the average participation for the selected variables by age measured as deviations from the cuto↵ age in each survey year. The red vertical lines show the standardized cuto↵ age. Data source: General Household Survey 2014-2017. Figure 6: Social cohesion indices and age Social Cohesion Attitudes on Immigrants Interpersonal Trust Trust in government .2 .2 .2 .2 .1 .1 .1 .1 0 0 0 0 -.1 -.1 -.1 -.1 -.2 -.2 -.2 -.2 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 Index Satisfaction with government Life satisfaction Racial diversity Social Participation .2 .2 .2 .2 .1 .1 .1 .1 0 0 0 0 -.1 -.1 -.1 -.1 -.2 -.2 -.2 -.2 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 Age in years (deviation from threshold) Note: Each circle represents the average value for the selected indices by age measured as deviations from the cuto↵ age in each survey year. The red vertical line shows the standardized cuto↵ age. Data source: South African Social Attitudes Survey 2008-2017. 29 Figure 7: Impact on social cohesion indices: reduced form estimates Social Cohesion Attitudes on Immigrants Interpersonal Trust Trust in government Satisfaction with government Life satisfaction Racial diversity Social Participation -.1 -.05 0 .05 .1 .15 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 30 Figure 8: Impact on about attitudes towards immigrants All people are welcomed Immigrants are welcomed Welcome all immigrants Welcome some immigrants Xenophobia is not an issue Immigrants increase crime Immigrants good for economy Immigrants take SA jobs Immigrants expand ideas Immigrants bring diseases Immigrants bring skills Immigrants use up resources Foreigners are positive Foreigners are friendly No nationality discrimination Immigrants: ahead at my expense Immigrants exclude me Immigrants theat to my traditions Immigrants dont understand me Welcomes immigrants Gov should welcome immigrants -.4 -.2 0 .2 .4 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 31 Figure 9: Impact on interpersonal trust Most people can be trusted Ask neighbor: cup of sugar Ask neighbor: money People would take advantage of you Racial groups wont trust each other -.2 0 .2 .4 .6 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 32 Figure 10: Impact on trust in institutions Electoral Commission (IEC) South African TV (SABC) Defense Forces Trade Union leader National government Courts Parliament Police Local government Churches Political parties Politicians SARS Trade Union -.4 -.2 0 .2 .4 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 33 Figure 11: Impact on satisfaction with government Handling corruption Water supply \& sanitation Electricity Removal of refuse Affordable housing Health care STIs, HIV Cutting crime Job creation Land reform Social grants provision Education -.2 -.1 0 .1 .2 .3 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 34 Figure 12: Impact on life satisfaction Life Living standards Health Life achievements Personal relationships Safety Community Finances Religion Satisfied in general Political leaders Daily activities Family members Democracy Economic situation -.2 0 .2 .4 .6 .8 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 35 Figure 13: Impact on racial diversity Not a Member of discriminated group No Racial Discrimination Members of Racial Group not discriminated Racism is not an issue No racial discrimination Since '94, race relations improved Acquaintances from abroad Knows Blacks Knows Coloured Knows Indians Knows Whites Black friends Coloured friends Indian friends White friends Contact with other race groups we Contact with other race groups, c Coloured are positive Coloured are friendly Indians are positive Indians are friendly Whites are positive Whites are friendly Whites are trustworthy Blacks are positive Blacks are friendly Blacks are trustworthy Since '94 race relations improved Favors racial quotas in sports Favors pref hiring of Blacks Gov promoting racial equality Gov promoting future racial equality Racial diversity in schools Language diversity in schools Affirmative Action: creates more skills Affirmative Action unifies society -.5 0 .5 1 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 36 Figure 14: Impact on social participation Contacted government Worked in action group Protested Signed petition Contacted media Contacted traditional leader -.06 -.04 -.02 0 .02 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 37 Figure 15: Robustness checks: Impact on social cohesion indices (smaller bandwidth) Social Cohesion Attitudes on Immigrants Interpersonal Trust Trust in government Satisfaction with government Life satisfaction Racial diversity Social Participation -.3 -.2 -.1 0 .1 .2 Note: Each circle reports the estimate at the threshold (Age a ¯) with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Sample was limited to ages within three years of the eligibility cuto↵. Data source: South African Social Attitudes Survey 2008-2017. 38 Figure 16: Heterogenous e↵ects: Impact on social cohesion indices by immigration flow (refugees) Social Cohesion Attitudes on Immigrants Interpersonal Trust Trust in government Satisfaction with government Life satisfaction Racial diversity Social Participation -.04 -.02 0 .02 .04 Note: Each circle reports the estimate at the threshold (Age a ¯) interacted with the number of refugees by year (in ten thousands) together with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 39 Figure 17: Heterogenous e↵ects: Impact on social cohesion indices by immigration flow (asylum seekers) Social Cohesion Attitudes on Immigrants Interpersonal Trust Trust in government Satisfaction with government Life satisfaction Racial diversity Social Participation -.04 -.02 0 .02 .04 Note: Each circle reports the estimate at the threshold (Age a ¯) interacted with the number of asylum seekers by year (in hundred thousands) together with the confidence intervals at the 95%. Each regression includes splines and controls for gender, race and province of residence fixed e↵ects. Data source: South African Social Attitudes Survey 2008-2017. 40 Table 1: Economic Development in South Africa and neighboring countries (2018) (1) (2) (3) (4) (5) (6) (7) (8) Botswana Eswatini Lesotho Mozambique Namibia South Africa Sub-Saharan Africa Zimbabwe GDP per capita 3,763 2,897 722 340 4,516 6,556 1,420 1,219 Infant mortality 52 67 93 117 45 48 79 60 Under-5 mortality 77 98 123 175 69 68 128 97 Life expectancy 57 53 52 49 58 59 54 55 Note: Authors’ calculations based on World Bank’s World Development Indicators. GDP per capita is measured in constant 2010 US$. Infant and under 5 mortality is per 1,000 live births. Life expectancy (in years) is at birth. 41 Table 2: First stage: Old Age Pension and age (1) (2) (3) (4) Old age pension Old age pension Old age pension Old age pension Age a ¯ 0.669⇤⇤⇤ 0.670⇤⇤⇤ 0.590⇤⇤⇤ 0.591⇤⇤⇤ [0.009] [0.009] [0.013] [0.012] Controls No Yes No Yes Observations 46010 46010 4711 4711 R2 0.722 0.726 0.420 0.434 ¯ c Y 0.000 0.000 0.000 0.000 Bandwidth (-10,9) (-10,9) (-1,0) (-1,0) Note: Robust standard errors (in brackets) are clustered at the primary sample unit level. All regressions include linear splines and controls for gender, race as well as province of residence and survey year fixed ¯c refers to the mean of the dependent variable (column title) among the non-eligible population e↵ects. Y (Age< a¯t ). ⇤p < 0.10, ⇤ ⇤ p < 0.05, ⇤ ⇤ ⇤p < 0.01. 42 Table 3: Impact on social cohesion indices: Reduced form (1) (2) (3) Social Cohesion Attitudes on Immigrants Interpersonal Trust Age ¯ a 0.017 0.006 0.043 [0.015] [0.019] [0.035] Observations 4419 4419 4419 R2 0.026 0.039 0.031 ¯ c Y -0.005 -0.042 0.014 (4) (5) (6) Trust in government Satisfaction with government Life satisfaction Age a ¯ -0.033 0.035 0.086⇤⇤⇤ [0.034] [0.031] [0.033] Observations 4419 4417 4418 R2 0.072 0.035 0.052 ¯ c Y -0.004 -0.007 -0.044 (7) (8) Racial diversity Social Participation Age ¯ a 0.028⇤ -0.044⇤ [0.017] [0.027] Observations 4419 4419 R2 0.073 0.047 ¯ c Y 0.000 0.047 Note: Robust standard errors (in brackets) are clustered at the primary sample unit level. All regressions include linear splines and controls for gender, race as well as province of residence and survey year fixed ¯c refers to the mean of the dependent variable (column title) among the non-eligible population e↵ects. Y (Age< a¯t ). ⇤p < 0.10, ⇤ ⇤ p < 0.05, ⇤ ⇤ ⇤p < 0.01. 43