Food poverty, hunger and household production in rural Eastern Cape households

ABSTRACT

More than two decades since the advent of democracy in South Africa, the place of small-scale agriculture in rural development, poverty alleviation and food security remains ambiguous and highly contested. However, there is now some new evidence that official income poverty estimates in South Africa may be underestimating the contribution of rural, land-based livelihoods when measuring household well-being. This paper aims to explore this possibility further by identifying how household production activities are associated with improved food security among rural Eastern Cape households in the former homelands. The analysis is based on data from Statistics South Africa’s 2008/9 Living Conditions Survey and its annual General Household Surveys. In adopting a food poverty lens, the findings suggest that hunger levels are lower among farming households in the Eastern Cape even though a higher percentage of these households (relative to non-farming households) live below the national food poverty line. The paper concludes by discussing some implications for policy.

KEYWORDS:

• Food security
• poverty
• agriculture

1. Introduction

More than two decades since the advent of democracy in South Africa, the place of agriculture in rural development, food security and poverty alleviation remains ambiguous and highly contested. This is due, in part, to the uncertainty surrounding land reform processes (and outcomes), a gradual process of de-agrarianisation, a poverty literature which has tended to underestimate the contribution of home production to household livelihoods and a number of fundamental (and often ideological) disagreements on the proper or ‘potential’ role of agriculture in poverty reduction.

Within the income poverty literature there is a particular concern that the size of the small-scale agricultural sector tends to be underestimated in South Africa. Meth (2006: 17), somewhat pithily, suggests that the under-estimation of subsistence agricultural activities in the country’s national household surveys explains the failure of these surveys to identify the livelihood strategies of the ‘relatively large number of people with the apparent ability to survive on thin air’. From a measurement perspective, one of the preoccupations in the South African poverty literature has been the large numbers of households which report ‘zero-income’ in household surveys and in the census and the ways in which poverty analysts must account for these households when estimating poverty levels and trends (Vermaak, 2012). While some (such as Meth, 2006) have speculated that home agricultural production may be a survival strategy in these zero-income households, a recent analysis of subjective poverty (Posel & Rogan, 2016) has provided some new evidence that official income poverty estimates in South Africa may be underestimating the contribution of rural, land-based livelihoods when measuring household well-being.

At the same time, the broader development literature is divided on the outcomes, if any, associated with household production in South Africa. On the one hand, a number of studies have found evidence indicating the importance of small-scale agriculture to food security (Pauw, 2007; Aliber & Hart, 2009; Baiphethi & Jacobs, 2009; Pienaar & von Fintel, 2014), nutrition (Kirsten et al., 1998; Hendriks, 2003; van Averbeke & Khosa, 2007) or household income (Baiphethi & Jacobs, 2009; Pauw, 2007). On the other hand, others (Sender, 2002, 2012; De Swardt, 2003; Bradstock, 2005; Misselhorn, 2005; Palmer & Sender, 2006) have argued that small-scale farming is not a viable, sufficient or desirable solution to rural poverty or household food insecurity. This latter body of work has suggested that the idea of efficient small-scale household farming activities is a fallacy (Sender & Johnston, 2004) and that, even at a subsistence level, households that farm are no less dependent on market sourced foods than are non-farming households (Palmer & Sender, 2006). This stance seems to be supported by the popular notion that the penetration of large supermarket chains into rural parts of the country has impacted on small-scale agriculture (du Toit & Neves, 2007; Murisa, 2013).

These debates have a direct relevance for policy as illustrated by a recent suggestion (see Sender, 2014) to place a moratorium on support for ‘food self-sufficiency’ interventions in South Africa’s Eastern Cape province. This recommendation for the withdrawal of support to smaller farmers has been suggested in favour of a strategy which ‘backs the winners’ in the form of large-scale and export oriented agriculture in order to revive rural wages (Palmer & Sender, 2006; Sender, 2012).¹ Proponents of this approach often cite a lack of evidence for the contribution of small farmers to either agricultural output or employment (Sender, 2012). The main thrust of this recommendation is that interventions or support packages aimed at home producers or community gardens are over-romanticised and tend to serve as a distraction to some of the real challenges associated with rural development (Sender, 2014).

The aim of this paper is to contribute to these debates by exploring whether and how household production activities are associated with improved food security with a particular focus on the Eastern Cape.² In addressing this objective I use a food poverty lens to understand how households in rural areas who are identified as ‘food poor’ are able to achieve food security and whether agricultural production plays a role.

The remainder of the paper is structured as follows. The following section reviews the literature on small-scale or subsistence farming in South Africa and links this body of work with a parallel literature on household food security. Section three identifies the definitions, data sources and methods that I use to analyse household production, food poverty and food security. In section four, the results are presented in two parts. I begin with a descriptive analysis of food poverty levels and food security (proxied by self-reported hunger levels) in both South Africa and the Eastern Cape. Next, I present the results of a series of multivariate analyses which explore whether and how household production activities are associated with food security. Finally, section five discusses the findings in relation to the existing literature as well as some of the recent policy recommendations on small-scale and subsistence agriculture.

2. Household agricultural production in South Africa

Despite the widespread claims of de-agrarianisation in South Africa, one of the few empirical studies (Aliber & Hart, 2009) of subsistence agriculture has suggested that the actual number of these farmers has been consistent and may actually have increased (between 2001 and 2007). This estimate comes from the erstwhile bi-annual Labour Force Surveys and suggests that roughly four million black South Africans are involved in subsistence agriculture. However, there is some inconsistency, even in the empirical literature, on this point and a second study, based on data from the National Income Dynamics Study (NIDS), has provided evidence for a decrease in subsistence activities between 2008 and 2012 (Daniels et al., 2013). An important caveat here is that definitions of subsistence agriculture or household production are not consistent in the literature or across different household surveys. The other caveat, of course, is that the two studies described above are also analysing different time periods.

In terms of provincial estimates, most studies identify the Eastern Cape, Limpopo and KwaZulu-Natal as the provinces with the highest levels of household production or small-scale agriculture. Aliber & Hart (2009), for example, show that the Eastern Cape has the highest provincial share of black ‘own account’ farmers. There are only a handful of studies, however, which have considered trends in small-scale agriculture in the province. Bank (2005), in citing a number of surveys and case studies conducted over a twenty year period, suggests that there has been a decrease in household production in the Eastern Cape over this period. He argues that the decline of family farming in the province is the result of structural shifts in the economy, rural livelihoods and household formation over a longer period of time. Andrew & Fox (2004), on the other hand, argue that there has not necessarily been a decrease in small-scale agriculture but rather a change in cultivation practices (i.e. an increase in home gardens)—albeit the evidence is from a single case study (see also Fay, 2013).

2.1. Household production and food security

While the extent of (and changes in) household production among South African and Eastern Cape households is not clear, there does appear to be a consensus on the reasons that households engage in farming activities. In particular, there is now a well-documented trend towards household production as an additional food source (as opposed to a main source of food or as a source of income) (Baiphethi & Jacobs, 2009). Relatedly, there are two recent studies which show that these activities are associated with improved food security. First, Pienaar & von Fintel (2014) find that small-scale farming activities in the former homelands have had an impact on both household income and food security. They show that households that report the sale of farm products as their main income source are significantly less likely to experience hunger relative to households that rely on wages. Moreover, their analysis also suggests that, in the former homeland areas, farming as a main food source is significantly correlated with lower hunger risks relative to non-homeland households. Within the homeland areas, larger farming activities as well as cultivation in school and community gardens seem to have a particularly important role in food security and the authors suggest that this might be due to knowledge pooling and risk sharing (Pienaar & von Fintel, 2014). The key conclusion from the study is that household production played at least some role in the convergence in household food security between homeland and non-homeland regions in South Africa in the 2000s. However, they show also that social grants (and particularly the old age pension) and government loans (specifically for farming) are important protectors against hunger among farming households in the former homelands (Pienaar & von Fintel, 2014).

Second, von Fintel & Pienaar (2015) extend their own work on food security by examining the role that social grants play in supporting small-scale farming to reduce hunger. In this second paper they find that the state’s old age pension enables small-scale farming and that this, in turn, significantly reduces the risk of child hunger in these households (von Fintel & Pienaar, 2015). The main contribution of this analysis is, therefore, that household production appears to be the channel through which social grant income is used to protect against food insecurity in some of the poorest households in the country.

2.2. Household agricultural production and income

With regard to home production activities and income generation, the evidence is more mixed. Most notably, two studies (Pauw (2007) and Palmer & Sender (2006)) based on data from the 2000 Income and Expenditure Survey (IES) have arrived at very different conclusions about the role and importance of small-scale agriculture in South Africa. While most of the key findings from these two papers are consistent, one difference is that Pauw (2007) suggests, based on the finding that black commercial farmers have income levels similar to those of black non-agricultural households (and higher than subsistence and farm worker households), that the key strategy for rural development should be some ‘degree of commercialisation’. In contrast, Palmer & Sender (2006) have argued that most farming households do not earn any income from their production and those that supplement their diets (i.e. through own consumption) are just as reliant on market-sourced food as non-farming households. Palmer & Sender (2006) therefore argue that, for the poorest rural households, household production is not a way to escape from poverty.

Adding to this specific debate, Aliber (2009) using the 2005/6 IES offers evidence that the gap in household food share expenditure between rural and urban households (at each income decile) is likely explained by ‘self-provisioning’. He finds that rural households in the lower half of the income distribution spend 15% less on food than their urban counterparts (Aliber, 2009). Provincial differences are also significant with Eastern Cape households reporting some of the lowest levels of per capita expenditure on food in the country (Aliber, 2009). Moreover, while income poverty is very high among black agricultural households, the difference between agricultural and non-agricultural poverty (among black households) is actually much smaller in the Eastern Cape and Limpopo (compared with other provinces and South Africa more broadly) (Pauw, 2007).

3. Data and methods

In order to explore further these connections between income poverty, food security and household production, the analysis in this paper makes use of data from the 2008/9 Living Conditions Survey (LCS) and, at the time of analysis, the most recently available (2012) General Household Survey (GHS). These two nationally representative household surveys both collect relatively comprehensive information on household agricultural activities, food security and household income or expenditure. It is therefore possible, using these two data sources, to compare directly two measures of household well-being (i.e. household food security and income poverty) with agricultural activities.

Household food security is measured using two questions in the GHS on how often either an adult or child in the household has experienced hunger in the past 12 months. The questions on hunger from the LCS are worded identically to the questions from the GHSs. The only difference is that there are five response options in the GHSs (1 = Never; 2 = Seldom; 3 = Sometimes; 4 = Often; 5 = Always) and only four in the LCS. The option of ‘sometimes’ is missing from the response list in the LCS. In measuring income poverty, households are identified as food poor if total household per capita income is below Statistics South Africa’s food poverty threshold of R305 (in March 2009 prices). This is the level below which Statistics South Africa estimates that a household cannot meet even the basic minimum food requirements of all of its members (Statistics South Africa, 2012). In 2008, about 26.3% of all South Africans lived in a household which was identified as food poor (Statistics South Africa, 2012). The Eastern Cape stands out from the country as a whole and is the province with the second highest incidence of food poverty (36% of the province’s population lives in a household which is unable to meet its basic food requirements).

The GHS and the LCS provide fairly similar estimates and suggest that between two million and 2.4 million South African households (African only) are engaged in some type of agricultural production.³ Both data sources also identify this as 21% of all African households. As expected, Eastern Cape households, and those in the tribal authority areas (i.e. deep rural and largely consisting of the former homelands) are far more likely to be engaged in agriculture. Between 38% and 42% of African households in the province reported agricultural production in the GHS and the LCS, respectively. Perhaps most significantly, both data sources also identify a clear majority of households in the tribal areas of the Eastern Cape as being involved in agriculture (58% and 65%, respectively).

4. Findings

4.1. Household production, food poverty, and hunger

Table 1 begins by analysing the risk of acute hunger⁴ and, for greater context, also identifies the percentage of households below the national poverty line.⁵ As expected, the descriptive statistics suggest that the risk of poverty is far higher in the Eastern Cape (62% of households) and particularly in the tribal areas (78%) compared with South Africa as a whole (48% of South African households are poor). This finding is well established in the literature as the Eastern Cape has consistently been identified as one of the poorest and most materially deprived of the nine provinces (Statistics South Africa, 2012).

Table 1. Household well-being in the Eastern Cape, 2008/9 and 2012.

	Poverty headcount (GHS-z = 577)	% of African HHs w/adult hunger (GHS)	% of African HHs w/child hunger (GHS)	% of African HHs w/adult hunger (LCS)	% of African HHs w/child hunger (LCS)
Eastern Cape	62.25 (1.65)	2.94 (0.34)	1.34 (0.23)	10.90 (0.64)	7.97 (0.56)
Eastern Cape tribal areas	78.40 (1.43)	2.01 (0.38)	1.21 (0.28)	13.72 (0.95)	10.86 (0.87)
South Africa	48.06 (0.69)	3.21 (0.15)	1.60 (0.10)	7.58 (0.21)	5.00 (0.17)

Notes: The data are weighted.

Source: Own calculations from the 2012 GHS and the 2008/9 LCS.

The link between income poverty and hunger, however, is not clear. While income poverty rates are far higher in the Eastern Cape (and particularly in the tribal areas) the risk of either adult or child hunger is actually lower in the tribal areas of the province. For example, only 2% of African households in the tribal areas reported a period of adult hunger in the past 12 months compared with 3.2% of African households in the country as a whole. The evidence from the GHS would therefore suggest that rural households in the Eastern Cape are far more likely to be income poor but are not necessarily more likely to experience acute hunger.

One concern with the measurement of hunger, however, is that the incidence measured by the GHS is considerably different than the risk of hunger as measured by the LCS. For example, the estimate (from the LCS) for adult hunger in the Eastern Cape tribal areas is nearly seven times higher (e.g. 2% compared with 13.7%) than the estimate from the GHS (and the two surveys were undertaken only about three years apart). More importantly, however, the LCS data suggest that the risks of both adult and child hunger is significantly higher in the tribal areas of the Eastern Cape (compared with the province and country as a whole). This is the opposite finding from the GHS which shows that African households in the rural parts of the province were relatively food secure despite their higher levels of income poverty.⁶

Table 2 deepens the analysis by linking food poverty and hunger levels with household production. A first glance at the table would suggest that living in a household below the official food poverty line is not a clear proxy for hunger or food insecurity. The prevalence of food poverty is far higher than the incidence of acute hunger in all geographic areas and according to both data sources. The interesting findings, however, are in the differences in the risk of food poverty and hunger between farming and non-farming households and between different geographic areas. For example, the risk of food poverty is higher⁷ for farming households than for non-farming households across all geographic areas. The gap, however, is much smaller in the tribal areas of the Eastern Cape where the differential in the risk of poverty between farming and non-farming households is not statistically significant (according to both the GHS and LCS).

Table 2. Food poverty and hunger among African farming households, 2008/9 and 2012.

	Poverty headcount (P0) (z = 305)		Percentage of households experiencing ‘acute’ hunger
	Farming households	Non-farming households	Farming households	Non-farming households
	GHS 2012
Eastern Cape	44.49 (1.70)	27.96 (1.50)	1.79 (0.42)	3.80 (0.48)
Eastern Cape tribal areas	46.37 (1.80)	40.46 (2.61)	1.86 (0.46)	2.66 (0.68)
South Africa	40.67 (0.79)	23.13 (0.55)	3.73 (0.31)	3.22 (0.18)
South African tribal areas	44.49 (0.88)	37.33 (1.03)	3.94 (0.34)	3.51 (0.30)
	LCS 2008/9
Eastern Cape	42.41 (1.91)	25.22 (2.87)	12.49 (1.04)	10.52 (0.84)
Eastern Cape tribal areas	44.05 (1.98)	41.27 (3.45)	13.48 (1.17)	15.85 (1.71)
South Africa	37.56 (1.17)	22.90 (0.89)	10.56 (0.51)	7.16 (0.24)
South African tribal areas	41.39 (1.24)	38.53 (1.48)	11.11 (0.60)	9.63 (0.47)

Notes: The data are weighted. Standard errors are in brackets.

Source: Own calculations from the 2012 GHS and the 2008/9 LCS.

The main finding from the table, however, is that the risk of hunger is actually lower for farming households in the Eastern Cape (relative to non-farming households). This finding holds across both the GHS and LCS data sets for Eastern Cape tribal areas while the LCS data suggest that there are no significant differences (i.e. 95% level confidence intervals do not overlap) in the risk of acute hunger between farming and non-farming households in the province as a whole. Both data sources therefore suggest that the Eastern Cape (and the tribal areas in particular) is somewhat unique in that the incidence of hunger is actually lower in farming households, even though the risk of food poverty is similar or higher than in non-farming households.

Perhaps the best way to identify whether the same households which are food poor are able to avoid hunger is to compare income poverty with hunger directly. While the estimates in the previous tables compared aggregate measures of poverty and hunger, Table 3 examines the overlap between income poverty (at the food poverty line) and hunger by identifying the extent to which households which are below the poverty line are able to avoid hunger. Given the sensitivity associated with how hunger is measured, the table presents three measures⁸ of hunger (i.e. acute, medium and expanded hunger). Starting with acute hunger, the key finding is that, in the Eastern Cape, food poor households that report agricultural activities are at a significantly lower risk than non-farming households (3.3% and 10%, respectively). In other words, among households with the lowest levels of measureable income, households that engage in home production are three times less likely to suffer from acute hunger (in the province as a whole).

Table 3. Hunger levels among food poor African households, 2012.

	Percentage of households experiencing ‘acute’ hunger		Percentage of households experiencing ‘medium’ hunger		Percentage of households experiencing ‘expanded’ hunger
	Farming	Non-farming	Farming	Non-farming	Farming	Non-farming
Eastern Cape	3.32 (0.87)	10.03 (1.45)	26.05 (2.12)	37.39 (2.54)	35.63 (2.33)	44.58 (2.63)
Eastern Cape tribal areas	3.23 (0.90)	4.38 (1.40)	26.35 (2.22)	36.49 (3.56)	36.37 (2.44)	44.71 (3.64)
South Africa	6.57 (0.63)	7.79 (0.56)	21.48 (0.98)	24.16 (0.87)	31.71 (1.09)	33.62 (0.95)
South African tribal areas	6.44 (0.65)	6.00 (0.65)	20.42 (1.02)	23.67 (1.23)	30.39 (1.15)	33.94 (1.34)

Notes: The data are weighted. Standard errors are in brackets.

Source: Own calculations from the 2012 GHS.

When ‘medium’ levels of hunger are considered, the differences between farming and non-farming households in the Eastern Cape become more evident. Among food poor households in the province that farm, just over a quarter reported a medium risk of hunger compared with about 37% of non-farming households (and the risks of hunger in the tribal areas for both farming and non-farming households are almost identical). At the same time, there is no significant difference in the risk of medium hunger between farming and non-farming households in the rest of South Africa or in the tribal areas across the country. Finally, the third column of the table suggests a similar finding for the risk of ‘expanded’ hunger levels. Among food poor households, farming households in the province have a lower (but not significantly so) risk of hunger when defined more broadly.

4.2. Household production, food poverty, and hunger in a multivariate context

Given the way in which farming households in the Eastern Cape seem to differ in their ability to avoid hunger despite their higher food poverty levels, this final section examines the risk of hunger in a series of simple multivariate estimates. Similar to the analysis by Pienaar & von Fintel (2014), I estimate a series of linear probability models (ordinary least squares) in order to identify whether and how household production may be correlated with food security. In the first set of specifications, I consider whether households that engage in agricultural activities report lower levels of hunger. The second set then narrows the sample to households which fall below the food poverty line.

Table 4 begins by showing that, as expected, low income is significantly associated with hunger. The positive coefficient in the first column of the table (1) confirms that households which are below Statistics South Africa’s food poverty line are also more likely to experience hunger. Moreover, after controlling for income (i.e. food poverty), there is no significant difference in the risk of hunger for households that engage in agricultural activities (2). In other words, farming households are no more or less likely to report experiencing hunger (after controlling for income). Since there are provincial and geographic differences in the risk of hunger and because income sources may be an important correlate of hunger (over and above the actual level of income), the third specification (3) adds in controls for geographic location as well as the main sources of household income.

Table 4. The correlates of household hunger (medium) among African households, 2012.

	(1)	(2)	(3)	(7)
Food poor	0.142*** (0.007)	0.143*** (0.007)	0.113*** (0.008)	0.112*** (0.008)
Home production		−0.003 (0.006)	−0.0143* (0.007)	0.002 (0.009)
Eastern Cape			0.034*** (0.008)	0.015 (0.011)
Tribal area			−0.024*** (0.007)	−0.032*** (0.007)
Main source of income
Business			0.036*** (0.012)	0.035*** (0.012)
Remittances			0.043*** (0.012)	0.041*** (0.012)
Pensions			−0.046** (0.018)	−0.047*** (0.018)
Grants			0.094*** (0.008)	0.105*** (0.010)
Farm sales			0.326** (0.158)	0.314** (0.159)
Other			0.006 (0.013)	0.006 (0.013)
No income			0.374*** (0.048)	0.375*** (0.048)
Interactions
EC*Home production				−0.025 (0.020)
EC*Tribal				0.053*** (0.019)
Home production*grants				−0.034** (0.015)
_cons	0.091*** (0.003)	0.091*** (0.003)	0.074*** (0.004)	0.074*** (0.004)
F stat	378.04	190.57	54.55	43.05
Prob > F	0.00	0.00	0.00	0.00
R^2	0.035	0.035	0.058	0.059
N (unweighted)	19 920	19 920	19 920	19 920

Notes: The data are weighted. Standard errors in parentheses. The omitted category is ‘salary/wages’ as the main source of household income.

Source: Own calculations from the 2012 GHS.

***Significant at the 99.9% confidence level.

**Significant at the 95% confidence level.

*Significant at the 90% confidence level.

There are several interesting findings from the full (main effects) specification in column three. The first is that, after controlling for province, area type and the main sources of income, farming households are significantly less likely to report hunger. There are several likely explanations for the home production coefficient becoming significant in the third specification. First, the reference category for the second group of variables (main source of household income) is income from salaries or wages. As expected, access to wages is an important protector against hunger and households that rely more on income from businesses (often informal), remittances and social grants are all significantly more likely to experience hunger. Once the regression controls for access to wages, the coefficient on home production therefore becomes significant and negative. This suggests that one of the key disadvantages for farming households, in terms of hunger risks, is the relative lack of wage income (see also Palmer & Sender, 2006).

One of the other interesting findings from the third specification in Table 4 is that, after controlling for the other variables, households that report the sale of farm products as their main source of income are significantly more likely to report hunger. This is the opposite finding from Pienaar & von Fintel (2014) who show a significant protective effect from the sale of farm products. The difference could be explained by the fact that the specification controls also for income and province while conditioning on home production. The most likely explanation for the finding, however, is that the reference category is wage income (and farming carries more risk than working for a regular wage) and, after controlling for other factors (i.e. income), the risk of hunger is greater for households that rely on farm income.⁹ Therefore the regression results suggest that engaging in surplus home production¹⁰ as a main source of household income is not a strong protector against hunger, holding all else constant.

The fourth regression (4) adds in a number of interaction terms in order to identify whether there are specific risks of hunger over and above those identified in the main effects regressions (1–3). The interactions suggest two further possible correlates of hunger. Over and above the variables included in the main effects specifications, households in the tribal areas of the Eastern Cape face a particular risk of hunger (but only after controlling for home production). However, households that engage in home production and report that social grants are the main source of income face a significantly lower risk of hunger. This interaction is particularly interesting since it suggests that combining home production with a regular (albeit small) source of income may be independently correlated with a lower level of hunger (over and above the main effects specifications). While the estimation strategy that I employ here cannot shed further light on this claim, it is interesting that the findings reported by von Fintel & Pienaar (2015) identify home production as a mechanism through which social grants improve food security.

Given the obvious importance of income in protecting against hunger, Table 5 extends the analysis by looking specifically at the poorest households (i.e. those below the food poverty line). The first two columns of the table (1 and 2) include all African households below the food poverty line while the next two columns (3 and 4) consider only Eastern Cape African households (below the food poverty line). The main effects specification for the entire sample (1) suggests that, after controlling for a number of factors, food poor households that engage in home production are significantly less likely to report hunger. As highlighted in the previous table, households that rely on wages more than any other income source have a significantly lower risk of hunger. Once again, households that report that the sale of farm products is the primary source of household income are significantly more likely to report hunger (relative to households that report wage income). Therefore, while home production seems to offer some protection against hunger among the poorest households in the country, relying on farming as a primary income source does not. While the first regression also suggests that Eastern Cape households face a higher risk of hunger (relative to households from the rest of the country), the interaction term between the Eastern Cape and home production (2) has a significant negative coefficient. This might suggest that home production has an additional protective effect against hunger for poor households in the Eastern Cape.

Table 5. The correlates of hunger among food poor African households, 2012.

	South Africa		Eastern Cape
	(1)	(2)	(3)	(4)
Home production	−0.0281* (0.0146)	−0.00733 (0.0156)	−0.113*** (0.0332)	−0.118*** (0.0375)
Eastern Cape	0.107*** (0.0182)	0.129*** (0.0343)
Tribal area	−0.0401*** (0.0149)	−0.0473*** (0.0157)		0.00344 (0.0424)
Main source of income
Business	0.123*** (0.0388)	0.121*** (0.0388)		−0.0411 (0.0915)
Remittances	0.0411** (0.0199)	0.0380* (0.0199)		−0.0938 (0.0609)
Pensions	−0.0544 (0.0438)	−0.0521 (0.0439)		−0.0966 (0.143)
Grants	0.114*** (0.0176)	0.113*** (0.0176)		0.00908 (0.0562)
Farm sales	0.497** (0.228)	0.483** (0.229)
Other	0.0162 (0.0244)	0.0154 (0.0243)		−0.109 (0.0798)
No income	0.389*** (0.0553)	0.388*** (0.0553)		0.404*** (0.121)
Interactions
EC*Home production		−0.109*** (0.0406)
EC*Tribal		0.0398 (0.0446)
_cons	0.174*** (0.0142)	0.173*** (0.0145)	0.374*** (0.0255)	0.394*** (0.0535)
F stat	15.64	13.27	11.70	4.72
Prob > F	0.00	0.00	0.01	0.00
R^2	0.042	0.044	0.015	0.041
N (unweighted)	5727	5727	882	882

Notes: The data are weighted. Standard errors in parentheses. The omitted category is ‘salary/wages’ as the main source of household income.

Source: Own calculations from the 2012 GHS.

***Significant at the 99.9% confidence level.

**Significant at the 95% confidence level.

*Significant at the 90% confidence level.

In order to explore this particular finding further the next two specifications (3 and 4) estimate the same correlates on a restricted sample (food poor African households from the Eastern Cape). Among poor households in the province, those that engage in home production are significantly less likely (3) to report hunger (relative to non-producing households). This association holds when the regression controls also for the main source of household income (4). This is a key finding since it shows that, irrespective of the main source of income for food poor households in the Eastern Cape, household production is significantly associated with lower levels of hunger.

5. Discussion

5.1. Household production and hunger

This paper aimed to explore household well-being and agricultural production activities through a food poverty lens. The results have suggested that, while households in the Eastern Cape Province are significantly poorer than households in South Africa as a whole, they do not necessarily report higher levels of hunger. In fact, the risk of adult hunger is significantly lower in the tribal areas of the province (relative to South Africa) even though poverty rates are far higher. This is the same finding which has been highlighted in the income poverty literature in South Africa (Meth, 2006; Posel & Rogan, 2016) and which formed one of the motivations for the comparison of poverty and hunger and their links with household production.

In order to narrow the analysis and to explore whether the gap between income poverty and hunger risks is explained, at least in part, by household production, the analysis focussed on a comparison between farming and non-farming households. Perhaps the main conclusion from the comparison is that hunger levels are actually lower among farming households in the Eastern Cape even though a higher percentage of these households (relative to non-farming households) live below the national food poverty line. There is also some persuasive evidence that Eastern Cape farming households are somewhat unique in their ability to avoid hunger compared with households in other provinces.

This link between household production and food security among poor households in the Eastern Cape was explored in greater detail through a series of linear probability models. First and foremost, the regressions showed that, while one of the key disadvantages of farming households, in terms of hunger risks, is the lack of wage income, there is clear evidence that household production is correlated with lower levels of hunger. This finding suggests that, while Sender’s (2014) argument regarding the importance of rural wages is undeniable, rural Eastern Cape households, in the absence of wage income, seem to protect themselves from hunger through household production. When the analysis is narrowed specifically to the poorest households in South Africa, the role of agriculture in protecting food security becomes even clearer and the evidence from the multivariate analysis suggests that this is particularly the case for Eastern Cape households that engage in household production.

While it is tempting to draw a number of inferences from these particular findings, the strongest claim that can be made, based on the data from the GHS, is that household production is a livelihood source for the majority of rural households in the Eastern Cape and there is clear evidence that, in the absence of wage income, household production is correlated with lower levels of hunger. It is also important to add that South Africa’s national household surveys, and particularly the income modules, tend to understate the importance of land-based livelihood activities to rural households. The income poverty literature, in particular, has underestimated the role of household production for own consumption in promoting household well-being.

5.2. Implications for policy

What do these results mean in terms of the best approach to supporting small-scale farmers and household producers? While keeping in mind the risk of over-romanticising home and community gardens, the data from the GHS (in support of the analysis by Fay (2013)) would suggest that continued support for household production without prospects for accumulation (see Shackleton et al., 2001; Cousins, 2013) is a relatively safe course of action. The vast majority of households in the province (and particularly ‘food poor’ households in the tribal areas) engage in production to supplement their diets and there are demonstrable links (irrespective of the direction of causality) between these activities and lower levels of hunger.

With respect to the more ambitious policy of promoting commercial agriculture among emerging farmers, there is more uncertainty. On the one hand, very few households in the province (or the country as a whole) that engage in production activities report that their main reason for farming is for any type of income. Moreover, the analysis also showed that when households depend on farm sales as their main source of income, the risk of hunger is significantly higher. This conclusion is supported by a large literature which demonstrates the inherent difficulties and risks faced by small farmers and particularly those which do not have access to the necessary infrastructure (e.g. transport and storage facilities). For example, even a generally favourable assessment (Tregurtha, 2009) of the Eastern Cape’s Siyakhula/Massive Food Programme found that poor market penetration (and a general lack of agricultural output markets) and structural problems with the supply of inputs and machinery were largely responsible for the programme not meeting its objectives.

More detailed (and robust) analysis than is possible when using the GHSs (or other large scale cross-sectional surveys) is still required but the fact that the results presented in this paper are closely in line with some of the findings from other recent analyses (Fay, 2013; Pienaar & von Fintel, 2014) suggests that we should take seriously the role of household production in supporting food security among poor rural households in South Africa. This conclusion in no way suggests that wage income is not important. In fact (the revival of) rural wages are likely to be crucial to the longer term development solutions in the rural parts of South Africa. However, food security is an immediate concern and the analysis of the available household survey data has shown that households are able to meet at least some of their basic needs by engaging in various types of household production even if the prospects for accumulation might be limited.

Acknowledgements

The author would like to thank John Reynolds for comments on an earlier version of this paper. The ideas, opinions, conclusions or recommendations expressed in this paper remain strictly those of the author.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

This work was done with the support of a grant awarded to the Institute of Social and Economic Research (ISER) from the REDI 3 × 3 project on “Employment/Unemployment, Income Distribution and Inclusive Growth”. The REDI 3 × 3 projects is funded by the South African National Treasury via the Southern Africa Labour and Development Research Unit (SALDRU).

Notes

1 Crucially, this proposed strategy also hinges on a greater level of state intervention in the commercial agricultural sector together with improved wages and a higher union density in the sector.

2 The Eastern Cape is the province with the second highest percentage of African households reporting agricultural activities (38%) and is home to roughly 22% of the 2.4 million African households in South Africa which are involved in some type of agriculture (own calculations from the 2012 General Household Survey).

3 Household participation in agricultural activities is derived from Q4.5–4.6 in the GHSs and by combining Q13.1 and Q23.1 in the LCS.

4 Acute hunger is defined as any adult/child in the household often or always going hungry because there was not enough food at any point in the past 12 months.

5 The poverty line used in the table is based on Statistics South Africa’s official upper bound poverty line of R577 (in 2009 prices). This poverty line is, by far, the most commonly used of the official national poverty lines and the poverty statistics which are cited in the academic and popular literatures are generally based on this threshold.

6 It is difficult to reconcile this large and somewhat unexpected discrepancy in the measure of hunger. While this requires further investigation, for now, I wish only to highlight it as a possible limitation to the way that subjective measures can be compared across different survey instruments.

7 This is not likely to be a causal relationship. In other words, households are not likely to be poor because they farm—rather households that are poor may choose to farm as a way of coping or diversifying their livelihood activities (see also von Fintel and Pienaar 2014).

8 Households suffering medium hunger levels reported that either an adult or child experienced hunger ‘sometimes, often or always’ over the past 12 months. Expanded hunger levels are those in which a household member reported hunger ‘seldom, sometimes, often or always’ over the past 12 months.

9 I also use a wider definition of hunger than Pienaar and von Fintel (2014) in that I combine adult and child hunger into a single binary outcome variable and identify households that ‘sometimes’ experience hunger as being food insecure.

10 NB: Only 1% of all African households that engage in agriculture in South Africa report that they farm as the main source of income in their household and only 5.5% supplement their household incomes with farm earnings (own calculations).