The study area
Assosa zone, the study area, is one of the three administrative zones in Benishangul-Gumuz region of western Ethiopia. Administratively, the study area is divided into seven districts, namely; Assosa, Homosha, Bambasi, Menge, Kurmuk, Sherkole and Odabildi-Guli districts. The zone has a total population of 283,707 people, out of which 144,616 and 139,091 are male and female, respectively. Furthermore, 86.28% of the population lives in rural area and 13.72% lives in urban area. The population density of the study area is 28 persons per kilometer square (BGRDGA (Benishangul Gumuz Region Development Gap Assessment) 2010). Mixed farming (crop production and livestock rearing) system is the main sources of livelihood for the majority of the population in the area. Crop production is dominated by rain fed agriculture while irrigation is practiced on small scale level. The major livestock reared in the area are cattle, donkey, goats, sheep and poultry (AZBARD (Assosa Zone Agriculture and Rural Development Office) 2015).
Sampling technique and sample size
The study employed three-stage random sampling method to select sample households. In the first stage, out of 7 districts in Assosa zone, three districts (namely Assosa, Bambasi and Sherkole) were randomly selected. In the second stage, a total of 12 peasant associations (PAs) were randomly selected using probability proportional to the number of PAs in each sampled districts. The reason for selecting PAs was that, in the study area almost all the households relied on agriculture and the emphasis of this study was on assessing the extent of food insecurity of households working on agriculture and their coping mechanisms. In the third stage, a total of 276 sample household heads were randomly selected based on probability proportional to size of the households in the selected PAs. The sample size for this study was determined by using Yamane formula (Yamane 1967).
$$ n=\frac{\mathrm{N}}{1+\mathrm{N}{(e)}^2}=\frac{40530}{1+\left(40530x{0.06}^2\right)}=276 $$
(1)
Where n = designates the sample size, N = designates total number of estimated household heads in the study area (40530) and e = designates maximum variability or margin of error (6%).
Data set and collection methods
For this study, primary data collected from sample households using interview schedule through the enumerators and the researchers was used. Particularly, primary data on the types and quantities of every food item consumed by the household head and his/her family members was collected using Weighed records method for 7 consecutive days from each sampled households. The reason for collecting the data from a single household for seven consecutive days was that food security is a sensitive issue that is affected by different unforeseen factors (religious, holidays, etc.) which can be captured by taking weighed data (Muche and Esubalew 2015). In addition to this, primary data on household’s socio-demographic and socio-economic factors as well as on households’ food insecurity and shortage coping mechanisms was obtained through interview schedule. Besides, focus group discussions and key informants interview were also employed to supplement the research finding with qualitative information.
Method of data analysis
To analyze the collected data, the study employed descriptive statistics, food insecurity index and Tobit model. Descriptive statistics such as mean, percentage and frequency were used to describe households’ food kilocalorie intake status and to explore the coping mechanisms to food insecurity in the study area. Furthermore, the study used Foster, Greer and Thorbecke (FGT) food insecurity index in the computation of the incidence, depth and severity of food insecurity. This model is widely applicable in poverty analysis. It is a class of additively decomposable measure of poverty and food insecurity. Foster and Shorrocks (1991, 1988) branded the decomposable components of FGT measures as consistent poverty indices and argued that they make analysis of the poverty dominance easier. Particularly in food security analysis, the model is essential in analyzing the sources of change in food insecurity due to changes in the components i.e. to know the change in food insecurity is due to the incidence, or increasing deprivation of the food insecure, or because of kilocalorie short-fall below the food security line have become more unequal, or some combination of the above. Thus, in this study the model enables to estimate the three food insecurity indicators, namely the number of households below the food security line (headcount), the extent of the short-fall of the kilocalorie of the food insecure from the food security line (food insecurity gap) and the exact pattern of distribution of the kilocalorie of the food insecure households (squared food insecurity gap). Accordingly, the Foster et al. (1984) measure used in estimation of food insecurity index components is given as:
$$ \mathrm{FGT}\left(\upalpha \right)=\left(1/\mathrm{n}\right){\sum}_{\mathrm{i}=1}^{\mathrm{q}}{\left[\left(\mathrm{c}-\mathrm{yi}\right)/\mathrm{c}\right]}^{\upalpha} $$
(2)
Where: FGT (α) is the FGT food insecurity index; n is the number of sample households; yi is the measure of per adult equivalent food kilocalorie intake of the ith household; c represents the cut off between food security and food insecurity households (expressed here in terms of caloric requirements of 2100 kcal*); q is the number of food-insecure households; and α is the weight attached to the severity of food insecurity. Regarding estimation of the model, when the weight attached to α = 0 the measure is simply the headcount ratio (incidence); when α = 1 the measure is food insecurity gap (depth of food insecurity); and when α = 2 the measure is squared food insecurity gap (severity of food insecurity).
Moreover, Tobit model was estimated to analyze determinants of extent of households’ food insecurity in the study area. Studies confirmed that, when a particular dependent variable assumes some constant value for some observations and a continuous value for the rest observations, the appropriate model will be a Tobit model developed by Tobin (1958) (Wooldridge 2002; Sisay and Edriss 2013; Agyeman et al. 2014; Bukenya 2017). Tobit is an extension of the probit model and it is one approach to deal with the problem of censored data (Johnston and Dinardo 1997). Thus, in this study the dependent variable was a censored variable in which it assumed a constant or threshold value of 2100 kcal/AE/day* for food secure households and the actual food energy intake in kilocalorie for food insecure households. Suppose, however, that Yi is observed if the latent variable Yi* < 2100 kcal and is not observed if Yi* > 2100 kcal. Then the observed Yi will be defined as:
$$ Yi=\left\{\kern0.75em \begin{array}{c}{Yi}^{\ast }=\beta Xi+ Ui\kern3.75em if\ {Yi}^{\ast }<2100\ kcal\kern1em \\ {}2100\ kcal\kern6em if\ {Yi}^{\ast}\ge 2100\ kcal\kern1em \end{array}\right. $$
(3)
Where: Yi* is the latent (unobserved) variable, Yi is the observed variable, Xi is vector of explanatory variables, Ui is a vector of error terms and β is a vector of parameters to be estimated.
*Note that 2100 kcal/AE/day is the threshold value of food security stated by FDRE (1996).
Operational definition of variables in the study
Extent of food insecurity
It is a limited dependent variable, taking the threshold value (2100 kcal) if the total food energy intake is greater than or equal to the threshold value and assumed the actual food energy intake for those households whose energy intake level is less than the threshold value. The quantity of food items consumed was converted to gram and the caloric content was estimated by using the nutrient composition table of commonly eaten foods in Ethiopia. Moreover, the estimated food energy was converted into adult equivalent and reached at figure of food calorie in kilo calorie/day/AE. Accordingly, household food calorie intake per day per adult equivalent (HFCi) was measured as:
$$ \mathrm{HFCi}=\frac{\mathrm{Total}\ \mathrm{calorie}\ \mathrm{consumed}\ \mathrm{by}\ \mathrm{a}\ \mathrm{household}}{\mathrm{Household}\ \mathrm{size}\ \mathrm{in}\ \mathrm{Adult}\ \mathrm{equivalent}\ast 7} $$
(4)
Nature of settlement of the household heads
This is a dummy variable used to indicate origin of household’s. The variable took the value of 1 if respondents were settlers and 0 if natives. As depicted in Asfir (2016), unlike settlers, native households in the study area were highly resistant to accept new technologies. However, studies argued that adoption of new technologies improves agricultural production and productivity (Tsegaye and Bekele 2012) which in turn reduces households’ exposure to incidence of food shortage and insecurity. In this study, this variable was hypothesized to affect extent of households’ food insecurity negatively.
Sex of head of household
It is a dummy variable taking the value 1 if the sex of household is male and 0, otherwise. As to Baten and Khan (2010) finding, female-headed households can find it difficult than men to gain access to valuable resource, which helps them to improve production and gain more income, this in turn increases their probability of being food insecure. Thus, in this study, it was expected to affect extent of households’ food insecurity negatively.
Age of head of household
It is a continuous variable measured in years. Many studies argued that young households’ heads are stronger and energetic than elderly households as they are expected to cultivate larger-size farm and obtain high yield (Abafita and Kim, 2014; Babatunde 2007). Hence, in this study age of the household head was expected to affect extent of food insecurity negatively.
Educational level of head of household
It is a continuous variable measured in years of schooling of the household head. Education, which is a social capital, has a positive impact on household ability to take good and well-informed production and nutritional status (Babatunde 2007). Besides, Amaza et al. (2006) argued that households with higher years of schooling are less likely to be food insecure as it enables them to produce more and consume more. Thus, higher years of schooling was expected to affect extent of food insecurity negatively.
Family size
It is a continuous variable which refers to the number of family members of the household. Studies argued that larger family size tends to exert more pressure on households consumption than the labor it contributes to production (Stephen and Samuel 2013; Muche et al. 2014). Therefore, in this study, larger household size was expected to affect extent food insecurity positively.
Dependency ratio
It refers to the proportion of economically inactive labor force (less than 15 and above 65 years old) to the active labor force (between 15 and 65 years old (Velasco 2003). Due to scarcity of resources, higher dependency ratio imposes burden on the active and inactive member of household to fulfill their immediate food demands (Muche et al. 2014). Besides, higher dependency ratio indicates that the labor force is small, with a constraint on the household per capita income and consumption, which also influences the wellbeing of the household members (Nugusse et al. 2013). In this study, it was expected to positively affect extent of households’ food insecurity.
Livestock ownership (excluding oxen and donkey)
It is a continuous variable measured by the number of Tropical Livestock Unit (TLU). Livestock are important source of food and income for rural households. Households with more livestock produce more milk, milk products and meat for direct consumption. Besides, livestock enable the farm households to have better chance to earn more income from selling livestock and livestock products which assist them to purchase stable food during food shortage and invest in purchasing of farm inputs that increase food production, and ensure household food security (Mitiku et al. 2012; Gemechu et al. 2015). Livestock possession mitigates vulnerability of households during crop failures and other calamities (Abafita and Kim, 2014). Thus, this study hypothesized that owning more TLU of livestock was expected to have negative effect on the extent of food insecurity of households.
Number of oxen and donkey owned
It is a continuous variable measured in numbers owned. Oxen and donkey serve as a source of traction power in many developing countries, thereby significantly affecting household’s crop production. Animal traction power enables households to cultivate their land; others land through renting, share cropping, and execute agricultural operations timely that will enhance households access to food items (Muche et al. 2014). Accordingly, in this study more number of oxen and donkeys owned by a household was expected to affect the extent of food insecurity negatively.
Cultivated land size
It is a continuous variable which refers to the total land cultivated by a household in the past one year production period. A larger size of cultivated land implies more production and availability of food grains (Mitiku et al. 2012). Therefore, higher production and the increased availability of grains produced help to insure food security status of households (Asmelash 2015). Hence, the size of cultivated land was expected to have negative impact on extent of food insecurity.
Access to irrigation
It is a dummy variable taking the value 1 if the farmers have access to irrigation and 0, otherwise. Irrigation, as one of the technology options available, enables smallholder farmers to directly produce consumable food grains and/or diversify their cropping and supplement moisture deficiency in agriculture so that it helps to increase production and food consumption (Van der Veen and Tagel 2011). Thus, in this study, it was expected to have negative impact on extent of households’ food insecurity.
Farm income
This is a continuous variable which measures the amount of income obtained from crop production and livestock rearing measured in US Dollar. According to Beyene and Muche (2010) finding, higher farm income earning enables farmers to purchase different nutritious food items to satisfy their family food demand. Thus, for this study, farm income was hypothesized to affect extent of households’ food insecurity negatively.
Off/non-farm income
It is a continuous variable which measures the amount of cash income obtained by any household member from off-farm and non-farm activities measured in US Dollar. Studies argued that households with higher off-farm and non-farm income are less likely to be food insecure as it enables them to purchase different food items to satisfy their family needs (Beyene & Muche, 2010; Abafita and Kim 2014). Thus, off/non-farm income was expected to affect extent of food insecurity negatively.
Cost of inputs
It is a continuous variable measured in US Dollar by converting the amount of the agricultural inputs used (such as fertilizers, seeds, pesticides, chemicals, and other agricultural implements.) into monetary value based on their market price. Investing higher amount of money on farm inputs helps farmers to increase their crop production and livestock breeding (Arene and Anyaeji, 2010). In this study, it was expected to affect extent of households’ food insecurity negatively.
Access to training
It is a dummy variable that takes value 1 if a household gets access to agricultural related training and 0, otherwise. Formal agricultural training on modern technologies (proper types and rates of fertilizer application, improved varieties of seeds, agro-chemicals, etc.) helps farmers to get better production, and then this most likely leads to obtain more income to fulfill their family requirements by enhancing their agricultural production skills, knowledge and experiences (Yishak et al., 2014). Therefore, in this study, it was expected to affect extent of households’ food insecurity negatively.
Frequency of extension contact
It is a continuous variable measured in number of visits by extension agent per year. More frequent extension contact enhances households’ access to better crop production techniques, improved input as well as other production incentives, and thishelps to improve food energy intake status of households (Hussein and Janekarnkij 2013; Nugusse et al. 2013). Accordingly, in this study more number of extension contacts were expected to affect extent of households’ food insecurity negatively.
Access to credit
It is a dummy variable, which takes the value 1 if the household had access to credit and 0 otherwise. Availability of credit eases the cash constraints and allows farmers to purchase inputs such as fertilizer, improved crop varieties, and irrigation facilities; which in turn enhance food production and ultimately increase household food energy intake (Stephen and Samuel 2013). In this study, it was expected to affect extent of households’ food insecurity negatively.
Remittance and aid
It is a dummy variable, which takes the value 1 if the household had access to remittance and aid in the past one year and 0 otherwise. Both remittance and aid,from governmental and non-governmental organizations are important to smooth consumption in the case of shock and shortage for the time of emergency (Okyere et al. 2013; Mesfin 2014). Thus, for this study, it was expected to negatively affect extent of households’ food insecurity.
Distance to market
it is a continuous variable measured in kilometer (km). Proximity to the market may create opportunity of more income by providing off/non- farm employment opportunities, which determine income level of rural households. In addition, the closer the farmer is to the market the more likely the farmer gets valuable information, purchase agricultural inputs and final products required for family consumption. Therefore, this variable was expected to positively determine households’ extent of food insecurity.