The challenges facing empirical estimation of household food (in)security in South Africa

Abstract

This paper sets out to describe current principle theoretical positions in food security research, describe and compare empirical South African food security studies with key international methodologies and pose research challenges for food security research in the country. Scant available food security studies for South Africa seem to indicate that food insecurity and hunger exist and are likely to increase due to increasing food price trends, greater reliance on cash food purchases and spiralling vulnerability as a result of HIV/AIDS morbidity and mortality. Empirical research is urgently needed to determine the coping strategies of households under ‘normal’ conditions, identify vulnerable households, and monitor the impact of various shocks and stresses on household food security.

1. International trends in food security research

Food security is a term that emerged following the 1980s shift in food policy debate from food supply to food demand and the emergence of new emphases on food entitlement, vulnerability, risk and access (Maxwell & Slater, 2003). Maxwell & Slater (2003: 532, citing Maxwell, 1990 and Hindle, 1990) explain that food security arose partly as a ‘proxy for poverty’ during structural adjustment. While food security does not capture all dimensions of poverty, food security may indicate poverty and is an important indicator of wellbeing (FANTA, 2002). Maxwell (1996), records the existence of more than 250 definitions of food security by 1996. Definitions of food security commonly include food supply, access, adequacy, utilisation, safety and, in some cases, cultural acceptability of food for all people at all times (see Maxwell, 1996). Food insecurity is simply the lack of food security that, at the extreme, is experienced as hunger.

The proliferation and progressive development of food security definitions mirrors international shifts in thinking about food security (Maxwell, 1996), and hints at some of the difficulties in adequately describing and measuring the varying and multiple contributing elements leading to hunger and malnutrition, vulnerability to hunger and malnutrition, and possible future vulnerability. The progressive development of definitions also reflects a shift of emphasis from objective to subjective measurements including cultural acceptability of foods and fears of experiencing hunger. Dowler (2003: 573) has asserted that ‘food poverty’ is a synonym for food insecurity, resonating current wider thinking about food security as more than physical efficiency of food consumption but also including the ability ‘to acquire or consume an adequate quantity or quality of food in socially acceptable ways, or the uncertainty that one will be able to do so’.

Initially, household food security was viewed as a measure to link national, regional and community level food supply to household food consumption and individual nutritional status and relate agricultural policy to issues of nutrition (Gittelsohn et al., 1998). However, food insecurity is no longer seen as a failure of food production at the national level but as livelihood failure (Devereux & Maxwell, 2001) as evidenced by the inclusion of food security as one outcome of sustainable livelihoods (Department for International Development, 2000). This shift in perspective away from emphasis on national food production and the increasing awareness that sustainable livelihoods are crucial elements in future food self-sufficiency that lead to a number of progressive developments towards more household-focused food security measurement methodologies. This poses many empirical challenges, as household level influences and determinants of food security are far more numerous and varied than measurements at national level. How to capture concisely and effectively the influence of varied and often inter-related influences in representative samples is a challenge food security researchers have not yet overcome.

A further complication to food security measurement is operationalising some of the terminology used in defining food security (for examples: adequacy and sustainable) (Gittelsohn et al., 1998). Maxwell & Fankenbergers' (1992, cited by Hoddinott, 1999a) annotated bibliography of food security studies, listed 194 papers covering conceptualisation and definitions of food security and a further 172 studies focusing on food security indicators. Hoddinott (1999a) reports that more than 450 indicators of food insecurity had been identified by the turn of the century. As South Africa is nationally food secure (National Department of Agriculture 2003; HSRC et al., 2004), this paper focuses on empirical measurement of household food security. Indicators of food security are discussed in the next section.

2. Household food security indicator classification

Measuring food security is costly and complicated. A wide variety of methodological approaches have been applied to food security studies (Scherr & Vosti, 1993), determined by the purpose of the analysis, availability of data, and the background and preference of analysts (Riely, 2000). The multidimensional character of food security makes measurement complex. Most household food security measurements are a collection of direct and indirect indicators reflecting food supply, food access and outcome indicators, i.e. proxies of adequate food consumption (see Chung et al., 1997; Hoddinott, 1999b; and United Nations Standing Committee on Nutrition, 2004 for advice on selecting indicators).

Household food supply indicators include agricultural production, pest management, regional conflict, market access and institutional support structures (Frankenberger, 1992: 84–9). Food access indicators include food entitlement and socio-economic indicators that indicate the ability of households to cope with various stresses induced by economic and social change/s (Frankenberger, 1992: 89–95). Measures of food access are useful to assess the severity of food shortfalls, characterise the nature of household food insecurity (for example, seasonal versus chronic), monitor changes in circumstances, assess the impact of various interventions (Hoddinott & Yohannes, 2002), and the capacity of households affected by social and economic shocks and disasters to withstand the effects of these shocks.

Most food security outcome indicators measure more than food consumption – for example, malnutrition indicators also capture the influence environmental aspects, bio-utilisation factors, health and sanitation on nutrition, growth and development (Frankenberger, 1992: 96). Frankenberger (1992: 96) has divided outcome indicators into direct and indirect indicators. Direct indicators measure the experience of food insecurity itself (for example: food consumption surveys, household perceptions of food insecurity or hunger and food frequency measurements). Direct indicators are best measured by observing households over time, along with interviewing members of that household in depth (Wolfe & Frongillo, 2000). Indirect household food security indicators are used where access to such information is either unavailable or too costly to collect. Indirect indicators include: dietary diversity, storage estimates at critical times of the year, subsistence potential for largely agrarian households (ratio of dietary energy requirement and food availability), and nutritional assessments (anthropometric indicators).

As Riely (2000: 4) states: ‘It is possible to develop a method to measure an outcome, such as malnutrition, that does not yield any understanding of the underlying causes of that outcome’. Moreover, multiple measures are required to effectively capture multiple dimensions of food security or to provide information for different programme approaches (Wolfe & Frongillo, 2000). Therefore, most food security measurement approaches include a mix of direct and outcome indicators to develop some scale or index of household food security to either differentiate between food and food insecure households or to monitor the impact of various interventions. The following section outlines the current international approaches to food security measurement.

3. Current international empirical food security approaches

Given the complexity of food security and the range of factors and elements that affect food supply, access, adequacy, utilisation, safety and cultural acceptability, measurement of food security is appreciably complex, extensive and expensive. The suite of variables included in measurement and monitoring of regional household food security is most often limited by data and/or resource availability. For example, a Food and Agriculture (FOA) (Croppenstedt, 2003) ten-country case study of the roles of agriculture in society included a module on food security. Six country case studies (China, Dominican Republic, Ethiopia, Ghana, India and South Africa) included discussions of household food security using a variety of data sources and empirical methodologies. Case studies for China (Xiurong et al., 2003), the Dominican Republic (Lizardo, 2003), Ethiopia (Adenew, 2003) and India (Srinivasan, 2003) used national survey data from consecutive national household or individual survey rounds. India's case study was based on the 55th national survey, enabling observation of changes over more than half a century. The Ghanaian case study was based on Maxwell's (2000, cited by Nyanteng & Asuming-Brempong, 2003) study of Accra households. In contrast, the South African case study (Kirsten et al., 2003) combined the poverty and food security modules as the country lacked conclusive and comparative household food security data. The roles of agriculture case studies included income and consumption elasticities, anthropometric studies and/or observations of household demographic (household composition, employment and migrancy), income, consumption, and nutrition changes over time.

Riely (2000) explains that even given the same set of indicators, different methods of constructing indexes can lead to different conclusions (even within the same approach to measurement). Given the lack of consensus regarding food security measurement, new food security indicators and measurement scales are continually being developed and tested. Currently, key empirical food security methodologies include the following systems or combinations of these systems to develop a food security index or score: household vulnerability approaches, dietary diversity measures, experiential tools and coping strategy assessment tools.

Household vulnerability analyses are used by various international aid agencies [Food and Early Warning System (FEWS), Food and Income Vulnerability Information Mapping Systems (FIVIMS), and the World Food Programme (WFP)] to facilitate identification of long-term mitigation activities, targeting food aid and emergency activities and monitoring interventions (Riely, 2000). While vulnerability assessment methods vary considerably, Riely (2000) explains that they share a common conceptual framework rooted in traditional economic models of household consumption. Vulnerability approaches attempt to determine the change in consumption (total food budget) and income sources as a result of a crisis compared to baseline ‘normal’ periods for each socio-economic group identified through key informant interviews. These approaches focus on the dynamism faced by households and their responses to these conditions and social and economic shocks (Gittlesohn et al., 1998; Riely, 2000). Household typologies are defined by socio-economic and geographical characteristics and the ability of each type of households to cope with risks and resource constraints is investigated. FEWS NET's (undated) food economy approach includes a livelihood construct for various rural and urban zones.

Rolling national vulnerability assessments were carried out to assess and monitor vulnerability of households to food insecurity in Malawi, Mozambique, Swaziland, Zambia and Zimbabwe during the 2002/2003 humanitarian crisis (Mdladla et al., 2003) and have been used for regional analysis of the impact of HIV/AIDS on food secuirty [for example, the South African Development Community Food, Agriculture and Nutrition Vulnerability Assessment Committee (SADC), 2003]. Stakeholders agree that the scope and frequency of the assessments is not sustainable and Vulnerability Assessment Committees aim to develop a livelihoods monitoring system using a proxy variable approach based on the Food Economy Zone Approach. Whether application at a national scale is possible is still subject to debate (Mdladla et al., 2003). Mdaldla et al. (2003), have indicated that bi-annual or annual national surveys are not sustainable in terms of resources.

The Save the Children's Fund's Household Economy Approach is another example of vulnerability assessments but uses rapid assessment methods for more detailed assessment of food aid needs over a longer period of time than FEWS, FIVIMS and WFP assessments (Riely, 2000). The Household Economy Approach enables development of a baseline picture of how households in a particular area survive in normal years (Boudreau, 1998). This information is collected through focus groups from each wealth group in a community using a mix of qualitative and quantitative approaches (O'Donnell, 2004). Secondary data are used to identify changes in the larger economy that will affect production and exchange options available to rural households that is used to predict food and income shortfalls given their ability to cope (Boudreau, 1998). Further development of the Household Economy Approach led to the development of the Individual Household Questionnaire used in Zimbabwe and Zambia in 2003 in an attempt to link AIDS with food security (O'Donnell, 2004). A database of total household income and food access was derived. Cross-tabulations were used to determine relationships to HIV/AIDS proxy indicators (as examples, household characteristics, presence of a chronically ill adult, dependency ratio, number of orphans and labour poverty).

The Individual Household Approach is a recent development that follows the same style as the Household Economy Approach using semi-structured interviews but collects individual household data rather than wealth group data (O'Donnell, 2004). The tool was developed specifically for poverty studies and calculates disposable income per adult equivalent after defined minimum food and non-food needs have been met. The approach has been tested in Swaziland and Mozambique in studies to investigate the impact of HIV/AIDS on food and livelihood security. In these tests, the tool proved reasonably inexpensive, practical and robust at quantifying HIV and other effects on rural livelihoods (O'Donnell, 2004).

Assessments of dietary diversity are widely used by the International Food and Policy Research Institute (IFPRI) and the Food and Nutrition Technical Assistance (FANTA) Project (FANTA, 2002; Hoddinott & Yohannes, 2002). In some cases food groups are counted (Hoddinott & Yohannes, 2002, citing: Kant et al., 1991; Hatloy et al., 1998; Swindale & Ohri-Vachaspati, 1999), while other studies count each food item consumed by household members over a period (Hoddinott & Yohannes, 2002, citing: Krebs-Smith et al., 1987; Drewnowski et al., 1997 and Hatloy et al., 1998). For example, Gittlesohn et al. (1998) counted the reported consumption frequency of 70 food items in a study in Nepal. Hoddinott & Yohannes (2002) compared data from ten countries (Bangladesh, Egypt, Ghana, India, Kenya, Malawi, Mali, Mexico, Mozambique and the Philippines) to examine whether dietary diversity was a good measure of household food access. The researchers found that increased dietary diversity of individual food items and food groups was strongly associated with increased household per capita energy availability for rural and urban households. Hoddinott & Yohannes (2002) concluded that dietary diversity showed promise as a means of measuring household food access, and monitoring changes and interventions impact/s.

More recent food security indicators have focused on development of experiential-based measures using qualitative surveys (Rose & Charlton, 2002) to measure the experience of food security over time among sample households. Experiential systems (such as the United States Federal Food Security Measure) measures include anxiety, perceptions and social acceptability of food insecurity (Wolfe & Frongillo, 2000). Hunger is seen as a severe stage of food insecurity, rather than as a distinct or separate condition from the more general experience of food insecurity. Experiential tools are simple and quick to administer (requiring four to ten minutes per household), enabling rapid and repeat interviews over time. The tool captures both the severity and level of food insecurity. The major weakness of experiential tools is in setting cut-offs for classification of households, such as in the Federal Food Security Measure where households are classified into four groups according to severity of experience (Wolfe & Frongillo, 2000). Following development of a definitive measure of food security through this methodology, the Federal Food Security Measure has been incorporated with other measures and integrated into the United States Current Population Survey to develop a continually monitored national food security measure (Wolfe & Frongillo, 2000).

The reliability of assessments based on people's experience of food insecurity or hunger is always subject to intentional misreporting due to participant self-interest. There is no saying if such data is more reliable than the misreporting and recall errors encountered in more rigorous dietary recall assessments and income and expenditure surveys (Wolfe & Frongillo, 2000). However, Nanama & Frongillo (2003) found that an experiential-based tool was valid for assessing household food security in a study in northern Burkina Faso when compared with wealth, dietary and anthropometric measures for 126 households.

The most recent measurement tool is the Coping Strategy Index (CSI) developed by CARE International and the World Food Programme (Maxwell et al., 2003). This tool is based on an earlier methodology developed by Maxwell (1995). Maxwell's (1995) original tool aimed at distinguishing and measuring short-term household food insecurity and included vulnerability measures and deliberate actions households take when faced with food insufficiency. The revised tool has been proposed as a rapid measurement of food security and the impact of food aid programmes in emergencies (Maxwell et al., 2003). The index was developed in Uganda and Ghana, and used recently for early warning and food security assessments in Kenya, Ethiopia, Eritrea, Zimbabwe, Zambia, Malawi and Burundi (Maxwell et al., 2003). The concept is based on assessment of responses to the question of what households do when they do not have enough food and do not have the money to buy food. A simple set of questions is used to determine how often households employ a list of coping strategies identified by community-level focus groups. Information on the frequency and severity (in terms of a particular local context) of coping strategies is then weighted according to the community-defined severity of identified coping strategies and combined into a single score (CSI). Coping strategies include behaviours reflecting dietary changes, increasing short-term household food availability (including use of credit), decreasing the number of people to feed, and rationing portion sizes.

Comparative studies have shown coping strategy indexes to be good proxies for food intake (energy adequacy), food budget shares, food frequency, income status and the presence or absence of a malnourished child in the household (Maxwell et al., 2003 citing: Maxwell et al. 1999; Christiansen & Boisvert, 2000). Maxell et al.'s (2003) coping strategy index has also identified changes in household conditions as a result of emergency food aid operations. Maxwell et al.'s (2003) approach combines experience of hunger with coping strategy assessments. One shortcoming of Maxwell et al.'s (2003) approach is that the assessment cannot be repeated for the same community as respondents may alter their responses to the coping strategy behaviour questions in subsequent rounds. Loevinsohn & Gillespie (2003) and Gillespie et al. (2001) warn that ‘coping’ implies adoption of reversible management strategies, while many responses of distressed households are irreversible (such as sale of major assets and taking children out of school), implying the need for caution in interpreting the Coping Strategy Index.

The choice of indicators depends on resources available for data collection and analysis (financial and field staff), the speed with which the analysis is required and the types of secondary data available. Chung et al. (1997) assert that household income and dietary intake are too cumbersome to be of practical use in targeting food aid and that practical food security indicators should be statistically reliable but straightforward to collect and analyse. To illustrate this point, Chung et al. (1997) conducted a comparative study of qualitative and quantitative data with benchmark measurements derived from dietary recall, anthropometric measures and blood data (serum levels of vitamin A and iron adequacy). The qualitative data was collected using: case studies, participatory mapping of vulnerable households, food charts and seasonality charts. Quantitative economic and nutrition surveys using three rounds of data collection from 324 households in south-central India. The study showed that modest but significant associations between the tested indicators and benchmark indicators may generate program savings that in some cases could be substantial.

Where financial and human resources are limited, qualitative methodologies may be the most viable option for food security analysis (Chung et al., 1997). Qualitative methodologies often facilitate simpler and quicker data collection but require skilled staff to conduct and interpret the studies. Qualitative methodologies use smaller, in-depth investigations with little potential for statistical analysis (Wolfe & Frongillo, 2000). However, comparisons between qualitative study narratives are limited due to considerable variation in what participants report; but smaller samples limit the possibility of generalising findings to larger segments of the population (Chung et al., 1997). Quantitative surveys require large sample sizes for statistical analysis and more resources to collect and analyse the data and may not include assessment of social and temporal dimensions of food insecurity. Wolfe & Frongillo (2000) suggest that qualitative measurements may be adequate for planning and designing food security interventions, while quantitative data is required for the monitoring and assessment of food aid. A number of trade-offs will always be necessary to meet the needs of specific measurement purposes within resource limitations.

4. Food security estimates for South Africa

Empirical South African food security research is very limited. The lack of comparative studies and time-series data sets prohibits accurate estimation of food security and food security trends in South Africa. To date no nationally representative food security surveys have been conducted for the South African population and food security monitoring systems have not yet been developed. This section of the paper reviews available food security-related studies conducted with South African samples to illustrate the diversity of methodologies employed and the variable results obtained.

The South African Vitamin A Consultative Group (SAVACG) conducted the first national study of nutrition in South Africa in 1994. The anthropometric study of 18 219 households with preschool children aged 6–71 months (Labadarios & Middelkoop, 1995, cited by Steyn, 1996) (Table 1). The study showed that nearly one in four children was stunted and one in ten were underweight. Only 2.6 percent of sample children were wasted. Stunting (reduced height for age) was identified as the major nutritional problem in South Africa and was found to be especially high among rural children.

Table 1 Summary of nutritional indicators for children aged 6–71 months, 1994

	Per cent of children in sample
Nutritional indicator	Northern Cape	Western Cape	Eastern Cape	KwaZulu- Natal	Mpuma- langa	Limpopo	Gauteng	Eastern Cape
Underweight^a	15,6	7,0	11,4	4,2	7,3	12,6	5,6	11,4
Stunting^b	22,8	11,6	28,8	15,6	20,4	34,2	11,5	28,8
Wasting^c	2,5	1,3	3,2	0,7	1,7	3,8	1,2	4,5
^aUnderweight children have low weights for age, reflecting current inadequate food intakes, past episodes of under-nutrition and/or poor health conditions that affect child nutrition.
^bStunting refers to low height for age, reflecting long-term under-nutrition and sustained poor environmental and socio-economic status of households.
^cWasting refers to low weight for height that is a result of weight loss associated with recent starvation or severe disease.
Source: Steyn (1996, citing SAVACG, 1994).

In the absence of full, nationally representative data on food and nutrient consumption, Rose et al. (2002) used the October 1995 Income and Expenditure Survey data (n = 28 704) representing the food spending and home production patterns of 28 704 households to estimate household availability of seven broad food groups for a variety of provincial, rural-urban and racial groups used price index data and food composition databases to approximate energy and nutrient consumption. It was found that daily energy consumption per adult female equivalent was lower in rural areas (ranging between 78 and 119% compared with the standard of 2200 calories) than in urban areas (92 to 130% of energy requirements for female adult equivalents).

In a second study using the same data set, Rose & Charlton (2002) developed two food insecurity measures: food poverty (whether household food expenditure was adequate to purchase a low cost food plan) and low energy availability (whether available food (purchased or home production) was less than the sum of household member's required energy intakes). Households were classified into four categories: food poverty only; low energy availability only; food insecure on both measures; and food secure. Both bivatiate and multivariate analyses showed that food insecure households were more likely to be in rural areas, have low incomes or large household sizes and be headed by Africans or individuals of mixed ancestry. Rose & Charlton (2002) report that the food insecurity findings of this study concur with the National Food Consumption Survey (Labadarios, 2000) findings in terms of the experience of hunger and food insecurity. Rose & Charlton's (2002) findings for food poverty and low energy availability are presented in Table 2. The findings infer that low energy availability may hold a key to understanding food insecurity in South Africa and more than likely concurs with findings that stunting is the major nutritional problem in the country as stunting is a result of continual inadequate intake of food. Rose & Charlton's (2002) study may indicate the efficiency of using national level household data in identifying food insecurity as an alternative to expensive anthropometric measurements but does not enable understanding of household responses to food shortages.

Table 2 Food poverty and low energy availability among households in three provinces, 1995

	Food secure	Food poverty only	Low energy availability only	Food poverty and low energy availability
	Per cent of population (urban and rural)
Eastern Cape	40,6	5,5	11,1	42,8
KwaZulu-Natal	41,2	3,8	15,1	39,9
Limpopo	37,4	13,5	5,5	43,7
South Africa	41,5	3,9	15,9	38,7
Source: Rose, 2004.

HSRC (2004, citing Statistics South Africa, 1994 and 1999) have summarised the results of pertinent food security-related questions from the October Household Surveys of 1994–99 (Table 3). The results show that across 94 and 1995, and 1996–98 the number of households not having ‘enough money to feed children’ and who ‘could not afford to feed children’ may indicate that household food security deteriorated over the survey rounds. Although the results should be interpreted with caution, as the questions across survey rounds were not consistent (HSRC, 2004), Bradshaw et al., 2000 and StatsSA, 2003 (both cited by HSRC, 2004) confirm that malnutrition (in terms of rates of underweight and stunted children) worsened in South Africa during the late 1990s. The surveys illustrate a way of developing and monitoring experience of food shortages and hunger. However, the surveys were discontinued and replaced with the bi-annual Labour Force Survey that does not continue measurement of these parameters.

Table 3 Subjective Assessment of Food Insecurity in South Africa by Province and Area of Residence, 1994–99

	Did not have enough money to feed children in household		Could not afford to feed children in the household		Children aged under 7 went hungry because there was not enough money to buy food
	1994	1995	1996	1997	1998	1999
National	41,0	31,7	27,0	25,5	31,1	23,4
Rural	49,2	–	29,1	29,9	37,2	27,6
Urban	34,1	–	25,5	22,3	26,4	19,1
Source: HSRC (2004, citing Statistics South Africa, 1994–99).

The 1999 South African National Food Consumption Survey of children aged 1–9 years was the first primary study to determine nutritional status and attempt an estimation of experiential food security in South Africa through development of a hunger scale. A cross-sectional sample of children was drawn from 1996 national census data (Labadarios, 2000; VIC, 2001). The final sample included 2894 children from 156 national randomly selected enumerator areas. Anthropometric measurements were recorded and a five-part questionnaire was administered to sample households. The survey instrument included multiple measurement approaches in sections covering:

•	Socio-demographic household factors related to the environment in which the child lived.
•	A 24-hour recall of food consumption for the child.
•	A qualitative food frequency questionnaire for the past six months.
•	An inventory of food procurement and household food stocks.
•	A hunger scale questionnaire providing information on actual hunger experienced (or not) by the child.

The hunger scale used for the National Food Consumption Survey was determined from responses to eight questions relating to whether the household had sufficient money to buy food, enough food, consumed adequate portion sizes and/or whether children ever went hungry (Gerike et al., 2000). A score was used to classify household risk of hunger derived from the sum of affirmative answers to the questions regarding experience of hunger (Gerike et al., 2000). A sum of five or more positive responses to the eight questions indicated a food shortage, and these households were considered ‘hungry’ (Table 4). The survey questions neglected home production of food and only questioned children's experience of hunger for the previous five and 30 days, and therefore does not take seasonality of agricultural production, income and expenditure into account as is common with other experiential measures of food security (for examples the Federal Food Measurement System (Wolfe & Frongillo, 2000) and de Swart (2003). Moreover, the study only included children of 1–9 years old and so does not give a complete picture of food insecurity in South Africa.

Table 4 Prevalence of under-nutrition, food insecurity and hunger among children in three provinces and South Africa, 1999

	Eastern Cape (per cent of sample) n = 465	KwaZulu-Natal (per cent of sample) n = 465	Limpopo (per cent of sample) n = 465	South Africa (per cent of sample) n = 2613
Stunting^a	20,5	18,5	23,1	21,6
Underweight^a	7,1	6,0	15,0	10,3
Wasting^a	1,8	4,3	7,5	3,7
Household food insecure^b	88–94	63–73	62–79	60–73
Individual insecurity and child hunger^b	64–92	38–61	37–72	36–67
Child hunger level^b	29	27	32	30
^aPercentages for  < − ,22SDs.
^bAdjusted scores for factor analysis.
Source: Labadarios & Nel, 2003, p 202; Gerike et al., 2003, p 837.

The National Agricultural Marketing Council's (NAMC, 2002) study of the impact of market deregulation on household food security lacks empirical estimates of household food security. The National Agricultural Marketing Council's (NAMC, 2002) investigation into the impact of market deregulation conducted a survey of household impacts among 60–160 non-metropolitan households in each of five provinces (Gauteng, Eastern Cape, Western Cape, KwaZulu-Natal and Limpopo). The study investigated physical and economic access to food to determine if households were consuming fewer key food products that could have been affected by market deregulation. No attempt was made to quantify food secure versus food insecure households. The study found that the effects of market deregulation on household income appeared virtually non-existent. However, a key finding of the study that informs empirical measurement development in South Africa is the finding that income was as the single most important determinant of a household's ability to meet food security needs, (NAMC, 2002).

Aliber & Modiselle (2002) conducted a National Department of Agriculture commissioned study to investigate the impact of price volatility for low-income households. The methodology developed was tested in six locations but the small sample size (30 households in all) and the lack of a random sampling frame limited the generalisation of the findings. The study included a structured questionnaire, focus group interviews and data collection from shops. The structured questionnaire included six sites in North West/Gauteng and KwaZulu-Natal, two rural and one urban site for each. The survey collected information on: food intake over 24 hours, dietary diversity, a coping strategy index, economic wealth and wellbeing and own production.

Households reported significant increases in food prices over the previous six months. However, apart from some brand switching, price increases did not have clear impacts on food purchases during the six-month study period. Dietary diversity was greater for urban households and relatively wealthier households. Among rural households, own production significantly improved dietary diversity but dietary diversity was not significantly influenced by price changes. Aliber & Modiselle (2002) reported that the coping strategy questions included in the survey revealed the power of such questions. The coping strategy responses showed that a high proportion of respondent households experienced anxiety over future food, dietary quality and sufficiency but engaged in a variety of coping strategies. A seeming binomial distribution was found between households that never experienced the food security problems described in the survey and those who experienced chronic food insecurity. Between these two peaks were a smaller number of households that experienced these problems less frequently. Aliber & Modiselle (2002) recommended shortening the survey and repeat cross-sectional and panel studies to monitor the impact of price volatility on low-income households. A sampling approach similar to the rotating bi-annual Labour Force Survey was recommended using sentinel sites.

Mekuria & Moletsane (1996) and Leroy et al. (2001) conducted two small studies of household food security in the Limpopo Province. Mekuria & Moletsane (1996) surveyed 197 randomly selected households from five districts in the Limpopo Province. Structured household questionnaires collected data on: household characteristics, demography, resource availability and use, food production, availability, consumption, income and expenditure on food, marketing, access to institutional support (credit, extension and research), household coping strategies and perceptions of food insecurity. Values of household food purchases, farm and non-farm income and the value of liquid assets for the purchase of food were used as proxies for household food security. Mekuria & Moletsane (1996) classified households as food insecure (58% of sample households), vulnerable to food insecurity (17%) and food secure (25%) based on energy consumption. Respondents saw drought and lack of income as causes of food insecurity. Coping strategies commonly employed by the sample households in the face of food insecurity were: dependence on community/social networks, own food production, seeking employment, and asset accumulation.

Leroy et al. (2001) applied a formal, structured questionnaire to 52 farmers in the Venda region. Survey data was verified through informal surveys. The following data were collected: household characteristics, food consumption and purchases, farm characteristics, production and farm and non-farm income. Energy requirements for 80 percent of sample households were lower than minimum energy, protein and fat requirements estimated from edible portions of available food and derived using food composition tables.

De Swart (2003) and Fraser et al. (2003) conducted surveys in the Eastern Cape. Fraser et al. (2003) surveyed 128 households from villages at Victoria East and Middledrift districts of the Eastern Cape for investigation of poverty levels and expenditure patterns. In-depth surveys and a one-month observation were carried out with eight purposively selected households from the overall samples from each district to investigate the strategies employed by households to alleviate poverty and to acquire food. The study found that households that had access to reliable income sources (grants, salaries and wages) experienced transitory food insecurity. These households purchased food from supermarkets in urban centres and claimed that food prices at these stores were lower than local prices, especially at month-end when special prices were offered. Such households experienced food shortages towards the end of the month. The ultra-poor experienced chronic food insecurity and reportedly never had enough food. These households diversified income sources for survival and relied on piecework for wages or food parcels. The ultra-poor found that piece wages were too small to warrant shopping trips to urban centres and so these households purchased small quantities at relatively higher prices from local stores. Family networks were an important resource in times of need. Household gardens provided food and a means of bartering for other food items in times of need. Poor households were often forced to liquidate assets following poor harvests, reduce the number of meals per day or gather traditional vegetables in times of hunger.

During 2002, de Swart (2003) surveyed over 2000 households (comprising 10 544 people) in the rural Eastern Cape (Mount Frere), rural Western Cape (Ceres) and urban Cape Town (Khayelitsha and Nyanga) using a household livelihood survey. The data collected included: livelihood assets, needs and strategies; health status and anthropometrical measures; and food and nutritional status. De Swart's (2003), study asked whether households had taken the following actions during the year prior to the survey: borrowed food; asked for credit at the store; worked for food or took no action. Between 37 and 55% of sample households indicated that general food consumption was lower than in the previous year. Households in Mount Frere experienced the greatest food insecurity, followed by urban households in Cape Town. Eighty-three per cent of households in Mount Frere indicated that there was too little food available, compared to 81% of urban sample households in Cape Town and 69% of the Ceres household sample. This may indicate that rural and urban food insecurity is problematic for a majority of poor sample households. The study found that rural households in Mount Frere were more food insecure than households in the other two sites. Given the food security problems in Mount Frere, further investigation of food security was undertaken in this area. The food situation was assessed in three periods during the year using surveys of the food security situation, food quantities consumed and a one-month food consumption diary (de Swart, 2003). De Swart's (2003) study indicated worsening food security, increased experience of hunger and an increasing dependence on money to purchase food.

The only food security related longitudinal panel data available is the KwaZulu-Natal Income Dynamics Study (KIDS) (May et al., 2000) database that includes the subset of 1100 households for KwaZulu-Natal from the 1993 Project for Statistics on Living Standards and Development (PSLSD). These households were re-surveyed in 1998 and a third round of surveys was being undertaken in 2004. While KIDS included a range of food security related indicators and data, no attempt was made to estimate household food security. The 2004 survey has been adapted to enable estimation of household food security. The 1998 KIDS survey showed an increase in the number of households who have incomes below the minimum household subsistence threshold, particularly for sample households in rural areas (Kirsten et al., 2003). The KIDS study showed that 22% of households surveyed in 1998 were poor in 1994, 19% had experienced a decrease in their level of material wellbeing and 11% had experienced and improvement, indicating that 41% of households had remained poor or experienced erosion of material wellbeing between 1994 and 1998. Moreover, Jinabhai et al. (2004) have shown that the prevalence of stunting and wasting among KIDS sampled households increased between 1994 and 1998 (Table 5). Lund (2004) compared the KIDS survey findings with the Risk and Vulnerability in Employment Study (Lund & Mkhize, 2003, cited by Lund, 2004) that used a mix of qualitative methodologies including livelihood time lines. Lund (2004) emphasised the value of longitudinal studies in monitoring long-term livelihood trends.

Table 5 Prevalence of under-nutrition (KIDS 1994 and 1998)

		Prevalence of stunting (per cent of sample)		Prevalence of wasting (per cent of sample)
Age of children	Survey round	Boys	Girls	Boys	Girls
4–5 years	1994	8,0	6,0	4,8	3,0
	1998	23,6	23,4	5,2	7,1
8–11 years	1994	20,6	17,0	9,7	7,0
	1998	28,4	26,8	7,3	7,2
Source: Jinabhai et al., 2004.

Given the dearth of comparative and conclusive empirical estimates of food security, the prevalence of food insecurity in South Africa is not known, nor is it possible to evaluate changes in food availability and access over time. Scant available data suggests that between 58.5 and 73% of South African households may experience food insecurity, 15.9% consume less than adequate energy, stunting rates (1999 national figures) could be about 22%, wasting may occur in 3.7% of the population and approximately 30% of households may experience hunger.

Hunger and low energy intakes result in unacceptably high rates of stunting and wasting. However, malnutrition rates are inadequate to determine food insecurity as this outcome indicator seems to under-estimate the extent of food insecurity but malnutrition rates seem to compare with Rose & Charlton's (2002 study cited in Rose, 2004) estimates of low energy availabilities. Analyses of the KIDS data seem to indicate increased malnutrition and poverty between 1994 and 1998 among the sample households in KwaZulu-Natal, although generalisations are not necessarily valid given the panel sample selection.

Available studies summarised here used a variety of survey instruments to investigate food insecurity, including: malnutrition, experiences of hunger, energy availabilities and food poverty. While some studies have included a mixture of qualitative and qualitative and experiential methodologies and data, and others applied a single approach or multiple proxy indicators to evaluate food insecurity. The National Food Consumption Survey used a mix of methodologies (section 3) but the time frame applied for experiential data was too short (five and 30 days). The study could have been significantly strengthened by inclusion of a seasonality chart to reflect seasonal experience of hunger. The seasonality of food insecurity and hunger has not been investigated, factors affecting household vulnerability to food insecurity have not been established, nor have the coping strategies employed by households facing food insecurity been documented apart from Mekuria & Moletsane's (1996) small study in Limpopo, Aliber & Modiselle's (2002) sample of 30 households, and Fraser et al.'s (2003) 16-household case study. Moreover, samples are often too small or narrowly targeted and the data not widely representative, fragmentary and not comparable between studies. No effort has been made to standardise data sets, employ comparable and tried and tested methodologies as used by neighbouring SADC countries, or plan for longitudinal studies. Without knowing the true extent of food insecurity or the trends in food security over time, there is little hope of effective policies and targeted programmes to address food insecurity in South Africa.

5. Future trends likely to influence food security estimation

Possible shocks and stresses to household food security in South Africa are likely to impact negatively on a large proportion of households already facing vulnerability to food insecurity, increasing hunger and malnutrition. Continually rising food prices and increasing poverty threaten household food security by reducing the purchasing power and available resources for purchasing food and eroding incomes. Watkinson & Kakgetla (2002, cited by HSRC, 2004) report that increasing food prices have a disproportionate and devastating impact on the rural ultra-poor (households who spend more than 50% of the income on food). While the NAMC (2002) investigation of the impact of market deregulation concluded that there was virtually no erosion of household income, Aliber & Modiselle (2002) have reported few clearly observable negative influences of rising food prices over a six-month period, yet Vink & Kirsten (2002) report that evidence from deregulation and liberalisation is experienced mainly by low-income earners in urban and semi-urban areas, small holder farmers in remote areas and unskilled farm workers. Vink & Kirstens' (2002) findings are confirmed by a Food Price Monitoring Committee (2004) investigation that found that food price inflation affected different income groups in various ways but poor households experienced higher inflation rates than wealthier households. While increasing food prices cannot be solely attributed to trade liberalisation, a steady increase in food prices has eroded incomes over time, particularly for those segments of the population already vulnerable to food insecurity. Continued monitoring of food prices and the impact on households vulnerable to food security is essential as is continual evaluation of the impact of various public programmes (increased social grants, food parcels and seed packages) implemented to address rising food prices.

Increasing reliance on food purchases increases vulnerability to food insecurity as greater reliance on wage income discourages home production of food. The widespread reliance on wage income, even in deep rural areas such as those surveyed by De Swart (2003) and Fraser et al. (2003) increases vulnerability to household food insecurity. While seeking wage incomes is a risk-aversion strategy that diversifies rural livelihoods, replacing or significantly decreasing agricultural production against wage income may seriously increase vulnerability to food insecurity in the long term as food prices increase and wage opportunities decrease. Hendriks (2003) and Kirsten et al. (1998) have shown that agricultural production may not reduce malnutrition rates significantly unless home production exceeds subsistence. Once agricultural production leads to income through sales, an income replacement effect is seen in households when the cash generated from agricultural sales is used to supplement cereal-based and nutrient deficient diets with meat, poultry and fats (Hendriks, 2003). However, Aliber & Modiselle (2002) found that rural households with food gardens experienced improved dietary diversity.

HIV/AIDS is a shock to household food security and household food security cannot be properly understood without considering the effects of HIV/AIDS (SADC, 2003). HIV/AIDS, food and nutrition security can become increasingly entwined in a vicious cycle – HIV heightens vulnerability to food insecurity, that can in turn increase susceptibility to HIV infection (Gillespie et al., 2004). While households affected by HIV are generally more sensitive to livelihood shocks (Mdladla et al., 2003) and as food security is a livelihood outcome, they are also more sensitive to food-related shocks. HIV can (rightly) be treated as a shock to household food security (Baylies, 2002, cited by SADC, 2003). HIV is, however, a unique, slow-moving and devastating shock (Gillespie et al., 2001) that strips households of livelihood assets, described by Mdladla et al. (2003) as a ‘creeping emergency’ that progressively erodes the lives and livelihoods of affected households. HIV/AIDS can leave households significantly impoverished (Kadiyala & Gillespie, 2003). Kadiyala & Gillespie (2003) go as far as postulating that food insecurity and HIV/AIDS may converge over time as increased HIV-related expenditures of richer households have a spill-over effect on the food security status of poorer households, especially with regard to hired labour from poorer families.

De Waal & Whiteside (2003) have termed HIV/AIDS a ‘new variant famine’. HIV sufferers have very specific and raised nutritional needs (Haddad & Gillespie, 2001), but the progression of the disease and worsening nutritional status are linked by negative synergies (Piot & Pinstrup-Andersen, 2002). HIV-affected households face increasing risk of food insecurity and malnutrition increase as sick family members cannot work, income decreases, health-care expenditures increase, care burdens increase and there is less time for caring for children (Piot & Pinstrup-Andersen, 2002). As O'Donnell (2004) points out, the damaging interactions between HIV and malnutrition limit the viability of consumption-based coping strategies as HIV sufferers cannot afford to reduce food consumption as is a common early response to food crises. The asset bases and capacity for coping is reduced in HIV/AIDS affected households. AIDS kills the most productive household members (Kadiyala & Gillespie, 2003). HIV-related morbidity and mortality reduce labour resources, increase the care burden for affected households, reduce dependency patterns, and simultaneously increase dependency rations within households (de Waal & Whiteside, 2003; O'Donnell, 2004). Responses may not be coping strategies at all but struggles to adjust to the shocks related to HIV/AIDS and struggles to survive (O'Donnell, 2004). While Rugalema (1999, cited by Drimie, 2002) has challenged the usefulness of the term ‘coping strategies’ in the wake of the AIDS pandemic, claiming that the concept has limited value in explaining household experiences of HIV/AIDS as many strategies employed by such households in an attempt to avert household collapse, are not planned and obscure the true cost of coping.

An SADC (2003), study of the impact of HIV/AIDS on households in Malawi, Zambia and Zimbabwe showed that HIV-related morbidity and mortality increased vulnerability to food insecurity and lead to earlier engagement in distress strategies in response to food insecurity, especially among poorer households. SADC (2003, citing: Rugalema, 1999 and de Waal & Tumushabe, 2003), explain that households affected by HIV/AIDS may escape destitution through adopting various response strategies but may not be able to avoid a long-term downward trend in food security and may never fully recover from the shock. Gillespie et al. (2001) also claim that the effect on households affected by HIV/AIDS may be permanent as the coping strategies used by affected households are more likely to be irreversible, compared to the strategies employed by households facing other shocks (Kadiyala & Gillespie, 2003). Rising HIV infection rates could lead to considerably more permanently food insecure households. As good nutrition is vital to maintain health immune systems, malnutrition from sustained food insecurity further threatens resilience of food insecure households (Gillespie et al., 2001), while the reduced capacity to respond may expedite the onset of famine conditions and compromise the ability to recover in the long term.

6 Conclusions

This paper has reviewed current theoretical approaches to food security, outlined categorisation of food security indicators and summarised key current international approaches to empirical estimation of food security (household vulnerability assessments, determination of dietary diversity, experiential measures, and coping strategy indexes). Findings from limited available food security research in South Africa have been presented. While food security-related studies are not directly comparable and sampling frames limit generalisations, available findings have shown that food insecurity and hunger exist in unacceptable levels. Food insecurity is likely to increase with rising food prices, increased reliance on cash food purchases and erosion of household coping strategies due to the impacts of HIV/AIDS.

Why has South Africa not invested in reliable representative longitudinal studies of food insecurity? Perhaps the studies reported here have only recently raised awareness of household food security in a nationally food secure nation? As national food security is not problematic, maybe donor attention to the food crisis in South Africa's neighbouring states has enabled vulnerability assessments to be carried out there, while South Africa lacks the resources to carry out similar analyses? Perhaps empirical estimation of the real extent of hunger and food insecurity in South Africa would have implications government does not want to face in the light that the right to food is enshrined in the constitution? Quantification of the problem may require large-scale investment in appropriate and wide-scale interventions. Current trends in empirical estimation of household food insecurity may reduce the cost of measuring food insecurity but given South Africa's diversity and inequalities, any estimation would require a considerable sample to be nationally representative and inclusive. Qualitative assessments are more resource intense and require more skill than quantitative surveys. However, given the lack of consensus in international empirical methodologies, it is not surprising that multiple approaches and innumerable variables are often included in surveys, increasing respondent burdens, resource requirements and complicating efficient analysis and interpretation.

Nevertheless, empirical research is urgently needed to investigate the coping strategies (or more realistically and HIV/AIDS sensitively phrased as ‘responses’) of rural households in South Africa to develop baseline knowledge of how households respond to household food security stresses and shocks. This knowledge is vital in assessment of vulnerability to food insecurity and monitoring the impact of various shocks on households. Such studies would need to consider the impact of HIV on households through inclusion of ethics-sensitive proxies such as: dependency ratios; presence of chronically ill household members and orphans; and recent deaths in the household.

Large-scale, representative time-series databases including various household food security indicators and covering a range of influencing factors are needed for standardised secondary analysis to evaluate vulnerability to food security, monitor changes in vulnerability over time, and target appropriate interventions. Longitudinal studies with panel samples would be first prize in detecting changes over time, but in the absence of appropriate and relevant secondary databases, smaller in-depth local studies of food insecurity are recommended using multidimensional approaches to investigate the experience of hunger, hunger cycles and food insecurity and how households cope with the stresses and shocks that threaten household vulnerability to food insecurity. Such studies at least enable targeting of appropriate policies and strategies to the most vulnerable, enable testing of methodologies that could lead to standardised survey question sets for inclusion in larger, comparative surveys, and begin to document understandings of household responses to food poverty and hunger in South Africa.

Additional information

Notes on contributors

Sheryl L Hendriks

Director, Food Security Programme, University of KwaZulu-Natal, Pietermaritzburg. The author expresses appreciation for the helpful suggestions made by the two anonymous reviewers.