Factors influencing household food security among smallholder farmers in the Mudzi district of Zimbabwe

Abstract

This article examines factors influencing household food security among smallholder farmers in Mudzi district of Zimbabwe. Data for this study were collected from 120 randomly selected households, using a structured questionnaire. Analytical techniques employed included descriptive statistics of respondents' characteristics and linear regression analysis to identify factors influencing their household food security. The results show that household dietary diversity is influenced by the age and education of the household head, household labour and size, livestock ownership, access to market information and remittances. Linear regression on another indicator, the household food insecurity access score, shows that labour, education of the household head, household size, remittances, livestock ownership and access to market information all affect household food security. The study therefore recommends that government and other development agencies enhance food security among smallholders through promoting labour-saving technologies, enhancing the flow of remittances to rural areas, facilitating access to market information, and farmer education.

Keywords:

• household food security
• household dietary diversity score
• household food insecurity access score
• smallholder farmers
• Zimbabwe

1. Introduction and background

Food insecurity has attracted increasing attention from policy-makers and programme implementers since the world food crisis of 1972–74 and the 2006–08 food price shock, in view of the high incidence of micronutrient malnutrition and hunger (Gebre, 2012; Sasson, 2012). Evidence from these two periods suggests that poor households (largely from developing countries) are particularly hard hit during such crises, because food is insufficient and the prices of staple cereals rise sharply beyond their reach (Gebre, 2012; Tyner, 2013). An estimated 870 million people worldwide are undernourished, of whom 850 million live in developing countries. The percentage of malnourished people in sub-Saharan Africa is estimated to have risen from 17% to 27% between 1990 and 2011 (FAO et al., 2012). Despite the progress made through international and in-country efforts to reduce food insecurity and hunger, people in most developing countries, especially in sub-Saharan Africa, still face challenges in accessing food (Diouf, 2005; Omotesho et al., 2006). The unacceptably high rate of undernourishment among these people suggests that food insecurity will remain a major problem confronting policy-makers, especially those from developing countries. Declining soil fertility, low adoption of production-enhancing and yield-enhancing technologies, dysfunctional input and output markets, and poor access to extension services have been identified as the drivers of household food insecurity in sub-Saharan Africa (Mapfumo, 2009; Vanlauwe et al., 2010). Climate change in the form of increased weather variability and prolonged mid-season and in-season droughts exacerbate the situation (Below et al., 2010). Recent research by Dubbeling et al. (2010) shows that unless considerable investments are made to increase per-capita food production in sub-Saharan Africa, the number of people facing hunger, undernourishment and malnourishment will increase significantly.

According to the FAO (1996) and Coates et al. (2007), food security is a state in which all people at all times have both physical and economic access to sufficient safe and nutritious food to meet their dietary needs and food preferences for a healthy and active life. Households are food insecure when they have uncertain or limited access to food through socially acceptable channels – a problem affecting many households worldwide, Zimbabwe included. Food insecurity has severe implications for nutrition, health and productivity. Members of a household that fails to obtain nutritious and preferred foods for a healthy and productive life may develop multiple chronic conditions, even obesity (Sanchez, 2002). Carter et al. (2010) note that food insecurity results in poor (self-rated) physical and mental health. As such, a poorly nourished population is a constraint to human and economic development.

According to the WFP (2012), smallholder farmers in the drier areas of Zimbabwe are the worst affected by food insecurity. In the Mudzi district of Mashonaland East province, where this study was conducted, food insecurity is particularly severe. The International Centre for Tropical Agriculture (CIAT), in collaboration with the Department of Agricultural Extension and Technical Services, is implementing a research-led development project in this district, which aims to improve food security, nutrition and income among smallholder groundnut producers. The project is premised on the understanding that the major causes of household food insecurity and income poverty among farmers producing high-value market crops are poor access to market information and a lack of knowledge about markets. Poor access to market information results in unequal negotiation with buyers and contributes to the continued exploitation of the farmers by itinerant traders. To improve the farmers' access to market information, CIAT formed Rural Group Enterprises, which allow farmers to bulk their produce and sell it collectively. The collective action allows them to access market information more readily and make informed selling decisions. Members of these enterprises are trained in business skills such as market research, negotiating with buyers, analysing marketing margins, and collective marketing. With improved access to market information and better knowledge about markets, it is hoped that farmers would realise higher incomes and improve their food security.

The literature suggests another major socio-economic determinant of food insecurity: household resource endowments (Sanusi et al., 2007; Idrisa et al., 2008; Gebre, 2012; Mitiku et al., 2012; Nyikahadzoi et al., 2012). Swift (1989) argues that when households are able to generate a surplus over and above their basic food requirements, they can divert the excess resources into different kinds of assets to sustain them in times of crisis. Under these circumstances, food security can be closely linked to a household's resource endowments. This would mean that a household with poor resource endowments would face a higher risk of being food insecure. Although livelihood assets are important determinants of food security, factors closely related to human resource development, such as education, healthcare, clean water, population growth, urbanisation and the displacement of people, also have a significant influence on food security (Dercon & Krishnan, 2000; Dercon & Hoddinott, 2003). Other factors, such as the size of the household, the availability of household labour, the age and education of the household head, household indebtedness, access to credit, the performance of input and output markets, household expenditure, on-farm and off-farm income, livestock assets, the size of arable land, agricultural inputs and extension services that could improve subsistence production, all affect household food security. This short review shows that multiple, interrelated factors can affect food security and they also vary from one context to another. This implies that policies to address food security should pursue a range of strategies (Barrett et al., 2001; Von Braun et al., 2005).

Notwithstanding the fact that farming households in the Mudzi district experience recurrent droughts and acute food shortages, little is known about the specific household-level predictors of food insecurity in this district. A detailed understanding of the determinants of food security is critical both for informing policy decisions and for assessing research-led development projects such as the one promoted by CIAT and its partners. To this end, the analysis was carried out at a household level. A national-level analysis of food security generally compares the availability of food and the food requirement. However, the availability of food at a household level is not always correlated with the supply of food at a national level. The dynamics of the food market are determined by both demand and supply, which affect prices and eventually the food security situation of the population. It is therefore important to analyse the situation at household level, because this highlights the different household characteristics that are specific to food security and demonstrates their implications for food-related issues. This paper therefore seeks to identify the factors that influence household food security, in order to inform the ongoing CIAT project to improve household food security in the district.

This study builds on previous studies by Nyikahadzoi et al. (2012) and Gebre (2012). The problem of food insecurity has several dimensions, which range from the global, regional, country, local and household levels to the individual level. To date, district-level research on these problems has been limited, as most studies in Zimbabwe focus on the national level. Since food security in rural Zimbabwe is largely based on primary production from the household's own fields, assessing the determinants of food security at this level will be particularly relevant for informing policy-making in Zimbabwe. Moreover, an analysis of the determinants of food security can also show how households' existing (physical and non-physical) assets determine their ability to secure food. Such information will again be critical for guiding interventions to curb food insecurity.

Literature on the complex and interrelated causes of household food insecurity in Zimbabwe and the local responses during crisis situations is also limited. This study attempts to address this gap through an empirical investigation of the household-level determinants of food security in order to guide the development of appropriate interventions to combat food insecurity. Specifically, it runs linear regressions with the household dietary diversity score (HDDS) and the household food insecurity access score (HFIAS) as proxies for household food security. Variables hypothesised to affect household food security are presented in Table 1.

Table 1: Independent variables

Variable	Description	Expected sign
Headage	Age of household head	–
Headeduc	Education level of household head (number of years)	+
Genderhh	Gender of household head (1 = male, 0 = female)	+/–
Household labour	Number of household members aged 16 to 65	+
Household size	Household size	–
Market info	Access to market information (1 if household has access, 0 otherwise)	+
Remittances	Household receives remittances (1 if true, 0 if not)	+
Livestock wealth	Weighted index for livestock wealth (total livestock units)	+

The rest of the article is arranged as follows. Section 2 discusses the study approach, which includes a description of the site, sampling and data collection. This is followed by a review of the underlying analytical approaches in Section 3. Section 4 then presents and assesses the results, while conclusions and recommendations follow in Section 5.

2. The study approach

2.1 Study sites

The study was conducted in the Mudzi district, which forms part of the Mashonaland East province of Zimbabwe (see Figure 1). The district is linked to the main groundnut market in Harare by a 250 km tarred road. The study sites lie in natural farming region IV, which is a semi-arid zone at an altitude of 500 to 900 m above sea level (Vincent & Thomas, 1960). Agro-ecologically, this natural farming region is a low-potential zone, with a high incidence of droughts and frequent, long mid-season and in-season dry spells. The mean annual rainfall in the district ranges from 450 to 500 mm, while the mean annual temperature is 23°C. The predominant soil type is the ferric luvisols, which are ideal for groundnuts. Owing to the high aridity, the yield of maize (the country's staple food crop) in this district is about 0.5 tonnes/ha. Groundnuts (Arachis Hypogaea) are the single most important legume crop grown in the area and the bulk of the population depends heavily on this crop for survival. The project is being implemented in Wards 12 and 16 in this district.⁶

Figure 1: Map of Zimbabwe showing the Mudzi district

2.2 Sampling and data collection

This study uses cross-sectional household data collected during a baseline survey under the project ‘Increasing Smallholder Farm Productivity, Income, and Health through Widespread Adoption of Integrated Soil Fertility Management (ISFM) in the Great Lake Regions and Southern Africa’. The simple random sampling technique was used to select the wards from a list obtained from the district. Within the selected wards, the interview households were randomly chosen from household lists provided by resident agricultural extension officers. A total of 120 households were then selected for the survey. The sample size of 120 exceeds the a priori minimum size (n = 74) derived from power analysis using G*Power (Franzel et al., 2007). G*Power is a stand-alone program for statistical tests (t, F, z and exact tests), which offers improved effect size calculators, graphic options and all types of power analysis. The sample therefore provides an acceptable statistical power (0.80) for moderate correlation (r = 0.30) at the two-tailed 0.05 level of significance according to the G*Power formula.

Data for this study were collected in December 2011 through the face-to-face administration of household survey questionnaires with both semi-structured and structured questions. The survey collected information on household demographics and socio-economic characteristics, a household dietary diversity assessment involving 24-hour recall of food consumption from major food groups, food insecurity, crop production and marketing. Data on household food security and nutrition were collected using the tools developed by the Food and Nutrition Technical Assistance Project for use in low-resource settings (Swindale & Bilinsky, 2006; Coates et al., 2007).

3. Analytical approaches

The collected data were entered, cleaned and then analysed using Stata version 11.20. The study applied multivariate linear regressions analysis to identify factors that affect the HDDS and the HFIAS. These food security measures were developed using the Food and Nutrition Technical Assistance Project guidelines (Coates et al., 2007). Coates et al. (2006) noted that these two indicators of food insecurity cover the universal experiences of food insecurity, namely insufficient quantity, inadequate quality, and uncertainty and anxiety about food.

In this study the proxies for the demographic characteristics of the household are the age, gender and level of education of the household head, the availability of household labour, and the size of the household. The socio-economic variables are access to market information, access to extension services, livestock-related wealth (measured in total livestock units) and access to remittances. We carried out a correlation test to see whether any of the independent variables are related. Using a cut-off point of 0.7 for the correlation tests, we dropped highly correlated variables from the models. This cut-off point was chosen since colinearity begins to distort model estimation and prediction significantly at values greater than 0.7.

3.1 Independent variables

Table 1 lists the variables used in the models, their description and their expected effect on the dependent variable.

3.1.1 Age of the household head

The age of the household head is a proxy for farming experience, on the assumption that the household's knowledge of food security issues will increase as the household head grows older and more experienced. However, studies of this variable produce contrasting results. Most studies confirm that as the age of the household head increases, the household acquires more farming experience, becomes more risk averse, and diversifies its production (Bogale & Shimelis, 2009; Mitiku et al., 2012). Households headed by older people are thus more likely to be food secure than those with younger heads. In contrast, Gebre (2012) shows that as the age of the household head increases, households become less productive and rely more on gifts and remittances. Hence households with older heads are more likely to be food insecure than those with younger heads. We expect a positive relationship between the age of the household head and food security.

3.1.2 Education of the household head

The level of education of the household head influences the household's access to and use of information and builds its capacity to enhance food security. Makombe et al. (2010) and Idrisa et al. (2008) demonstrate that this variable has a positive effect on food security. That is, households with better-educated heads are more likely to receive information and use it in their decisions than those with less-learned heads. The former households are assumed to have better management techniques, which can help them secure a year-round supply of diversified and even preferred food. We thus expect a positive relationship between food security and the level of education of the head of the household.

3.1.3 Gender of the household head

The gender of the household head represents the food security orientation between male-headed and female-headed households. Horrell & Krishnan (2007) argue that male-headed households are better positioned to source on-farm labour than their female-headed counterparts. As such, we expect a positive relationship between male-headed households and food security.

3.1.4 Household labour

The availability of household labour reflects the human capital resources available to a household for agronomic activities. Bogale & Shimelis (2009) noted that households with large pools of labour are more likely to be food secure, as they can carry out farming activities on time. They can also commit large tracts of land to food crops. As such, we expect a positive relationship between household labour and food security.

3.1.5 Household size

Household size represents the consumption-level needs of a household and shows the burden it faces to feed its members. Bigger households obviously have a higher ‘burden to feed’ (Bogale & Shimelis, 2009; Gebre, 2012; Mitiku et al., 2012) and are thus more likely to be food insecure. As such, we expect a negative relationship between household size and food security.

3.1.6 Market information

Access to market information is the main factor inducing market participation (Alene et al., 2008). Households with access to market information know what is demanded by the market, when to sell, whom to sell to, and what price to accept. They are therefore likely to get higher cash incomes, which they can use to buy diversified and even preferred foodstuffs. Indeed, households with access to market information are more likely to be food secure than their counterparts. We thus expect a positive relationship between access to market information and household food security.

3.1.7 Remittances

According to Nyikahadzoi et al. (2012), remittances represent an alternative source of income. Households with access to remittances can purchase more appropriate and nutritious foods than low-income groups can. As such, households with access to remittances are more likely to be food secure than those without this source of income. We thus expect a positive relationship between access to remittances and food security.

3.1.8 Livestock wealth

Livestock is an indicator of the social and economic standing of a farmer. Resource-endowed farmers can sell their livestock during food shortages. They can also get milk, milk products and meat from their livestock or sell them for cash. Livestock contributes draft power and helps the household meet subsistence, income and nutritional requirements (Bogale & Shimelis, 2009). As such, livestock ownership has a positive effect on food security. We thus expect households with livestock to be more food secure than those without these assets.

3.2 Dependent variables

3.2.1 Household dietary diversity score

The HDDS is defined as a qualitative free recall of all food or drink consumed by any household member during the last 24 hours (FAO, 2011). According to Uraguchi (2012), the HDDS indicates the number of food groups and items that households consume in a 24-hour period for 7 days. It is usually measured by summing the number of food groups consumed over a reference period (Kassie et al., 2008).

The HDDS is an important proxy indicator of food security, as documented by research studies in various countries. These studies include Hatløy et al. (1998) in Mali, Hoddinott & Yohannes (2002) in 10 developing countries, Ogle et al. (2001) in Vietnam, Mirmiran et al. (2006) in Iran, and Ekesa et al. (2008) in Kenya. One advantage of the HDDS is that it is highly correlated with factors such as the adequacy of a household's intake of calories, protein and other nutrients (Goshu et al., 2013). Dietary diversity indicators based on food groups predict nutrient adequacy better than those based on individual foods (Ruel, 2002). This is because a household might quite possibly meet the energy requirements of its members, but they may be unable to live active, healthy lives because their intake of other nutrients is deficient (Goshu et al., 2013). Thus it is important to include indicators of nutritional quality when analysing household food security. Dietary diversity shows how varied the foods typically consumed by a household are (Smith & Subandoro, 2007), as well as the household's economic capacity to consume a variety of foods (Hoddinot & Yohannes, 2002).

Since the HDDS reflects dietary quality, in practice it is likely to be inversely related to malnutrition. A high HDDS indicates a diversified household diet, which is more likely to allow balanced micronutrient intakes (Arimond et al., 2010). In contrast, a low HDDS is associated with a high intake of starch staples; this contributes to nutritional problems, as these staples are low in micronutrients (Steyn et al., 2006).

According to the FAO (2008), the HDDS is composed of 14 food groups. The score is calculated by counting the number of food groups consumed at individual or household level (excluding food consumed outside the household). Each food group is given a weight of one. The literature identifies a strong and positive relationship between the HDDS and food security (Nyikahadzoi et al., 2012). A more diversified diet is highly correlated with an adequate intake of calories and protein (Steyn et al., 2006; Kennedy et al., 2007).

Evidence from multi-country analysis suggests that household-level dietary diversity is strongly associated with per-capita consumption (a proxy for income) and energy availability, suggesting that it could be a useful indicator of household food security. Research to quantify this association shows that a 1% increase in dietary diversity is associated with a 1% increase in per-capita consumption (Hoddinot & Yohannes, 2002). However, the score does not enable an estimate of how much food is lacking because it cannot be used directly to quantify the amount of food consumed. The HDDS approach also does not shed light on the causes of consumption patterns (e.g. income, prices and own production). Moreover, the reliability and accuracy of the HDDS have been questioned on methodological grounds (Uraguchi, 2012). Uraguchi criticised the HDDS for two main reasons: first, there is no universally accepted standard for the main food groups or food types to use; and second, while the HDDS can show changes in the dietary energy consumption of households, it has not been easy to empirically demonstrate the significance of the HDDS in nutrient adequacy.

Despite these weaknesses, the HDDS remains useful and a good proxy for the nutrient adequacy of a household's diet, hence its use as a measure of household food security.

In this study, the sampled households were asked to recall all foods eaten in the past 24 hours. The food and drink mentioned by the respondents were then classified into one of 14 standardised food groups. The responses to the questions were either yes (score = 1) or no (score = 0). The groups are as follows: A, cereals; B, vitamin-rich vegetables; C, roots and tubers; D, dark green, leafy vegetables; E, other vegetables; F, fruits rich in vitamin A; G, other fruits; H, meat, poultry and offal; I, eggs; J, fish and seafood; K, pulses, legumes and nuts; L, milk and milk products; M, oils and fats; and N, sugar and honey.

Following Swindale & Bilinsky (2005), the HDDS is calculated as follows:

3.2.2 Household food insecurity access score

The HFIAS is a continuous measure of the degree of food insecurity (in terms of access) in the household in the past 30 days. According to Deitchler et al. (2011), the HFIAS reflects the three universal domains of household food insecurity: anxiety about household food insecurity, insufficient quality of food supplies, and insufficient quantity of such supplies. This indicator captures the members of the household's perception of their diet, regardless of its nutritional composition (Coates et al., 2007). It focuses on consumption-related strategies and captures the household's behavioural and psychological responses to (perceived) food insecurity. The HFIAS is based on the assumption that households' experiences of food insecurity cause predictable reactions and responses that can be captured and quantified through a survey and then summarised into a score.

During the survey, the respondents were asked nine occurrence questions, reflecting an increasing level of food insecurity. The occurrence questions can be summarised as follows: (Q1a) anxiety about food adequacy; (Q2a) eating less-preferred foods; (Q3a) eating foods of a limited variety; (Q4a) inability to eat even less-preferred foods; (Q5a) eating smaller meals than needed; (Q6a) eating fewer meals in a day; (Q7a) failing to obtain food of any kind; (Q8a) going to bed hungry; and (Q9a) going the whole day or night without eating anything. Specifically, the respondents were asked whether a particular condition (Q1a to Q9a) associated with the experience of food insecurity occurred at all during the past 30 days, with a ‘yes’ answer being given a value of one and a ‘no’ answer given a value of zero. Each severity question was followed by a frequency-of-occurrence question, which asked how often the reported condition occurred during the previous 30 days (1 = rarely, 2 = sometimes, and 3 = often). The minimum HFIAS is zero and is obtained when a household responds ‘no’ to all of these questions. The maximum score is 27, which is obtained when a household responds ‘yes’ to an occurrence question and ‘often’ to the nine frequency-of-occurrence questions.

Following the guidelines by Coates et al. (2007), the HFIAS is computed as the sum of the frequency of occurrence during the past 30 days for the nine food insecurity-related conditions as follows:

At a household level, a high HFIAS score shows that a household is very food insecure, and vice versa.

4. Results and discussion

The results of this study are presented in Tables 2 to 6. Table 2 provides descriptive statistics of the sample, Table 3 presents the dietary scores and terciles, and Table 4 presents the affirmative responses to the HFIAS. Tables 5 and 6 show the linear regression results for the determinants of the HDDS and the HFIAS, respectively.

Table 2: Sample characterisation

Variable	Mean or proportion	Standard deviation
Age of household head	52.23	14.9800
Gender of household head	77.50	0.0384
Education of household head (% with secondary)	45.83	0.0457
Household size	5.91	2.4200
Household labour	2.86	1.6900
Livestock wealth	56.59	53.4900
Remittances (proportion)	40.00	0.0450
Market information (proportion)	64.17	0.0440

Table 3: Household dietary diversity scores and dietary terciles for smallholder farmers in the Mudzi district

	Proportion of households (%)
Household dietary diversity score
Proportion of households with HDDS of 2 to 4	15.00
Proportion of households with HDDS of 5 to 7	73.33
Proportion of households with HDDS of 8 to 9	11.67
Household dietary diversity score terciles
Low (groups 1 to 4)	15.00
Medium (groups 5 to 8)	80.00
High (groups 9 to 12)	5.00

Table 4: Percentage distribution of responses to the HFIAS during the past 30 days

Indicator	% affirmative response
Anxiety about food insecurity	69.20
Inability to eat preferred foods	84.20
Availability of only a limited variety owing to a lack of resources	83.20
Inability to eat even less-preferred food	81.70
Availability of only smaller amounts of food	52.50
Reduced number of meals	51.70
Having no food in the house	26.70
Going to bed without having eaten any food	18.30
Spending the day and night without any food	24.20

Table 5: Determinants of household dietary diversity in the Mudzi district

Explanatory variable	Coefficient	Standard error	p-value
Head_gender	0.0760	0.3076	0.805
Head_age	–0.0173	0.0102	0.092*
Household_labor	0.2342	0.0951	0.015**
Head_education^a	0.4631	0.2290	0.046**
Household_size	–0.1194	0.0671	0.078*
Remittances	0.8294	0.2573	0.002***
Livestock_wealth	0.0059	0.0024	0.018**
Market_information^b	0.5021	0.2828	0.079*
Constant	5.2653	0.7465	0.000***

Notes: N = 120, F(8, 111) = 7.37, R² = 34.69. Significant levels: ***1%, **5%, *10%. ^a,bVariables influenced by the project.

Table 6: Determinants of household food insecurity access score

Explanatory variable	Coefficient	Standard error	p-value
Head_gender	1.7614	1.5400	0.2550
Head_age	0.0432	0.0511	0.4000
Household_labor	–0.8298	0.4761	0.0840*
Head_education^a	–3.3972	1.1466	0.0040***
Household_size	0.7517	0.3358	0.0270**
Remittances	–3.8782	1.2881	0.0030***
Livestock_wealth	–0.0342	0.0122	0.0060**
Market_info^b	–3.1026	1.4162	0.0310**
Constant	14.9893	3.7378	0.0000***

Notes: N = 120, F(8, 111) = 8.15, R² = 37.00. Significant levels: ***1%, **5%, *10%. ^a,bVariables influenced by the project.

4.1 Sample characterisation

Table 2 shows that the average age of the household head is 52.23 years, while the mean household size is 5.91. The majority (78%) of the households are headed by men. Most household heads in the Mudzi district are educated, with 45% having attained secondary education. Respectively 40% and 64% of the households have access to remittances and market information.

The results in Table 3 show that the HDDS score for the Mudzi district ranges between 2 and 9. The modal HDDS is in groups 5 to 7, with 73.33% of the sampled households. The proportion of households in the lower tercile (groups 1 to 4) is 15%, in the medium tercile (groups 5 to 8) 80%, and in the upper tercile (groups 9 to 12) is 5%. Thus, based on a 24-hour recall, most households consume between five and eight food groups.

Table 4 shows that, based on the food experiences of the last 30 days, a high proportion of households in the Mudzi district had been anxious about food insecurity (69.2%), their inability to eat their preferred food (84.2%), the limited variety of food (83.2%), and even their inability to eat less-preferred food (81.7%). However, only 26.7% had been completely without food in the house, 18.3% had gone to bed without eating, and 24.2% had spent at least a day and night without eating any food at all.

4.2 Model estimates for the determinants of the HDDS

Linear regression results in Table 5 indicate that five variables – namely the availability of household labour, the education of the household head, remittances, livestock wealth and access to market information – have a positive influence on food security. However, the age of the household head and household size negatively influence food security.

4.3 Model estimates for the determinants of the HFIAS

Table 6 shows that households with more labour, education, access to remittances, market information and livestock wealth are less likely to be food insecure than their counterparts. However, larger households are more likely to be food insecure.

4.4 Discussion of results

The results show that household food security in the Mudzi district is influenced by the age and educational level of the head of the household, the size of the household, its access to market information, its livestock wealth and any remittances. However, the CIAT project can only meaningfully enhance food security among smallholder farmers by influencing household labour, education and access to market information.

4.4.1 Household labour

The results in Tables 5 and 6 indicate that a household's labour resources have a significantly positive effect on its dietary diversity but a significantly negative effect on its food insecurity access score. This means that households with sufficient labour resources are less likely to be food insecure and are therefore less anxious about food insecurity than those with insufficient labour resources. The probable explanation is that most farmers in Mudzi do not own draft cattle, which means that household labour is needed for farming activities. Those with sufficient labour conduct farming activities on time, which leads to high yields. These findings are consistent with the results of Bogale & Shimelis (2009). The implication is that the CIAT project can usefully promote labour-saving technologies in crop production and post-harvest handling practices.

4.4.2 Education of the household head

The results show that the level of education of the head of the household has a positive influence on the dietary diversity of the household and a negative influence on its food insecurity. This means that households with literate heads are less likely to be food insecure or anxious about food security than their counterparts. These results are consistent with the findings of Bashir et al. (2012), Gebre (2012), Amaza et al. (2009) and Idrisa et al. (2008). This implies that the ongoing CIAT project can contribute to household food security by capacitating smallholder farmers through the training of trainers in groundnut agronomy, marketing and farming as a business. Such training will assist farmers in adopting production-enhancing technologies and implementing good marketing practices, including the use of better groundnut varieties, the correct use of fertilisers, the recommended plant spacing, and the use of collective marketing.

4.4.3 Market information

The results conform to the findings of Nyikahadzoi et al. (2012) and show that access to market information has a positive influence on the household's dietary diversity and a negative influence on food insecurity. This implies that the CIAT project could usefully create and strengthen Rural Group Enterprises (collective marketing groups) through training, because these could improve information sharing among farmers. They could also improve farmers' knowledge about markets through carrying out participatory market surveys and sharing information. Such information could assist farmers in timing their sales, bulking produce and increasing bargaining power. In addition, the project could facilitate the creation of an institutional platform in the Mudzi district for collecting and sharing market intelligence and linking up with buyers.

5. Conclusions and recommendations

In this paper we empirically identified the determinants of household food security among smallholder groundnut farmers in the Mudzi district in Zimbabwe. The results show that the CIAT project is proposing to address the correct variables. Any project aimed at improving food security in the district should ideally facilitate the creation of an institutional platform for disseminating precise and reliable market intelligence, share information on strengthening Rural Group Enterprises, and link up with players in the input and output markets. Admittedly this is beyond the scope of the research and development community. Improving food security among smallholder farmers should therefore be a comprehensive agricultural development effort involving many stakeholders along the commodity value chains. Their joint efforts should focus on promoting labour-saving technologies in both crop production and post-harvesting handling practices and providing the means to access market information. Furthermore, farmers should be encouraged to engage in good livestock husbandry, as owning livestock will ultimately improve their food security status.

Acknowledgements

The authors would like to acknowledge the financial assistance of the International Fund for Agricultural Development and CIAT. They are also grateful to the language editor appointed by Development Southern Africa for editing the article.

Notes

⁶A ward is the second smallest administrative unit after the village and consists of several villages. Wards are given numbers rather than names, as some cut across several communities.