Quantifying the Impact of COVID-19 Relief Vouchers Schemes on Food Security: Empirical Evidence Insights from South Africa

Abstract

Livestock production systems are essential for sustaining household food security, especially in the drylands of Africa. This study assesses the impact of South Africa’s targeted Large Stock Unit (LSU) social protection program on the acute food insecurity effects of the COVID-19 pandemic among selected smallholder livestock farmers. An embedded research approach was utilized in four local municipalities purposively selected in the Northern Cape Province, where 217 households were selected using a stratified proportionate random method. A structured questionnaire was employed, while secondary data on beneficiary farmers were collected from implementing agencies. A full information maximum likelihood (FIML) Endogenous Switching Regression (ESR) model was adopted to capture variations due to self-selection bias among respondents. ESR model results show that the decision maker’s age, the household head’s education level, the land holdings’ size, average relative livestock losses, the orientation of production, and the level of external support impact food security. The study concludes that based on average treatment effects analysis, beneficiaries of the LSU program are better off in the household food security relative to their non-beneficiary counterparts. These findings validate the need for enhancing support initiatives during COVID-19 shocks for households to attain food security using their main livelihood sources as the gateway. Increasing the diversity of livelihood strategies in these vulnerable communities needs to be scaled up to protect households from acute food insecurity. Targeted support programs, including direct financing and binding networks, may also be supported through youth-sensitive training programs to enhance mitigations and resilience against COVID-19 acute food insecurity. A policy can tap into existing local structures and province-wide institutional platforms for the long-term sustainability of the LSU support initiatives and mitigation of COVID-19 food security vulnerabilities.

1. Introduction

The impact of COVID-19 on the lives of the billions of people affected by the pandemic is not limited to the direct threat that the virus poses to their health [1]. As articulated by Amare et al. [2], in addition to the direct health impacts, the pandemic has widespread effects on employment, poverty, food security, nutrition, education, and health, while disrupting the overall functioning of food systems. For increased coverage, this study captures any food item that falls into the seven food categories (vegetables; fruits; grains, legumes, nuts, and seeds; meat and poultry; fish and seafood; dairy foods; eggs). In response to the effects of COVID-19, governments have imposed a range of measures to contain the spread of the virus and “flatten the curve”. These immediate interventions include social distancing, restrictions on mobility, and temporary closure of most workplaces, generally known as a “lockdown” [3]. This matrix of strategies in the wake of COVID-19 is progressively destabilizing supply chains at all levels and creating instability in food supply and prices.

Food security is defined as a condition where all people always have physical, social, and economic access to sufficient, safe, and nutritious food that meets their food preferences and dietary needs for an active and healthy life [1]. COVID-19 undermines food security directly by disrupting food systems and indirectly through the impacts of lockdowns on household incomes and physical access to food. As such, COVID-19 could affect households’ food security through different pathways. For instance, COVID-19-related lockdowns and social distancing measures can adversely affect incomes by reducing economic and livelihood activities, which directly affects food security [4]. These lockdowns and restrictions also disrupt food supply chains and community services, including education-linked programs (e.g., school feeding) and social protection programs, which ultimately negatively affect food prices. It is thus problematic to note that the pandemic’s effects on prices show no consistent and predictable pattern. This was also reported by Barrientos and Hulme [5] using a more theoretical lens in relation to social support programs.

National and state-level restrictions and lockdowns affect food transportation within the country, with clear implications for food supply and food prices [6]. This is expected to significantly affect acute food insecurity, particularly in poorer and more vulnerable rural farming households. Developing countries with poor healthcare and more limited social protection systems are more likely to be badly affected. The pandemic is likely to push 49 million people into extreme poverty in 2020 [7]. More than 45% (23 million) of these people are in sub-Saharan Africa, implying that the region will be hit hardest in terms of increased extreme poverty and acute food insecurity vulnerabilities [8].

As informed by the aforementioned observations, there are at least three ways the pandemic may affect household food security, especially for smallholder farmers who are the unit of analysis (unit of analysis is the individual livestock farming household and, as such, the level of food security analysis is performed at the household level). First, individuals in some households may contract the virus, which will have direct economic impacts, such as loss of earnings, and indirect effects, such as the need to meet medical costs [9]. Even in the absence of direct infection, fear of contracting the virus could reduce income-generating activities. Second, government restrictions on movement and gatherings to slow the spread of COVID-19 has disrupted livelihood activities, reducing household incomes [2]. Third, access to food has been affected by market disruptions and food value chains [9]. However, empirical evidence on the magnitude of the impact of the pandemic on household food security remains limited, partly because the pandemic is still unfolding, and detailed household survey data are not available yet.

There is still very little formal analysis of the impact of COVID-19 policy interventions on food systems and household acute food insecurity in South Africa. The direct impacts of the lockdown policies, combined with associated macroeconomic shocks to household consumption, exports, and investment, have knock-on effects that spread through the entire economy [10]. Reduced activity in one sector has consequences for suppliers of intermediate inputs to that sector, who face lower demand, and for the users of the industry’s output, who face supply disruptions. Although several special issues are expected to be available over time, most current information comes from web-based material, grey literature, news, social media accounts, and first-hand observations. Projections show that, globally, the pandemic is likely to push 88–115 million people into extreme poverty by 2020, a third of these being from sub-Saharan Africa [10].

Globally, the pandemic is projected to double the number of people facing acute food insecurity by the end of 2020 [8]. Acute food insecurity is a situation when a person’s inability to consume adequate food exposes their lives and/or associated livelihoods to immediate danger [1]. The COVID-19 epidemic set off a chain of events that runs from rapid increases in the share of the population infected to increasingly widespread sickness with a positive and non-trivial probability of death to public health policies designed to contain the pandemic. Unfortunately, these public health policies greatly affect economic activity and income. These shocks are sufficiently large to push many households into positions of acute food insecurity, especially in the absence of countervailing government intervention policies. Increased acute food insecurity results primarily from the severe shock to household income and the means to purchase food rather than from a supply shock, such as drought [6].

Within this context, this study makes two contributions. First, it adds to the small but growing literature on the impact of the pandemic on household food security in sub-Saharan Africa, South Africa in general, and the Northern Cape Province in particular. Second, the study assesses the efficacy of policy interventions in mitigating these damaging impacts. Aggarwal et al. [4] note that since the pandemic outbreak, more than 150 countries and territories have implemented or announced plans to implement different policy interventions, such as COVID-19 relief vouchers and social protection measures, yet little is known about the efficacy of these interventions. Similarly, in the context of the COVID-19 pandemic, the mechanisms through which the social protection and safety nets benefit recipients are vague [1]. The main objective of the study is to assess the impact of South Africa’s COVID-19 relief vouchers program, the Large Stock Unit (LSU), on the adverse effects of the COVID-19 pandemic on the food security of farming households.

COVID-19 Relief Vouchers in South Africa

The principle of a large stock unit (LSU) is widely used in South Africa to determine the carrying capacity of farms [11]. It has dominantly been adopted in systems where the state aims to grant aid and, in the process, encourage farmers to protect grazing resources. In the LSU system, several dimensions, including the breed, gender, weight, age, and the animal’s physiological stage, are factored into the final decision on the LSU scheme grant to be allocated to the farmer [12]. Broadly, the veld quality at the farm is also a key determinant in the final LSU grant to be allocated to the farmers. All this is also done with some magnitude of variation depending on the contexts of the communities.

During the COVID-19 period, a number of relief grants were provided to households by the state to cushion them against the shocks induced by the pandemic [13]. Around April 2020, the LSU scheme was launched with the aim of supporting smallholder livestock farmers in competitively producing areas to sustain their activities. There were variations in the amount of the grant allocated to respective farmers, mainly guided by the number of LSUs available at the farm. The voucher system was adopted to reduce incidences of diverting the grant to other uses, given the exposure to which the households were vulnerable during the COVID-19 pandemic [12,13].

There were, however, several critiques about the manner that was adopted for administering the COVID-19 relief scheme for small-scale farmers, as it was designed and positioned to be adaptive over time [14]. The major notable challenge presented by stakeholders in South Africa was grounded on the government’s limited comprehension of farming and the food systems. In response to the COVID-19-related lockdown restrictions in South Africa, there was evidence that small-scale farmers lost access to markets and critical related services, as alluded to by the Institute for Poverty, Land, and Agrarian Studies (PLAAS) [15]. In extreme cases, some farmers lost access to their farms and livestock. Stakeholders felt that the government launched the relief programs (including the LSU scheme) specifically for small-scale farmers and provided subsidies for farming inputs to address the crisis. The adequacy of these COVID-19 relief vouchers has been presented; however, there is also a need to analyze the benefits associated with these interventions, with a focus on the LSU voucher scheme, as proposed by the current study. Given the success of similar schemes in, for example, Ethiopia and South Africa, lessons from the analysis will also shape future voucher-based interventions in times of shocks such as COVID-19 and beyond, with possibilities of including the private sector, as in the case of the Solidarity Fund Initiative [6,9].

2. Theoretical Framework

The systems thinking theoretical model was utilized to isolate the factors included in the ESR model. These theoretical arguments can effectively act as a means for explaining the various interconnected processes that determine the impact of COVID-19 policy interventions on food security. Three distinct subsystems were isolated: social (e.g., age, education), economic (e.g., savings), and institutional (e.g., credit, preparedness). The clusters were informed by previous approaches used by Béné [1], who reported social factors such as the age and farming experience of decision-making units as being critical in explaining acute food insecurity outcomes during COVID-19 in the wake of policy interventions. Amare et al. [2] also reported the importance of institutional arrangements in Nigeria in a COVID-19 study and extracted the critical role of credit for individuals and business organizations in supporting food security agendas. They argued that these intricate relationships were ultimately responsible for sustaining value chains even during times of shocks such as COVID-19, thus reducing acute food insecurity vulnerabilities.

Interestingly, guided by the systems thinking ideology, Aggarwal et al. [4] interrogated the market disruptions occurring during COVID-19 and traced the impacts of market support through credit lines as a strong candidate for determining the food security outcomes. They reported the critical role of a blended matrix of social, economic, and institutional drivers in shaping the success of COVID-19 policy interventions on acute food insecurity. Interestingly, a similar proposition was made by Arndt [6], who explored the connection between social and economic factors in determining the impact of COVID-19 lockdowns on income distribution and acute food insecurity. They also noted that there was a strong positive relationship between income and acute food insecurity as influenced by the COVID-19 lockdown restrictions. Informed by the above experiences within the COVID-19 realm and various policy interventions targeting acute food insecurity, the systems thinking framework becomes the most appropriate for isolating the variables of interest, as outlined in Table 2. This framework, as highlighted in Figure 1, allows for explaining the overlaps across the identified subsystems and associated variables in a way that helps to reshape the LSU policy intervention that was put in place to mitigate the effects of COVID-19 on acute food insecurity.

As depicted, COVID-19 policy interventions were influenced by a number of factors, such as social, economic, and institutional factors. Social, economic, and institutional factors determined the acute food insecurity of an individual household. The policy intervention was expected to reduce acute food insecurity. The main impact evaluation method, namely endogenous switching regression, was employed to quantify the impact of COVID-19 policy intervention on acute food insecurity. This study used a framework developed by the authors, as presented in Figure 1.

Three important clusters of variables, namely social (e.g., age of household head, experience in agriculture, the level of education, etc.), economic (e.g., savings), and institutional (e.g., credit, support, etc.) were incorporated in the framework in relation to acute food insecurity, as measured by the HFIAS. The framework, therefore, emphasized the impact of the COVID-19 LSU policy intervention on acute food insecurity, the impact of social, economic, and institutional factors on access to the LSU policy intervention variable, and how the aforementioned factors also affect the status of acute food insecurity among the households. A detailed description of the variables is depicted in Table 1.

In the past five years, a lot of effort has been channeled towards developing frameworks for guiding processes for reducing acute food insecurity during shocks. In this vein, several frameworks for tracing acute food insecurity have been proposed [2,4,9]. These are important in understanding how various policy interventions may affect acute food insecurity exposure during shocks and stresses such as agricultural droughts and COVID-19, thus identifying key leverage points for supporting and enhancing shock-absorbing programming activities by the state and non-state actors [1,2,9].

Patnaik and Das [16] observed inherent similarities across these frameworks. They noted that the orientation of these frameworks places the household as the unit of analysis, surrounded by numerous other pillars such as the social conditions, the economic space within the household, the institutional arrangements that the decision-making units may take advantage of, and how the various policy interventions during shocks can affect the extent of acute food insecurity. Bahta [17], however, reported that, in most instances, it is the resilience-building capabilities of the households, within the realm of the aforementioned social, economic, and institutional factors, that are of long-term importance in defining policy intervention pathways during shocks. However, among all these opinions, it is clearly noted that it is the responsibility of the government to respond through policy interventions to ensure that households are not vulnerable to acute food insecurity during shocks such as COVID-19.

The adopted framework is justified given that it pushes for analyzing the impact of COVID-19 LSU policy intervention through the lens of the variation in the extent of acute food insecurity across households in Northern Cape Province, South Africa, while acknowledging that the differences are caused by an array of cross-cutting and intertwined factors (Figure 1).

3. Materials and Methods

The study adopts a farm-level analysis using smallholder livestock farmers as the research unit since they are the principal decision-makers. Targeting them provided deep insights concerning the fundamental issues surrounding the LSU program and the possible relationship to food security conditions at the household level during the COVID-19 pandemic phase.

3.1. Study Site Description

The study was conducted in the Northern Cape Province, located in the northwest part of South Africa at coordinates 29.0467° S, 21.8569° E. The province has variable altitudes, the lowest being 0 m and the highest at 2156 m [18]. The province borders the Western and Eastern Cape provinces in the south and the Free State and North West provinces in the east and shares international borders with Namibia and Botswana. The province’s total surface area (361,830 km²) accounts for 30% of the country’s total area, and thus is the largest province by surface area; however, the province has the smallest population, representing 2% of the country’s population while also having the lowest population density of 2 persons/km² [18]. The Northern Cape’s administrative government is divided into 5 districts with 26 local municipalities. The Frances Baard District Municipality is located in the northeastern region and has 387,741 residents, who represent approximately 32.5% of the population, with the highest density. Figure 2 below shows the map of the Northern Cape Province and Frances Baard District Municipality.

The study area is a major livestock-producing zone and is vulnerable to acute food insecurity since it experiences temperatures averaging 18–40 °C and average annual rainfall as low as 20 mm in the west to approximately 300 mm in the east. Smallholder farmers mainly produce goats, sheep, and cattle for subsistence, with intermittent sales during droughts and other shocks. Mining activities are also common in rural municipalities as a safety net against failure in agriculture. Prolonged and persistent droughts are also common, and this makes dependence on agricultural livelihoods problematic [20]. Regardless of the aforementioned bottlenecks, agriculture still, directly and indirectly, employs about 45,000 people in the province and, therefore, accounts for 16% of the total formal employment base and 40.3% of the province’s economic activities.

The Northern Cape Province is a hub of South African agriculture and produces approximately 6% of the country’s total agricultural production, 12% of groundnuts, 10% of barley, and 16% of wheat. The province is characterized by diverse agricultural practices, with the table grape industry employing a significant number of people, and raisins are also a popular product in the province. Sheep farming is also a strategic subsector and accounts for nearly 25% of the province’s economy. The other major livestock enterprises include goats and cattle, which account for 9% each. Most of the commercial farms in the province are export-ready, and the commercialization of goat enterprises is an untapped business opportunity for new farmers. The province accounts for approximately 12% of the country’s commercial farms, with 4,829 farms. As a supporting base, the province dominates in terms of the country’s commercial agricultural land and accounts for a proportion of about 37% [18]. Smallholder farmers, however, remain marginalized from mainstream commercial activities and are, therefore, highly exposed to acute food insecurity vulnerabilities, especially during times of shocks such as COVID-19.

3.2. Sampling and Data Collection

An embedded mixed-methods approach was used for the study. Martey [21] also recommends this approach because it embraces both the qualitative and quantitative methods and, therefore, captures the depth and breadth of the interconnectedness of food security under shocks such as COVID-19 from the perspective of support programs initiated to reduce the effects on households. A cross-sectional survey was conducted using multi-stage samples of 217 participants to measure the impacts of a COVID-19 social support program on household food security. The selection of the Northern Cape Province, Frances Baard district municipality, and local municipalities was motivated by their dominance in agricultural activities, with a strong bias towards smallholder livestock production systems. These areas are also classified as acute food insecurity disaster areas by the South African government and, therefore, present the needed conditions for analyzing household acute food insecurity under the COVID-19 pandemic after the provision of a support program (LSU) targeting the main livelihood source, which is livestock. The Department of Agriculture, Forestry, and Fisheries (DAFF) provided a database of smallholder farmers, which showed that the Department of Agriculture, Forestry, and Fisheries (DAFF) directly assisted 553 (63%) smallholder livestock farmers through the LSU program, and the rest of smallholder livestock farmers, 325 (37%), did not benefit from the COVID-19 LSU intervention [22]. These farmers were included in the study as part of the population for the study, which included 878 livestock farmers. Table 1 below summarizes smallholder livestock farmers according to their local municipality.

Table 1. Summary of farmer composition based on the location.

Municipality	No. of Livestock Farmers in Municipality	Proportion of Farmers (%)	Farmers in Sample
Dikgatlong	351	40%	87
Magareng	120	14%	30
Phokwane	266	30%	65
Sol Plaatje	141	16%	35
Total	878	100%	217

Source: Northern Cape Department of Agriculture, Forestry, and Fisheries [20] and authors’ compilations.

Table 1. Summary of farmer composition based on the location.

Municipality	No. of Livestock Farmers in Municipality	Proportion of Farmers (%)	Farmers in Sample
Dikgatlong	351	40%	87
Magareng	120	14%	30
Phokwane	266	30%	65
Sol Plaatje	141	16%	35
Total	878	100%	217

Source: Northern Cape Department of Agriculture, Forestry, and Fisheries [20] and authors’ compilations.

The respondents were selected using the multi-stage sampling approach earlier presented. Guided by Glen et al. [23], this process facilitated the separation of large populations into targeted clusters for effective analysis of characteristics and outcomes while accounting for heterogeneity among farmers, if there was any. This was achieved by accommodating the differences in, for example, the population of farmers across municipalities by using a proportionate sampling technique to generate a representative sample. The sample size formula by Cochran [24] was used and adjusted to 217 smallholder farmers using the Cochran [24] correctional method because the initial value of 288.83 that was generated from the unadjusted computation exceeded 5% of the population, and as a rule of thumb, there was a need for adjustment. This procedure then generated a more reliable sample size that reflected the structure of the population. Using face-to-face interviews, primary data were gathered from smallholder LSU program beneficiary farmers in four district local municipalities, namely Dikgatlong, Magareng, Phokwane, and Sol Plaatje, using a structured questionnaire. The core data themes targeted demographics of the households, the various determinants of LSU share received, food consumption patterns such as the number of meals per day or the number of nights spent without eating food, and indicators of probable acute food insecurity. The data were used to trace households’ differential food security conditions based on the value of LSU support that was accessed. This will be integral to informing the (re)formulation of context-appropriate food security stabilizing strategies in the wake of COVID-19 (and other pandemics) induced shocks while acknowledging the role of LSU programs in supporting the core livelihood strategy. The primary data collected from the smallholder livestock farmers were then captured in the STATA 13 statistical program, cleaned, coded, and analyzed as guided by the research objectives.

Social, economic, and institutional factors were included in the analysis and related to acute food insecurity, as proxied by the HFIAS (Figure 1). The social factors are presented as the core drivers of the local level conditions for the specific household, as well as the extent to which they relate to its socio-cultural spaces as it copes with acute food insecurity. These define the internal subsystems of the decision-making processes of the individual household [25]. The economic factors are grounded on the ability of the household to generate and/or utilize value as a way of cushioning against shocks [6]; for example, the savings a household can use in times of production and marketing disruptions. This was measured by whether or not the household had savings that they could utilize during emergencies such as COVID-19. The institutional factors are grounded on the rules which define how the household relates to the various pillars along the livestock value chain [17]. These included access to credit facilities and were measured depending on whether the household had access to adequate credit facilities or not.

The total LSU scheme value received by each beneficiary household in the Northern Cape Province was calculated and operationalized in Equation (1):

$$\mathrm{Total}\mathrm{LSU}\mathrm{scheme}\mathrm{value}=\mathrm{Average}\mathrm{amount}\mathrm{per}\mathrm{LSU}\times \mathrm{Number}\mathrm{of}\mathrm{LSU}$$

(1)

where the “total LSU scheme value” is the total amount in Rand (South African Currency) received under the LSU special COVID-19 grant in April 2020; “LSU” is the magnitude of the scheme units, and “average amount per LSU” is the fixed R195 that was provided for each unit. Since livestock production was reported by the Department of Agriculture, Land Reform, and Rural Development of South Africa (DALRRD) [26] as being the mainstay of households in the Northern Cape Province, the study had a strong hypothesis that the COVID-19 phase LSU scheme package was well targeted to boost livestock production and could catalyze the food security status of these farmers, both directly and indirectly. The variable was measured directly from records provided by DALRRD in relation to each farmer’s access to the LSU support program.

The dependent variable, which depicts the extent of acute food insecurity, was computed using the household acute food insecurity access scale (HFIAS) as reported by Coates et al. [27] and presented in Equation (2).

$$\mathrm{HFIAS}={{\displaystyle \sum }}_{\mathrm{i}=1}^n{X}_i{F}_i$$

(2)

where “$\mathrm{HFIAS}$” is the score, “$X_i$” is the observed acute food insecurity occurrence, and “$F_i$” is frequency of the associated occurrence. The “HFIAS” was utilized in the study as guided by Musara and Musemwa [28], who noted that the scale is a strong measure of acute food insecurity. It is a continuous recall access metric that isolates the extent of the household’s acute food insecurity status over the past 30 days. Eight distinct categories of occurrences were isolated in the study.

“1 = Anxiety about food (in) adequacy; 2 = Eating foods of a limited variety; 3 = Eating less-preferred foods; 4 = Inability to eat even the less-preferred foods; 5 = Eating smaller meals than needed; 6 = Eating fewer meals in a day; 7 = Going to bed hungry; 8 = Failing to obtain food of any kind during the whole day or night”.

The progression from 1 to 8 shows the increasing levels of acute food insecurity. A binary response was depicted as being yes (1) and no (0), depending on whether any of the stated 8 categories were encountered by the household. To further operationalize the measure, a severity question based on the frequency of occurrence was assigned as a follow-up. A scale was developed as, “1 = rarely, 2 = sometimes, and 3 = often”, and this then implied that the range for the HFIAS was 0–24.

The measure was justified and utilized in the study since it captures numerous dimensions of food consumption-related activities during the COVID-19 conditions while targeting the household’s responses to acute food insecurity in relation to the LSU scheme and other factors in the household. A household whose HFIAS score is high also has high levels of acute food insecurity.

The instrumental variable included in the ESR model, the coping strategy diversity index, was computed using the Shannon–Weiner diversity index to generate a ‘$\mathrm{CDI}$’, which was estimated using Equation (3).

$$\mathrm{CDI}= -{{\displaystyle \sum }}_{\mathrm{i}=1}^k\left(P_{{\mathrm{strategy}}_i}\right).\ln \left(P_{{\mathrm{strategy}}_i}\right)$$

(3)

where “$\mathrm{CDI}$” is a measure of the coping strategies diversity against acute food insecurity in the farming community during COVID-19, “$k$” is the total number of coping strategies used by households in the community during the COVID-19 period, and $P_{{\mathrm{strategy}}_i}$ is the proportion of strategies used by a household relative to the total number of strategies used in the farming community. The diversity of coping strategies is critical in defining the balance in managing acute food insecurity and access to the COVID-19 LSU policy intervention. The instrumental variable was selected as guided by the socio-economic–institutional arrangements in the study and supported by a resilience study conducted by Bahta [17], where it was reported that, based on the agricultural drought resilience index, the number of coping strategies was critical in determining the exposure to acute food insecurity. Di Falco [29] also alluded to the need to control for coping strategy diversity in the model when they conducted a climate change adaptation strategies impact study. This theoretical grounding supports the selection of the instrumental variable. Within the same arguments and given the importance of livestock in the livelihoods of the communities under study, it was also critical to control for the livestock disposal rate, since this also directly determined the value of the LSU policy intervention, as well as the status of acute food insecurity.

To assess the impact of the LSU beneficiary differentials on acute household food insecurity under COVID-19, the study compared the acute food insecurity outcomes between two LSU categories of ‘non-beneficiary’ and ‘beneficiary’ households. The approach used in dissecting the food security concept allowed the study to operationalize the food security outcome, which was measured using the HFIAS. This outcome measure accounts for a wider food security measurement range, thus allowing the classification of households according to their level of acute food insecurity status conditions (stated as food insecurity vulnerabilities in the study) as guided by the classifications developed by Coates et al. [27]. The data on the incremental LSU scheme benefit attributed to the COVID-19 pandemic period was collected from the support programming implementing agencies who also informed the aforementioned classification of beneficiaries. The concept of classifying COVID-19 grant benefits is also applied by other researchers in similar studies, such as Abay [9] in a Productive Safety Net Program (PSNP) study conducted in Ethiopia. This is critical when reflecting on the benefits associated with livelihood-enhancing programs, especially in times of shock. As such, the nature of these interventions needs to be factored into the analysis to circumvent the problems of selection bias triggered by the truncated observed distributions of the outcomes [30].

The study determines the possible factors influencing the farmers’ classification beyond the number of livestock units owned and the resultant impact on the household food security during the COVID-19 pandemic-affected period. These variables are presented and described in Table 2 and supported by the framework in Figure 1. The endogenous switching regression (ESR) model was adopted because it accounts for the selection bias problem. The ESR is a generalization of Heckman’s selection correction method and accounts for the selection of the unobservable by capturing selectivity as an omitted variable limitation [31].

Table 2. Description of variables included in the ESR model and descriptive statistics.

			Mean for Beneficiary Categories
Variable	Variable Description	Unit	Total Sample	Beneficiary	Non-Beneficiary	Test Value
Dependent variable
HFIAS	A continuous variable of the household acute food insecurity access scale	number	7.43	8.54	6.28	−1.263 *
Status	Dummy variable showing the beneficiary status of the household (0 = No, 1 = Yes)	dummy		0.63	0.37	1.457 *
Explanatory variables
Age	Continuous variable for age of household head	years	51.37	53.57	48.27	−2.796 ***
Education	Continuous variable for the adult mean number of years in education	number	7.80	7.46	8.28	1.361
Household	Continuous variable of the active family members	number	5.18	5.15	5.22	0.185
Experience	Continuous variable of cumulative experience years in livestock farming	years	10.87	11.87	9.44	−2.015 **
Land size	Continuous variable of the total land holdings for the family	hectares	2.43	2.66	2.09	−1.563
Credit	Dummy variable for access to credit facilities (0 = No, 1 = Yes)	dummy	0.13	0.13	0.13	−0.011
Savings	Dummy variable for usable savings during emergencies (0 = No, 1 = Yes)	dummy	0.19	0.21	0.16	−1.055
Livestock loss	Continuous variable for percentage of loss in livestock in last 12 months	%	2.13	2.06	2.189	−0.448
Orientation	Dummy variable for orientation of agricultural production (0 = subsistence, 1 = market)	dummy	1.50	1.54	1.46	−0.789
Preparedness	Dummy variable for perceived preparedness to shocks (0 = not prepared, 1 = prepared)	dummy	0.72	0.69	0.77	1.195
Support	Dummy variable for whether the farmer got any form of external support	dummy	0.25	0.26	0.23	−1.707 *
Instrumental variables
Coping diversity	Dummy variable for diversity of coping strategies (0 = low, 1 = high)	dummy	1.90	2.27	1.71	−1.849 **
Livestock sales	Dummy variable of selling livestock as the main coping strategy (0 = No, 1 = Yes)	dummy	0.72	0.85	0.69	−1.928 **

Source: Authors’ own computation. Notes: *; ** and *** indicate p-values significant at 1%, 5%, and 10% levels, respectively; t-test was used for continuous variables and chi-square for categorical variable. ANOVA test was used to test whether there is a statistically significant difference among the 3 subgroups.

Table 2. Description of variables included in the ESR model and descriptive statistics.

			Mean for Beneficiary Categories
Variable	Variable Description	Unit	Total Sample	Beneficiary	Non-Beneficiary	Test Value
Dependent variable
HFIAS	A continuous variable of the household acute food insecurity access scale	number	7.43	8.54	6.28	−1.263 *
Status	Dummy variable showing the beneficiary status of the household (0 = No, 1 = Yes)	dummy		0.63	0.37	1.457 *
Explanatory variables
Age	Continuous variable for age of household head	years	51.37	53.57	48.27	−2.796 ***
Education	Continuous variable for the adult mean number of years in education	number	7.80	7.46	8.28	1.361
Household	Continuous variable of the active family members	number	5.18	5.15	5.22	0.185
Experience	Continuous variable of cumulative experience years in livestock farming	years	10.87	11.87	9.44	−2.015 **
Land size	Continuous variable of the total land holdings for the family	hectares	2.43	2.66	2.09	−1.563
Credit	Dummy variable for access to credit facilities (0 = No, 1 = Yes)	dummy	0.13	0.13	0.13	−0.011
Savings	Dummy variable for usable savings during emergencies (0 = No, 1 = Yes)	dummy	0.19	0.21	0.16	−1.055
Livestock loss	Continuous variable for percentage of loss in livestock in last 12 months	%	2.13	2.06	2.189	−0.448
Orientation	Dummy variable for orientation of agricultural production (0 = subsistence, 1 = market)	dummy	1.50	1.54	1.46	−0.789
Preparedness	Dummy variable for perceived preparedness to shocks (0 = not prepared, 1 = prepared)	dummy	0.72	0.69	0.77	1.195
Support	Dummy variable for whether the farmer got any form of external support	dummy	0.25	0.26	0.23	−1.707 *
Instrumental variables
Coping diversity	Dummy variable for diversity of coping strategies (0 = low, 1 = high)	dummy	1.90	2.27	1.71	−1.849 **
Livestock sales	Dummy variable of selling livestock as the main coping strategy (0 = No, 1 = Yes)	dummy	0.72	0.85	0.69	−1.928 **

Source: Authors’ own computation. Notes: *; ** and *** indicate p-values significant at 1%, 5%, and 10% levels, respectively; t-test was used for continuous variables and chi-square for categorical variable. ANOVA test was used to test whether there is a statistically significant difference among the 3 subgroups.

3.3. The Endogenous Switching Regression (ESR)

The study utilized a two-step estimation procedure to fit the ESR model. In the first stage, the livestock farmers’ inclusion in the scheme category (‘non-beneficiary or ‘beneficiary’) was estimated using the Probit model, thereby generating inverse Mills ratios while accounting for the unobserved heterogeneity [32]. Given that the classification of the households depends on the interaction of the farmers in the community and the farm’s specific attributes (e.g., farm size), the farmer’s self-selection determines the LSU beneficiary classification based on the aforementioned characteristics and not by random assignment.

In the second stage, the outcome equation was estimated by including the inverse Mills ratios as an additional explanatory variable to control the selection bias problem. The study utilized the full information maximum likelihood (FIML) estimation method. This approach was also successfully used by Di Falco et al. [29], and it concurrently estimates the Probit selection equation and the regression equations, thereby yielding consistent standard error values. By doing this, the outcome function (food security status) is discretely estimated for the two LSU COVID-19 policy intervention beneficiary classifications (’non-LSU beneficiary’ and ‘LSU beneficiary’). This captures the endogenous nature of these classification outcomes. Based on the outcome variable and the explanatory variables ‘${\mathrm{X}}_{\mathrm{i}}$’, the two classification regimes were estimated as:

$${\mathrm{C}}_i = 1\left(z_i\gamma + u_i >0\right),$$

(4)

$$\mathrm{Regime}1:Y_{0\mathrm{i}}{ = \mathrm{X}}_{0\mathrm{i}}{\beta }_0{+\mathsf{\epsilon }}_{0\mathrm{i}}{\mathrm{when}\mathrm{C}}_i=0\left(\text{‘}\mathrm{non}\text{-}\mathrm{LSU}\mathrm{beneficiary}\mathrm{category}\text{’}\right)$$

(5)

$$\mathrm{Regime}2:Y_{1\mathrm{i}}{=\mathrm{X}}_{1\mathrm{i}}{\beta }_1{+\mathsf{\epsilon }}_{1\mathrm{i}}{\mathrm{when}\mathrm{C}}_i=1(\text{‘}\mathrm{LSU} \mathrm{beneficiary}\mathrm{category}\text{’})$$

(6)

In the above sequence, the selection Equation (6) represents the regime, $z_i$ is avector of regressors explaining the probability of a farmer classified in the LSU beneficiary category, and $u_i$, ${\mathsf{\epsilon }}_{0\mathrm{i}}$ and ${\mathsf{\epsilon }}_{1\mathrm{i}}$ are the error terms. Since the farmer’s classification outcome may be endogenous and influenced by the sample selection criteria, the mean value for the correlation of the error terms ${\mathsf{\epsilon }}_{0\mathrm{i}}$ and ${\mathsf{\epsilon }}_{1\mathrm{i}}$ is non-zero [33]. In this way, if the parameters (${\beta }_1{\mathrm{and}\beta }_2$) are estimated using OLS estimation procedures, it would lead to sample selection bias, which according to Lee [34], is also called the problem of missing variables.

The study also observed that, due to the sampling methods that were used, the conditions of beneficiary and non-beneficiary cannot be observed at the same time, and therefore, the covariance between ${\mathsf{\epsilon }}_{\mathrm{i}}{}_1$ and ${\mathsf{\epsilon }}_{\mathrm{i}}{}_0$ cannot be determined with accuracy. Informed by this assumption, the mean values for ${\mathsf{\epsilon }}_{1\mathrm{i}}$ and ${\mathsf{\epsilon }}_{0\mathrm{i}}$ account for the the inverse Mills ratio $\lambda (·)$ and defined in Equation (7) as:

$${\lambda }_1 =\frac{Ø\left(z_i\gamma \right)}{f\left(z_i\gamma \right)}\mathrm{if}\left({\mathrm{C}}_i=1\right)\mathrm{and}{\lambda }_0 =\frac{Ø\left(z_i\gamma \right)}{1-f\left(z_i\gamma \right)}\mathrm{if}\left({\mathrm{C}}_i=0\right)$$

(7)

where $Ø$ and $\varphi$ are the probability distribution function (pdf) and cumulative distribution function (CDF) for the standard normal variable, respectively. Given the study area conditions, it is possible to encounter endogeneity for the LSU classification on the food security outcome. To address this, a deep review of the literature was done to assist in selecting covariates to include in the analysis [35,36]. The study also borrows from Abdulai and Huffman [33], and valid instrumental variables were included in the selection equation to guarantee identification. For identification, the hypothesis followed in the study was that the probability of a farmer being classified in the LSU beneficiary category is an increasing function of the prior exposure to similar social grants, as shown by the two selection instruments captured in the model as the diversity of acute food insecurity coping strategies and the whether the farmer depends on selling livestock as the major coping strategy. The rejection test was used to determine the validity of these selected instruments, as informed by Abdulai and Huffman [33], by testing whether the instruments affected the classification and not the household food security outcomes for the non-LSU beneficiary households. Results from the analysis show that they can be utilized as valid selection instruments.

The coefficients of the ESR model were used to determine the Average Treatment Effect (ATE) by estimating the impact of the LSU beneficiary classification on food security. The Average Treatment Effect on the Treated (ATT) was utilized to compute the difference between the food security status of the households in the LSU beneficiary category and their values, assuming they were placed in the non-LSU beneficiary category. This was estimated as the difference between Equations (5) and (6). The Average Treatment Effect on the Untreated (ATU), a measure of the difference between the household food security of the non-LSU beneficiary category and the associated counterfactuals, was also adopted. These estimates will sufficiently account for the selection bias, unlike the absolute differences in the two LSU categories, as shown in Table 2. The conditional expected value of the outcome variables in Equations (8) and (9) were presented as:

$$E\left(Y_{1\mathrm{i}}|{\mathit{x}}_i{,\mathrm{C}}_i=1\right){=x}_i{\beta }_1{+\mathsf{\rho }}_1{\lambda }_1\left(z_i\gamma \right)$$

(8)

$$E\left(Y_{0\mathrm{i}}|{\mathit{x}}_i{,\mathrm{C}}_i=1\right) {=x}_i{\beta }_0{+\mathsf{\rho }}_0{\lambda }_0\left(z_i\gamma \right)$$

(9)

Borrowing from the application by Mujeyi et al. [37], the ATT was also computed by Equation (10):

$$E\left(Y_{1\mathrm{i}}| {\mathit{x}}_i{,\mathrm{C}}_i=1\right)-\mathrm{E}\left(Y_{0\mathrm{i}}|x_i{,\mathrm{C}}_i=1\right){=x}_i\left({\beta }_1{-\mathsf{\beta }}_0\right) {+\mathsf{\rho }}_1{\lambda }_1{-\mathsf{\rho }}_0{\lambda }_0$$

(10)

In Equations (4)–(6), the component $E\left(Y_{0\mathrm{i}}|x_i{,\mathrm{C}}_i=1\right)$ is the expected value of food security ‘$Y_{\mathrm{i}}$’, assuming that the household had not been classified in the LSU beneficiary category. Di Falco et al. [29] report that this is the estimated unobserved contribution of the LSU beneficiary differential, while the component $E\left(Y_{1\mathrm{i}}|{\mathit{x}}_i{,\mathrm{C}}_i=1\right)$ denotes the actual expected value for the household food security status.

4. Results

4.1. Descriptive Analysis

Table 2 displays descriptive summaries derived from sampled farming households and highlights their absolute differences among the LSU beneficiary and non-LSU categories. The results in Table 2 show that the sampled farmers had, on average, 2.43 ha of land, which did not vary significantly according to the LSU beneficiary classification. In determining the relative magnitude of the instrumental variable, the acute food insecurity coping strategy diversity index was computed, and the extent of dependency on livestock sales to fend off acute food insecurity during the COVID-19 period was estimated. The diversity index variable was significantly higher for the households in the LSU beneficiary cluster (2.27) relative to their counterparts in the non-LSU beneficiary cluster (1.71); the total sample average for this variable was 1.90.

Regarding livestock disposal, a binary response was used to show whether the household agreed (1) or disagreed (0) with the notion that they depended on livestock disposal strategies during periods of shocks, including COVID-19, to sustain food security. The response rates for these respective conditions were 85% and 69%, with a sample average of 72%. The age of the household head was 53 years for the LSU beneficiary category, 48 years for the non-LSU category, and 51 years for the total sample. There was no significant difference between the non-LSU beneficiary clusters for the household size variable, and the results in Table 2 also show that the average duration in the schooling of the respondents was eight years, which was not significantly different between the LSU beneficiary statuses. Additionally, households in the LSU beneficiary category were reported to have sizes averaging five; those in the non-LSU beneficiary category had five family members, while the total sample average was five.

Table 2 also shows that there were significant differences in the level of the farmers’ experience in managing livestock production systems across the 2 LSU beneficiary clusters, with those in the non-LSU beneficiary cluster having an average of 9 years and those in the LSU beneficiary category having 11 years, while the sample mean was 10 years. The level of external support affects the extent to which households become vulnerable to acute food insecurity during shocks such as COVID-19. Results from Table 2 show that households that received various forms of external support during the COVID-19 period were domiciled in the LSU beneficiary category with a score of 0.26, while the score was 0.23 in the non-LSU beneficiary category. The total sample average for the support variable was 0.25.

Results from Table 2 show that the average HFIAS for the total sample is reported to be 7.43. In absolute terms, the farming households who were classified under the LSU beneficiary cluster reported significantly (p < 0.1) higher HFIAS by a magnitude of 2.26, i.e., (8.54–6.28). It is also reported that 63% of the sampled households were beneficiaries of the COVID-19 LSU intervention, while the balance of 37% was classified as non-beneficiaries.

4.2. Endogenous Switching Regression (ESR)

Using STATA 13 software and the full information maximum likelihood method, the extent of LSU support and the associated HFIAS outcome equations were jointly estimated.

Table 3. Full information maximum likelihood results for the food security ESR model.

	Selection Equation (LSU Support)	Outcome Equation (HFIAS)
		LSU Beneficiaries	LSU Non-Beneficiaries
Variable	Coefficient	Coefficient	Coefficient
Age	2.03 ** (1.012)	−1.849 *** (0.076)	0.329 (0.237)
Education	−0.136 ** (0.065)	0.409 (0.369)	0.421 * (0.221)
Household	0.192 (0.130)	0.494 (0.548)	0.124 (0.553)
Experience	−0.415 (0.567)	−1.516 (2.561)	−0.609 (0.884)
Land size	−0.205 (0.163)	0.435 (0.423)	−0.032 *** (0.007)
Credit	−0.273 ** (0.129)	−0.277 (0.646)	−0.589 (0.417)
Savings	0.153 (0.359)	0.361 (0.493)	0.289 (0.233)
Livestock loss	−1.375 (1.539)	0.205 (0.215)	1.004 *** (0.349)
Orientation	0.166 *** (0.044)	−0.414 *** (0.151)	3.569 * (1.966)
Preparedness	−1.375 (1.575)	0.022 (0.07)	0.016 (0.021)
Support	−3.522 * (1.876)	−4.354 ** (2.059)	−0.0762 (0.075)
Coping diversity	0.040 *** (0.014)	0.739 ** (0.293)	−0.604 (0.889)
Livestock sales	−0.134 *** (0.055)	−0.322 (0.257)	1.278 *** (0.579)
Constant	−2.616 * (1.499)	0.151 ** (0.069)	1.850 *** (0.079)
rho0	0.083 (0.287)
rho1	−0.478 (0.296)
/lns0	−0.415 *** (0.151)
/lns1	0.446 ** (0.225)
/r0	−1.111 *** (0.359)
/r1	0.379 (0.382)
Wald chi2 (11)	76.69 ***
Log likelihood	−382.684
LR test	6.32 **
No. of obs.	217

Source: Authors’ computation. Notes: *; ** and *** indicate p-values significant at 1%, 5%, and 10% levels, respectively; z-values estimated on robust standard errors in parenthesis.

presents results from ESR capturing the selection equation and the equations for the two regimes (equations 5 and 6), which are highlighted in the earlier sections of the paper. The first column shows the selection equation derived from the Probit model.

Results from Table 3 show that the instruments are uncorrelated with the dependent variable (HFIAS) and are highly significant (p < 0.01) in both selection models. It can therefore be concluded that these instruments are valid. A strong positive co-relationship with the LSU categorization outcome indicates that the smallholder livestock farmers with higher coping diversity for absorbing the shocks of the COVID-19 pandemic had higher probabilities of being located in the LSU beneficiary cluster and were, hence, more food secure. On the contrary, smallholder farmers who depend on livestock sales were less likely to be in the LSU beneficiary category due to their higher disposal rates, which the onset of the COVID-19 pandemic compounded. Based on the selection criterion shown in the first column of Table 3, the most important factors affecting the classification into either the ‘non-LSU beneficiary’ or ‘LSU beneficiary’ categories are the household head’s age, education level attained by the household head, the orientation of production at the farm, access to credit, and the magnitude of external support. Since the factors influencing the classification were not the paper’s mainstay, these results are not discussed in this paper but can be accessed in an associated supporting paper.

The results in the second and third columns of Table 3 show the two regime equations of non-LSU beneficiary and LSU beneficiary classification. The age of the household head is a highly important determinant in the LSU beneficiary regime in relation to the HFIAS. Older decision-making units in the LSU beneficiary category are relatively more food secure as opposed to their counterparts in the non-LSU beneficiary cluster and are, therefore, less exposed to the acute food insecurity vulnerabilities of COVID-19. The results in Table 3 also show that for the non-LSU beneficiary regime, the education level has a positive and significant effect on the acute food insecurity status of households. This implies that decision-making units with higher education but in the non-LSU beneficiary cluster have significantly higher chances of food insecurity. The land size variable has a negative and significant effect on farmers’ food insecurity status in the non-LSU beneficiary cluster. As per prior expectations, when the land size available for the farmer increases, the acute food insecurity conditions also decrease. Table 3 shows that the value of percentage livestock losses is also reported as an important factor when the considerations about the food security outcome (HFIAS) as triggered by COVID-19-induced LSU interventions are made. For the farmers in the non-LSU beneficiary cluster, the higher the livestock loss, the higher the acute food insecurity status. This signals the positive impact of the COVID-19 policy intervention implemented in the study area.

The production orientation was also captured in the study area as informed by the response from the farmers. The orientation of production is a significant consideration for both beneficiary and non-beneficiary cluster households. However, the direction of effect for the two regimes is different and is negative in the former group and positive in the latter. The implication is that households in the non-LSU beneficiary category who are more oriented toward market-based production have significantly higher (p < 0.05) acute food insecurity outcomes. On the contrary, those farmers in the LSU beneficiary category who are market-oriented have lower incidences of acute food insecurity. Results in Table 3 show that the level of external support, as shown by a transformed variable capturing direct food support, access to assets support, and funding support from relatives, shows a significant and negative relationship with the acute food insecurity status of households in the Northern Cape Province of South Africa.

The treatment effects estimates for the COVID-19 policy intervention classification (non-LSU beneficiary and LSU beneficiary) on the food security for the sampled households are presented in

Table 4. Average treatment effect of LSU differentials on food security.

Treatment Effect Index	Household Food Insecurity Access Scale (HFIAS)
	Estimate	Robust Std. Err.	z Value
Average Treatment Effect on the Treated (ATT)	0.288	0.106	2.72 **
Average Treatment Effect on the Untreated (ATU)	−0.204	0.109	−1.87 *
Average Treatment Effect (ATE)	0.556	0.128	4.34 ***

Source: Authors’ computations. Notes: *; ** and *** indicate p-values significant at 1%, 5%, and 10% levels, respectively.

.

The Average Treatment Effect on the Treated (ATT) estimates the difference between the observed food security values for the LSU beneficiary households (the treatment) and their values, assuming they were in the non-LSU beneficiary cluster. The findings of the ATT show that the food security for the treated group of livestock farming households is positive at 0.288 and statistically significant (p < 0.05). The implication is that the farmers in the LSU beneficiary category would have been acutely food insecure had they received the COVID-19 policy intervention support in the non-LSU beneficiary category. Using the Average Treatment Effect (ATE) outcomes, results from Table 4 show that the households in the non-LSU beneficiary categories would have achieved food security benefits if they had been classified under the LSU beneficiary category. The Average Treatment Effect on the Untreated (ATU) is a measure of the difference between the food security of those in the non-LSU beneficiary category and the counterfactuals, and the estimates account for the selection bias. Results show that the ATU is negative, with a value of −0.204. The results show that the farmers in the LSU beneficiary category would have been worse off if they had not received COVID-19 policy intervention benefits. This also implies that those in the non-LSU beneficiary category of benefits would have benefited if they had been provided with LSU support during the COVID-19 pandemic. All of this points towards the positive food security impact of the policy intervention in the study area.

5. Discussion

This section presents the discussion of the study findings. Also, it links experiences in the study area and ideas from the literature body regarding the impacts of various support programs initiated at the onset of COVID-19, and other similar shocks, to arrest the effects on acute food insecurity vulnerabilities.

From a policy perspective, the findings on the HFIAS differences point toward the need to structure interventions around the LSU scheme to reduce household acute food insecurity vulnerabilities. This was also observed by Bahta and Enoch [38] in a smallholder vegetable policy intervention program in South Africa, where beneficiary households with higher support were also relatively less vulnerable to acute food insecurity. Ng’ombe [25] also supported this argument and noted income gains from a direct policy intervention on livelihoods. This can be inferred as a very strong argument for the need to scale up the COVID-19 policy intervention in greater South Africa.

In terms of the age variable differences, the observed patterns in the study area can be associated with the synergetic effects of older decision makers’ ability to effectively utilize the LSU scheme provisions to specifically target food security-enhancing functions through livestock, for example, supplementary feeding projects in the community cooperatives. Poczta-Wajda et al. [39] also observed that, in Poland, the older smallholder farmers who received policy intervention packages were relatively better off in food security. They argued that the conservative nature of these groups of farmers could be critical in sustaining cautious utilization of the scheme packages, thus resulting in beneficial effects in the short run. A different long-run perspective within the age variable realm was presented by Mujeyi et al. [37] in a climate-smart agriculture study, where younger farmers who received more policy intervention support during risky operating conditions exhibited better outcomes in fending off acute food insecurity vulnerabilities. These inconclusive observations make the current study necessary in the discussion around the effect of the age variable on the magnitude of the outcomes, given variations in the extent of support extended to households under unpredictable conditions such as COVID-19 and other shocks. From a policy perspective, this calls for the need to design age-sensitive interventions, creating a meeting point between the existing indigenous technical knowledge and modern science if policy interventions such as the LSU COVID-19 scheme are to be effective.

The study’s findings pinpoint the importance of higher levels of LSU support in sustaining food security, regardless of the respondents’ education. This also agrees with findings by Patnaik and Das [16] in a study conducted in India, and by Baiyegunhi et al. in Nigeria [40], where policy intervention support designed around single parameters did not support adaptive capacity for smallholder farmers. Instead, a broader recommendation captured a review of the education curriculum, functional capacity building, and targeted scheme grants in the wake of the persistent flooding in the area. A similar trajectory can also be embraced in the current study setting if the benefits of education are to be intertwined with the physical capital in terms of accessing the primary factors of production using, for example, the LSU package during pandemics such as COVID-19. This motivates the extended distribution of the scheme to other communities of South Africa in an effort to reduce exposure to acute food insecurity

The findings in relation to the land size can be attributed to the observations from the study area showing that the nature of the grazing systems is such that the livestock grazes on communal land; thus, regardless of the LSU received, farmers can still utilize the land for other agricultural activities and be food secure. This result is fundamental in explaining the nature of the COVID-19 support that has to be provided to the farmers, which must not be purely over-focused on livestock enterprises as the core livelihood option, but also expand to other activities at the farm. Thus, the policy practitioners may need to repackage the LSU support framework and blend it with other peripheral support tools to be effective in the Northern Cape Province of South Africa. There is evidence from similar previous studies which also show these patterns [41,42].

The deaths of livestock–food security patterns in the study area can directly indicate the farmers’ tendencies to reallocate the LSU benefits to other food security supporting functions such as more livestock. Amare et al. [2] also reported similar findings using panel data in Nigeria. They reported that smallholder farmers have a higher proclivity to virement scheme benefits to more preferred livelihood strategies as guided by the prevailing conditions and the associated opportunity costs. From a policy perspective, this may also inform the need to restructure the basis of LSU amounts allocated to the farmers and move away from the available LSU as the sole determinant. The inadequacy of the LSU benefits can be one factor triggering the observed results for the livestock losses variable in the current study. Issues such as the breeds and the expected feeding costs, among others, need to be factored into the LSU support models. This is especially so during pandemics such as COVID-19, when households are further constrained by other needs such as health expenses and may not be able to bridge the LSU inadequacy gap, especially when they are in the non-LSU beneficiary cluster. If this is done, then the policy intervention can be more effective in addressing the vulnerability of households to acute food insecurity.

Suresh et al. [36] supported the use of farmer responses in terms of production orientation since their responses are directed by intrinsic needs and expectations of their production systems. This finding is in direct agreement with findings presented by Abay et al. [9], who reported reduced acute food insecurity gaps for the top-end beneficiaries of a scheme grant program in Ethiopia in terms of value. The significant long-term possibility of the study’s observation can be reversed given the market volatility because of COVID-19-induced disruptions. There are higher chances that the over-dependence on the market mechanisms cannot be sustained in the long term under COVID-19 conditions ceteris paribus. A similar conclusion was made by Aggarwal et al. [4] in a market disruptions study during COVID-19 in Malawi and Liberia, where policy interventions were effective in supporting livelihoods. This again further points to the need to blend the current LSU package with other market-supporting strategies to enhance food security among those participating in markets, even during pandemics.

The LSU beneficiary households who also received higher external support from relatives and community members showed higher signs of food security. This reinforces the possibility that these various forms of support can have a synergetic effect on the food security status, especially during pandemics such as COVID-19, when self-reliance for most households would be compromised. Arndt et al. [6] also presented an income distribution and food security argument in a COVID-19 study done in South Africa, where they laid down the need for adequate policy intervention packages to reduce the effect of lockdowns. If this approach is adopted, then adequate LSU packages can also be provided to the smallholder livestock farmers to sustain food security in the wake of COVID-19 while depending on their core livelihood strategy based on livestock systems. This matrix of outcomes can be caused by the households’ ability to combine various coping strategies and thus preserve other assets, including livestock, as suggested by Sinyolo [43] and Bahta [17].

The general results, as presented in Table 3, show that the provision of LSU among smallholder livestock farmers significantly affects households’ food security. This is a strong indicator of the need to formulate adequate and appropriate policy schemes that target the most vulnerable households and reduce the incidences of acute food insecurity. Given the characteristics of households in the study area, this is fundamental in considering realigning these schemes to the specific needs of the households and going beyond the lumped LSU package, which can be appropriate for some households but not others. The finding is consistent with similar studies conducted during the COVID-19 pandemic, which also reported benefits from access to various forms of support [6,9].

Of note is the report by Abay [9] in Ethiopia under the Productive Safety Net Program (PSNP), which noted that households who were supported through COVID-19 policy intervention were relatively more food and nutrition secure than their non-intervention receiving counterparts. Based on a panel data series in Nigeria, the findings presented by Amare et al. [2] also established a strong relationship between food security and access to high levels of COVID-19 targeted policy interventions among acute food insecurity vulnerable households. From a policy implication perspective, this indicates the role of COVID-19 policy targeted interventions initiated in smallholder farming communities as long-term safety nets for acute food insecurity. Given that food security stability is the fundamental objective for these households, it becomes indispensable to foster the provision of adequately higher LSU support since they directly benefit households’ food security during COVID-19.

6. Conclusions

South Africa is among several sub-Saharan African (SSA) countries that are aiming to revive the smallholder livestock sub-sector and build acute food insecurity resilience among producers so that they remain food secure even during times of shocks such as COVID-19. An increasing proclivity towards commercialization and networking policies that promote productivity gains provides enormous opportunities for aggressive vertical and horizontal integrations in competitively advantageous production zones such as the Nothern Cape Province of South Africa. This paper has significantly contributed to this debate by using the ESR model to examine the impact of COVID-19 targeted LSU policy support on food security among smallholder livestock farming households in the Northern Cape Province of South Africa as a gateway to mitigating COVID-19 effects. The sampling method used in the study was robust and included 217 smallholder livestock farmers.

First, the ESR model results support the conclusion that a number of factors, including the age of the principal decision-making unit in the household, the education status of the household head, the size of the land holdings for the household, average relative livestock losses, the orientation of production and the level of external support from relatives and community members influence the acute food insecurity outcome. These different factors are fundamental in explaining the differences in acute food insecurity among various households. This shows the interaction of multiple clusters of factors that are located in the social, economic, and institutional subsystems in determining the status of acute food insecurity for the household and thus supports the application of the systems thinking theory as adopted in this study. These findings are consistent with earlier studies and offer empirical evidence insights into policy interventions during shocks such as COVID-19 that significantly reduced acute food insecurity in vulnerable communities. Understanding these household and community-wide characteristics is therefore concluded to be important in enhancing the impact of the COVID-19 policy intervention in reducing incidences of food security vulnerabilities. It is also concluded that targeting these dimensions within the context of non-LSU beneficiary and LSU beneficiary classifications provides substantial insights into the design of the COVID-19 policy interventions in terms of adequacy, appropriateness, and implementation strategies.

Second, the results show that policy intervention programs, specifically South Africa’s LSU policy intervention, mitigated the impacts of COVID-19 on acute food insecurity. Specifically, based on the average treatment effects analyses findings, the study concludes that, on average, the beneficiary farmers in the LSU beneficiary category are better off in the food security component, as shown by the positive coefficient (0.288) for the ATE category and for the ATT category (0.288). The protective role of the LSU program was significantly (p < 0.05) higher for the beneficiaries. In this regard, it is therefore concluded that there are positive food security benefits that can be derived from the COVID-19 policy intervention in South Africa. In general, this is a strong conclusion on the positive contribution and effectiveness of the COVID-19 policy intervention in mitigating acute food insecurity during the pandemic, where 63% of the farmers benefited from the LSU policy intervention.

Finally, the study results also show that it is strategic for efficient policy intervention programs to be functional even during periods of stability (i.e., where there are no shocks such as COVID-19) so as to cushion vulnerable households from acute food insecurity. This can motivate the policy practitioners to scale up the LSU support incentives among smallholder farmers in the marginal areas of South Africa, especially during pandemics such as COVID-19, and curb acute food insecurity vulnerabilities. Reports by Amare [2] in Nigeria, Aggarwal [4] in Liberia and Malawi, Abay [9] in Ethiopia, and elsewhere by Béné [1] suggest that the support which farmers get from COVID-19 policy interventions can lead to lower incidences of acute food insecurity.

The study findings are fundamentally relevant in South Africa, where there are higher risks of the pandemic (or other shocks such as droughts) resurfacing, and thus provide empirical insights to the government as it weighs alternative policy interventions for supporting acute food insecurity in vulnerable households and build their resilience.

The study recommends that improving strategies that pay direct attention to enhancing age-sensitive production capacities, developing stronger networking among farmers, and building decision-making resilience within community structures can reduce acute food insecurity by boosting access to resources and markets during shocks such as COVID-19. Building on the aforementioned conclusions, it is recommended that multi-faceted locally designed interventions target reducing cattle losses and encourage market-oriented livestock production systems while building on the socio-economic and institutional meeting points. Specifically, there is a need to design a double-edged strategy that directly supports production and marketing strategies at their interface, especially in the wake of uncertainties presented by COVID-19. These interactive platforms should also embrace a range of age groups, including the youths, to expand the horizon of the food security benefits while utilizing well-targeted and adequate LSU interventions in the livestock-dominant zones.

Although the theoretical approach presented here is designed to address many of the complex relationships between acute food insecurity and COVID-19 policy interventions, it is not entirely comprehensive. Going forward, a national-scale analysis of the LSU scheme can be conducted to provide the patterns in the heterogeneous contexts of various smallholder farmers. Similar approaches, using a diverse set of acute food insecurity metrics to represent the social, economic, and institutional economic and other criteria, can be used to evaluate the impact of COVID-19 policy interventions on acute food insecurity among households. Using a combination of metrics from various disciplines should also provide scope for decision makers to understand the broader picture and prioritize the variables to interact with to efficiently utilize limited resources to support farmers during shocks such as COVID-19.

The study acknowledges that most of the steps suggested above require data collection through processes that may be costly to conduct. It is therefore suggested to utilize available secondary data from food surveys and existing livestock value chain analysis reports while focusing the primary data collection on critical data gaps using, for example, participatory techniques. In this regard, future research can also use secondary panel data to trace the temporal effects of various COVID-19 policy intervention phases within the LSU model’s framework while expanding the orientation to accommodate other core livelihoods policy supporting interventions in agriculture in an effort to build acute food insecurity resilience. The authors are confident that this article can be instrumental in contributing to enhancing policy intervention dimensions in acute food insecurity research among livestock-producing communities while promoting dialogue on the research and policy platforms.