Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 May 1.
Published in final edited form as: Soc Sci Med. 2023 Mar 31;324:115883. doi: 10.1016/j.socscimed.2023.115883

The impact of a randomized cash transfer intervention on mortality of adult household members in rural South Africa, 2011–2022

Molly Rosenberg a,e, Erika Beidelman a, Xiwei Chen b, David Canning c, Lindsay Kobayashi d,e, Kathleen Kahn e,f,g, Audrey Pettifor e,h, Chodziwadziwa Whiteson Kabudula e
PMCID: PMC10124166  NIHMSID: NIHMS1889353  PMID: 37023659

Abstract

Background:

Increasing socioeconomic resources through cash transfer payments could help promote healthy longevity. However, research in this area is limited due to endogeneity in cash transfer exposures and limited geographic representation.

Methods:

We leveraged the HPTN 068 randomized cash transfer trial, conducted from 2011–2015 in a rural setting in South Africa. We assessed long-term mortality follow-up (until March 2022) on older adult members (n=3568) of households enrolled in the trial from the complete Agincourt Health and socio-Demographic Surveillance System census of the underlying source population. The trial intervention was a monthly cash payment of 300 Rand conditional on school enrollment of index young women. The payments were split between the young woman (1/3) and their caregiver (2/3). Young women and their households were randomized 1:1 to intervention vs. control. We used Cox PH models to compare mortality rates in older adults living in intervention vs. control households.

Findings:

The cash transfer intervention did not significantly impact mortality in the full sample [HR (95% CI): 0.94 (0.80, 1.10)]. However, we observed strong protective effects of the cash transfer intervention among those with above-median household assets [HR (95% CI): 0.66 (0.50, 0.86)] and higher educational attainment [HR (95% CI): 0.37 (0.15, 0.93)].

Interpretation:

Our findings indicate that short-term cash transfers can lead to reduced mortality in certain subgroups of older adults with higher baseline socioeconomic status. Future work should focus on understanding the optimal timing, structure, and targets to maximize the benefits of cash transfer programs in promoting healthy aging and longevity.

Keywords: cash transfer, income, mortality, South Africa, older adults, aging

INTRODUCTION

Although global life expectancy has increased by more than six years since 2000 (World Health Organization, 2020), high-income countries have generally reached much higher life expectancies than their low- and middle-income country (LMIC) counterparts (Jetter et al., 2019; Preston, 1975). For example, in 2019 the United States had an estimated overall life expectancy at 78.5 years, while in South Africa, it was over a decade younger at 65.3 years (World Health Organization, 2020). At the same time, LMIC settings like those in sub-Saharan Africa are the very areas where aging populations are rapidly growing. The population of adults over 65 living in African countries is projected to more than triple from 2019 to 2050 (United Nations Department of Economic and Social Affairs Population Division, 2019). The collision of these aging population dynamics with underlying LMIC environments that pose challenges to healthy longevity create the conditions for public health crises. There is an urgent need to address disparities in life expectancies between high- and low-income settings, and to identify interventions that can prolong healthy lives and promote healthy aging globally.

Cash transfer interventions could plausibly help to reach this goal (Freeman et al., 2020). Cash transfers provide supplemental income to recipients through direct monetary payments. As low socioeconomic status (SES) is consistently linked to increased mortality in both high- and low-income settings (Bosma et al., 2001; Chetty et al., 2016; Kabudula et al., 2017a; Kinge et al., 2019; Steenland et al., 2004a; Steenland et al., 2004b), increasing SES through cash transfer payments could help promote healthy longevity (Figure 1)(Forde et al., 2012b; Sun et al., 2021). Cash transfers increase healthcare access and utilization (Ranganathan & Lagarde, 2012; Salinas-Rodríguez & Manrique-Espinoza, 2013); increase access to healthy food and clean water leading to improved nutritional status and food/water security; and increase investments in household infrastructure, such as quality sanitation, roofing, and kitchen stoves (Sun et al., 2021). These established cash transfer benefits provide plausible pathways to improved health and long-term survival. However, much of the current evidence base for cash transfer interventions is drawn from programs designed to improve child health and well-being (e.g. Brazil’s Bolsa Familia, Mexico’s Progresa, and South Africa’s Child Support Grant). It is less clear how much of these benefits ‘trickle up’ to reach other adult household members.

Figure 1.

Figure 1.

Open in a new tab

Conceptual model outlining hypothesized pathways between cash transfer intervention and adult mortality

Although much of the prior evidence supports pathways to a protective effect of cash transfers on survival, there are plausible pathways in the opposite direction as well. It is possible that increased income could facilitate access to exposures that negatively impact survival, such as high fat and calorically dense foods (Fernald et al., 2008; Forde et al., 2012a; Leroy et al., 2013), alcohol, and tobacco. However, a growing number of studies have evaluated outcomes in this area, and they tend to agree that the concern around cash transfers leading to increased consumption of these ‘temptation goods’ is not backed by evidence (Evans & Popova, 2014; Haushofer & Shapiro, 2013).

The existing evidence base for the relationship between cash transfer interventions and mortality is mixed and limited in scope. The bulk of the evidence that exists is weighted towards a protective effect. Findings from Latin America show significant reductions in adult mortality attributable to national cash transfer programs (Barham & Rowberry, 2013; Pescarini et al., 2020). However, preliminary evidence from Mexico evaluating a pension program (70 y Más) found a 5% increase in older age mortality attributable to the program (Feeney, 2017). A more robust set of evidence shows protective impacts of cash transfers on maternal, infant, and child mortality outcomes (Okeke & Abubakar, 2020; Ramos et al., 2021; Rasella et al., 2021), though these effects may be achieved through different pathways (e.g. childhood vaccinations, prenatal care) than more general adult mortality outcomes.

There are clear gaps in the current evidence base. Evaluating the causal nature of the relationship between cash transfers and mortality is challenging due to the inherent endogeneity of many cash transfer exposures, as few truly randomized cash transfer interventions exist, and due to the long-term follow-up required to observe longitudinal trends in mortality outcomes. The literature is also comprised of studies exclusively from Latin America, leaving a large research gap for other LMIC settings, especially those in sub-Saharan Africa. Finally, there is a growing body of evidence suggesting that cash transfer interventions have heterogeneous effects on health across population subgroups (e.g., sex, SES, education) (Cooper et al., 2020). It is important to formally incorporate subgroup analyses to be able to make policy recommendations about how to best design and target cash transfer programs to promote healthy longevity.

In this study, we leverage a randomized cash transfer intervention (HPTN 068), which was administered in rural South African communities from 2011–2015, to evaluate its association with long-term mortality outcomes of adults aged 40+ living in the same households as trial participants. We augment this unique randomized cash transfer intervention with high-quality longitudinal mortality data through March 2022 from the Agincourt Health and socio-Demographic Surveillance System (HDSS), the sampling frame underpinning the trial. Using these merged datasets, we estimate the causal effect of the cash transfer intervention on adult mortality, evaluate time trends in these effects, explore whether effects differed by duration and nature (direct vs. spillover) of transfer receipt, and evaluate whether effects differed across key population subgroups, including those defined by age, sex, SES, and education.

METHODS

Study setting and data sources

Data for this study are from two major linked data sources: the Agincourt HDSS (Kahn et al., 2012) and the HIV Prevention Trial Network (HPTN) 068 study (Pettifor et al., 2016a; Pettifor et al., 2016b). HPTN 068 is the source of the randomized cash transfer exposure for this study, while the Agincourt HDSS is the source for the adult mortality outcome. The Agincourt HDSS is a surveillance system, established in 1992, in the rural Mpumalanga province of northeastern South Africa, near the border with Mozambique. The Agincourt HDSS regularly collects vital statistics, health, and sociodemographic data on the full community cohort of ~117,000 people living within a geographically well-defined area covering 31 villages in the rural Agincourt sub-district. Vital events data (births, deaths, migrations) were collected during annual household visits until 2020. Since 2020, the data collection has expanded to include one household visit and two telephone surveys per year. In these visits, trained fieldworkers conduct interviews and update information for all members of the household. Other data collection modules (e.g., SES, employment, education) are collected through modules embedded within the household visit interview, but at less frequent intervals.

The Agincourt HDSS covers a rural, Black South African population in an area of the country to which Black South Africans were forcibly moved to during the Apartheid period (1948–1994) of legislated racial segregation (Christie & Gaganakis, 1989). The area is characterized by a history of poor-quality education and high unemployment. As with the broader South African population, residents in the Agincourt HDSS study have access to a broad social safety net. Two of the largest national cash transfer programs are designed to offset the costs of raising children to age 18 [Child Support Grant (Triegaardt, 2005)], and to support older adults age 60 and above [Old Age Pension (Case & Deaton, 1998)].

The HPTN 068 study was a Phase 3 randomized control trial designed to test the effect of a conditional cash transfer on HIV incidence in young South African women. The study was embedded within the Agincourt HDSS study site and used the Agincourt HDSS as a random sampling frame. In 2011, n=2533 young women from the Agincourt HDSS were enrolled in the trial. Eligibility was restricted to young women aged 13–20 who were enrolled in school (grade 8–11), and not married or pregnant at enrollment (Pettifor et al., 2016b). Participants were also required to have the documentation needed to open a bank or post office account, and co-reside with a parent/caregiver, also with documentation for a bank or post office account. The refusal rate was 17.3% (Pettifor et al., 2016a). Young women were randomized with their parent/caregiver using sequentially numbered sealed envelopes with block randomized assignments 1:1 to the cash transfer intervention or control arm. The cash transfer intervention was paid out monthly, conditional on the young woman attending 80% of school days in the past month. The monthly cash value of the transfer was 300 Rand (approximately ~$36USD at the time). The payment was split so that 1/3 went directly to the young woman, and 2/3 to the parent/caregiver. The amount of the cash transfer was chosen to be similar to the monthly payment value of the South African Child Support Grant (R280 in 2012) (Pettifor et al., 2016a). In 2011, the average monthly income in Mpumalanga Province was R6467 (Mpumalanga Provincial Treasury, 2015). The R300 transfer represented a 5% increase to this monthly average. Young women participants and their parent/caregivers who were assigned to the control arm did not receive payments.

HPTN 068 participants were enrolled in the trial and received their intervention or control conditions for up to three years. Participants were no longer eligible for participation after completing grade 12. Thus, participants enrolled in grades 8,9, or 10 at baseline could contribute a full three years (R10,800 possible), while participants enrolled in grade 11 could contribute up to two years (R7200 possible). The amount of the cash transfer was also determined by school attendance patterns, as monthly payment was conditional on 80% school attendance. However, most participants received their payments each month, with the average participant attending 95% of monthly school days. The study protocol was reviewed and approved by the University of North Carolina-Chapel Hill Institutional Review Board and the University of the Witwatersrand Human Research Ethics Committee (Medical).

Cohort construction

To understand the impact of the cash transfer intervention on the mortality patterns of older adults living in households that did and did not receive the cash transfer, we constructed a cohort using the Agincourt HDSS census platform of all members of the HPTN 068 households who were aged 40 years or older in 2011 at the time of trial enrollment. Members began contributing person-time at the date of the baseline HPTN 068 interview with the index young woman, which occurred between March 5th, 2011 and December 17th, 2012. Members accumulated person-time until they died (primary outcome), out-migrated from the study area (n=352), or otherwise were administrative censored on March 4, 2022. The maximum amount of time contributed by a participant was thus ~11 years. People were excluded from analysis if their end outcome in the Agincourt HDSS was coded as a death or migration, but no accompanying end date was available (n=4).

Key measures

The primary outcome we assessed was mortality, as recorded in the complete annual Agincourt HDSS census. Any death that occurred in the interval between two census rounds was recorded in the subsequent census with the date of death. Survival time was then calculated from the date of the baseline HPTN 068 interview to the date of death (or date of censoring in cases where a death did not occur). We assessed whether effects of the cash transfer program on mortality differed across short-, medium-, and longer-term time horizons by categorizing the timing of mortality post-enrollment date in categories of up to 3-, 5-, and 7-years, and across the full follow-up period.

The primary exposure was the randomly allocated HPTN 068 cash transfer arm. To create this variable, HPTN 068 trial data were merged with Agincourt HDSS census data and the HPTN 068 arm assignment was applied to all household members of enrolled households.

We examined the effect of the cash transfer intervention by two key differences in its distribution. First, we categorized older adults by primary caregiver status to track those most likely to be direct vs. indirect recipients of the cash transfer. The R200 household portion of the HPTN cash transfer was paid directly to primary caregivers of the index young woman. Other household members did not directly receive the intervention but would have received any benefit through household spillover. This primary caregiver variable is maintained in Agincourt HDSS records. However, these data are only available for household members who were related to the young woman as parents or grandparents. Members with other household relationships were coded as non-primary caregivers if a primary caregiver was otherwise identified. They were coded as missing if a primary caregiver was not otherwise identified. Second, we categorized the amount of time each household received the intervention into durations of <1, 1–2, and 2–3+ years (max 3.3 years). The duration of the HPTN 068 transfer was primarily determined based on the school grade in which the young women were enrolled at baseline. Trial participants had to be enrolled in school to contribute to the trial, so those who enrolled in later grades were eligible to remain in the trial for shorter follow-up periods. To create an appropriate comparison group in the control arm for these duration categories, we assigned control households to duration categories based on grade of the index young woman at baseline. Young women in Grade 8–9 were assigned to 2–3+ years, Grade 10 to 1–2 years, and Grade 11 to <1 year.

We used several covariates to contextualize the cohort and to identify the effectiveness of the cash transfer intervention across certain subgroups. Most of these covariates were collected in the HPTN 068 baseline survey: age (in years, at time of HPTN 068 enrollment), gender (man/woman), and household SES (measured with an asset index categorized in quartiles). Two of the covariates were collected in the Agincourt HDSS census: former Mozambican refugee status (yes/no) (Polzer, 2004) and highest educational attainment (none, some primary or secondary education, secondary education or above). Covariates were assessed at the time of baseline HPTN 068 enrollment, or from the Agincourt census year closest in time to the date of enrollment in the HPTN 068 study. The categorical educational attainment variable was derived from the number of years of education reported in the Agincourt HDSS in 2012. In the case of missing educational data for 2012, we used the most recent report of education between age 25 and the year 2012, under the assumption that educational attainment is relatively static after age 25. For observations with no educational attainment data in this range from which to draw (15% of our sample), we used multiple imputation to fill in the remaining missing observations. Twenty datasets were imputed using a Markov Chain Monte Carlo algorithm with multiple chains. Missing data on educational attainment were imputed using gender, former refugee status, age at baseline, household SES (quartile), and educational attainment from years 1999 to 2019.

Statistical analysis

To evaluate the association between the cash transfer intervention and mortality, we specified Cox proportional hazards models. We conducted an intention-to-treat analysis. Because the cash transfer intervention was almost universally paid out to intervention arm participants, we expect this estimate to approximate a per-protocol analysis (Pettifor et al., 2016b). Time origin was the HPTN 068 baseline interview, with follow-up until the date of death, outmigration, or March 4, 2022. We calculated unadjusted hazard ratios for mortality outcomes after the full follow-up time, and at 3-, 5-, and 7- years. Adjustment for covariates was not necessary as socio-demographic variables were balanced through randomization. Stratified analyses comparing treatment to control arms were conducted by the key covariates defined above. Intervention duration was not incorporated into these stratified models. The subgroup analysis by education used imputed datasets, and Cox proportional hazard regression was conducted on and then pooled across all 20 imputed datasets. To account for clustering at the household level, we constructed Cox mixed effects models, specifying household as a random effect.

Role of the funding source

The study sponsors had no role in the study design, data analysis, interpretation of results, manuscript drafting, or in the decision to submit the manuscript for publication.

RESULTS

Overall, we identified n=3568 (n=1785 intervention vs. n=1783 control) adult members of households with HPTN 068 enrollment contributing person-time to this analysis, living across 2116 households (Table 1, Figure 2). The duration of follow-up time ranged from 1 to 132 months (median 126 months). This cohort was majority women (60%), and over a third (35%) identified as a former refugee from Mozambique, reflective of the underlying Agincourt population of this age range. The mean age of cohort members was 54 years (range 40–108 years), with nearly half (43%) between the ages of 40–50 years. Consistent with the limited educational opportunities experienced under Apartheid, the cohort reported low levels of educational attainment, with about a quarter (28%) reporting no formal education. After using multiple imputation to impute the missing education data (15%), the mean (SE) years of education was 5.65 (0.08), compared to 5.98 (0.08) pre-imputation. As expected, due to the randomized design, there were no statistically significant differences between arms in the socio-demographic covariates we assessed (Table 1).

Table 1.

Sociodemographic characteristics of n=3568 adult household members of HPTN 068 trial participants, by cash transfer intervention arm

Total (n = 3568) Cash Transfer (n = 1785) No Cash Transfer (n = 1783)
Mean (SD) Mean (SD) Mean (SD) P-value 1
Household Asset Score 15.4 (7.3) 15.5 (7.4) 15.3 (7.1) 0.9
Missing N=7 N=0 N=7 0.4
Age (years) 54.4 (12.3) 54.3 (12.3) 54.6 (12.2)
Freq. (%) Freq. (%) Freq. (%) P-value 2
Gender 0.8
Male 1417 (40) 713 (40) 704 (39)
Female 2151 (60) 1072 (60) 1079 (61)
Former Mozambican refugee 0.4
Yes 1242 (35) 634 (36) 608 (34)
No 2324 (65) 1149 (64) 1175 (66)
Missing N=2 N=2 N=0
Age at HPTN enrollment 0.3
< 50 years old 1539 (43) 747 (42) 792 (44)
50 to 79 years old 1522 (43) 776 (43) 746 (42)
≥ 80 years old 507 (14) 262 (15) 245 (14)
Marital Status 0.9
Divorced/Separated 227 (7) 115 (7) 112 (7)
Married/Cohabiting 2117 (63) 1059 (64) 1058 (63)
Single 453 (14) 217 (13) 236 (14)
Widowed 539 (16) 269 (16) 270 (16)
Missing N=232 N=125 N=107
Household asset index 0.3
Quartile 1 690 (19) 351 (20) 339 (19)
Quartile 2 924 (26) 458 (26) 466 (26)
Quartile 3 947 (27) 494 (28) 453 (26)
Quartile 4 1000 (28) 482 (27) 518 (29)
Missing N=7 N=0 N=7
Years of education, pre-imputation 0.07
No education 849 (28) 444 (29) 405 (27)
< 12 years 1733 (57) 867 (57) 866 (57)
≥ 12 years 436 (14) 198 (13) 238 (16)
Missing N=550 N=276 N=277
Years of education, post-imputation
No education 884 (25) 461 (26) 423 (24)
< 12 years 2123 (60) 1058 (59) 1065 (60)
≥ 12 years 561 (16) 261 (15) 300 (17)
Missing N=0 N=0 N=0
Relationship to young woman 0.3
Father 947 (27) 472 (26) 475 (27)
Mother 1149 (32) 551 (31) 598 (34)
Grandfather 140 (4) 79 (4) 61 (3)
Grandmother 472 (13) 244 (14) 228 (13)
Other 860 (24) 439 (25) 421 (24)
Primary caregiver 3 0.9
Yes 1376 (49) 665 (49) 711 (48)
No 1385 (51) 708 (51) 677 (52)
Missing N=807 N=412 N=395
Cumulative deaths
By 3-years 177 (5) 84 (5) 93 (5) 0.5
By 5-years 303 (9) 143 (8) 160 (9) 0.3
By 7-years 423 (12) 197 (11) 226 (13) 0.1
By Study End 640 (18) 313 (17) 327 (18) 0.6
Cash Transfer Duration -
<1 year - 380 (23) -
1–2 years - 664 (39) -
2+ years - 641 (38)
Missing - N=100 -
Open in a new tab
1

Wilcoxon rank sum test

2

Pearson’s Chi-squared test

3

Primary caregiver status assessed for participants who were parents or grandparents of the HPTN 068 young women only.

Participants with other relationships were coded as missing if a primary caregiver was not otherwise noted for the young woman.

Figure 2.

Figure 2.

Open in a new tab

Flow chart illustrating the steps to arrive at the final analytic cohort n=3568

There was some variation in the duration and way the cohort members were exposed to the cash transfer. Most individuals in the cohort were exposed to the intervention for 1–2 years (39%) or 2–3 years (38%). Fewer were exposed to the cash transfer for <1 years (23%). About half of the sample in both arms (49%) were the primary caregiver of the index young woman.

There were 640 deaths observed over the follow-up period. About a quarter of deaths (28%) were observed in the first 3 years of follow-up, nearly half (47%) in the first 5 years of follow-up, and two-thirds (66%) in the first 7 years of follow-up. The cash transfer intervention did not have a significant effect on mortality outcomes in the full sample over the full follow-up period (Table 2). The hazard of mortality was slightly lower in the intervention arm compared to the control arm, but the effect size was small with confidence intervals spanning the null [HR (95% CI): 0.94 (0.80, 1.10)]. Similar effect sizes were observed for each of the follow-up time subdivisions, with the strongest effect estimate observed for mortality at 7 years [HR (95% CI): 0.86 (0.71, 1.05)]. Similar effect sizes were also observed across the different durations of cash transfer exposures, with the strongest effect estimate observed for durations of 1–2 years [HR (95% CI): 0.71 (0.50, 1.02)]. There were also no differences in effects observed by primary caregiver status.

Table 2.

Impact of the cash transfer intervention on mortality outcome in full follow-up of full sample, and in subgroups by 3-, 5-, and 7-years of follow-up and by duration of cash transfer intervention

Intervention arm MR per 1000 PM Control arm MR per 1000 PM HR (95% CI)
Full sample 1.62 1.73 0.94 (0.80, 1.10)
Mortality timing
3 years 1.24 1.37 0.90 (0.67, 1.21)
5 years 1.36 1.52 0.89 (0.71, 1.12)
7 years 1.40 1.62 0.86 (0.71, 1.05)
Primary caregiver 1,2
Yes 0.96 1.05 0.91 (0.62, 1.35)
No 1.67 1.70 0.98 (0.74, 1.31)
Duration of cash transfer 2
<1 year 1.54 1.61 0.96 (0.64, 1.44)
1–2 years 1.13 1.58 0.71 (0.50, 1.02)
2+ years 1.61 1.65 0.97 (0.73, 1.23)
Open in a new tab
1

Primary caregivers were most likely the direct recipients of the household portion of the cash transfer intervention. Non-primary caregivers likely received exposure to the cash transfer through indirect spillover mechanisms.

2

At 7 years of follow-up

PM = person-month HR = hazard ratio

The cash transfer intervention had strong protective effects in some of the socio-demographic subgroups we assessed (Table 3). There were no significant differences in effect sizes by gender, marital status or age category, although the point estimates across nearly all these categories trended on the protective side of the null. The exception to this trend was the ‘divorced/separated’ marital status category with a point estimate of 1.45 (95% CI: 0.64, 3.40). For household assets, we found strongly protective effects of the cash transfer intervention on mortality in those with above median assets [HR (95% CI): 0.60 (0.50, 0.86)], but not those with below median assets [HR (95% CI): 1.15 (0.87, 1.52)]. The protective effects were of similar magnitude across the highest two asset quartiles [HRQ3 (95% CI): 0.63 (0.43, 0.93); HRQ4 (95% CI): 0.68 (0.46, 1.01)]. Similarly, we found strongly protective effects for those with the highest educational attainment [HR (95% CI): 0.37 (0.15, 0.93)], but not for those with lower levels of educational attainment.

Table 3.

Impact of the cash transfer intervention1 on 7-year mortality outcomes, in subgroups by gender, age, household assets, and education

Intervention arm MR per 1000 PM Control arm MR per 1000 PM HR (95% CI)
Gender
Male 1.54 1.91 0.81 (0.60, 1.08)
Female 1.30 1.43 0.91 (0.70, 1.18)
Age
40–49 years 0.64 0.82 0.78 (0.50, 1.19)
50–79 years 1.18 1.43 0.82 (0.60, 1.12)
80+ years 4.67 5.24 0.89 (0.65, 1.22)
Marital Status
Divorced/Separated 1.54 1.03 1.45 (0.64, 3.40)
Married/Cohabiting 1.23 1.38 0.89 (0.68, 1.16)
Single 1.83 2.25 0.81 (0.50, 1.32)
Widowed 2.18 2.52 0.86 (0.58, 1.27)
Household asset quartile
Q1 1.72 1.84 0.93 (0.62, 1.39)
Q2 1.69 1.23 1.37 (0.94, 2.02)
Q3 1.10 1.72 0.63 (0.43, 0.94)
Q4 1.19 1.74 0.68 (0.46, 1.01)
Below median 1.71 1.48 1.15 (0.87, 1.52)
Above median 1.14 1.73 0.66 (0.50, 0.86)
Education (years)
No education 2.26 2.56 0.81 (0.59, 1.12)
< 12 years 1.23 1.26 0.93 (0.71, 1.22)
≥ 12 years 0.47 1.21 0.37 (0.15, 0.93)
Open in a new tab

PM = person-month HR = hazard ratio

1

Intervention vs. control arm assignment

DISCUSSION

We found new evidence from an extension to a rigorous, randomized trial that a cash transfer intervention led to reduced mortality in certain subgroups of a cohort of older adults in a rural South African setting. Older adults in households that received the intervention tended to live longer than those in control households, though the protective effect in the overall cohort was small in magnitude with confidence intervals overlapping the null. However, we observed strong protective effects in individuals from households with above median SES, and in individuals with higher educational attainment. It is important to contextualize these subgroups in the larger and limited socio-economic environment of the study site. The Agincourt HDSS site covers an area with uniformly low economic opportunities (Collinson, 2010; Kabudula et al., 2017b; Sartorius et al., 2013; Wilkinson et al., 2017), and it covers an older adult population who were limited from accessing educational opportunities during Apartheid (Gómez-Olivé et al., 2018). Thus, even the upper ends of the SES and educational spectrum are living in a low-income and limited educational environment. That we observe strong protective effects on mortality in these subgroups with a modestly sized, 2–3 year cash transfer intervention is promising evidence of the utility of cash transfers to promote healthy aging outcomes.

Our findings generally align with the benefits of cash transfers on adult mortality found across Latin America (Barham & Rowberry, 2013; Pescarini et al., 2020). We extend the generalizability of this prior work to sub-Saharan Africa with our rural South African setting. Our findings are consistent with pathways from cash transfers to delayed mortality that run through some of the established benefits of cash transfers: reductions in high blood pressure prevalence (Behrman & Parker, 2013; Fernald et al., 2008), increases in vaccinations (Salinas-Rodríguez & Manrique-Espinoza, 2013), and increased healthcare utilization in older adults (Riumallo-Herl & Aguila, 2019). Our findings are in contrast to the single study we identified that found increased mortality from a pension cash transfer program in Mexico (Feeney, 2017). These differences may be explained by underlying differences in the South African versus Mexican study populations, and by differences in the income investment decisions that could arise due to age differences in the target beneficiaries, and gender profile of the direct recipients.

Importantly, the strong protective effects we observed of the cash transfer intervention on mortality were restricted to subgroups who generally exhibited higher indicators of SES. There are several plausible explanations for why we did not observe a more robust effect for all household members. First, the cumulative value of the grant could have been insufficient to benefit the poorest and least educated households. Anchored at ~5% of the average monthly income in the area, the cash transfer may have been too small to lift the poorest households out of poverty and to a place where the extra income could be allocated towards longer-term investment in health and well-being. Second, the most educated members of the cohort may have had higher levels of financial and/or health literacy. Their educational backgrounds could plausibly provide them with the knowledge base to effectively use the additional income to maximize health and well-being benefits. Finally, the direct recipients of the household portion of the grant were primarily women caregivers, often mothers or grandmothers of the trial participants. Women tend to invest household money in children and other household members instead of directly in themselves (Burns et al., 2005; Duflo, 2003). Thus, it is possible that we do not observe strong effects in the direct recipients because they were more likely to be women who invested the money outside of themselves. At the same time, it is possible that the spillover effects from this investment were too diluted to significantly benefit other older household members. Spillover effects are likely stronger in households with fewer members. If higher SES households trended towards lower household size, then this would provide another potential explanation for our findings.

The broad social protection landscape in South Africa is important context to consider in interpreting our results. Households in the study area had access to a family of government cash transfer programs, with the Child Support Grant (Triegaardt, 2005) and the Old Age Pension (Case & Deaton, 1998) programs having the largest reach. Both of these programs have age eligibility requirements for beneficiaries and are means tested. However, in the low-income area of our study population, most households with age-eligible members would qualify under the means test. Most recent estimates from Mpumalanga province show that 54% of households are receiving income from governmental cash transfers (Statistics South Africa, 2020), and 74% of households with age-eligible children are receiving the Child Support Grant (Statistics South Africa, 2018). In this context, it is possible that the marginal effect of the cash transfer examined in our study may not have been sufficiently large to have beneficial impacts above the governmental cash transfer income received. Future work should focus on isolating the independent and synergistic effects of cash transfer income from multiple sources.

Elements of our study design support robust causal inference about the effect of a cash transfer intervention on adult mortality over a long-term time horizon. We leveraged a randomly allocated cash transfer experiment and were thus able to address the concerns about endogeneity of cash transfer exposures that are common in observational evaluations. We were able to link this cash transfer experiment to high-quality and complete longitudinal mortality data of members of households involved in the trial because of the Agincourt HDSS platform that underpinned both the trial and an annual census on all community members in the study site. The underlying mortality data allowed us to assess longer-term follow-up (>10 years) than is usually possible in cash transfer evaluations. This unique linkage of high-quality datasets is a major strength of our work.

Though the randomized exposure is a major strength, variation in the ‘dose’ of cash transfer exposure received by household members influences the interpretation of our results. The older adult members of our study cohort were not the target beneficiaries of the cash transfer, so exposure would either occur through direct receipt of the cash transfer on behalf of the young woman beneficiary, or through indirect spillover mechanisms. We observed no difference in effect by primary caregiver status. However, variations in household composition and income sharing behaviors indicate that some members of our cohort will have received more direct income than others in ways we were not able to capture.

Additionally, some aspects of our study design and data structure limit the generalizability of our results. First, outmigration from the study site was a source of loss-to-follow-up as mortality outcomes were not monitored after migration. Although overall out-migration in this study was low (<10%), if migrant mortality patterns differed appreciably from those who remained in ways related to the cash transfer exposure, our results would be biased. Second, our study was conducted among older adults in a low-income, rural community in northeastern South Africa. Future work should extend these findings to other geographic settings.

In sum, we found promising evidence that a short-term cash transfer exposure can benefit people living in households that receive it by delaying mortality. However, the protective effects were restricted to members of our cohort who were already in the upper distributions of socio-economic status and education. Our findings should prompt researchers and policymakers to investigate whether cash transfer program values need to be scaled to larger amounts for lower-SES groups. Future work should focus on understanding the optimal amount, timing, structure, and targets to maximize the benefits of cash transfer programs in promoting healthy aging and longevity.

Data sharing statement:

The data analyzed in this paper are available upon request because they were obtained from a third party and contain sensitive health and personal information. The HPTN 068 data access is managed by FHI360. Data requests can be made by contacting Erica Hamilton at EHamilton@fhi360.org.The Agincourt Health and Demographic Surveillance System data are available upon request as described on their data overview website (http://www.agincourt.co.za/index.php/data/) and by contacting their data manager (DataManager@agincourt.co.za).

HIGHLIGHTS.

  • High-income countries have higher life expectancy than lower-income countries

  • Cash transfers could plausibly prolong healthy lives in low-income settings

  • We evaluate a randomized cash transfer exposure on mortality in rural South Africa

  • Cash transfers delayed death for those in higher socioeconomic status (SES) groups

  • Cash transfers may need to be scaled to larger amounts for lower-SES groups

Funding:

This work was supported by the National Institute on Aging of the National Institutes of Health (grant number R01AG069128). The Agincourt Health and socio-Demographic Surveillance System is supported by the National Department of Science and Innovation, the Medical Research Council and University of the Witwatersrand, South Africa, and previously the Wellcome Trust, UK (grants 058893/Z/99/A; 069683/Z/02/Z; 085477/Z/08/Z; 085477/B/08/Z). The HPTN 068 was supported by the National Institutes of Health (grant numbers R01MH087118 and R24 HD050924) and overall support to the HPTN through National Institute of Allergy and Infectious Diseases (NIAID), the National Institute of Mental Health (NIMH), and the National Institute on Drug Abuse (NIDA) of the National Institutes of Health (NIH; award numbers UM1AI068619 [HPTN Leadership and Operations Center], UM1AI068617 [HPTN Statistical and Data Management Center], and UM1AI068613 [HPTN Laboratory Center]. The study sponsors had no role in the study design, data analysis, interpretation of results, manuscript drafting, or in the decision to submit the manuscript for publication.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Trial registry and protocol: The HPTN 068 parent trial is registered at clinicaltrials.gov (NCT01233531). The study protocol is archived on the HPTN website: https://www.hptn.org/research/studies/hptn068.

Authorship statement:

Molly Rosenberg: Conceptualization, funding acquisition, investigation, methodology, data curation, project administration, resources, writing-original draft, writing-review and editing, approval of final manuscript as submitted

Erika Beidelman: Formal analysis, data curation, writing-review and editing, approval of final manuscript as submitted

Xiwei Chen: Formal analysis, writing-review and editing, approval of final manuscript as submitted

David Canning: Funding acquisition, interpretation of study results, approval of final manuscript as submitted

Lindsay Kobayashi: Conceptualization, funding acquisition, project administration, writing-review and editing, approval of final manuscript as submitted

Kathleen Kahn: Funding acquisition, interpretation of study results, writing-review and editing, approval of final manuscript as submitted

Audrey Pettifor: Funding acquisition, data curation, interpretation of study results, writing-review and editing, approval of final manuscript as submitted

Chodziwadziwa Kabudula: Funding acquisition, data curation, interpretation of study results, writing-review and editing, approval of final manuscript as submitted

Declaration of interests:

The authors declare no conflicts of interest.

REFERENCES

  1. Barham T, & Rowberry J (2013). Living longer: The effect of the Mexican conditional cash transfer program on elderly mortality. Journal of Development Economics, 105, 226–236. [Google Scholar]
  2. Behrman JR, & Parker SW (2013). Is health of the aging improved by conditional cash transfer programs? Evidence from Mexico. Demography, 50, 1363–1386. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bosma H, Dike Van De Mheen H, Borsboom GJ, & Mackenbach JP (2001). Neighborhood socioeconomic status and all-cause mortality. Am J Epidemiol, 153, 363–371. [DOI] [PubMed] [Google Scholar]
  4. Burns J, Keswell M, & Leibbrandt M (2005). Social assistance, gender, and the aged in South Africa. Feminist economics, 11, 103–115. [Google Scholar]
  5. Case A, & Deaton A (1998). Large cash transfers to the elderly in South Africa. The Economic Journal, 108, 1330–1361. [Google Scholar]
  6. Chetty R, Stepner M, Abraham S, Lin S, Scuderi B, Turner N, et al. (2016). The association between income and life expectancy in the United States, 2001–2014. JAMA, 315, 1750–1766. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Christie P, & Gaganakis M (1989). Farm schools in South Africa: The face of rural apartheid. Comparative education review, 33, 77–92. [Google Scholar]
  8. Collinson MA (2010). Striving against adversity: the dynamics of migration, health and poverty in rural South Africa. Glob Health Action, 3, 5080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cooper JE, Benmarhnia T, Koski A, & King NB (2020). Cash transfer programs have differential effects on health: a review of the literature from low and middle-income countries. Social science & medicine, 247, 112806. [DOI] [PubMed] [Google Scholar]
  10. Duflo E (2003). Grandmothers and granddaughters: old-age pensions and intrahousehold allocation in South Africa. The World Bank Economic Review, 17, 1–25. [Google Scholar]
  11. Evans DK, & Popova A (2014). Cash transfers and temptation goods : a review of global evidence. The World Bank. [Google Scholar]
  12. Feeney K (2017). Cash Transfers and Adult Mortality: Evidence from Pension Policies [Unpublished doctoral dissertation]. University of California, Berkeley. [Google Scholar]
  13. Fernald LC, Gertler PJ, & Hou X (2008). Cash component of conditional cash transfer program is associated with higher body mass index and blood pressure in adults. J Nutr, 138, 2250–2257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Forde I, Chandola T, Garcia S, Marmot MG, & Attanasio O (2012a). The impact of cash transfers to poor women in Colombia on BMI and obesity: prospective cohort study. International Journal of Obesity, 36, 1209–1214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Forde I, Rasanathan K, & Krech R (2012b). Cash transfer schemes and the health sector: making the case for greater involvement. pp. 551–553): SciELO Public Health. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Freeman T, Gesesew HA, Bambra C, Giugliani ERJ, Popay J, Sanders D, et al. (2020). Why do some countries do better or worse in life expectancy relative to income? An analysis of Brazil, Ethiopia, and the United States of America. International journal for equity in health, 19, 1–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Gómez-Olivé FX, Montana L, Wagner RG, Kabudula CW, Rohr JK, Kahn K, et al. (2018). Cohort profile: health and ageing in Africa: a longitudinal study of an indepth community in South Africa (HAALSI). Int J Epidemiol, 47, 689–690j. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Haushofer J, & Shapiro J (2013). Household response to income changes: Evidence from an unconditional cash transfer program in Kenya. Massachusetts Institute of Technology, 24, 1–57. [Google Scholar]
  19. Jetter M, Laudage S, & Stadelmann D (2019). The intimate link between income levels and life expectancy: global evidence from 213 years. Social Science Quarterly, 100, 1387–1403. [Google Scholar]
  20. Kabudula CW, Houle B, Collinson MA, Kahn K, Gómez-Olivé FX, Tollman S, et al. (2017a). Socioeconomic differences in mortality in the antiretroviral therapy era in Agincourt, rural South Africa, 2001–13: a population surveillance analysis. The Lancet Global Health, 5, e924–e935. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kabudula CW, Houle B, Collinson MA, Kahn K, Tollman S, & Clark S (2017b). Assessing changes in household socioeconomic status in rural South Africa, 2001–2013: a distributional analysis using household asset indicators. Social Indicators Research, 133, 1047–1073. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kahn K, Collinson MA, Gomez-Olive FX, Mokoena O, Twine R, Mee P, et al. (2012). Profile: Agincourt Health and Socio-demographic Surveillance System. Int J Epidemiol, 41, 988–1001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kinge JM, Modalsli JH, Øverland S, Gjessing HK, Tollånes MC, Knudsen AK, et al. (2019). Association of household income with life expectancy and cause-specific mortality in Norway, 2005–2015. JAMA, 321, 1916–1925. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Leroy JL, Gadsden P, González de Cossío T, & Gertler P (2013). Cash and in-kind transfers lead to excess weight gain in a population of women with a high prevalence of overweight in rural Mexico. The Journal of nutrition, 143, 378–383. [DOI] [PubMed] [Google Scholar]
  25. Mpumalanga Provincial Treasury. (2015). Socio-economic review and outlook of Mpumalanga.
  26. Okeke EN, & Abubakar IS (2020). Healthcare at the beginning of life and child survival: evidence from a cash transfer experiment in Nigeria. Journal of Development Economics, 143, 102426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Pescarini JM, Craig P, Allik M, Amorim L, Ali S, Smeeth L, et al. (2020). Evaluating the impact of the Bolsa Familia conditional cash transfer program on premature cardiovascular and all-cause mortality using the 100 million Brazilian cohort: a natural experiment study protocol. BMJ Open, 10, e039658. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Pettifor A, MacPhail C, Hughes JP, Selin A, Wang J, Gómez-Olivé FX, et al. (2016a). The effect of a conditional cash transfer on HIV incidence in young women in rural South Africa (HPTN 068): a phase 3, randomised controlled trial. The Lancet Global Health, 4, e978–e988. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Pettifor A, MacPhail C, Selin A, Gómez-Olivé FX, Rosenberg M, Wagner RG, et al. (2016b). HPTN 068: A Randomized Control Trial of a Conditional Cash Transfer to Reduce HIV Infection in Young Women in South Africa—Study Design and Baseline Results. AIDS Behav, 1–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Polzer T (2004). We are all South Africans now: The integration of Mozambican refugees in rural South Africa.
  31. Preston SH (1975). The changing relation between mortality and level of economic development. Population studies, 29, 231–248. [PubMed] [Google Scholar]
  32. Ramos D, da Silva NB, Ichihara MY, Fiaccone RL, Almeida D, Sena S, et al. (2021). Conditional cash transfer program and child mortality: A cross-sectional analysis nested within the 100 Million Brazilian Cohort. PLoS medicine, 18, e1003509. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ranganathan M, & Lagarde M (2012). Promoting healthy behaviours and improving health outcomes in low and middle income countries: a review of the impact of conditional cash transfer programmes. Prev Med, 55, S95–S105. [DOI] [PubMed] [Google Scholar]
  34. Rasella D, Alves FJO, Rebouças P, de Jesus GS, Barreto ML, Campello T, et al. (2021). Long-term impact of a conditional cash transfer programme on maternal mortality: a nationwide analysis of Brazilian longitudinal data. BMC medicine, 19, 1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Riumallo-Herl C, & Aguila E (2019). The effect of old-age pensions on health care utilization patterns and insurance uptake in Mexico. BMJ Global Health, 4, e001771. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Salinas-Rodríguez A, & Manrique-Espinoza BS (2013). Effect of the conditional cash transfer program Oportunidades on vaccination coverage in older Mexican people. BMC Int Health Hum Rights, 13, 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Sartorius K, Sartorius B, Tollman S, Schatz E, Kirsten J, & Collinson M (2013). Rural poverty dynamics and refugee communities in South Africa: A spatial–temporal model. Population, Space and Place, 19, 103–123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Statistics South Africa. (2018). Children’s education and well-being in South Africa, 2018: Education series, volume VII. Education Series.
  39. Statistics South Africa. (2020). General Household Survey 2019.
  40. Steenland K, Henley J, Calle E, & Thun M (2004a). Individual-and area-level socioeconomic status variables as predictors of mortality in a cohort of 179,383 persons. Am J Epidemiol, 159, 1047–1056. [DOI] [PubMed] [Google Scholar]
  41. Steenland K, Hu S, & Walker J (2004b). All-cause and cause-specific mortality by socioeconomic status among employed persons in 27 US states, 1984–1997. American journal of public health, 94, 1037–1042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Sun S, Huang J, Hudson DL, & Sherraden M (2021). Cash transfers and health. Annual review of public health, 42, 363–380. [DOI] [PubMed] [Google Scholar]
  43. Triegaardt JD (2005). The Child Support Grant in South Africa: a social policy for poverty alleviation? International Journal of Social Welfare, 14, 249–255. [Google Scholar]
  44. United Nations Department of Economic and Social Affairs Population Division. (2019). World Population Prospects 2019, Volume I: Comprehensive Tables (ST/ESA/SER.A/426).
  45. Wilkinson A, Pettifor A, Rosenberg M, Halpern C, Thirumurthy H, Collinson MA, et al. (2017). The employment environment for youth in rural South Africa: A mixed-methods study. Dev South Afr, 34, 17–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. World Health Organization. (2020). Life expectancy at birth (years).

RESOURCES