COVID-19, Vulnerability, and Health in South Africa: Evidence on the Changing Role of Social Cash Transfers

Abstract

The COVID-19 pandemic generated severe health and livelihood shocks in South Africa, yet much of the existing evidence has focused on mental health rather than broader measures of overall wellbeing. This paper examines changes in self-reported health before and during the pandemic and assesses whether social cash transfers mitigated the adverse health effects associated with vulnerability. Using linked data from the National Income Dynamics Study and the NIDS-Coronavirus Rapid Mobile Survey, we compare pre-pandemic and pandemic-period health outcomes and estimate probit models of poor self-reported health. The results show that the prevalence of poor self-reported health more than doubled during the pandemic and remained elevated across survey waves. Household hunger, chronic illness and income loss were associated with a higher probability of reporting poor health, with the effects of hunger and chronic illness intensifying over time. Household receipt of social grants was associated with a lower probability of poor health at the onset of the pandemic, but this protective association disappeared in later waves. These findings suggest that while cash transfers were associated with short-term protection, cash support alone may be insufficient to address the broader and cumulative health impacts of a prolonged crisis. Future pandemic and crisis responses should therefore combine income support with nutrition interventions, continuity of primary healthcare, chronic medication access and targeted support for vulnerable groups.

1. Introduction

The COVID-19 pandemic and the associated lockdowns instituted to reduce transmission generated a multidimensional shock to health, livelihoods and social wellbeing in South Africa. The restrictions in economic activity resulted in the highest contraction in GDP since the 1960s, with quarterly GDP declining by a staggering 16% between the first and second quarters of 2020 (Statistics South Africa, 2020a). The economic contraction was devastating, with an estimated 2.8 million South Africans losing employment in just a few months of the pandemic (Posel et al., 2021). These labour-market and income shocks were unevenly distributed and were especially likely to affect individuals and households already exposed to poverty, food insecurity, chronic illness and insecure employment (Omotayo & Ogunniyi, 2024).

The coronavirus pandemic also severely impacted health, with over 4 million South Africans having contracted COVID-19 and an estimated 100,000 having died as of December 2022 (Republic of South Africa, 2022). Beyond the deaths and infections, health was affected in other ways. First, the pandemic negatively impacted mental health, arising from factors such as job losses, reductions in household income, and social isolation occasioned by the lockdowns and travel restrictions. Several studies on South Africa report an increase in depressive symptoms because of job losses (Posel et al., 2021), increased burden of childcare (Nwosu, 2021), and increased food insecurity and hunger (Hunt et al., 2021). However, the health consequences of the pandemic were not confined to mental health. They also included disruptions to chronic disease management, reduced use of routine healthcare services, nutritional stress, reduced mobility and increased exposure to social and economic insecurity (Burger et al., 2020; Mboweni & Risenga, 2022).

Second, due to lockdown restrictions and fear of contracting the virus, many non-Covid patients with chronic or other conditions experienced difficulties visiting health facilities and accessing medication. Cullinan (2020) estimates that public health clinic attendance decreased by between 30% and 70% during the initial “hard” lockdown, with Burger et al. (2020) reporting evidence of many HIV patients failing to collect their medicines during the early stages of the pandemic. Estimates from Statistics South Africa indicate that among chronic patients who reported difficulties accessing medication, 55% attributed this to fear of contracting COVID-19, 26% cited worries about arrests and fines, and the remainder cited transport constraints (Statistics South Africa, 2020b). Recent review evidence confirms that the pandemic disrupted prevention, diagnosis, routine care and self-care for chronic conditions in South Africa and across Africa (Mboweni & Risenga, 2022).

Another pathway through which COVID-19 affected health was through reduced physical activity due to restrictions on movement and fear of infection. Reduced physical activity may have adversely affected physical health and exacerbated already high levels of non-communicable diseases in South Africa.

As highlighted above, the COVID-19 pandemic likely affected health in multiple ways. However, much of the literature on COVID-19 and health has focused primarily on psychological outcomes, with relatively less attention given to overall health status. This leaves an important gap in the South African literature: while existing studies provide strong evidence on employment, poverty, hunger and mental health effects, less is known about how overall self-reported health evolved during the pandemic, which vulnerability factors were most strongly associated with poor health, and whether social protection measures mitigated these effects. In this paper, we adopt the broad definition of health provided by the World Health Organization (WHO), which defines health as a state of complete physical, mental and social well-being (WHO, 2021). Given that all dimensions of health may have been affected during the pandemic, it is important to assess overall population health.

Health status is conventionally measured through clinical examination, but this is rarely feasible in large-scale surveys, particularly during a pandemic. In such contexts, self-reported health (SRH) is widely used as a proxy for overall health. SRH typically involves a single question asking respondents to rate their health on a five-point scale ranging from “very poor” to “excellent.” Epidemiological studies have shown that SRH is a strong predictor of underlying health status, healthcare utilisation and mortality (Miilunpalo et al., 1997; Lorem et al., 2020). It has also been shown to capture nutritional and environmental health risks (Gallagher et al., 2016), physical health difficulties (Yaya et al., 2020), and long-term mental health outcomes (Ishida et al., 2020). During periods of widespread health shocks, such as the COVID-19 pandemic, SRH may be particularly useful in capturing how individuals reassess their overall health status (van de Weijer et al., 2022).

Despite its strengths, SRH is not without limitations. Responses may be sensitive to question framing, and individuals may interpret response categories differently (Crossley & Kennedy, 2002). Nonetheless, when measured consistently across time and individuals, SRH provides a useful and comparable indicator of health dynamics. In this study, we rely on the consistency of SRH measurement across waves of the National Income Dynamics Study (NIDS) and the NIDS-Coronavirus Rapid Mobile Survey (NIDS-CRAM).

This paper examines changes in overall health, measured by SRH, before and during the COVID-19 pandemic in South Africa. It uses the fifth wave of the core NIDS survey to characterise pre-pandemic health and waves 1, 4 and 5 of NIDS-CRAM to examine health outcomes during the pandemic. The analysis pays particular attention to the possible aggravating role of vulnerability factors and the mitigating role of social cash transfers.

The paper is guided by three research questions. First, how did self-reported health in South Africa change during the COVID-19 pandemic relative to the pre-pandemic period? Second, which vulnerability factors were associated with poor self-reported health during the pandemic? Third, did household receipt of social grants mitigate the adverse health effects associated with hunger, chronic illness and income loss?

The literature on COVID-19 has identified a wide range of vulnerability factors, including demographic, socio-economic and epidemiological characteristics (Yu, 2023; Schotte & Zizzamia, 2021). In South Africa, these vulnerabilities are particularly pronounced due to high levels of inequality, widespread labour-market precarity, a high burden of chronic illness and uneven access to healthcare. Individuals with chronic conditions, those living in poverty, those in informal employment and those experiencing food insecurity are likely to be at elevated risk of poor health outcomes during pandemics.

The positive relationship between income and health is well established (Pritchett & Summers, 1996). Stable access to income enables households to meet basic needs, maintain adequate nutrition and access healthcare. During the pandemic, widespread income losses were likely to affect health through increased stress, food insecurity and reduced ability to access care. Government interventions in the form of income support would therefore be expected to mitigate these effects. Evidence from South Africa suggests that social grants helped to reduce vulnerability during the pandemic, although the extent and persistence of their protective effect on health remains an open empirical question (Omotayo & Ogunniyi, 2024).

This paper makes three contributions. First, it extends the South African COVID-19 literature beyond labour market outcomes, poverty and mental health by focusing on overall self-reported health. Second, it links pre-pandemic NIDS data with NIDS-CRAM data to examine changes in health status over time. Third, it contributes to policy debates on social protection by assessing whether social cash transfers were associated with better health outcomes during a large-scale crisis.

The findings have direct implications for society and economic policy. If poor health increased substantially during the pandemic, and if vulnerability factors such as hunger and chronic illness intensified this risk, then policy responses cannot rely on cash transfers alone. While social grants may protect households from immediate income shocks, they may need to be complemented by nutrition support, continuity of primary healthcare, access to chronic medication and targeted support for vulnerable groups. The paper therefore speaks to the design of integrated crisis-response systems that combine social protection with public health interventions.

The rest of the paper is organised as follows. Section 2 reviews the literature on COVID-19, health, vulnerability and social protection. Section 3 discusses South Africa’s social protection response to the pandemic. Section 4 describes the data and methods. Section 5 presents the results. Section 6 discusses the findings and their policy implications. Section 7 concludes.

2. Literature Review

2.1. COVID-19 and Health Outcomes

The COVID-19 pandemic has generated a rapidly expanding body of literature examining its health impacts across both developed and developing countries. A substantial share of this literature has focused on mental health outcomes, documenting increases in anxiety, depression and psychological distress associated with lockdowns, income losses and social isolation (Posel et al., 2021; Nwosu, 2021; Hunt et al., 2021).

Recent international evidence suggests that the health consequences of the COVID-19 pandemic were particularly severe in countries characterised by high levels of inequality and socioeconomic vulnerability. In Brazil, nearly one-third of the adults reported worsening self-rated health during the early stages of the pandemic, with deterioration strongly associated with pre-existing health conditions and mental health problems (Szwarcwald et al., 2021). Similarly, a comparative study in Brazil, Chile, Ecuador, Mexico, Peru and Spain found that poor self-perceived health was consistently elevated among socioeconomically disadvantaged groups, with Peru recording the highest prevalence of poor self-rated health while Spain recorded the lowest prevalence during the first wave of the pandemic (Salas Quijada et al., 2023). These findings suggest that, beyond the direct effects of COVID-19 infection, the pandemic amplified pre-existing health and socio-economic inequalities in diverse national contexts, highlighting the importance of examining overall health outcomes.

Evidence from South Africa suggests that job loss, food insecurity and increased care burdens were significant drivers of deteriorating mental health during the pandemic. While mental health effects are well documented, the pandemic likely affected multiple dimensions of wellbeing, including physical health, access to healthcare and social functioning. Disruptions to healthcare systems, delays in treatment and reduced utilisation of routine health services affected physical health and broader social wellbeing, particularly among vulnerable populations (Burger et al., 2020). These multidimensional health consequences highlight the need for a broader measure of health that captures individuals’ overall assessment of their wellbeing, rather than focusing on specific physical or psychological conditions alone.

In this context, self-reported health (SRH) is a useful measure of overall health because it captures individuals’ assessment of their physical, mental and social wellbeing. SRH is a strong predictor of morbidity, healthcare utilisation and mortality (Miilunpalo et al., 1997; Lorem et al., 2020) and is particularly useful during large-scale health shocks, when individuals reassess their health in response to changing risks and constraints (van de Weijer et al., 2022). Despite its widespread use, relatively few studies have examined how SRH evolved during the COVID-19 pandemic in the South African context.

2.2. Vulnerability Factors and Health During the Pandemic

A growing body of literature identifies socioeconomic vulnerability as a key determinant of health during public health crises. Factors such as income loss, food insecurity, pre-existing chronic illness and gender-related inequalities influence both individuals’ exposure to health risks and their capacity to cope with economic and healthcare disruptions (Schotte & Zizzamia, 2021; Yu, 2023). These vulnerabilities are particularly pronounced in South African because of persistently high inequality, labour market informality and a substantial burden of chronic illness.

Income loss represents one of the key channels through which pandemics affect health. Reduced household income limits access to food, healthcare and other essential goods, thereby increasing vulnerability to poor health outcomes (Pritchett & Summers, 1996). During the COVID-19 pandemic, income losses were widespread, particularly among informal and low-skilled workers, contributing to sharp increases in food insecurity and hunger (Posel et al., 2021; Hunt et al., 2021). Similar patterns have been documented elsewhere. A comparative study in Brazil, Chile, Ecuador, Mexico, Peru and Spain found that poorer and more socially vulnerable groups consistently reported worse self-perceived health during the first wave of the pandemic (Salas Quijada et al., 2023). Likewise, evidence from Eastern Ethiopia showed that poorer, less educated and rural households experienced significantly greater pandemic-related hardships, job losses and income shocks, illustrating how pre-existing socioeconomic vulnerabilities amplified the adverse effects of the pandemic (Muir et al., 2023).

Food insecurity represents a critical pathway through which economic shocks translate into adverse health outcomes. Prolonged food insecurity is associated with poorer physical and mental health and reduced resilience during crises. Evidence from low- and middle-income countries suggests that social assistance programmes generate stronger health outcomes when cash transfers are complemented by nutrition and related support services (Aurino & Giunti, 2022). In South Africa, food insecurity increased sharply during the early stages of the pandemic, before declining following the expansion of social assistance programmes (Wills et al., 2020).

Pre-existing chronic illness further increases vulnerability during pandemics by raising the risk of severe illness while simultaneously increasing dependence on uninterrupted healthcare services. Evidence from Brazil shows that worsening self-rated health during the pandemic was strongly associated with poor pre-pandemic health, mental health problems and disruptions to healthy lifestyles (Szwarcwald et al., 2021). Similarly, in South Africa, disruptions to routine healthcare and medication access may have exacerbated existing conditions and contributed to worsening overall health (Burger et al., 2020; Mboweni & Risenga, 2022).

Gender also influences vulnerability during public health crises through differences in labour market outcomes and caregiving responsibilities. In Latin America, women consistently reported poorer self-perceived health during the first wave of the pandemic (Salas Quijada et al., 2023). Similar patterns emerged in South Africa, where women experienced greater childcare responsibilities, higher rates of job loss in certain sectors and increased domestic burdens during lockdowns (Nwosu, 2021).

2.3. Social Protection and Health

The Role of Cash Transfers Social protection systems have been widely recognised as critical in mitigating the socio-economic impacts of the COVID-19 pandemic. In South Africa, the government implemented a large-scale social assistance response, including top-ups to existing grants and the introduction of the Social Relief of Distress (SRD) grant (National Treasury, 2020). These measures expanded coverage to millions of individuals, including working-age adults who were previously excluded from the social assistance system.

South Africa’s response reflected a broader global trend, where more than 220 countries introduced or expanded social protection measures to cushion households against income shocks (Gentilini et al., 2022). Cash transfers can improve health by stabilising household income, reducing financial stress, improving food security and facilitating access to healthcare (Pritchett & Summers, 1996; Manley et al., 2022).

Evidence from South Africa suggests that social grants helped cushion households against poverty and income losses during the pandemic (Köhler & Bhorat, 2020; Barnes et al., 2021). Similar evidence has emerged from other middle-income countries. For example, Brazil’s Bolsa Família programme was associated with improvements in child health, while Mexico’s Progresa/Oportunidades programme demonstrated the benefits of integrating cash transfers with preventive healthcare and nutrition services (Rasella et al., 2013; Aurino & Giunti, 2022). Together, these studies suggest that cash transfers are most effective when complemented by broader health and social interventions.

Further South African evidence suggests that household grant receipt was associated with a lower likelihood of reporting poor self-rated health during the pandemic (Omotayo & Ogunniyi, 2024). However, whether this association persisted as the pandemic evolved remains unclear. Moreover, while cash transfers can alleviate income constraints, they may not address broader constraints to health, including healthcare access, medication availability and structural health system constraints. These considerations motivate the present study, which examines whether the association between social grants and self-rated health changed over the course of the pandemic.

2.4. Research Gap

Despite the growing international literature on COVID-19, three important gaps remain. First, while there is extensive evidence on labour market outcomes, poverty and mental health, relatively little research has examined overall self-reported health as a comprehensive measure of wellbeing during the pandemic. Second, although vulnerability factors such as income loss, hunger and chronic illness have been studied individually, there is limited evidence on how these factors jointly influenced overall health outcomes during the pandemic. Third, while social protection responses have been widely analysed in terms of income and poverty effects, less is known about the dynamics and whether social cash transfers mitigated the broader health impacts of the pandemic, particularly in countries with large unconditional social grant systems such as South Africa.

This paper addresses these gaps by examining changes in self-reported health before and during the COVID-19 pandemic in South Africa. We analyse the role of key vulnerability factors, and assess whether social grants mitigated the adverse health effects associated with these vulnerabilities.

3. South Africa’s Response to COVID-19

South Africa entered the COVID-19 pandemic with one of the most extensive social assistance systems in the developing world. This existing institutional infrastructure enabled the government to respond relatively quickly to the economic and social disruptions caused by the pandemic. Before the pandemic, the social grant system already covered a large share of poor and vulnerable households through the Child Support Grant, Old Age Pension, Disability Grant, Foster Child Grant and Care Dependency Grant. This pre-existing system became the foundation for South Africa’s emergency social protection response.

The COVID-19 shock was severe. Early evidence from NIDS-CRAM Wave 1 showed that 53% of adult respondents lived in households that had run out of money to buy food in April 2020, while about four in ten respondents reported that their household had lost its main source of income following the announcement of the hard lockdown on 27 March 2020 (Wills et al., 2020). These findings pointed to an immediate risk of widespread food insecurity, income loss and deteriorating wellbeing.

In response, the South African government announced a fiscal stimulus package of R502 billion in April 2020 (National Treasury, 2020). A key component of this package was direct income support through social assistance and wage protection. Approximately R40 billion was allocated to wage protection through the Temporary Employee-Employer Relief Scheme (TERS), while R50 billion was dedicated to social assistance through top-ups to existing grants and the introduction of the Social Relief of Distress (SRD) grant.

The TERS programme was designed to protect formal-sector workers from income losses and prevent retrenchments during the lockdown. It provided income support to employees whose employers had fully or partially closed operations because of the pandemic. The programme was financed through the Unemployment Insurance Fund and reached its highest coverage during the early months of the lockdown. Estimates from NIDS-CRAM indicate that more than 4 million individuals received TERS at least once between April 2020 and May 2021, with the number of active recipients highest during the most stringent phases of lockdown (Köhler & Hill, 2021). The distribution of TERS receipt across income groups is presented in

Table 1. Receipt of TERS by decile.

Household Income Decile (Based on NIDS Wave 5)	% of Adults Accessing TERS in Any Wave of CRAM	Average Pre-Pandemic Salary/Wage (per Month in 2020 Prices)
1	15.7%	R1065
2	12.6%	R1474
3	10.9%	R1460
4	15.2%	R2401
5	14.7%	R2456
6	27.5%	R3560
7	23.0%	R4492
8	23.2%	R5190
9	19.7%	R8691
10	18.7%	R24,487
All	19.0%	R8340

Source: Adapted from Moses and Woolard (2023).

, highlighting how the programme reached workers across the income distribution.

In parallel, the government expanded social assistance. Existing grant beneficiaries received temporary top-ups, and a new SRD grant was introduced for unemployed working-age adults who were not receiving other forms of income support. This was a significant policy development because working-age unemployed adults had historically been largely excluded from South Africa’s social assistance system. The SRD grant therefore represented an important, though temporary, expansion of the social safety net (Seekings, 2020).

The emergency response combined temporary increases in existing grants with the introduction of a new cash transfer targeted at previously uncovered groups. During the early stages of the pandemic, beneficiaries of the Child Support Grant received temporary top-ups, while recipients of the Older Person’s Grant and Disability Grant also received additional monthly payments. In addition, the Social Relief of Distress (SRD) grant of R350 per month was introduced in May 2020 for unemployed working-age adults who were not receiving other forms of government income support. Although the temporary top-ups were subsequently phased out, the SRD grant remained the principal pandemic-specific cash transfer for much of the study period. Consequently, overall social assistance coverage expanded considerably during the pandemic, particularly among working-age adults who had previously been excluded from South Africa’s grant system. However, while the expansion substantially increased the number of beneficiaries, the relatively modest value of the transfers meant that they were primarily intended to cushion immediate income losses rather than fully compensate households for prolonged unemployment or broader disruptions to livelihoods and healthcare.

The scale of the response was substantial. By June 2020, approximately 4.3 million applications for the SRD grant had been approved, rising to about 6.1 million by March 2021. Figure 1 shows the distribution of SRD grant receipt across income deciles, illustrating the extent to which the programme expanded coverage among lower-income groups.

Overall, the combination of grant top-ups, the SRD grant and TERS meant that millions of individuals and households received some form of direct cash support during the pandemic. Compared with many other African countries, South Africa’s response was notable for both its scale and its reliance on existing social protection systems.

The design of the response also has direct relevance for this study. Social grants may influence health through several channels. First, they may reduce income stress and thereby lower psychological distress. Second, they may improve household food security, which is closely linked to physical health. Third, they may enable households to meet basic healthcare-related costs, including transport to clinics and medication-related expenses. In this sense, social cash transfers may mitigate some of the adverse health effects associated with income loss, hunger and vulnerability.

However, cash transfers may also have important limits as health-protection instruments. While they may ease income constraints, they cannot fully compensate for disrupted healthcare services, reduced clinic attendance, interruptions in chronic medication access or broader weaknesses in the health system. For individuals with chronic illnesses for example, income support may be helpful but insufficient if routine care and medication access are interrupted. Similarly, cash support may reduce food insecurity but may not fully substitute for community nutrition programmes or other targeted health interventions.

This study therefore examines not only whether vulnerability factors such as hunger, chronic illness and income loss were associated with poor self-reported health, but also whether household receipt of social grants mitigated these associations. The central policy question is whether South Africa’s cash-based response helped protect health during the pandemic, and whether this protective association persisted as the pandemic evolved.

4. Data and Methods

4.1. Data Sources

This study uses data from the National Income Dynamics Study (NIDS) and the NIDS-Coronavirus Rapid Mobile Survey (NIDS-CRAM). NIDS is South Africa’s first nationally representative household panel survey and was conducted approximately every two years between 2008 and 2017. The survey tracks individuals and households over time and provides detailed information on socio-economic status, demographic characteristics, labour market outcomes and health.

The NIDS-CRAM survey was designed as a rapid follow-up survey to measure the social and economic impacts of the COVID-19 pandemic. It followed a representative subsample of adults from the 2017 NIDS Wave 5 sample using computer-assisted telephone interviews. NIDS-CRAM consisted of five waves conducted between May 2020 and May 2021. Wave 1 was conducted in May–June 2020, Wave 2 in July–August 2020, Wave 3 in November–December 2020, Wave 4 in February–March 2021, and Wave 5 in April–May 2021 (Ingle et al., 2021).

This study uses NIDS Wave 5 as the pre-pandemic baseline and NIDS-CRAM Waves 1, 4 and 5 as the pandemic-period surveys. These three NIDS-CRAM waves are used because self-reported health was collected only in these waves. The linked structure of the data allows us to compare health outcomes before and during the pandemic and to examine how individual and household characteristics were associated with poor health during different stages of the pandemic.

4.2. Outcome Variable

The main outcome variable is self-reported health. Respondents were asked to describe their current health status using five categories: “excellent”, “very good”, “good”, “fair” and “poor”. Following the existing literature, we construct a binary indicator of poor self-reported health that takes the value of one if the respondent reports their health as either “fair” or “poor”, and zero if the respondent reports their health as “good”, “very good” or “excellent” (Oyenubi et al., 2021).

Self-reported health is used because it provides a broad measure of health that captures physical, mental and social dimensions of wellbeing. Although subjective, it is widely used in health economics and epidemiology and has been shown to predict morbidity, healthcare use and mortality. Its consistent measurement across NIDS and NIDS-CRAM makes it suitable for analysing health dynamics before and during the pandemic.

4.3. Explanatory Variables

The analysis focuses on three sets of explanatory variables: vulnerability factors, social protection variables and demographic controls.

The main vulnerability factors are household hunger, chronic illness and household income loss. Household hunger is measured using an indicator for whether any household member experienced hunger. Chronic illness is measured using an indicator for whether the respondent reported a chronic health condition, such as diabetes, hypertension, tuberculosis, HIV/AIDS or another long-term condition. Household income loss captures whether the respondent reported that household income had declined during the pandemic.

Social protection is measured using household receipt of any social grant. This variable captures whether the respondent lived in a household in which at least one member received a social grant. This includes existing grants as well as pandemic-related grant support. The variable is used to assess whether social grants were associated with a lower probability of poor self-reported health.

The control variables include age, gender, race, marital status, tertiary education, dwelling type and household size. Baseline income quintiles from NIDS Wave 5 are also included to capture pre-pandemic socio-economic status. Using baseline income is important because it reduces the risk that pandemic-period income shocks mechanically determine both current income status and health outcomes.

4.4. Descriptive Statistics

Table 2. Summary statistics.

	Wave 1			Wave 4			Wave 5
Variable	Obs	Mean/%	Std. Dev.	Obs	Mean/%	Std. Dev.	Obs	Mean/%	Std. Dev.
Poor SRH	7026	29.6%	0.456	5578	31.3%	0.464	5838	28.2%	0.450
Respondent received grant	7037	21.5%	0.411	5624	43.0%	0.495	5852	43.9%	0.496
Household member received grant	7054	74.7%	0.435	5629	80.2%	0.399	5861	79.0%	0.407
Household experienced hunger	7016	26.5%	0.441	5615	19.2%	0.394	5857	19.1%	0.393
Chronic health condition	7044	21.6%	0.412	-	-	-	-	-	-
Age	7073	40.58	15.62	5629	41.09	15.78	5862	41.28	15.73
Female	7067	61.0%	0.488	5629	61.7%	0.486	5862	61.7%	0.486
African	7073	85.5%	0.352	5629	87.0%	0.337	5862	86.5%	0.342
Coloured	7073	8.8%	0.284	5629	7.8%	0.268	5862	8.4%	0.277
Asian	7073	1.1%	0.104	5629	0.8%	0.087	5862	0.8%	0.088
White	7073	4.6%	0.209	5629	4.4%	0.206	5862	4.3%	0.204
Married	5671	46.0%	0.498	5629	43.6%	0.496	5862	43.7%	0.496
Tertiary education	5375	33.8%	0.473	4047	34.4%	0.475	4239	35.1%	0.477
Formal dwelling	7067	77.9%	0.415	5579	76.8%	0.422	5803	76.9%	0.421
Traditional dwelling (hut)	7067	11.9%	0.324	5579	11.3%	0.317	5803	11.6%	0.320
Informal dwelling (shack)	7067	10.2%	0.303	5579	11.9%	0.324	5803	11.5%	0.319
Household size	7039	5.56	3.309	5629	5.30	3.204	5862	5.28	3.234

Source: Author estimates based on NIDS-CRAM Waves 1, 4 and 5 datasets.

presents summary statistics for the main variables used in the analysis across NIDS-CRAM Waves 1, 4 and 5. Approximately 30% of respondents report poor self-reported health in Waves 1 and 4, with a slight decline to 28% in Wave 5.

The proportion of respondents receiving social grants increases substantially over time, reflecting the expansion of social assistance during the pandemic. Household-level grant receipt rises from about 75% in Wave 1 to approximately 80% in later waves. At the same time, the incidence of household hunger declines from 26% to about 19%, consistent with the mitigating effects of social support.

The sample is predominantly female, with an average age of approximately 41 years. A large majority of respondents identify as African, and average household size is about five to six members. These characteristics are broadly consistent across waves.

4.5. Descriptive and Transition Analysis

The empirical analysis begins with descriptive statistics showing the prevalence of poor self-reported health before and during the pandemic. We compare health outcomes across the five waves of the core NIDS survey and the relevant NIDS-CRAM waves. This allows us to examine whether the incidence of poor health increased during the pandemic relative to the pre-pandemic period.

We then estimate health transition matrices. These matrices show the movement of individuals between good and poor health states across survey waves. Entry into poor health is defined as the proportion of individuals who were in good health in period t but reported poor health in period t + 1. Similarly, exit from poor health is defined as the proportion of individuals who were in poor health in period t but reported good health in period t + 1. This approach allows us to examine whether transitions into poor health became more common during the pandemic.

4.6. Regression Specification

To examine the association between vulnerability factors, social grants and poor self-reported health, we estimate probit regression models separately for NIDS-CRAM Waves 1 and 5. The probit model is appropriate because the dependent variable is binary and takes the value of one if the respondent reports poor health and zero otherwise.

The basic empirical specification is:

$$\begin{array}{ll}P({PoorHealth}_{i,t}& =1)\\ & =\Phi (\alpha +{\beta }_1{Hunger}_{i,t}+{\beta }_2{Chronic}_i+{\beta }_3{IncomeLoss}_{it}\\ & +{\beta }_4 {Grant}_{i,t}+\gamma X_{i,t}+\delta Q_i+{\epsilon }_{i,t})\end{array}$$

where ${PoorHealth}_{it }$ is an indicator equal to one if individual i reports fair or poor health in wave t, and zero otherwise. ${Hunger}_{it }$ captures household hunger, ${Chronic}_{i }$ captures whether the respondent has a chronic health condition, ${IncomeLoss}_{it }$ captures household income loss, and ${Grant}_{it }$ captures household receipt of any social grant. $X_{it }$ is a vector of demographic and household controls, while $Q_{i }$ represents baseline income quintiles from NIDS Wave 5. $\Phi$ denotes the cumulative normal distribution function.

Because probit coefficients are not directly interpretable as changes in probability, we report marginal effects. These show the estimated percentage-point change in the probability of reporting poor health associated with each explanatory variable.

5. Results

5.1. Prevalence of Poor Self-Reported Health

We begin by examining patterns of self-reported health before and during the COVID-19 pandemic. Using the binary indicator of poor health (equal to one if health is reported as “fair” or “poor”), we find a substantial deterioration in overall health during the pandemic.

Compared to the decade preceding the pandemic, the prevalence of poor self-reported health increased markedly. The average incidence of poor health rose from approximately 11.9% in the pre-pandemic period to about 26% during the pandemic. This represents more than a doubling in the risk of poor health.

Figure 2 presents trends in poor self-reported health across the five waves of the core NIDS survey and the relevant NIDS-CRAM waves. A key feature of the figure is that poor health increased sharply at the onset of the pandemic and remained persistently elevated throughout the pandemic period. Although there is a slight decline in the final wave of NIDS-CRAM, the overall level of poor health remains significantly higher than pre-pandemic levels.

This finding suggests that the health effects of the pandemic extended beyond the initial shock and persisted even as some economic activity resumed. Given the strong association between self-reported health and morbidity and mortality, the sustained increase in poor health has important implications for public health and social policy.

5.2. Health Transitions Before and During the Pandemic

To better understand the dynamics underlying the observed increase in poor health, we examine transitions between health states using transition matrices.

Table 3. Transitions into and out of poor health in the NIDS (2008 to 2017) datasets.

Health status in previous wave		2010/11				2012
		Good	Poor			Good	Poor
2008	Good	93%	7%	Wave 2	Good	91%	9%
	Poor	69%	31%		Poor	65%	35%
		2014/15				2017
		Good	Poor			Good	Poor
2012	Good	90%	10%	Wave 4	Good	92%	8%
	Poor	65%	36%		Poor	69%	31%
		2017
		Good	Poor
2008	Good	90%	10%
	Poor	70%	30%

Notes: Author estimates based on NIDS Waves 1–5 data, conducted between 2008 and 2017. Panel weights used.

presents transitions into and out of poor health across the five waves of the NIDS survey prior to the pandemic. The results indicate a high degree of stability in health status during this period. Individuals who reported good health in any given wave had at least a 90% probability of remaining in good health in the subsequent wave. Conversely, among those reporting poor health, approximately two-thirds transitioned back into good health in the following period.

These results suggest that prior to the pandemic, poor health was relatively transitory, and most individuals were able to recover to good health over time.

Table 4. Health transitions in the NIDS Wave 5 and NIDS-CRAM datasets (from 2017 to 2021).

Health status in previous wave		May/Jun 2020				Feb/Mar 2021
		Good	Poor			Good	Poor
2017	Good	71%	29%	May/Jun 2020	Good	74%	26%
	Poor	62%	38%		Poor	56%	44%
		Apr/May 2021				Apr/May 2021
		Good	Poor			Good	Poor
Feb/Mar 2021	Good	83%	17%	2017	Good	73%	27%
	Poor	49%	51%		Poor	58%	42%

Notes: Author estimates based on the linked NIDS wave 5 and NIDS-CRAM waves 1,4 and 5 datasets.

presents comparable transition matrices for the pandemic period using NIDS Wave 5 and NIDS-CRAM Waves 1, 4 and 5. In contrast to the pre-pandemic period, we observe a substantial increase in transitions into poor health. The probability that an individual in good health transitions into poor health rises to approximately 25% during the pandemic, compared to less than 10% before the pandemic.

Moreover, the elevated probability of transitioning into poor health persists across waves, indicating that the deterioration in health is not a short-lived phenomenon. At the same time, the probability of exiting poor health declines, suggesting that recovery becomes more difficult during the pandemic. Taken together, these findings indicate a fundamental shift in health dynamics during the pandemic, characterised by both increased entry into poor health and reduced exit from poor health.

5.3. Vulnerability Factors and Transitions into Poor Health

We next examine how transitions into poor health vary across key vulnerability factors.

Table 5. Events associated with poor health transitions in NIDS-CRAM.

	Transition Probabilities
Bottom 20%	31.97
Quintile 2	28.92
Quintile 3	25.38
Quintile 4	26.17
Top 20%	20.78
Employment loss in w1
No	26.71
Yes	32.19
Income loss in w1
No	24.28
Yes	31.08
Chronic illness
No	24.4
Yes	37.65
Gender
Male	25.95
Female	27.92

Notes: Author estimates based on Waves 1 and 5 of NIDS-CRAM. Income quintiles based on the NIDS W5 data.

reports the probability of transitioning into poor health between Wave 1 and Wave 5 of NIDS-CRAM for different subgroups. A clear socio-economic gradient is evident. Individuals in the lowest income quintile have the highest probability of transitioning into poor health (approximately 32%), while those in the highest quintile have a significantly lower probability (around 21%).

Household income shocks are also strongly associated with health transitions. Individuals in households that experienced income loss or employment loss have a higher probability of transitioning into poor health compared to those who did not experience such shocks. This finding is consistent with the idea that economic stress translates into adverse health outcomes through multiple channels, including reduced food security and increased psychological distress.

Chronic illness emerges as one of the strongest predictors of health deterioration. Individuals with pre-existing chronic conditions have a substantially higher probability of transitioning into poor health than those without such conditions. This likely reflects both increased vulnerability to COVID-19-related risks and disruptions to routine healthcare and medication access.

Gender differences are also observed, with females exhibiting a slightly higher probability of transitioning into poor health. This may reflect gendered inequalities in labour market outcomes, caregiving responsibilities and exposure to pandemic-related stress.

5.4. Regression Analysis

While the transition analysis provides descriptive evidence, we next estimate probit models to examine whether the observed relationships hold after controlling for multiple covariates.

Table 6. Probit regressions of SRH and various vulnerabilities.

	Wave 1		Wave 5
VARIABLES	Basic	Full	Basic	Full
Household vulnerability
Household hunger	0.299 ***	0.284 ***	0.520 ***	0.560 ***
	(0.037)	(0.049)	(0.049)	(0.061)
Chronic illness	0.199 ***	0.100 *	0.350 ***	0.315 ***
	(0.039)	(0.054)	(0.046)	(0.061)
Income quintile 2	−0.028	−0.034	−0.010	0.017
	(0.048)	(0.065)	(0.059)	(0.075)
Income quintile 3	−0.152 ***	−0.068	−0.124 **	−0.046
	(0.048)	(0.066)	(0.060)	(0.075)
Income quintile 4	−0.141 ***	−0.022	−0.065	0.055
	(0.050)	(0.069)	(0.061)	(0.078)
Income quintile 5	−0.322 ***	−0.106	−0.219 ***	0.084
	(0.057)	(0.083)	(0.071)	(0.096)
HH income loss	0.091 ***	0.097 **	0.117 **	0.084
	(0.033)	(0.043)	(0.054)	(0.066)
Social Protection
HH social grant		−0.119 **		−0.013
		(0.054)		(0.067)
Household and Demographic controls
Age		0.001		0.004 *
		(0.002)		(0.002)
Female		−0.059		−0.042
		(0.045)		(0.052)
African		0.602 ***		0.906 ***
		(0.076)		(0.101)
Married		0.110 **		0.010
		(0.046)		(0.053)
Tertiary education		0.006		−0.038
		(0.047)		(0.053)
Formal dwelling		−0.029		0.082
		(0.052)		(0.060)
Household size		0.023 ***		0.018 **
		(0.007)		(0.008)
Constant	−0.597 ***	−1.261 ***	−0.728 ***	−1.960 ***
	(0.039)	(0.135)	(0.046)	(0.166)
Observations	6799	4131	4691	3398

Standard errors in parentheses: *** p < 0.01, ** p < 0.05, * p < 0.1. Notes: The basic model include household vulnerability variables only while the full model additionally controls for household grant receipt and socio-demographic characteristics. The number of observations differs across specifications because the more comprehensive models include additional socio-demographic covariates. Estimation uses listwise deletion, so observations with missing values on any included variable are excluded from that specification.

and

Table 7. Marginal effects.

	Wave1		Wave 5
VARIABLES	Basic	Full	Basic	Full
Household vulnerability
Household hunger	0.101 ***	0.092 ***	0.167 ***	0.166 ***
	(0.012)	(0.016)	(0.015)	(0.017)
Chronic illness	0.067 ***	0.032 *	0.113 ***	0.093 ***
	(0.013)	(0.018)	(0.015)	(0.018)
Income quintile 2	−0.010	−0.011	−0.003	0.005
	(0.017)	(0.021)	(0.020)	(0.022)
Income quintile 3	−0.053 ***	−0.022	−0.040 **	−0.013
	(0.017)	(0.022)	(0.019)	(0.022)
Income quintile 4	−0.049 ***	−0.007	−0.022	0.016
	(0.017)	(0.023)	(0.020)	(0.023)
Income quintile 5	−0.106 ***	−0.034	−0.069 ***	0.025
	(0.018)	(0.027)	(0.022)	(0.029)
HH income loss	0.030 ***	0.031 **	0.038 **	0.025
	(0.011)	(0.014)	(0.017)	(0.019)
Social Protection
HH social grant		−0.039 **		−0.004
		(0.018)		(0.020)
Household and Demographic controls
Age		0.000		0.001 *
		(0.001)		(0.001)
Female		−0.019		−0.012
		(0.014)		(0.015)
African		0.170 ***		0.208 ***
		(0.018)		(0.016)
Married		0.036 **		0.003
		(0.015)		(0.016)
Tertiary education		0.002		−0.011
		(0.015)		(0.016)
Formal dwelling		−0.010		0.024
		(0.017)		(0.017)
Household size		0.008 ***		0.005 **
		(0.002)		(0.002)
Observations	6799	4131	4691	3398

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. Notes: The basic model include household vulnerability variables only while the full model additionally controls for household grant receipt and socio-demographic characteristics. The number of observations differs across specifications because the more comprehensive models include additional socio-demographic covariates. Estimation uses listwise deletion, so observations with missing values on any included variable are excluded from that specification.

present the probit estimates and corresponding marginal effects for Waves 1 and 5 of NIDS-CRAM. We focus on the marginal effects, which provide a more intuitive interpretation.

5.4.1. Wave 1 Results

In Wave 1, we find that household hunger, chronic illness and household income loss are all significantly associated with a higher probability of reporting poor health. In particular, individuals in households experiencing hunger are approximately 10 percentage points more likely to report poor health compared to those in food-secure households. Similarly, individuals with chronic conditions have a significantly higher likelihood of reporting poor health.

Socio-economic status also plays an important role. Individuals in higher income quintiles are less likely to report poor health, consistent with the well-established income-health gradient.

Importantly, receipt of household social grants is associated with a reduction in the probability of reporting poor health. The marginal effects suggest that individuals in grant-receiving households are approximately 4 percentage points less likely to report poor health. This provides evidence of a protective association between household grant receipt and self-reported health during the early stages of the pandemic.

5.4.2. Wave 5 Results

By Wave 5, the relationships between vulnerability factors and poor health become stronger. The marginal effect of household hunger increases substantially, indicating that the adverse health effects of food insecurity intensified over time. Similarly, the effect of chronic illness becomes significantly larger, suggesting that individuals with chronic conditions became increasingly vulnerable as the pandemic progressed.

In contrast, the protective association observed in Wave 1 is no longer evident by Wave 5. The coefficient on household grant receipt becomes statistically insignificant, indicating that grants are no longer associated with a lower probability of poor health. Additional interaction models (reported in Appendix A 

Table A3. Probit regressions with household grant interaction effects (Wave 1).

VARIABLES	(1)	(2)	(3)
Household hunger = 1	0.305 ***	0.284 ***	0.284 ***
	(0.098)	(0.049)	(0.049)
HH social grant = 1	−0.113 *	−0.110 *	−0.017
	(0.060)	(0.058)	(0.069)
Chronic illness	0.100 *	0.139	0.098 *
	(0.054)	(0.106)	(0.054)
Income quintile 2	−0.033	−0.033	−0.032
	(0.065)	(0.065)	(0.065)
Income quintile 3	−0.069	−0.068	−0.067
	(0.066)	(0.066)	(0.066)
Income quintile 4	−0.022	−0.021	−0.023
	(0.069)	(0.069)	(0.069)
Income quintile 5	−0.104	−0.105	−0.091
	(0.084)	(0.083)	(0.084)
HH income loss	0.097 **	0.097 **	0.262 ***
	(0.043)	(0.043)	(0.081)
Hunger [0]#HHgrant [0]	0.000
	(0.000)
Hunger [0]#HHgrant [1]	0.000
	(0.000)
Hunger [1]#HHgrant [0]	0.000
	(0.000)
Hunger [1]#HHgrant [1]	−0.027
	(0.111)
Chronic illness [0]#HHgrant [0]		0.000
		(0.000)
Chronic illness [0]#HHgrant [1]		0.000
		(0.000)
Chronic illness [1]#HHgrant [0]		0.000
		(0.000)
Chronic illness [1]#HHgrant [1]		−0.052
		(0.121)
HHincomeloss [0]#HHgrant [0]			0.000
			(0.000)
HHincomeloss [0]#HHgrant [1]			0.000
			(0.000)
HHincomeloss [1]#HHgrant [0]			0.000
			(0.000)
HHincomeloss [1]#HHgrant [1]			−0.228 **
			(0.095)
Female	−0.058	−0.058	−0.058
	(0.044)	(0.044)	(0.044)
African	0.606 ***	0.606 ***	0.603 ***
	(0.076)	(0.076)	(0.076)
Married	0.111 **	0.111 **	0.113 **
	(0.046)	(0.046)	(0.046)
Tertiary education	0.004	0.003	0.011
	(0.047)	(0.047)	(0.047)
Formal dwelling	−0.030	−0.031	−0.027
	(0.052)	(0.052)	(0.052)
Household size	0.023 ***	0.023 ***	0.024 ***
	(0.007)	(0.007)	(0.007)
Constant	−1.259 ***	−1.259 ***	−1.340 ***
	(0.143)	(0.141)	(0.145)
Observations	4131	4131	4131

Notes: Columns (1), (2) and (3) present interaction effects for hunger, chronic illness and household income loss, respectively. Standard errors in parentheses. # denotes interaction terms between household grant receipt and the respective vulnerability variable. *** p < 0.01, ** p < 0.05, * p < 0.1.

and

Table A4. Probit regressions with household grant interaction effects (Wave 5).

VARIABLES	(1)	(2)	(3)
Household hunger	0.759 ***	0.560 ***	0.555 ***
	(0.146)	(0.061)	(0.061)
HH social grant	0.014	0.003	0.015
	(0.071)	(0.071)	(0.072)
Chronic illness	0.315 ***	0.437 ***	0.315 ***
	(0.061)	(0.134)	(0.061)
Income quintile 2	0.020	0.017	0.018
	(0.075)	(0.075)	(0.075)
Income quintile 3	−0.049	−0.048	−0.046
	(0.075)	(0.075)	(0.075)
Income quintile 4	0.055	0.055	0.056
	(0.078)	(0.078)	(0.078)
Income quintile 5	0.098	0.082	0.089
	(0.096)	(0.096)	(0.096)
HH income loss	0.074	0.084	0.228 *
	(0.066)	(0.066)	(0.126)
Hunger [0]#HHgrant [0]	0.000
	(0.000)
Hunger [0]#HHgrant [1]	0.000
	(0.000)
Hunger [1]#HHgrant [0]	0.000
	(0.000)
Hunger [1]#HHgrant [1]	−0.240
	(0.160)
Chronic illness [0]#HHgrant [0]		0.000
		(0.000)
Chronic illness [0]#HHgrant [1]		0.000
		(0.000)
Chronic illness [1]#HHgrant [0]		0.000
		(0.000)
Chronic illness [1]#HHgrant [1]		−0.153
		(0.149)
HHincomeloss [0]#HHgrant [0]			0.000
			(0.000)
HHincomeloss [0]#HHgrant [1]			0.000
			(0.000)
HHincomeloss [1]#HHgrant [0]			0.000
			(0.000)
HHincomeloss [1]#HHgrant [1]			−0.197
			(0.147)
Age	0.005 *	0.005 *	0.005 *
	(0.002)	(0.002)	(0.002)
Female	−0.039	−0.045	−0.043
	(0.052)	(0.052)	(0.052)
African	0.890 ***	0.900 ***	0.895 ***
	(0.099)	(0.100)	(0.099)
Married	0.012	0.011	0.010
	(0.053)	(0.053)	(0.053)
Tertiary education	−0.032	−0.038	−0.036
	(0.053)	(0.053)	(0.053)
Formal dwelling	0.081	0.082	0.086
	(0.060)	(0.060)	(0.060)
Household size	0.019 **	0.018 **	0.018 **
	(0.008)	(0.008)	(0.008)
Constant	−2.016 ***	−2.006 ***	−2.015 ***
	(0.173)	(0.173)	(0.173)
Observations	3398	3398	3398

Notes; Columns (1), (2) and (3) present interaction effects for hunger, chronic illness and household income loss, respectively. Standard errors in parentheses. # denotes interaction terms between household grant receipt and the respective vulnerability variable. *** p < 0.01, ** p < 0.05, * p < 0.1.

) indicate that household grants exhibited only limited buffering effects against specific vulnerabilities. During Wave 1, grants modestly mitigated the adverse health consequences of household income loss, whereas no significant interaction effects were observed by Wave 5.

Figure 3 presents the average marginal effects of household grant receipt on the probability of reporting poor self-reported health in Waves 1 and 5, together with 95% confidence intervals. During Wave 1, household grant receipt was associated with a 3.9 percentage-point lower probability of reporting poor health (95% CI: −7.3 to −0.4 percentage points). By Wave 5, however, the estimated marginal effect was close to zero and the confidence interval included zero, indicating no statistically significant association between household grant receipt and self-reported health. The figure provides a visual summary of the regression results reported in Table 6 and illustrates that the protective association observed during the early stages of the pandemic was no longer evident by the later stage of the study period.

This finding suggests that while social grants may have provided short-term relief during the initial phase of the pandemic, the strength of their protective association with health diminished over time. One possible explanation is that prolonged disruptions to healthcare services, persistent food insecurity and cumulative stress outweighed the benefits of income support.

5.5. Robustness Checks

5.5.1. Robustness Checks Addressing Potential Endogeneity

We conducted a series of robustness checks to assess whether the estimated association between household grant receipt and poor self-reported health was sensitive to potential endogeneity arising from reverse causality and pre-existing household characteristics. Specifically, the main regression models were re-estimated controlling simultaneously for three pre-pandemic characteristics measured in the 2017 NIDS survey: (i) prior household grant receipt (Household grant receipt in 2017), (ii) baseline self-reported health (Self-reported health in 2017), and (iii) baseline household income quintiles (NIDS household income quintiles). These variables capture pre-existing socioeconomic disadvantage, long-term reliance on social protection and underlying health status, thereby reducing concerns that contemporaneous grant receipt simply reflects households that were already poorer or less healthy before the pandemic.

The results (

Table A1. Robustness Models Controlling for Pre-pandemic Health, Socioeconomic Status and Prior Grant Receipt.

VARIABLES	Wave 1	Wave 5
Household hunger	0.089 ***	0.168 ***
	(0.016)	(0.018)
Chronic illness	0.030 *	0.089 ***
	(0.018)	(0.019)
NIDS income quintiles = 2	−0.012	0.007
	(0.022)	(0.023)
NIDS income quintiles = 3	−0.027	−0.019
	(0.022)	(0.022)
NIDS income quintiles = 4	−0.010	0.016
	(0.024)	(0.024)
NIDS income quintiles = 5	−0.029	0.026
	(0.029)	(0.031)
HH income loss	0.032 **	0.026
	(0.015)	(0.020)
HH social grant	−0.041 **	−0.005
	(0.019)	(0.021)
HH grant receipt (2017)	0.008	0.009
	(0.018)	(0.018)
Poor self-reported health (in 2017)	0.058 **	0.092 ***
	(0.029)	(0.030)
Age	−0.000	0.001
	(0.001)	(0.001)
Female	−0.023	−0.021
	(0.015)	(0.016)
African	0.186 ***	0.270 ***
	(0.025)	(0.030)
Married	0.036 **	0.003
	(0.015)	(0.016)
Tertiary education	0.000	−0.008
	(0.016)	(0.016)
Formal dwelling	−0.005	0.031 *
	(0.018)	(0.018)
Household size	0.007 ***	0.006 **
	(0.002)	(0.003)
Observations	3907	3217

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

) show that the principal findings remain robust. Household grant receipt continued to be associated with a significantly lower probability of poor self-reported health during Wave 1, while remaining statistically insignificant in Wave 5. Importantly, prior household grant receipt was not significantly associated with poor self-reported health in either wave, suggesting that the observed Wave 1 protective association is not merely capturing long-term grant dependence. Similarly, baseline self-reported health remained a strong predictor of poor health during the pandemic, confirming persistence in health status over time, while baseline income quintiles were generally not statistically significant once household hunger, chronic illness and income loss were taken into account. Overall, these findings indicate that the principal conclusions are robust to controls for pre-pandemic health, socioeconomic status and prior grant receipt.

5.5.2. Grant Heterogeneity

We further examined whether the overall household grant effect concealed important differences across social assistance programmes by replacing the aggregate household grant indicator with indicators for households receiving the Child Support Grant (CSG), Older Person’s Grant (OAP) and the COVID-19 Social Relief of Distress (SRD) grant, while retaining the same baseline controls (

Table A2. Grant Heterogeneity Models Controlling for Pre-pandemic Characteristics.

VARIABLES	Wave 1	Wave 5
Household hunger	0.088 ***	0.167 ***
	(0.016)	(0.018)
Chronic illness	0.032 *	0.089 ***
	(0.018)	(0.019)
NIDS income quintiles = 2	−0.012	0.007
	(0.022)	(0.023)
NIDS income quintiles = 3	−0.027	−0.019
	(0.023)	(0.022)
NIDS income quintiles = 4	−0.010	0.016
	(0.024)	(0.025)
NIDS income quintiles = 5	−0.033	0.024
	(0.029)	(0.031)
HH income loss	0.033 **	0.024
	(0.015)	(0.020)
HH that receives CSG	−0.047 **	0.002
	(0.019)	(0.019)
HH that receives Old Age Pension	−0.009	0.003
	(0.017)	(0.019)
HH receives SRD grant	0.056	0.039 *
	(0.079)	(0.021)
HH grant receipt (in 2017)	0.010	0.009
	(0.018)	(0.018)
Poor self-reported health (in 2017)	0.056 *	0.085 ***
	(0.029)	(0.030)
Age	−0.000	0.001 *
	(0.001)	(0.001)
Female	−0.021	−0.016
	(0.015)	(0.016)
African	0.195 ***	0.266 ***
	(0.025)	(0.031)
Married	0.037 **	0.004
	(0.016)	(0.016)
Tertiary education	0.003	−0.006
	(0.016)	(0.016)
Formal dwelling	−0.006	0.031 *
	(0.018)	(0.018)
Household size	0.008 ***	0.005 *
	(0.003)	(0.003)
Observations	3848	3196

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

). The results indicate that the protective association observed during Wave 1 was driven primarily by households receiving the Child Support Grant, whereas receipt of the Older Person’s Grant was not independently associated with self-reported health. By Wave 5, the protective association of the Child Support Grant was no longer evident. Respondents receiving the SRD grant remained more likely to report poor health during Wave 5, likely reflecting the continued vulnerability of individuals targeted by this emergency programme rather than an adverse effect of the grant itself. Taken together, the grant heterogeneity analysis suggests that the early protective association identified in the aggregate models was not uniform across programmes and was strongest among households receiving the Child Support Grant.

5.5.3. Interaction Effects

Finally, interaction models were estimated to examine whether household grants moderated the adverse health effects of household hunger, chronic illness and income loss. These models presented in Table A3 and Table A4 provide a more direct test of whether social grants buffered the health consequences of vulnerability. The interaction effects were generally weak. The only statistically significant interaction was between household grant receipt and household income loss during Wave 1, suggesting that grants modestly mitigated the health consequences of pandemic-related income shocks during the initial phase of the crisis. No significant interaction effects were observed in Wave 5, indicating that this buffering role did not persist as the pandemic evolved. These findings reinforce the main conclusion that although social grants provided important short-term protection, they were insufficient to offset the cumulative health impacts of a prolonged crisis.

Overall, the robustness checks reinforce the main findings of the study. The protective association between household grant receipt and poor self-rated health during Wave 1 remained robust after controlling for pre-pandemic health status, prior household grant receipt and baseline socioeconomic status, suggesting that the results are unlikely to be driven by pre-existing differences between grant and non-grant households. Grant heterogeneity analyses further indicate that this protective association was driven primarily by households receiving the Child Support Grant. Consistent with the main analysis, the protective effect of grants weakened by Wave 5, while the adverse effects of vulnerability factors, particularly household hunger and chronic illness, became more pronounced.

6. Discussion

The results of this study provide new evidence on how overall health changed during the COVID-19 pandemic in South Africa and whether social protection receipt was associated with weaker adverse health patterns. Three key findings emerge.

First, self-reported health deteriorated substantially during the pandemic. The prevalence of poor health more than doubled relative to the pre-pandemic period and remained elevated throughout the pandemic. Much of the existing South African literature has understandably focused on mental health. Our findings suggest that the consequences of the pandemic were considerably broader, extending to overall health and wellbeing. The persistence of poor health also indicates that the health burden of the pandemic was not confined to the initial lockdown period. Rather, it reflected prolonged disruptions to livelihoods, healthcare access and everyday life.

The second finding concerns the role of household vulnerability. Hunger, chronic illness and income loss were all strongly associated with poor self-reported health, with these relationships generally becoming stronger as the pandemic progressed. The increasing association between chronic illness and poor health is unsurprising given the disruption of routine healthcare services and medication access documented elsewhere in South Africa (Burger et al., 2020; Mboweni & Risenga, 2022). At the same time, the consistently strong relationship between household hunger and poor health reinforces the importance of food security as a fundamental determinant of wellbeing. Although income loss is frequently used as an indicator of vulnerability, our results suggest that direct measures of material deprivation, particularly hunger, are more closely associated with overall health. This finding is consistent with evidence that food insecurity intensified during the early stages of the pandemic and became an important source of both physical and psychological distress (Hunt et al., 2021; Wills et al., 2020).

The relationship between social grants and health is more nuanced. At the onset of the pandemic, respondents living in grant-receiving households were less likely to report poor health, suggesting that income support was associated with an important short-term protective role. By the later stages of the pandemic, however, this association was no longer statistically evident. Importantly, this pattern remained consistent across a series of robustness checks that controlled for baseline health, pre-pandemic socioeconomic status, lagged grant receipt and grant-type heterogeneity. While these additional analyses cannot eliminate all concerns regarding endogeneity, they increase confidence that the main findings are not simply the result of contemporaneous reverse causality.

Why did the association weaken over time? One explanation is that the purchasing power of transfers gradually eroded as households faced prolonged unemployment, rising food prices and continuing economic uncertainty. At the same time, access to healthcare became an increasingly important determinant of health. Cash transfers can relieve financial pressure, but they cannot compensate for interrupted treatment, delayed diagnosis or reduced access to medication. Household resilience may also have diminished as savings were exhausted and informal coping mechanisms became increasingly strained. Collectively, these findings suggest that the benefits associated with income support were greatest during the initial stages of the crisis but became less apparent as the pandemic evolved and structural constraints became more binding.

This interpretation is broadly consistent with the international literature. Brazil’s Bolsa Família programme, for example, has been associated with improvements in child health through its integration with routine health services (Rasella et al., 2013). More recently, Aurino and Giunti (2022) show that, in countries such as Niger, Pakistan and Somalia, cash transfers generated stronger health and nutrition outcomes when combined with complementary interventions, including nutrition promotion, improved water services and access to healthcare. Although South Africa’s social assistance system differs from these programmes in both design and objectives, the key lesson is that income support is likely to be most effective when delivered alongside other health and social interventions.

These findings have important implications for policy. The rapid expansion of the Social Relief of Distress grant demonstrated that South Africa is capable of extending income support quickly during periods of crisis. Yet the results suggest that cash transfers alone are unlikely to sustain improvements in health during prolonged emergencies. Strengthening resilience will require closer coordination between social protection and the health sector. Maintaining primary healthcare services, ensuring uninterrupted access to chronic medication and protecting nutrition programmes could be just as important as providing financial assistance.

The relevance of these findings extends beyond South Africa. Many middle-income countries expanded existing cash transfer programmes during the COVID-19 pandemic in an effort to shield vulnerable households from income shocks (Gentilini et al., 2022). Although institutional arrangements differ, our findings suggest that emergency cash transfers may provide valuable short-term support during periods of crisis. Maintaining those gains over longer periods, however, is likely to depend on complementary investments in healthcare, nutrition and other essential services.

Several limitations should be acknowledged. Self-reported health is inherently subjective and may be influenced by reporting behaviour, although the consistent measurement of the outcome across survey waves helps to minimise this concern. The study is also observational in nature and therefore identifies conditional associations rather than causal effects. While the inclusion of baseline controls, lagged grant measures and additional robustness checks substantially reduces concerns regarding observable confounding and reverse causality, unobserved time-varying factors cannot be ruled out. Finally, because NIDS-CRAM relied on telephone interviews, individuals without reliable access to communication technologies may be under-represented.

Despite these limitations, the study contributes new evidence on the evolution of overall health during the COVID-19 pandemic and the association between social protection and health in South Africa. More broadly, the findings suggest that protecting health during future crises will require integrated policy responses that combine income support with accessible healthcare, nutrition programmes and other complementary services.

7. Conclusions

This paper examined changes in self-reported health before and during the COVID-19 pandemic in South Africa, focusing on the role of household vulnerability and the association between social cash transfers and health. Using data from the National Income Dynamics Study (NIDS) and the NIDS-Coronavirus Rapid Mobile Survey (NIDS-CRAM), the analysis compared pre-pandemic and pandemic-period health outcomes and assessed how socioeconomic conditions were associated with poor self-reported health.

The findings show that overall health deteriorated substantially during the pandemic and remained persistently poor throughout the study period. Household hunger, chronic illness and income loss were consistently associated with poorer health, with these associations generally strengthening as the pandemic progressed. Although household grant receipt was associated with a lower probability of poor health at the onset of the pandemic, this protective association was no longer evident in the later stages of the crisis.

These findings suggest that while social cash transfers remain an important instrument for cushioning households against short-term economic shocks, they are unlikely to sustain health improvements during prolonged crises when implemented in isolation. Strengthening population health will require more integrated responses that combine income support with accessible healthcare, continuity of chronic disease management and effective food and nutrition programmes.

More broadly, the COVID-19 pandemic demonstrated that health outcomes are influenced by both economic and non-economic vulnerabilities. Building more resilient social protection systems will therefore require closer integration between social assistance and public health interventions to better protect vulnerable populations during future health and economic emergencies.