Sustainability Science

, Volume 11, Issue 2, pp 193–214 | Cite as

Investigating the sensitivity of household food security to agriculture-related shocks and the implication of social and natural capital

  • Byela Tibesigwa
  • Martine Visser
  • Mark Collinson
  • Wayne Twine
Original Article
Part of the following topical collections:
  1. Climate Change Mitigation, Adaption, and Resilience


This paper examines the impact of agriculture-related shocks on consumption patterns of rural farming households using 3 years of data from South Africa. We make two key observations. First, agriculture-related shocks reduce households’ consumption. Second, natural resources and informal social capital somewhat counteract this reduction and sustain dietary requirements. In general, our findings suggest the promotion of informal social capital and natural resources as they are cheaper and more accessible coping strategies, in comparison to, for example, insurance, which remains unaffordable in most rural parts of sub-Saharan Africa. However, a lingering concern centres on the sustainability of these less conventional adaptation strategies.


Food security Natural capital Social capital Weather-related crop failure 


South Africa, the second largest economy in Africa, recently released a national report,1 coinciding with the 2014 World Hunger Day,2 showing that only 46 % of South Africans are food secure and that 26 % experience full-blown hunger (Shisana et al. 2014). Further, the current literature asserts that variability in weather and climatic conditions in South Africa, like elsewhere in the sub-Saharan Africa region, are expected to have considerable adverse impacts on the livelihoods of small-scale subsistence farming households (Kochar 1995; Mirza 2003; Christiansen and Subbarao 2005; Dercon and Krishnan 2000; DEA 2011). Further to this, there are ~1.3 million small-scale farming units in South Africa, and it is estimated that 70 % of South Africa’s poorest households reside in these areas and are said to be food self-reliant (DEA 2011).

The rural farming households are particularly vulnerable because they are mostly dependent on rain-fed agriculture and have low adaptive capacity due to low economic resources (IPCC 2007; Shields and Fletcher 2013). The vulnerability of rural farming households is further worsened by the fact that rural South Africa is mainly characterised by high population densities due to the historic settlement patterns imposed by Apartheid, high levels of poverty and under-developed labour markets (DEA 2011). Even more unfortunate, the majority of the food-insecure South Africans reside in resource-poor rural South Africa (Shisana et al. 2014), hence any weather-related shocks are likely to translate into even more severe food insecurities (FAO 2008; Nhemachena et al. 2010; Nelson 2010; Shields and Fletcher 2013). It is no surprise that one of the national policy concerns is to tackle food insecurities in the era of climate change (DEA 2011).

This paper contributes to the growing literature on the impact of agriculture-related shocks on small-scale subsistence farming households’ consumption patterns (see, e.g., Kochar 1995; Dercon and Krishnan 2000; Dercon 2004; Mogues 2006; Christiansen and Subbarao 2005; Di Falco and Bulte 2009; Osbahr et al. 2010; Porter 2011; Dillon 2012; Dinkelman 2013; Tibesigwa et al. 2015). The assessment is based on a unique panel spanning 3 years (2010–2012) from the Agincourt Health and Demographic Surveillance System (AHDSS) site in rural Mpumalanga, South Africa. The panel consists of rural households whose main sources of dietary needs are small-scale subsistence farming, natural resources such as edible wild fruits, vegetables and insects, and purchased food (i.e., groceries of basic food necessities, e.g., maize meal, cooking oil, salt). In our assessment, we are particularly interested in the interplay between social capital, natural capital and agriculture-related shocks. Thus, we explore the hypothesis that the agricultural shocks are likely to have a lower impact in the presence of social capital and/or natural capital, especially given that several studies suggest that they are pivotal coping strategies in rural South Africa (e.g., Reid and Vogel 2006; Hunter et al. 2007; Kashula 2008). Because the concepts of both social and natural capital are broad, we restrict our focus to membership of groups (formal social capital), reciprocal relations and interactions (informal social capital) and local indigenous natural resources (natural capital).

We depart from and build upon previous related studies in several ways. First, we use caloric and monetary consumption measures as outcomes, on the premise that monetary values are likely to introduce bias because small-scale subsistence farmers are more likely to sell in informal markets (e.g., streets or open markets) where price negotiation is likely to be prevalent. Using two measures allows us to explore the robustness of our results to different outcome measures. Second, unlike the current studies that use endogenous shocks (e.g., crop failure from pests or diseases) and treat such shocks as exogenous regressors, we use agriculture-related shocks caused by weather-related crop failure (poor rainfall or hail storms), hence providing a more exogenous measure. In “Results”, we test this assertion. In addition, we do not only measure whether households experience the shocks but also capture the magnitude of the shocks. Third, we control for the likely self-reported error from recall bias by using an alternative binary variable, where one represents a household that has experienced crop failure and zero otherwise. Lastly, we use a new study area and panel in our assessment, i.e., rural Bushbuckridge in the Mpumalanga Province of South Africa. Thus, the analysis offers new insights from an unexplored area whose population is characterised by substantially high food insecurity, and by dependence on natural resources and agriculture for rural livelihoods (Reid and Vogel 2006; Hunter et al. 2007).

In general, the results indicate that agriculture-related shocks have a negative and significant impact on household consumption and that informal social capital is pivotal in cushioning the most vulnerable households against the shocks. We also observe that the use of natural resources somewhat relieves the impact of such shocks. Quite surprisingly, we observe that formal social capital is significant amongst the least vulnerable, i.e., those who lost a small portion of their crops. Various robustness checks yielded satisfactory results and provided consistent findings. Overall, our results suggest that informal social capital and natural capital can be utilised to facilitate various measures to improve the adaptive capacity of poor rural households, thereby making them less vulnerable to shocks and stresses. The remainder of this paper is organised as follows: the subsequent section contains the body of selected literature relevant to this study, while “Empirical strategy” presents a detailed description of the data and study area, including the definition of variables and the estimation strategy. Thereafter, “Results” presents the descriptive and empirical analysis and the final section provides a conclusion, policy considerations and areas that require further exploration.

Agriculture-related shocks, household responses and related empirical studies

Sub-Saharan Africa remains vulnerable to chronic food insecurity (IPCC 2007; Hunter et al. 2009; Kotir 2011). Food security exists when all people, at all times, have physical and economic access to sufficient, safe and nutritious food which meets their dietary needs and food preferences for an active and healthy life (FAO 2008, p. 3). Table 9 in “Appendix” shows food security statistics from the Food and Agriculture Organisation (FAO).3 Chronic food insecurity is further exacerbated by the fact that almost 70 % of sub-Saharan Africans depend on rain-fed small-scale farming. Hence, any weather-related irregularities are likely to have adverse effects on the food security of many households in the region (Ellis and Freeman 2004; Hellmuth et al. 2007; Kotir 2011; Tibesigwa and Visser 2015). To cushion against such negative weather events, households in turn adopt various methods to boost their dietary or income needs.

The availability of local natural capital such as wild foods (e.g., bushmeat, edible insects, wild fruits and vegetables), fuelwood, and materials for crafts, which are often freely available in rural sub-Saharan Africa, plays an important role in buffering households from food or income shortages (Shackleton and Shackleton 2004; Hunter et al. 2007; Kashula 2008; McGarry and Shackleton 2009). For instance, in reviewing natural resource capital in Africa, Kashula (2008) notes that ~18–32 % of meals in Tanzania, Niger, Ethiopia, South Africa and Swaziland are sourced from natural capital. In another example, a study by Twine et al. (2003) found that, on average, households in a rural district of Limpopo Province of South African use approximately ZAR3959 (US$328 in current terms) worth of local natural resources annually, and that the value was highest in poorest villages. Evidence from another study by Hunter et al. (2007) has shown guxe (one of 41 species of local wild vegetables), marula (a popular local wild fruit) and other wild fruits to be important sources of food and income among rural households in South Africa. In particular, wild leafy greens such as guxe are regularly cooked as a relish that is eaten together with maize meal (a staple food in the region). The marula fruit are eaten raw and the marula nuts are eaten either raw or cooked together with guxe. Residents of Bushbuckridge regard the guxe plant as being commonly available and drought resistant, making it an ideal staple food as well (Shackleton et al. 1998). Likewise, Reid and Vogel (2006) found that, in the rural KwaZulu Natal region of South Africa, women use local grasses, reeds and beads to make crafts, brooms or mats to generate income, thereby decreasing their vulnerability to crop failure. Thus, in general, the role of natural capital in improving food security amongst households in rural resource-poor settings cannot be over-emphasised.

Social capital also plays an important role in food security (Misselhorn 2009). Although a subject of much debate, social capital can generally be defined as the ‘attributes of social relations from which members of formal or informal social networks may derive economic benefits and is often linked to trust, reciprocity and exchange within a community’ (Gilbert and McLeman 2010, p. 15). Formal social capital, as the name suggests, is more formally organised with a management structure and membership dues. Informal social capital, on the other hand, refers to a group or network of people who come together for a common good (Putnam 2001; Pichler and Wallace 2007). In developed countries, these structures are more formal in nature. In contrast, in developing regions such as sub-Saharan Africa, where communities are more integrated, both formal and informal structures exist, with the latter, however, being more prevalent. Such strong social cohesion enables communities to exchange resources in the form of credit or gifts, thus enabling vulnerable households to manage shocks or stresses (Misselhorn 2009; Lippman et al. 2013). For example, Deressa et al. (2009) observe that social capital, such as having relatives in close proximity and farmer-to-farmer extension, enhances households’ adaptation. In support of Deressa et al. (2009) and Osbahr et al. (2010) stress the importance of collective action and building of social capital as an adaptation tool within communities. Echoing a similar view, Tesso et al. (2012) state that households’ participation in local institutions and having relatives in the same area contribute to the resilience of vulnerable households.

This suggests that households’ experiences of shocks are likely to vary depending on the availability of natural capital or social capital. As alluded to earlier, this paper investigates the impact of agriculture-related shocks on small-scale rural farming households’ consumption patterns, with a particular focus on the role of natural and social capital. Next, we provide a literature review of previous studies and highlight our contribution to the current literature. Kochar (1995), in investigating the impact of crop income shocks on household consumption (wage income and borrowing) in India, found that households are able to mitigate against the negative shocks by increasing their participation in labour markets. Importantly, small negative crop shocks had a positive and significant effect, which is unexpected, while larger negative shocks appeared to be insignificant.

Dercon and Krishnan (2000) provide further evidence using a panel of households in rural Ethiopia. The authors found that the consumption patterns (food and non-food consumption in monetary values) were affected by agriculture-related shocks (crop failure from climate, pest, diseases and illnesses) and rainfall shocks. In addition, the authors found food aid initiatives to have relatively marginal effect on relieving households from shocks. Along similar lines, Carter and Maluccio (2003) used a household panel to examine the effects of shocks on child nutritional status (height for age Z-score of a child) in the KwaZulu Natal region of South Africa. Similar analysis can be found in the studies by Yamano et al. (2005), Akresh et al. (2011) and Dillon (2012). Slightly different from the aforementioned studies, measures the relationship between livestock assets, environmental shocks and social capital in north-east of Ethiopia. In another empirical investigation by Dercon (2004), using a panel of households from rural villages in Ethiopia, the study found that rainfall shocks, agriculture-related shocks (crop damages from frost, animal trampling, weed and plant diseases) and livestock suffering index (lack of water or fodder) have adverse effects on consumption (in monetary and caloric values).

On the other hand, Christiansen and Subbarao (2005) use repeated cross-sectional data from households in the same community in Kenya. The authors conclude that households that experienced rainfall shocks were more vulnerable, especially those in arid areas, and that illness shock had non-negligible effects on consumption (food expenditure per adult). In a like manner, Salvatori and Chavas (2008) measured the effects of rainfall shocks on agro-ecosystems productivity in southern Italy. In the same spirit, Di Falco and Bulte (2009) measured the effects of weather shocks and the role of social capital (kinship networks) in adaptation to climate change in rural Ethiopia. Similarly, Porter (2011) measured the effects of rainfall shock and agriculture-related shocks (crop failure due to illness and crop pests) on consumption (household consumption in monetary values) in rural Ethiopia. Porter (2011) finds the rainfall shock to be negatively related to consumption. However, agriculture-related shocks have a positive relationship with consumption, which is unexpected. The authors attribute this to the bias in self-reporting shocks or to the definition of the outcome variable, which did not include consumption from gifts of food.

Complementing and building from the above-mentioned studies, the current study investigates the impact of agriculture-related shocks on consumption patterns among rural households. As previously stated, we use a unique panel from the Bushbuckridge, a former Apartheid Bantustan or homeland region, in Mpumalanga Province, South Africa. The panel covers 3 years and contains information that offers valuable insights into the human–environment relationships. For the majority of the households in this area, rain-fed homestead farming and harvesting of natural resources play an important role in their livelihoods (Shackleton and Shackleton 2004; Twine and Hunter 2011). The region represents a typical rural setting in South Africa, characterised by poverty, high dependence on remittances and migrant labour, high human density and limited formal labour markets (Hunter et al. 2009; Twine and Hunter 2011). In synthesising the above review of current empirical studies, we observe that there appear to be mixed results. While some studies have found the effect of household shocks to be negative and significant, as expected, other studies have found the results to be insignificant, and others have had positive and significant results. This variation in results can be attributed to various factors. In an attempt to explain the likely causes of this variation, we also highlight our contribution to the current studies.

First, while rainfall shock is a strictly exogenous measure, agriculture-related shocks from crop failure may be either exogenous or endogenous. Crop failure is likely to be exogenous if it is weather-related, for example, poor rainfall, hailstorms, floods or frost. However, crop failure is likely to be endogenous if the source is from pests or diseases, as this is likely to be correlated with the effort one exerts on the farm. That is, if a household invests more effort by using more labour, pesticides or herbicides, then it is likely to experience minimal crop failure in comparison to a household that invests less effort. The endogeneity of crop failure from pests and diseases is likely to decrease with decreases in household income. For example, poor households are more likely to lack resources needed to purchase pesticides. In the current study, weather-related crop failure is an agriculture-related shock, and, as such, this is likely to be an exogenous measure. This assertion is tested in “Results”.

Second, we recognise the short-fall in self-reported variables, which may be biased as a result of the recall error. Accordingly, we use an alternative binary regressor, represented by one if crop failure was experienced and zero otherwise. Third, in general, although most small-scale farming households in our study site produce solely for their own consumption, those that sell their products do so in informal markets (e.g., streets or open markets) where buyers and sellers engage in price negotiation. Because of this negotiation process, there is likely to be a very high degree of variation in prices in these informal markets. Thus, using monetary values is likely to introduce measurement error in the variable and to bias the estimation results. Accordingly, in the current study, we use caloric and monetary consumption measures. Fourth, some of the past empirical models are likely to be influenced by unobservables due to the cross-sectional analysis. We control for unobservable heterogeneity by using panel data methods.

Empirical strategy

Econometric model

In describing the empirical model and the variables used for estimation, we follow the current literature and define a consumption function depicted by Eq. (1).
$$y_{it} = f({\text{SHOCK}}_{it} , X_{it} ),$$
where yit is per capita consumption belonging to household i at time t. SHOCKit is categorical in nature and captures a negative agricultural-related shock experienced by household i at time t, Xit are household characteristics (education and age of the head of the household, size of the household, household income, informal and formal social capital). We estimate Eq. (1) with fixed-effects model to account for potential time-invariant unobserved individual heterogeneity shown in Eq. (2), where \(\gamma_{i}\) is the unobservable heterogeneity, which captures the time-invariant effects, while \(\varepsilon_{it}\) is the random error term
$$y_{it} = \beta_{1} + \beta_{2} {\text{SHOCK}}_{it} + X_{it} \beta + \gamma_{i} + \varepsilon_{it} .$$

The advantage of fixed-effects model is that it controls for unmeasurable time-invariant factors that may likely influence yit and SHOCKit. An example of such unobservable influence is farm management or ability.

Study area, data and definition of variables

This study uses the first 3 years (2010–2012) of the sustainability in communal socio-ecological systems (SUCSES) panel data. SUCSES is nested in the Agincourt Health and Socio-Demographic Surveillance System (AHDSS), located in Bushbuckridge local municipality in the Mpumalanga Province of South Africa. The field-site covers 27 villages with a population of 90,000 inhabitants who are enumerated annually (Kahn et al. 2012). SUCSES investigates the relationship between rural livelihoods, the environment, and human well-being in a communal tenure system. Land tenure is communal, with households in a village sharing access to grazing and woodland resources in the village commons. Most households grow food crops in homestead gardens, while the minority have larger fields outside of the village. Land for these is allocated by the village headman. In the SUCSES study, a detailed questionnaire was used to collect diverse and rich information on livelihood capital (financial, physical, social, human and natural), activities (on-farm and off-farm economic activities, migration, and natural resource harvesting) and well-being outcomes (health, food and nutrition and heights and weights of children). The SUCSES panel consists of a random stratified sample of 590 households. The current study is based on an uneven panel of 1528 observations, with ~500 households per wave. In Fig. 1 we depict the geographical boundary of the AHDSS field-site, and for more details on the AHDSS, please see Kahn et al. (2012).
Fig. 1

Agincourt study site map

While we are interested in the impact of agriculture-related shocks, it is important to get a comprehensive measure of all food sources accessed by the household. Accordingly, we use three consumption outcomes: consumption from crop farming only, consumption from crop farming and natural resources gathered from the local environment and, lastly, a combination of consumption from crop farming and natural resources with groceries (i.e., food purchases). We use both caloric and monetary measures to obtain these consumption outcomes. Our first measure, monthly caloric consumption per capita from crop farming, is derived by adding together the calorie content of all crops harvested. This is then divided by the household size (number of household members). In this conversion, we use the FAO conversion tables.4 The second measure is monthly caloric consumption per capita from crop farming and natural resources. The measure extends the previous measure by including household consumption of natural resources. These natural resources include wild fruits, wild vegetables, edible insects, fish from local rivers and bushmeat obtained from the local environment. Our third and final outcome is monthly monetary consumption per capita from crop farming, natural resources and groceries (food purchased). Thus, unlike the previous measures, which capture partial household consumption, this measure portrays a more comprehensive picture of household consumption. Also, unlike the previous measures, here we include the total monthly expenditure on food purchased (groceries) and produced (farming and natural resources), and then divide by the total number of household members.

We favour caloric measures over monetary measures of consumption, because caloric measures reduce the bias associated with monetary measures. This follows from our earlier example of small-scale farmers and price negotiation. Thus, using monetary values (as opposed to caloric values) are likely to introduce measurement error. We anticipate that this bias is likely to decrease with increases in farm size. Furthermore, even if one uses self-reported monetary values, it is unlikely that the households will recall the prices of their products due to the likely high price variation over time. A similar argument holds for monetary expenditure measures. First, households with higher incomes are likely to consume from formal markets while those with lower income consume from informal markets. Second, and as before, even if one uses monetary values self-reported by the households, it is unlikely that they will recall the prices of their household food expenditure. This recall bias is likely to be skewed toward those who purchase in the informal markets in comparison to those who purchase in the formal markets.

Note that according to Danes et al. (1987): the term informal market sector has been used to describe income-generating economic activities in the Third World that are characterised by ease of entry, reliance on indigenous resources, family ownership and labour in the enterprises, small-scale operation, labour-intensive technology, and unregulated markets (Danes et al. 1987, p. 632). Some examples of informal markets include: street vendors, open markets (Danes et al. 1987; Figuié and Moustier 2009; Brown et al. 2010). Danes et al. (1987) state that ‘poor families are usually the people who participate in informal market activities’ (Danes et al. 1987, p. 633). Danes et al. (1987) further note that ‘Informal market activities differ from formal market activities because they are casual and non-permanent, lack company or government regulations, and take place on a small scale and in less capitalised establishments, relying on household labour (Danes et al. 1987, p. 632). Some examples of formal markets include supermarkets (Figuié and Moustier 2009), where prices are usually higher than the informal markets (Figuié and Moustier 2009). In our study, we follow these terms and define informal markets as unregulated open markets for agriculture produce characterised by price variation, and formal as regulated markets, e.g., supermarkets where prices are likely to be higher and more constant.

Returning to our definition of variables, the main regressor is agriculture-related shocks, defined as crop failure from (1) poor rainfall and (2) hail or strong winds. This information was obtained from the following question: “In the past season, how much crop loss did you experience due to the following causes?” The responses options were, ‘none’, ‘a little’, ‘some’, ‘most’ and ‘all’, referring to the proportion of crop that failed or was lost. We defined ‘a little’ as an observed but inconsequential amount, while ‘some’ was defined as a consequential amount but less than half of the crop, and ‘most’ was defined as more than half, but not all, of the crop. Although these are not precise categories, they are an expression of perceived relative severity of the shock. Since the survey was done between April and June each year, at end of the growing season (October–March), the recall period was under 6 months. Like all self-reported variables, our regressor is likely to be prone to measurement error (see Carter and Maluccio 2003). Measurement error becomes harmful if it is systematic (Greene 2002). We expect the error to be systematic because it is easier for a household with fewer food sources (e.g., a household with a small garden) to remember the amount of crops they lost than a household with more food sources. Our strategy to overcome this bias is to use an alternative binary variable, where one represents a household that has experienced crop failure, and zero otherwise.

We add various household characteristics, following the current literature. These include education and age of the head of the household. The size of the household, which captures the total number of household members, is also included. We account for different household income sources: labour income, agriculture income and natural resource income (firewood, wild fruits and vegetables, edible insects, fish from local rivers, bushmeat and medicinal plants) by means of dummy variables represented by 1 if the household receives the income and 0 otherwise. It is reasonable to assume that, in the event of agricultural-related shocks, households with multiple sources of income are less impacted and more able to adapt than are households whose livelihoods entirely depend on farming (see Kochar 1995; Christiansen and Subbarao 2005; Birhanu and Zeller 2009; Porter 2011). We also include social capital (informal and formal). It is expected that social capital will enable households to cope with stresses and shocks (see Misselhorn 2009; Deressa et al. 2009; Osbahr et al. 2010; Cavatassi et al. 2011; Tesso et al. 2012). Following Pichler and Wallace (2007), we define formal social capital as “participation in formally constituted organisations and activities” (Pichler and Wallace 2007, p. 423).

The term formal is attached because of existing structures that register them as organisations or associations. This is aligned with the literature on democracy and civil society, e.g., social clubs, churches or clubs (Pichler and Wallace 2007). Accordingly, our measure of formal social capital is household membership in the following associations: farmers’ association, grocery stokvel (saving club) or business association. Grocery stokvel is the most common type of formal social capital in our data. In contrast, informal social capital, which is more aligned with social network literature, is “the density, strength (i.e., the extent to which people give or provide services of different kinds) and extensiveness of social networks with colleagues, friends and neighbours” (Pichler and Wallace 2007, p. 427). Our measure of informal social capital is the ability of households to ask for assistance from relatives, neighbours or friends in matters related to household needs (e.g., food, money, transport, fuel, child and elderly care, clothes and uniforms) in times of household stresses. Our current data show that food is the most prevalent type of assistance that these rural households receive from their informal networks. Hence, unlike formal social capital, informal social capital here refers to the exchange of food and other household necessities and lacks a functioning structure.5


Data description

Table 1 shows the descriptive statistics. We find 53 as the average age of the heads of households. We also observe that the average household contains 8 household members (both permanent and migrants). The descriptive statistics also reveal that 57 % of the household earn some form of labour income, 12 % receive income from agricultural activities and 11.5 % receive income from selling natural resources. Additionally, Table 1 indicates that, on average, most households have experienced agriculture-related shocks. We find that 45.3 % of the households have access to formal social capital and that 60.3 % of the households have received assistance from close friends, relatives and neighbours. Lastly, 51.9 % have given some form of assistance to other households.
Table 1

Summary statistics




Log kcal consumption (crops) per capita



Log kcal consumption (crops, natural resources) per capita



Log monetary consumption (crops, natural resources and groceries) per capita



Agricultural-related shocka



Informal social capital



Formal social capital






Household size



Agriculture income source



Natural resource income source



Trade income source



aAgricultural-related shock: 0 is ‘none’ of the crops were destroyed, 1 is ‘a little’ of the crops were destroyed, 2 is ‘some’ of the crops were destroyed, 3 is ‘most’ of crops were destroyed, 4 is ‘all’ of the crops were destroyed

Further exploration of the data reveals that the majority of the households keep the agricultural output for their own consumption, with just 4.5 % of the households selling the crops they harvest. This supports current literature that states that small-scale farming in sub-Saharan Africa is often subsistence in nature, where the main reason for participating in farming is to supplement dietary needs. This also explains the low number of households with agriculture- related income in the descriptive statistics (Table 1). Table 2 shows the distribution of households’ experience of agriculture-related shocks. Table 2 shows that almost 78.1 % of the households have experienced such shocks, with the majority of them (31.5 %) having lost ‘most’ of their crops in the 2010–2012 period. Note that crop loss from poor rainfall (64.5 %) is more common than crop loss from hail storms and strong winds (10.9 %), probably because, in our study area, storms are geographically localised and of short duration.
Table 2

Percentage of households who have experienced agriculture-related shocks






‘None’ of the crops were lost to poor rainfall or hail storm





‘A little’ of the crops were lost to poor rainfall or hailstorm





‘Some’ of the crops were lost to poor rainfall or hailstorm





‘Most’ of the crops were lost to poor rainfall or hailstorm





‘All’ of the crops were lost to poor rainfall or hailstorm





We compare our outcome variables to the FAO statistics in Table 9. Our data show that the average food consumption, in monetary value, is US$1.26 per capita per day. This is somewhat consistent with FAO statistics from other parts of sub-Saharan Africa, which reveal a range between US$0.05 and 1.62 amongst individuals in the low income percentiles and US$0.09–3.04 for those in the middle income percentiles. We compare with the poor and middle income individuals because these individuals are likely to be similar to the individuals in our data. Further, our data show that the average food consumption, using caloric values, from crop farming alone and crop farming together with natural resources is 452.1 and 567.8 kcal per capita per day, respectively. These values also fall within the range of FAO statistics, when we compare with the share of dietary energy from own food production in Table 9. In particular, the statistics from FAO show that the caloric consumption from the production of own food ranges between 188.1–1485.2 kcal for low income percentiles and 139.4–1572.0 kcal for middle income percentiles.6

Regression results

Household consumption and agriculture-related shocks

Table 3 reports the baseline results, where we begin by analysing the effect of agriculture-related shocks on different per capita household consumption measures. Note that the Hausman test rejects the null hypothesis of the regressors being correlated with the error term, hence we only report the fixed-effects models. In Panel A (Columns 1–3), we include agriculture-related shocks but suppress household characteristics. In specific, Column 1 uses consumption from crop farming as the outcome and agriculture-related shocks as the only regressor, while Column 2 reports estimates for our second outcome: caloric intake from crop farming combined with natural resources. Finally, Column 3 shows estimates from consuming crops, natural resources and groceries. Overall, and most importantly, we observe qualitatively similar results: agriculture-related shocks are negatively associated with all the per capita household consumption measures. In particular, the expected percentage decrease in caloric intake between households who did not lose any crops compared to those who ‘lost most of their crops’ is about 33.9 %, and this decrease is 76.5 % amongst households who ‘lost all their crops’. Moving to Column 2, the percentage is 34.0 and 72.1 %, respectively, while in Column 3 we will expect a percentage decrease of 21.3 and 47.8 %, respectively, in per capita household consumption.
Table 3

Impact of negative agriculture-related shocks on household consumption

Dependent variable

Panel A without household characteristics

Panel B with household characteristics







ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

Shock, lost a little crop

0.132 (0.104)

0.0559 (0.0858)

0.0599 (0.0605)

0.152 (0.104)

0.0730 (0.0845)

0.0739 (0.0581)

Shock, lost some crops

0.0136 (0.0992)

−0.0497 (0.0773)

−0.0645 (0.0607)

0.0254 (0.0985)

−0.0470 (0.0764)

−0.0642 (0.0581)

Shock, lost most of the crops

−0.414*** (0.0848)

−0.416*** (0.0730)

−0.239*** (0.0559)

−0.388*** (0.0866)

−0.395*** (0.0721)

−0.222*** (0.0541)

Shock, lost all of the crops

−1.448*** (0.364)

−1.278*** (0.295)

−0.649*** (0.193)

−1.398*** (0.370)

−1.251*** (0.297)

−0.632*** (0.187)

Head of household age


0.169*** (0.0562)

0.119*** (0.0420)

0.103*** (0.0234)

Head of household age2


−0.00171*** (0.000521)

−0.00122*** (0.000368)

−0.000993*** (0.000205)

Number of household members


−0.884*** (0.245)

−0.984*** (0.174)

−0.835*** (0.134)

Income source: agriculture


0.238* (0.126)

0.254*** (0.0978)

0.246*** (0.0749)

Income source: natural resource


0.158 (0.105)

0.191** (0.0915)

0.0746 (0.0818)

Income source: labour


0.186** (0.0807)

0.152** (0.0632)

0.145*** (0.0461)


8.823*** (0.0518)

9.178*** (0.0416)

5.630*** (0.0317)

6.546*** (1.449)

8.342*** (1.141)

4.654*** (0.635)















Number of observations







Robust standard errors in parentheses

Reference category for shock (crop failure) is none

*** p < 0.01, ** p < 0.05, * p < 0.1

The negative relationship suggests that the shocks lead to a reduction in caloric intake for each of the household members. This result is in line with our expectation and is broadly consistent with previous studies that have observed a decrease in household welfare after experiencing a negative shock (e.g., Dercon 2004; Porter 2011). Also important, we observe that the magnitude of the shock matters, as the coefficients are negative and significant, at the 1 % level, amongst households who lost ‘most’ and ‘all’ of their crops and insignificant among those who lost ‘a little’ and ‘some’ of their crops. Consumption is therefore likely to be lower amongst these households in comparison to those who did not lose any crops. This suggests that the shocks affect the most vulnerable households.

More important, the size of the coefficients reduce as we move from Column 1 to Column 3, i.e., when we include consumption from natural resources (Column 2) and groceries (Column 3). This suggests that the shocks have stronger impact when we consider caloric intake from crop farming only (an activity mainly engaged into contribute to household dietary requirements in this rural setting), and this impact wears out once we include consumption from natural resources and groceries. This indicates that households’ consumption of natural resources and purchased food buffer households against agriculture-related shocks. Hunter et al. (2007) showed that wild foods, such as edible insects, may substitute previously purchased foods in households dealing with the livelihood impacts of an adult mortality due to HIV/AIDS. Our results are consistent with studies in other parts of sub-Saharan Africa where natural resources have been identified as a key strategy in increasing the livelihood viability of households in resource-poor rural settings (see, e.g., Omolo 2010).

An advantage of natural resources, is that they are freely available in rural communal areas. Another advantage is that they (e.g., guxe) are more resilient to weather variability in comparison to crop farming (Shackleton et al. 1998; Hunter et al. 2007). However, the use of natural resources as an adaptation method is vulnerable to unsustainable use. That is, while some of the resources, such as guxe herbs and marula fruit are often gathered in homestead yards and fields, most resources are harvested from communal woodlands surrounding the villages. Added to this, the current high population densities, chronic poverty, and weakening traditional governance institutions in the Bushbuckridge local municipality, our study area, are already a threat to sustainability (Twine 2005; Kirkland et al. 2007). This is further exacerbated by the fact that climate variability is expected to become more pronounced into the future, and the evidence suggests that small-scale subsistence farming households, particularly the poorest, can be expected to increase their dependence on natural resources.

According to (IPCC 2014), ‘adaptation is the process of adjustment to actual or expected climate and its effects. In human systems, adaptation seeks to moderate harm or exploit beneficial opportunities. In natural systems, human intervention may facilitate adjustment to expected climate and its effects’ (IPCC 2014, p. 838). However, adaptation efforts have somewhat neglected sustainable development, especially when addressing the most food-insecure and vulnerable populations (Eriksen et al. 2011). It is particularly important to associate adaptation with sustainability. Here, sustainable adaptation is defined as “adaptation that contributes to socially and environmentally sustainable development pathways, including both social justice and environmental integrity” (Eriksen et al. 2011, p. 8). The highest priority therefore, in resource-poor settings, is a win–win policy design that successfully links natural capital adaptation with sustainability. This is easier said than done, and has indeed proven to be a challenge in the current policy-making process (Burton and Development Programme United Nations 2005).

In Panel B (Columns 4–6), we proceed to run the same regressions, but here, we introduce household characteristics. The agriculture-related shocks estimated in Panel B mirror those we found in Panel A. In addition, and as expected, Panel B shows that the consumption levels decrease with increment in household size. This is evident in the negative and significant household size coefficient. Also, it is apparent that the age of the head of the household has a non-linear relationship with household caloric intake. Panel B further shows positive and significant coefficients on the household income sources (labour, agriculture and natural resources). This indicates that households who receive income from participating in labour markets are more likely to have higher consumption. Also, households with some income from agriculture activities or from selling natural resources are also more likely to have higher consumption levels. In summary, in this section we uncovered two key observations: first, the agriculture-related shocks reduce consumption levels and hurt the most vulnerable households. Second, having additional consumption from natural resources and groceries somewhat minimises the effects of the shocks.

Household consumption, agriculture-related shocks, and formal and informal social capital

In the previous section, we found that the agriculture-related shock affects the most vulnerable and that natural resources and additional food purchases act as a buffer against the shocks. Here, we introduce social capital. Our data contain detailed information on social capital: formal and informal. In addition, we are able to differentiate between informal social capital-receive, which is the ability to receive assistance, and informal social capital-give, which is the ability of households to give assistance. Our data show that 51.8 % of the households have given some form of assistance, while 60.3 % have received some form of assistance. Figure 2 shows the distribution of the informal social capital by household income quintiles, while Fig. 3 shows distribution of formal social capital. We observe that formal social capital, in Fig. 2, is higher amongst the higher income households, while informal social capital is equally distributed across all income levels.
Fig. 2

Distribution of informal social capital by income quintile

Fig. 3

Distribution of formal social capital by income quintile

We proceed to extend the baseline analysis in Table 3, by re-estimating the regressions and including the different measures of social capital as regressors. The results are reported in Table 4. Here, we continue to observe a pattern qualitatively similar to Table 3. In addition to the similarity with our baseline results, here, we find both informal and formal social capital to be insignificant, suggesting that they do not have any direct effect on caloric intake.
Table 4

Impact of negative agriculture-related shocks and social capital on household consumption

Dependent variable

Panel A without interactions

Panel B with interactions







ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

Shock, lost a little crop

0.139 (0.106)

0.104 (0.0939)

0.0825 (0.0590)

0.00927 (0.184)

−0.00477 (0.149)

−0.0697 (0.0989)

Shock, lost some crops

−0.0178 (0.0962)

−0.0856 (0.0872)

−0.0806 (0.0595)

0.158 (0.174)

0.00220 (0.168)

−0.0367 (0.0979)

Shock, lost most of the crops

−0.432*** (0.0841)

−0.419*** (0.0815)

−0.234*** (0.0547)

−0.331** (0.151)

−0.290** (0.144)

−0.258*** (0.0902)

Shock, lost all of the crops

−1.433*** (0.375)

−1.118*** (0.368)

−0.624*** (0.184)

−0.615 (0.431)

−0.748* (0.395)

−0.618 (0.377)

Informal social capital, receive

0.0201 (0.107)

−0.0811 (0.0950)

0.00330 (0.0674)

−0.0709 (0.215)

−0.141 (0.177)

−0.0527 (0.114)

Shock, a little* informal social capital, receive


0.335 (0.325)

0.370 (0.295)

0.366 (0.234)

Shock, some* informal social capital, receive


0.000673 (0.288)

0.215 (0.255)

0.0781 (0.183)

Shock, most* informal social capital, receive


0.0662 (0.273)

−0.145 (0.272)

−0.0111 (0.151)

Shock, all* informal social capital, receive


4.244*** (0.745)

2.799*** (0.771)

0.403* (0.226)

Informal social capital, give

0.0431 (0.0758)

0.0851 (0.0811)

0.0657 (0.0425)

0.251 (0.153)

0.246 (0.153)

0.0876 (0.0801)

Shock, a little* informal social capital, give


−0.162 (0.232)

−0.228 (0.205)

0.0536 (0.128)

Shock, some* informal social capital, give


−0.352* (0.205)

−0.253 (0.182)

−0.108 (0.120)

Shock, most* informal social capital, give


−0.302 (0.199)

−0.213 (0.179)

−0.0166 (0.111)

Shock, all* informal social capital, give


−1.445** (0.628)

−0.424 (0.646)

0.0311 (0.420)

Formal social capital

−0.00915 (0.0727)

−0.0662 (0.0610)

−0.0135 (0.0435)

−0.112 (0.129)

−0.138 (0.117)

−0.0736 (0.0801)

Shock, a little* formal social capital


0.415** (0.207)

0.438** (0.192)

0.226* (0.124)

Shock, some* formal social capital


0.0372 (0.202)

0.0584 (0.174)

0.0117 (0.126)

Shock, most* formal social capital


0.0870 (0.185)

−0.0116 (0.170)

0.0789 (0.104)

Shock, all* formal social capital


−0.771 (0.760)

−1.283 (0.818)

−0.330 (0.272)

Head of household age

0.155*** (0.0571)

0.141*** (0.0452)

0.115*** (0.0252)

0.160*** (0.0568)

0.146*** (0.0452)

0.115*** (0.0260)

Head of household age2

−0.00159*** (0.000512)

−0.00138*** (0.000392)

−0.00108*** (0.000218)

−0.00162*** (0.000505)

−0.00142*** (0.000387)

−0.00108*** (0.000223)

Number of household members

−0.891*** (0.248)

−1.107*** (0.184)

−0.860*** (0.139)

−0.855*** (0.252)

−1.084*** (0.187)

−0.855*** (0.141)

Income source: agriculture

0.289** (0.124)

0.401*** (0.151)

0.243*** (0.0777)

0.286** (0.123)

0.409*** (0.150)

0.248*** (0.0771)

Income source: natural resource

0.158 (0.106)

0.156* (0.0930)

0.0695 (0.0835)

0.183* (0.109)

0.169* (0.0960)

0.0680 (0.0828)

Income source: labour

0.158* (0.0808)

0.114 (0.0817)

0.130*** (0.0478)

0.152* (0.0798)

0.116 (0.0835)

0.132*** (0.0487)


7.029*** (1.484)

7.843*** (1.236)

4.329*** (0.684)

6.712*** (1.506)

7.551*** (1.278)

4.318*** (0.720)















Number of observations







Robust standard errors in parentheses

Reference category for shock (crop failure) is none

*** p < 0.01, ** p < 0.05, * p < 0.1

Continuing with social capital, Panel B introduces the interaction effects. In general, the results in Panel B echo the previous panel, with only two key differences: the shocks coefficients are smaller (in comparison to Panel A) and, although the coefficients of formal and informal social capital remain insignificant, the interaction coefficients are significant. On the whole, Panel B shows some interesting results. First, we observe that informal social capital is more effective among the most vulnerable households, i.e., those that lost the majority of the agriculture products. This is evident in the shock to the informal social capital-receive interaction coefficient, which is positive and significant amongst those who lost ‘all’ their crops. This indicates that the effects of the shocks are lessened amongst the most vulnerable when households receive assistance (informal social capital), which in turn increases their consumption levels. Stated differently, this suggests that, in times of stresses and shocks, when consumption is low, the transfer of food becomes a lifeline for the most vulnerable households. Second, there appears to be a trade-off between giving and receiving assistance. That is, although we observe that consumption increases when a household receives assistance, we find that, when assistance is offered to other households, this reduces consumption. This is shown by the shock to the informal social capital-give interaction coefficient, which is negative and significant amongst those who lost ‘all’ of their crops.

Third, and related to the above observation, there appear to be some entangled mechanisms, perhaps pointing to something even beyond a trade-off, to cultural or familiar expectations/pressure, such that households feel obligated to offer assistance even when they themselves are being assisted. The coefficient on the interaction between households that have lost most of their crops and receiving assistance has a positive effect on caloric intake; whereas, for similar households, being involved in giving assistance to other households significantly lowers caloric intake. This emphasises the heightened vulnerability of such households, having lost a large portion of their normal caloric intake, but also the important role of social ties in either buffering caloric-poor households against agricultural shocks or placing further strain on the resources of the household, depending on the direction of the caloric exchange. A potential explanation for this finding is the set-up in rural communities. Rural communities are characterised by close ties (Hofferth and Iceland 1998), and according to Coleman (1988), these ties consist of strong interpersonal relationships, with mutual obligations, expectations and reciprocity. The observed giving and receiving of assistance is also somewhat consistent with current literature. For example, a study by Hofferth and Iceland (1998) investigated the type, prevalence and extent of social exchanges and found that receiving and giving assistance is more common in rural than in urban areas. Also, Goudge et al.’s (2009b) qualitative study reported the following verbatim finding: ‘When I cannot get enough money to buy food it is difficult to go out and borrow because I know I will not be able to repay the money on time. I do go to the neighbours to borrow, say, mielie meal, but only to find that they are also running low which makes it difficult, but at times people do give without expecting me to return it’ (Goudge et al. 2009a: p. 246).

Fourth, in Columns 5 and 6, we find that giving assistance no longer reduces consumption (in contrast to Panel A), as shown by the insignificant shock to informal social capital-give interaction coefficients. Taken together, this suggests that having additional food from natural resources (Column 5) and groceries (Column 6) has a somewhat cushioning effect against shocks and food transfers (i.e., social capital-give) as well. Fifth, surprisingly, the formal social capital (membership in an association) becomes significant among the less vulnerable, i.e., those who lost a little of their crops. This is somewhat of a puzzle in that formal social capital is effective amongst the less vulnerable and ineffective among the most vulnerable. A plausible explanation is that this observation may be driven by the fact that households with more economic resources, who are likely to be less vulnerable, are more likely to afford membership fees and other requirements associated with being a member of a formal association. On the other hand, the most vulnerable, who are likely to have lower economic resources, are more likely to be excluded as a result of membership requirements. Nonetheless, formal social capital has been found to be significant in other settings. For example, Deressa et al. (2009) and Cavatassi et al. (2011) found formal social capital (farmers’ associations, networks for seed exchange) to be significant in predicting farmers’ adaptation decisions (soil conservation, crop varieties, planting trees, changing planting date, irrigation, no adaptation). Similarly, social capital has been linked with increased food security (see, e.g., Misselhorn 2009).

Our results suggest the need for further in-depth probing into food security of farm households in resource-poor rural settings. In this regard, we recommend the following promising areas of future research: first, this study could be expanded to assess the role of norms, values and symbolical organisation on social capital and food security. Second, another study could explore the type of assistance that will be more significant in improving food security in the presence of climate shocks, e.g., is it financial, physical or human capital. Third, recall that we use two outcomes: real consumption per capita and caloric consumption per capita. In both outcomes, to derive the per capita measures, we divide by the number of household members. We use this approach so as to obtain consistent measures which are comparable. This approach assumes that all household members receive equivalent food intake. We suggest that future studies consider adult equivalent measures. For further discussion on adult equivalent measures see Claro et al. (2010). Finally, in addition to social and natural capital, there are other more accessible alternative coping strategies, e.g., crop/livestock diversification and modification of cropping patterns. Because these areas are not covered in this study, we suggest that future studies explore these equally important strategies in relation to food security of farm households in resource-poor rural settings.

Robustness checks

Before we conclude, it is important to investigate whether our results remain consistent after we address potential estimation pitfalls. To this effect, in addition to testing the response of different consumption measures in the previous section, this section first tests whether measurement error in self-reported agriculture-related shocks influence our results. Second, we test whether the results hold after we introduce household income, which is likely to be endogenous, as a control. Third, we test our assertion of exogeneity of the agriculture-related shocks.

Measurement error in reported agriculture-related shocks

As previously explained, the agriculture-related shocks regressor is likely to face measurement error (Carter and Maluccio 2003), and this is harmful (Greene 2002) because the error is likely to be systematic. We say it is systematic because the error is likely to vary by household vulnerability. For instance, a vulnerable household with a small garden is more likely to remember how much crop they lost than is a less vulnerable household with many alternative food sources. We curb this bias by using a binary measure. This binary regressor takes the value of one if the household has experienced the shock and zero otherwise. Table 5 re-estimates the regressions using a binary agriculture-related shock. Our coefficient of interest shows that households who have experienced the shock are likely to have less consumption, which is consistent with our previous finding.
Table 5

Impact of negative agricultural related shock using binary shock regressor

Dependent variable

Panel A without interactions

Panel B with interactions







ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

Shock, crop failure

−0.191** (0.0775)

−0.207*** (0.0705)

−0.118** (0.0456)

−0.132 (0.133)

−0.158 (0.127)

−0.162** (0.0765)

Informal social capital, receive

0.0134 (0.110)

−0.0903 (0.0989)

−0.00267 (0.0682)

−0.128 (0.213)

−0.197 (0.175)

−0.0917 (0.119)

Shock* informal social capital, receive


0.181 (0.245)

0.137 (0.206)

0.115 (0.147)

Informal social capital, give

0.0168 (0.0766)

0.0619 (0.0816)

0.0518 (0.0427)

0.229 (0.154)

0.222 (0.155)

0.0738 (0.0806)

Shock* informal social capital, give


−0.303* (0.172)

−0.229 (0.153)

−0.0280 (0.0952)

Formal social capital

0.0356 (0.0735)

−0.0273 (0.0624)

0.0107 (0.0430)

−0.117 (0.132)

−0.135 (0.120)

−0.0698 (0.0802)

Shock* formal social capital


0.193 (0.160)

0.135 (0.143)

0.112 (0.0923)

Head of household—age

0.144** (0.0588)

0.131*** (0.0465)

0.111*** (0.0249)

0.142** (0.0585)

0.130*** (0.0469)

0.107*** (0.0253)

Head of household—age2

−0.00151*** (0.000528)

−0.00131*** (0.000403)

−0.00105*** (0.000216)

−0.00149*** (0.000525)

−0.00130*** (0.000405)

−0.00102*** (0.000219)

Number of household members

−0.863*** (0.249)

−1.078*** (0.184)

−0.836*** (0.138)

−0.858*** (0.249)

−1.074*** (0.184)

−0.836*** (0.139)

Head of household—education dummy

−0.110 (0.119)

−0.0299 (0.0920)

0.0130 (0.0674)

−0.110 (0.118)

−0.0292 (0.0916)

0.00860 (0.0677)

Income source: agriculture

0.290** (0.125)

0.402*** (0.153)

0.243*** (0.0790)

0.295** (0.125)

0.405*** (0.152)

0.247*** (0.0791)

Income source: natural resource

0.165 (0.111)

0.165* (0.0946)

0.0775 (0.0845)

0.188* (0.114)

0.182* (0.0966)

0.0811 (0.0844)

Income source: labour

0.191** (0.0829)

0.142* (0.0827)

0.145*** (0.0482)

0.186** (0.0828)

0.138 (0.0838)

0.146*** (0.0486)


7.294*** (1.525)

8.063*** (1.269)

4.403*** (0.673)

7.328*** (1.537)

8.077*** (1.303)

4.534*** (0.697)















Number of observations







Robust standard errors in parentheses

*** p < 0.01, ** p < 0.05, * p < 0.1

Adding household income as an additional control

Thus far, the estimations have included sources of income dummies as controls and have omitted household income. This is because introducing household income brings with it endogeneity. Here, we measure whether our results will be consistent once we include household income as an additional control. Our assertion that household income is likely to be endogenous emanates from past empirical studies. A potential source of endogeneity is reverse causality between income and the consumption outcome, in that income enters a consumption function, and, in like manner, consumption enters an income function through nutrition/health; for example, a healthier/more nourished person is more likely to earn more income. We use lagged income value as an instrument to mute the endogeneity in the income regressor.

Table 6 presents the results from the fixed-effects IV (FE2SLS) model. After controlling for household income, in Table 6, the coefficients of the agriculture-related shocks amongst those who lost ‘most’ of their crops remain robust in sign and significance. We also observe a statistically significant sign on the coefficient of those who lost ‘all’ of their crops (Panel A); however, once we introduce the interaction effects in Panel B, this significance disappears. The insignificance of the coefficient is replaced by the positive and significant interaction coefficient on shocks and social capital-receive. This suggests that although experiencing loss in harvest reduces consumption, the presence of informal social capital cushions households against crop failure and increases household consumption. Of special interest, in Table 6, is the household income coefficient, which is not statistically different from zero across the various consumption measures. This is somewhat surprising. Speculatively, this may suggest that household income is mainly budgeted for non-food consumption (e.g., school fees, transport and other essentials) rather than food consumption, while other household activities such as farming and gathering of natural resources provide food consumption.
Table 6

Impact of negative agriculture-related shocks on household consumption (household income control)

Dependent variable

Panel A without interactions

Panel B with interactions







ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

Shock, lost a little crop

0.155 (0.190)

0.104 (0.151)

−0.0339 (0.110)

0.0275 (0.292)

0.122 (0.236)

0.00907 (0.174)

Shock, lost some crops

0.0366 (0.174)

0.0410 (0.138)

−0.0263 (0.101)

−0.155 (0.260)

−0.0725 (0.210)

−0.0805 (0.155)

Shock, lost most of the crops

−0.549*** (0.149)

−0.486*** (0.119)

−0.300*** (0.0861)

−0.742*** (0.236)

−0.626*** (0.190)

−0.397*** (0.140)

Shock, lost all of the crops

−0.802* (0.428)

−0.776** (0.340)

−0.383 (0.248)

0.646 (0.713)

−0.0379 (0.575)

0.0679 (0.424)

Informal social capital, receive

0.00284 (0.206)

−0.239 (0.165)

0.0283 (0.119)

0.215 (0.404)

−0.209 (0.326)

0.119 (0.240)

Shock, a little* informal social capital, receive


−0.107 (0.770)

0.493 (0.621)

0.102 (0.458)

Shock, some* informal social capital, receive


−0.295 (0.559)

0.120 (0.451)

0.109 (0.332)

Shock, most* informal social capital, receive


−0.621 (0.528)

−0.436 (0.429)

−0.370 (0.314)

Shock, all* informal social capital, receive


6.546*** (2.213)

4.382** (1.785)

1.303 (1.316)

Informal social capital, give

0.00840 (0.127)

0.00232 (0.101)

0.00490 (0.0734)

−0.265 (0.251)

−0.215 (0.203)

−0.138 (0.149)

Shock, a little* informal social capital, give


0.408 (0.403)

0.103 (0.326)

0.115 (0.240)

Shock, some*Informal social capital, give


0.333 (0.353)

0.333 (0.285)

0.194 (0.210)

Shock, most* informal social capital, give


0.477 (0.316)

0.396 (0.255)

0.276 (0.188)

Shock, all* informal social capital, give


−2.795*** (0.941)

−1.326* (0.759)

−0.559 (0.560)

Formal social capital

0.0247 (0.130)

−0.105 (0.103)

0.0261 (0.0751)

0.0815 (0.244)

0.0222 (0.198)

0.121 (0.145)

Shock, a little* formal social capital


−0.132 (0.378)

−0.222 (0.306)

−0.218 (0.225)

Shock, some* formal social capital


0.187 (0.375)

−0.118 (0.303)

−0.0957 (0.223)

Shock, most* formal social capital


0.0437 (0.306)

−0.0253 (0.248)

0.00153 (0.182)

Shock, all* formal social capital


−0.303 (1.651)

−0.910 (1.331)

−0.938 (0.982)

Head of household age

0.144* (0.0834)

0.0745 (0.0663)

0.0604 (0.0483)

0.155* (0.0826)

0.0833 (0.0666)

0.0660 (0.0491)

Head of household age2

−0.00143** (0.000715)

−0.000813 (0.000568)

−0.000660 (0.000414)

−0.00150** (0.000707)

−0.000879 (0.000570)

−0.000709* (0.000420)

Number of household members

−1.312*** (0.456)

−1.367*** (0.362)

−1.173*** (0.264)

−1.351*** (0.451)

−1.410*** (0.364)

−1.200*** (0.268)

Household income

−3.40e−06 (1.53e−05)

4.63e−06 (1.22e−05)

−1.06e−05 (8.86e−06)

−1.79e−06 (1.52e−05)

5.36e−06 (1.22e−05)

−1.08e−05 (9.02e−06)

Income source: agriculture

0.284 (0.235)

0.246 (0.187)

0.279** (0.136)

0.309 (0.234)

0.275 (0.189)

0.308** (0.139)

Income source: natural resource

0.0106 (0.240)

0.110 (0.191)

0.204 (0.139)

0.0388 (0.238)

0.125 (0.192)

0.216 (0.141)

Income source: labour

0.370*** (0.140)

0.260** (0.112)

0.257*** (0.0813)

0.395*** (0.138)

0.281** (0.111)

0.268*** (0.0821)


7.828*** (2.377)

10.19*** (1.889)

6.564*** (1.377)

7.595*** (2.364)

10.03*** (1.907)

6.474*** (1.406)








Number of observations







Standard errors in parentheses

Instrumented for household income (panel B). Excluded instruments: lag household income

*** p < 0.01, ** p < 0.05, * p < 0.1

Nonetheless, the inclusion of household income shields against potential omitted variable bias, and still provides consistent results. A valid concern, however, is our choice of IV. Admittedly, lagged income value is unlikely to be a perfect IV. A priori, it is reasonable to suspect that the previous year’s (t − 1) income is likely to affect this year’s (t) consumption, which implies correlation with the error term. One potential channel is farm management effects. Specifically, some households are more likely to manage their farms better than others. If this happens in year t − 1, for instance, such that the households use income to purchase extensions, e.g., fertilisers, pesticides or labour to boost garden yields, these effects (boost in yields) are likely to be faced not only in year t − 1 but in year t as well. This may be through improved soil capability over time or even left through over extensions (from year t − 1) used in year t.

To investigate this premise, we use the t test and compare differences in mean agriculture output (consumption) in year t between those who purchased and those who did not purchase extensions (fertilisers, pesticides, herbicides, ploughing, implements and labour) in year t − 1. If the premise holds, then our expectation is that the agriculture output of households who purchase extensions would be higher than those who do not purchase extensions. Consistent with our expectations, the results of the t test revealed that households who use extensions had significantly higher crop yield (94,477.2 ± 7650.8) kcal compared to those who did not use any extensions (66,637.9 ± 6210.3); t(1034) = −2.7031, p = 0.0070. This statistically significant difference provides suggestive evidence that the lagged income value is likely to be correlated with the error term.

Debunking exogenous agriculture-related shocks: adaptation effects?

So far, we have asserted that our agriculture-related shocks from weather-related crop failure are somewhat more exogenous in comparison to crop failure from pests or diseases. Here, we probe this assertion. A concern is that, to some extent, it is plausible for households to cushion themselves against weather-related crop failure through adaptation. For example, in the presence of poor rainfall, households may opt to water/irrigate their gardens to reduce crop failure. Adaption is more likely to be present in higher-income and/or more-knowledgeable households (i.e., those with awareness of weather variability and adaptation methods) in comparison to lower-income/or less-knowledgeable households. Indeed, studies have found adaptation to be correlated with income and knowledge (e.g., Knowler and Bradshaw 2007; Deressa et al. 2009).

In testing this, first, we investigate whether the agriculture-related shocks systematically differ by income levels. If we find systematic differences, it would suggest that observable household characteristics, such as income, affect the shocks. Second, we include agriculture-related shockst+1 as an additional regressor conditional on the current shocks (agriculture-related shocks). The expectation is that we should not find significant coefficients on the agriculture-related shockst+1 (Duryea et al. 2007; Dinkelman et al. 2008). If this holds, it would be some indication that the agriculture-related shocks are not prone to some unobserved household influence (e.g., knowledge).

To that effect, Table 7 shows the distribution of the agriculture-related shocks (1 if the shock was experienced and 0 otherwise) by 2010 household income quartiles. Fortunately, the shocks are not systematic, suggesting that lower-income households are not more prone to shocks than are higher-income households. In Table 8, we re-estimate regressions but introduce agriculture-related shockst+1 as an additional regressor using a 2SLS model. Consistent with our expectation, the coefficient of agriculture-related shockst+1 is statistically insignificant across the three consumption outcomes, providing some evidence that the significant effects of agriculture-related shocks are unlikely to be due to unobservable influence.
Table 7

Distribution of agriculture-related shocks by income quartiles

Income quartile




















Table 8

Testing agriculture-related shocks

Dependent variable




ln kcal cons per capita (crops)

ln kcal cons per capita (crops and nat. resources)

ln real cons per capita (crops, nat. resources and groceries)

Shock, lost a little crop

0.178 (0.160)

0.103 (0.150)

0.116 (0.117)

Shock, lost some crops

0.0240 (0.163)

0.0896 (0.146)

0.0787 (0.103)

Shock, lost most of the crops

−0.323* (0.190)

−0.273 (0.174)

−0.183 (0.125)

Shock, lost all of the crops

−2.074*** (0.443)

−1.755*** (0.411)

−1.240*** (0.299)

Shock, lost a little crop (t + 1)

0.0587 (0.226)

0.000431 (0.209)

0.00106 (0.150)

Shock, lost some crops (t + 1)







Shock, lost most of the crops (t + 1)

0.0988 (0.202)

0.0989 (0.176)

0.0492 (0.115)

Shock, lost all of the crops (t + 1)

−0.404 (0.363)

−0.384 (0.319)

−0.198 (0.217)

Informal social capital, receive

0.177 (0.205)

0.128 (0.190)

0.114 (0.119)

Informal social capital, give

0.120 (0.126)

0.00745 (0.115)

0.0597 (0.0830)

Formal social capital

0.296** (0.137)

0.113 (0.128)

0.106 (0.0917)

Head of household age

0.108** (0.0440)

0.0893*** (0.0338)

0.0463** (0.0227)

Head of household age2

−0.000824** (0.000390)

−0.000659** (0.000299)

−0.000325 (0.000200)

Number of household members

−0.756*** (0.148)

−0.780*** (0.132)

−0.683*** (0.104)

Household income

−0.0948 (0.184)

0.00136 (0.165)

0.0996 (0.115)

Income source: agriculture

0.552** (0.274)





Income source: natural resource

0.344 (0.326)

0.319 (0.304)

0.221 (0.227)

Income source: labour

0.733 (1.251)

0.0114 (1.123)

−0.553 (0.786)


6.970*** (1.221)

7.871*** (0.918)

5.248*** (0.625)









Robust standard errors in parentheses

Reference category for shock (crop failure) is none

Instrumented for household income (panel B). Excluded instruments: lag household income

*** p < 0.01, ** p < 0.05, * p < 0.1


The current paper investigates the impact of crop failure from poor rainfall and hail storms, on rural household consumption patterns, in an attempt to discover coping mechanisms that currently exist. We observe three key findings. First, the magnitude of the shock matters, in that, households who lost all or most of their harvest are likely to consume significantly less. Second, although there appears to be no evidence of direct effects of informal social capital and formal social capital on consumption, the significant interaction effects show that receiving assistance has a cushioning effect on the consumption level of the most vulnerable, while giving assistance has the opposite effect, also among the most vulnerable. Surprisingly, we find formal social capital to be significant amongst the least vulnerable (i.e., with minimum crop loss). Third, apart from informal social capital, the use of natural resources also reduces the negative effects of the shock.

Climate variability is likely to become more frequent, resulting in increased weather-related events such as poor rainfall, floods or storms. Most rural households are already food insecure and depend on rain-fed homestead farming; hence, any weather-related event is likely to heighten food insecurity. Our findings from this study show that crop production, which is the mainstay of the majority of households in sub-Saharan Africa, is under threat from poor rainfall. While this issue has been previously investigated, the major concern of this study was the adaptive strategies that are effective in reducing the negative effects of shocks. Periodic fluctuations in rainfall are not new to a vast majority in rural sub-Saharan Africa. Our findings suggest that one way of improving the adaptive capacity of the rural poor is to strengthen social and natural capital, as they could provide easier, cheaper and more accessible alternative household coping strategies, in comparison to other, more technical and capital-intensive strategies, such as insurance. Yet, little is being done in most parts of sub-Saharan African countries to capture, utilise and promote these opportunities.

Currently, this untapped coping strategy is effectively being utilised in response to the AIDS epidemic, especially in resource-limited regions like those of sub-Saharan Africa (see Goudge et al. 2009a, b; Lippman et al. 2013). Their effectiveness has led to various interventions such as ‘treatment buddies’, while the more formal structures include community- and home-based care targeted at improving treatment response and coping mechanisms. Such valuable lessons can be drawn and adopted in the current context: household vulnerability to agriculture-related shocks. This is especially true because the current literature recognises that climate variability is likely to continue, which implies that weather-related crop failure is more likely to be a common occurrence. In the current rural setting, which is characterised by poverty, insurance is unlikely to be a short-term solution, thus calling for the promotion of more informal methods readily available in resource-poor settings. A remaining concern centres on the sustainability of these less-conventional adaptation strategies currently utilised by rural households. While informal social capital is more of a sustainable adaptation strategy, the use of natural capital is less likely to be sustainable. Our concern here is natural resource depletion, and how unsustainable resource use may render households more vulnerable to the impacts of exogenous agricultural shocks on food securities.

Sustainability is already threatened by high population densities, chronic poverty, and weakening traditional governance institutions (Twine 2005; Kirkland et al. 2007). Most resources are harvested from communal woodlands surrounding the villages, and few are gathered from the homestead yards. For example, local fuelwood shortages have resulted in an increase in the incidence of the felling of marula trees, which are prized for their fruit and theoretically protected by both traditional taboos and national legislation. With the increase in climate and weather variability which threaten agriculture productivity, local communities are likely to increase their dependence of natural resources. This calls for a win–win policy intervention that can successfully link natural capital adaptation strategies with sustainability. This requires the strengthening of tenure rights and local institutions, empowerment of local communities to manage their resources, and extension support for community-based natural resource management. However, with informal social capital, we are concerned with the likely negative effects on the most vulnerable households, i.e., their welfare and the trade-off between giving and receiving assistance. Because a plausible driver for this trade-off is the culture of strong ties and interpersonal relationships in rural communities, to achieve sustainability, policy designs will benefit by targeting the existing relationships.


  1. 1.

    The report is based on the first South Africa National Health and Nutrition Examination Survey (SANHANES-1) conducted by the Human Science Research Council (HSRC). The survey is expected to occur periodically and report on the health and nutritional status of South Africans.

  2. 2.

    According to the World Hunger and Poverty Facts and Statistics report, there is an increase in the level of hunger in Africa, with one in every four Africans suffering from hunger. One of the reasons for this increment is climate change.

  3. 3.

    The table includes statistics of the only available, sub-Saharan African countries: Chad, Côte d'Ivoire, Ghana, Kenya, Malawi, Mozambique, Niger, Sudan, Togo, Uganda and Zambia.

  4. 4.

    An example will clarify our approach. According to the FAO conversion tables, 100 g of pumpkins, one of the main crops in the area, contains 26 calories (kcal). Hence, a household that harvests 2000 g (2 kg) of pumpkins will earn a total of 52,000 kcal for the household. This process is repeated for each crop produced by the household; thereafter, we add all calories and divide by the total number of household members.

  5. 5.

    For a comprehensive review of social capital (informal and formal), see Wallace and Pichler (2009), Lovell (2009) and Bhandari and Yasunobu (2009).

  6. 6.

    The FAO statistics also show the total dietary energy consumption, which is an aggregation of energy from (1) purchased food, (2) own production, and (3) other sources. Here, we observe that amongst those in the poorest percentiles the caloric consumption ranges between 1251.7 and 1765.2 kcal, while in the medium percentile this range is between 2036.7 and 2418.6 kcal. See Table 9. Note that, due to data limitation, we are unable to show these values from our data, because we cannot observe caloric values from groceries and livestock farming.



We are thankful and acknowledge helpful comments from participants at the Africa Climate Development Initiative (ACDI) seminar series and the 5th World Congress of Environmental and Resource Economists (WCERE). We also thank the Environment for Development (EfD) Initiative and Economic Research Southern Africa (ERSA) for financial support. The SUCSES panel study was funded by the South African National Research foundation. This work was indirectly supported by the Wellcome Trust (Grant 085477/Z/08/Z) through its support of the Agincourt Health and Demographic Surveillance System.


  1. Akresh R, Verwimp P, Bundervoet T (2011) Civil war, crop failure and child stunting in Rwanda. Econ Dev Cult Chang 59(4):777–810CrossRefGoogle Scholar
  2. Bhandari H, Yasunobu K (2009) What is social capital? A comprehensive review of the concept. Asian J Soc Sci 37(3):480–510CrossRefGoogle Scholar
  3. Birhanu A, Zeller M (2009) Using panel data to estimate the effects of rainfall shocks on smallholder food security and vulnerability in rural Ethiopia. Centre for Agriculture in the Tropics and Subtropics, Discussion Paper No. 2/2009Google Scholar
  4. Brown A, Lyons M, Dankoco I (2010) Street traders and the emerging spaces for urban voice and citizenship in African cities. Urban Stud 47(3):666–683Google Scholar
  5. Burton I, Development Programme United Nations (2005) Adaptation policy frameworks for climate change. In: Lim B (ed) Developing strategies, policies and measures, Cambridge University Press, Cambridge, p 258Google Scholar
  6. Carter M, Maluccio J (2003) Social capital and coping with economic shocks: an analysis of stunting of South African children. World Develop 31(7):1147–1163CrossRefGoogle Scholar
  7. Cavatassi R, Lipper L, Narloch U (2011) Modern variety adaptation and risk management in drought prone areas: insights from the sorghum farmers of eastern Ethiopia. Agric Econ 42:279–292CrossRefGoogle Scholar
  8. Christiansen L, Subbarao K (2005) Towards an understanding of household vulnerability in rural Kenya. J Afr Econ 14(4):520–558CrossRefGoogle Scholar
  9. Claro RM, Levy RB, Bandoni DH, Mondini L (2010) Per capita versus adult-equivalent estimates of calorie availability in household budget surveys. Cad de Saúde Pública 26(11):2188–2195CrossRefGoogle Scholar
  10. Coleman J (1988) Social capital in the creation of human capital. Am J Sociol 94:S95–S120CrossRefGoogle Scholar
  11. Danes S, Winter M, Whiteford MB (1987) Level of living and participation in the informal market sector among rural Honduran women. J Marriage Fam 631–639Google Scholar
  12. DEA (2011) South Africa’s second national communication under the United Nations framework convention on climate change. Department of Environmental Affairs (DEA), Republic of South Africa, PretoriaGoogle Scholar
  13. Dercon S (2004) Growth and shocks: evidence from rural Ethiopia. J Dev Econ 74:309–329CrossRefGoogle Scholar
  14. Dercon S, Krishnan P (2000) Vulnerability, seasonality and poverty in Ethiopia. J Dev Stud 36(6):25–53CrossRefGoogle Scholar
  15. Deressa T, Hassan R, Ringler C, Alemu T, Yesuf M (2009) Determinants of farmers’ choice of adaptation methods to climate change in the Nile Basin of Ethiopia. Glob Environ Chang 19:248–255CrossRefGoogle Scholar
  16. Di Falco S, Bulte E (2009) Social capital and weather shocks in Ethiopia: climate change and culturally-induced poverty traps. London School of Economics, Working PaperGoogle Scholar
  17. Dillon A (2012) Child labour and schooling responses to production and health shocks in northern Mali. J Afr Econ 22(2):276–299CrossRefGoogle Scholar
  18. Dinkelman T (2013) Mitigating long-run health effects of drought: evidence from South Africa (No. w19756). National Bureau of Economic ResearchGoogle Scholar
  19. Dinkelman T, Lam D, Leibbrandt M (2008) Linking poverty and income shocks to risky sexual behaviour: evidence from a panel study of young adults in Cape Town. S Afr J Econ 76(1):S52–S74CrossRefGoogle Scholar
  20. Duryea S, Lam D, Levison D (2007) Effects of economic shocks on children’s employment and schooling in Brazil. J Dev Econ 84:188–214CrossRefGoogle Scholar
  21. Ellis F, Freeman H (2004) Rural livelihoods and poverty reduction strategies in four African countries. J Dev Stud 40(4):1–30CrossRefGoogle Scholar
  22. Eriksen S, Aldunce P, Bahinipati CS, Martins RDA, Molefe JI, Nhemachena C, O’Brien K, Olorunfemi F, Park J, Sygna L, Ulsrud K (2011) When not every response to climate change is a good one: identifying principles for sustainable adaptation. Clim Devt 3(1):7–20CrossRefGoogle Scholar
  23. FAO (2008) Climate change and food security: a framework document. The Food and Agriculture Organization of the United Nations, RomeGoogle Scholar
  24. Figuié M, Moustier P (2009) Market appeal in an emerging economy: supermarkets and poor consumers in Vietnam. Food Policy 34(2):210–217CrossRefGoogle Scholar
  25. Gilbert G, McLeman R (2010) Household access to capital and its effects on drought adaptation and migration: a case study of rural Alberta in the 1930s. Popul Environ 32:3–26CrossRefGoogle Scholar
  26. Goudge J, Russell S, Gilson L, Gumede T (2009a) Illness-related impoverishment in rural South Africa: why does social protection work for some households but not others? J Int Dev 21:231–251CrossRefGoogle Scholar
  27. Goudge J, Gilson L, Russell S, Gumede T, Mills A (2009b) Affordability, availability and acceptability barriers to heath care for the chronically ill: longitudinal case studies from South Africa. BMC Health Serv Res 9(75):1–18Google Scholar
  28. Greene W (2002) Econometric analysis, 5th edn. Prentice-Hall, Upper Saddle RiverGoogle Scholar
  29. Hellmuth M, Moorhead A, Thomson M, Williams J (eds) (2007) Climate risk management in africa: learning from practice. International Research Institute for Climate and Society (IRI), Columbia University, New YorkGoogle Scholar
  30. Hofferth S, Iceland J (1998) Social capital in rural and urban communities. Rural Sociol 63(4):574–598CrossRefGoogle Scholar
  31. Hunter L, Twine W, Patterson L (2007) Locusts are now our beef: adult mortality and household dietary use of local environmental resources in rural South Africa. Scand J Public Health 35(3):165–174CrossRefGoogle Scholar
  32. Hunter L, Patterson L, Twine W (2009) HIV/AIDS, food security and the role of the natural environment: evidence from the Agincourt Health and Demographic Surveillance Site in rural South Africa. IBS Population Program POP2009-01Google Scholar
  33. IPCC (2007) Climate change 2007: impacts, adaptation and vulnerability. Summary for policy makers. World Meteorological Organisation, GenevaGoogle Scholar
  34. IPCC (2014) Climate change 2014: impacts, adaptation, and vulnerability, Part A: global and sectoral aspects. Contribution of working group II to the fifth assessment report of the intergovernmental Panel on Climate ChangeGoogle Scholar
  35. Kahn K, Collinson MA, Gomez-Olive FX, Mokoena O, Twine R, Mee P, Tollman SM (2012) Profile: Agincourt Health and Socio-demographic Surveillance System. Int J Epidemiol 41(4):988–1001CrossRefGoogle Scholar
  36. Kashula S (2008) Wild foods and household food security responses to AIDS: evidence from South Africa. Popul Environ 29(3–5):169–185Google Scholar
  37. Kirkland T, Hunter LM, Twine W (2007) The bush is no more: insights on institutional change and natural resource availability in rural South Africa. Soc Nat Res 20(4):337–350CrossRefGoogle Scholar
  38. Knowler D, Bradshaw B (2007) Farmers’ adoption of conservation agriculture: a review and synthesis of recent research. Food Policy 32(1):25–48CrossRefGoogle Scholar
  39. Kochar A (1995) Explaining household vulnerability to idiosyncratic income shocks. Am Econ Rev 85(2):159–164Google Scholar
  40. Kotir JH (2011) Climate change and variability in sub-Saharan Africa: a review of current and future trends and impacts on agriculture and food security. Environ Dev Sustain 13(3):587–605CrossRefGoogle Scholar
  41. Lippman S, Maman S, MacPhail C, Twine R, Peacock D, Kahn K, Pettifor A (2013) Conceptualising community mobilisation for HIV prevention: implication for HIV prevention programming in the African context. PLOS 8(10):1–13Google Scholar
  42. Lovell S (2009) Social capital: the panacea for community? Geogr Compass 3(2):781–796CrossRefGoogle Scholar
  43. McGarry D, Shackleton C (2009) Children navigating rural poverty: rural children’s use of wild resources to counteract food insecurity in the eastern Cape, South Africa. J Child Poverty 15(1):19–37CrossRefGoogle Scholar
  44. Mirza M (2003) Climate change and extreme weather events: can developing countries adapt?. Clim policy 3(3):233–248CrossRefGoogle Scholar
  45. Misselhorn A (2009) Is a focus on social capital useful in considering food security interventions? Insights from KwaZulu-Natal. Dev South Afr 26:189–208CrossRefGoogle Scholar
  46. Mogues T (2006) Shocks, livestock asset dynamics and social capital in Ethiopia. DSGD discussion papers, No. 38Google Scholar
  47. Nelson G (2010) The costs of agricultural adaptation to climate change. World Bank Discussion Paper, No. 4Google Scholar
  48. Nhemachena C, Hassan R, Chikwizira J (2010) Economic impacts of climate change on agriculture and implications for food security in Southern Africa. Centre for Environmental Economics and Policy in Africa (CEEPA)Google Scholar
  49. Omolo N (2010) Gender and climate change-induced conflict in pastoral communities: case study of Turkana in northwestern Kenya. Afr J Confl Resolut Environ Confl 10(2):81–102Google Scholar
  50. Osbahr H, Twyman C, Adger W, Thomas D (2010) Evaluating successful livelihood adaptation to climate variability and change in Southern Africa. Ecol Soc 15(2):27Google Scholar
  51. Pichler F, Wallace C (2007) Patterns of formal and informal social capital in Europe. Eur Sociol Rev 23(4):423–435CrossRefGoogle Scholar
  52. Porter C (2011) Shocks, consumption and income diversification in rural Ethiopia. J Dev Stud 48(9):1209–1222CrossRefGoogle Scholar
  53. Putnam R (2001) Social capital: measurement and consequences. Can J Policy Res 2(1):41–51Google Scholar
  54. Reid P, Vogel C (2006) Living and responding to multiple stressors in South Africa—glimpse from KwaZulu-Natal. Glob Environ Chang 16:195–206CrossRefGoogle Scholar
  55. Salvatori D, Chavas J (2008) Rainfall shocks, resilience and the effects of crop biodiversity on agro-ecosystems productivity. Land Econ 84(1):83–96CrossRefGoogle Scholar
  56. Shackleton C, Shackleton S (2004) The importance of non-timber forest products in rural livelihood security and as safety nets: a review of evidence from South Africa. S Afr J Sci 100(11–12):658–664Google Scholar
  57. Shackleton SE, Dzerefos CM, Shackleton CM, Mathabela FR, Shackleton ASE (1998) Use and trading of wild edible herbs in the central lowveld savanna region, South Africa. Econ Bot 52(3):251–259CrossRefGoogle Scholar
  58. Shields J, Fletcher D (2013) What smallholder sweet potato farmers are doing to adapt to a changing climate: evidence from six agro-ecological zones of Uganda. Eur J Clim Chang 10:2668–3784Google Scholar
  59. Shisana O, Labadarios D, Rehle T, Simbayi L, Zuma K (2014) South African National Health and Nutrition Examination Survey (SANHANES-1), 2014th edn. HSRC Press, Cape TownGoogle Scholar
  60. Tesso G, Emana B, Ketema M (2012) Analysis of vulnerability and resilience to climate change induced shocks in North Shewa, Ethiopia. Agric Sci 3(6):871–888Google Scholar
  61. Tibesigwa B, Visser M (2015) Small-scale subsistence farming, food security, climate change and adaptation in South Africa: male-female headed households and Urban-Rural Nexus (No. 527)Google Scholar
  62. Tibesigwa B, Visser M, Turpie J (2015) The impact of climate change on net revenue and food adequacy of subsistence farming households in South Africa. Environ Develop Econ 20(3):327–353CrossRefGoogle Scholar
  63. Twine W (2005) Socio-economic transitions influence vegetation change in the communal rangelands of the South African lowveld. Afr J Range Forage Sci 22(2):93–99CrossRefGoogle Scholar
  64. Twine W, Hunter L (2011) Adult mortality and household food security in rural South Africa: does AIDS represent a unique mortality shock? Dev South Afr 28(4):431–444CrossRefGoogle Scholar
  65. Twine W, Moshe D, Netshiluvhi T, Siphugu V (2003) Consumption and direct-use values of savannah bio-resources used by rural households in Mametja, a semi-arid area of Limpopo province, South Africa. S Afr J Sci 99:467–473Google Scholar
  66. Wallace C, Pichler F (2009) More participation, happier Society? A comparative study of civil society and the quality of life. Soc Indic Res 93(2):255–274CrossRefGoogle Scholar
  67. Yamano T, Alderman H, Christiansen L (2005) Child growth, shocks and food aid in rural Ethiopia. Am J Agric Econ 87(2):273–288CrossRefGoogle Scholar

Copyright information

© Springer Japan 2015

Authors and Affiliations

  • Byela Tibesigwa
    • 1
  • Martine Visser
    • 1
  • Mark Collinson
    • 2
    • 3
  • Wayne Twine
    • 4
  1. 1.Environmental-Economics Policy Research Unit, School of EconomicsUniversity of Cape TownCape TownSouth Africa
  2. 2.MRC/Wits University Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the WitwatersrandJohannesburgSouth Africa
  3. 3.INDEPTH NetworkAccraGhana
  4. 4.School of Animal, Plant and Environmental SciencesUniversity of the WitwatersrandJohannesburgSouth Africa

Personalised recommendations