Study area
This study was conducted in Munguli village, Mwangeza ward in Mkalama District in Tanzania. The study area experiences a warm and dry climate with mean annual temperatures ranging from 25 to 30 °C. The mean annual rainfall ranges from 300 to 600 mm. The rains mainly fall between December and May [25]. The area consists of a mixture of savanna grasslands, woodlands, and deciduous bushlands and shrub lands [26]. Further, information obtained from the village office indicated that at the time of conducting this study, the area had a total of 419 households distributed in four sub-villages as follows: Kipamba (130 households), Munung’una (111), Midembwi (103) and Mwazururaji (75). Residents in these areas are traditionally foragers who are currently involved in various forms of livelihoods including foraging, beekeeping, farming, and trade. The term ‘forager’ is a more generic name for ‘hunter-gatherer’ [27]. In this study, the two terms are used interchangeably.
Data collection
Data were drawn from 200 households in a cross-sectional survey carried out between May and August, 2017. A household was defined as a group of people who sleep under the same roof and take meals together. The study involved a combination of focus group discussions, key informant interviews and household survey. Household selection was based on random sampling procedures based on the official list of households obtained from village leaders. The sample size was estimated through a proportionate sampling technique described in Miah [28] as follows:
$$n = \frac{{N\sum {N_{h} P_{h} Q_{h} } }}{{N^{2} D^{2} + \sum {N_{h} P_{h} Q_{h} } }}$$
(1)
$$n_{h} = \frac{{N_{h} }}{N}*n$$
(2)
where \(n =\) total sample size, \(n_{h} =\) Sample size for h stratum (village), \(N =\) total population (419), \(N{}_{h} =\) population size of h stratum, \(P_{h} =\) proportion of households involved in foraging, beekeeping or farming as their primary livelihood activity in h stratum (0.5), \(Q_{h} =\)\(1 - P_{h}\), D = d/z, \(d =\) precision (error). Using value for d = 0.05, z = 1.96 (95%, confidence interval), substituted in Eq. 1, the sample size obtained was 200 households. The proportionate sample size of households living in each sub-village was calculated using Eq. 2 as follows: Kipamba (62 households), Munung’una (53), Midembwi (49) and Mwazururaji (36). The research project was approved by the research committee at the Institute of Rural Development Planning. Interviews were conducted in Kiswahili, the National language in Tanzania. The responses were then translated into English. The survey was administered to the household head or a responsible adult in the household who could respond on behalf of the entire household. According to Barrett [29], food security is based on three pillars: availability, access, and utilization. This is a hierarchical classification because availability is necessary but not sufficient to guarantee access, while access itself is necessary but not sufficient to assure effective utilization. A large variety of indicators for assessing food security have been proposed [30]. This study focuses on the three pillars with a set of selected indicators as detailed in the following sub-sections.
Food availability
Food availability refers to the sufficiency of a food supply to meet people’s needs [31]. In this study, food availability was indirectly assessed by inquiring the extent to which wild food resources could be obtained for the household needs. The availability of important wild food resources was rated as 1 = very low, 2 = low, 3 = high or 4 = very high. Household food availability meant food obtained through cultivation in the field and collection from the wild environment, and was assessed through Months of Adequate Food Provisioning (MAHFP) 12 months preceding the survey. This indicator refers to the number of months per year that households report no food shortages.
Food access
To measure food access, the Household Food Insecurity Access Scale (HFIAS) based on a household’s recent experiences of food insecurity were used as indicators of Household Food Insecurity (HFI) as described earlier [32]. The HFIAS guideline was used to review questions in order to suit the local context and ensure that the questions were understood correctly. A total of nine food insecurity conditions were inquired. These conditions were whether four weeks preceding the survey the respondent or any member of the family worried about food, unable to eat preferred foods, ate a few variety of foods, ate food they did not want to eat, ate smaller meal and ate fewer meals in a day. Others were no food of any kind in the household, went to sleep hungry or went day and night without food. As described in the HFIAS guideline, these indicators provide summary information on the prevalence of households experiencing one or more behaviours in each of the three domains reflected in the HFIAS namely anxiety and uncertainty, insufficient quality and insufficient food intake and its physical consequences [32]. Each indicator was given 1 point of score for each form of food insecurity that a household experienced or zero if a given form of food insecurity was not experienced. An affirmative answer was then followed by a frequency-of-occurrence question to determine if the condition happened rarely (once or twice), sometimes (3–10 times), or often (> 10 times) during the reference period.
Dietary diversity
Dietary diversity was measured in terms of dietary diversity scores (DDS), a common indicator that counts the number of food groups consumed over a certain period of time [33]. It is one of a priori defined diet quality indices used to assess nutrient adequacy, and has been positively associated with the number of different foods consumed [34]. High food variety is regarded to be necessary for an adequate nutrient intake, to lessen the chances of deficient or excessive intake of single nutrient [35]. In this study, respondents were asked to recall all the dishes they had consumed in the previous 24 h. Food items were categorized into 12 different food groups. These were cereals, tubers, legumes, meat, egg, vegetables, fruits, oil, sweets, milk, fish and sugar or honey. Each food group counted toward the household score adding “1” if any family member consumed a food item from the group.
Statistical analysis
Livelihood options
In respect of the livelihood options, data were fitted to a Multinomial Logit (MNL) model to estimate the significance of the factors that influence a household’s choice of the main activity. The model is expressed in Eq. 3 as follows:
$${\text{Pr }}\left[ {Yi \, = \, j} \right] \, = \frac{{\exp \;(\beta_{j}^{i} \;X_{i} )}}{{\sum\nolimits_{j\; = \;0}^{j} {\exp \;(\beta_{j}^{i} \;X_{i} )} }}$$
(3)
where: \(\Pr [Y_{i} = j]\) is the probability of choosing foraging or farming with beekeeping as the reference category; j = is the number of possible activities; \(j = 0\) = beekeeping; \({\rm X}_{i}\) is the vector of the predictor variables and \(\beta_{j}\) is a vector of the estimated parameters. Since logit model uses logarithmic transformation to assume linearity of the outcome variables on the explanatory variables, the specific logit model to predict the odds of activity choice is given in Eq. 4.
$$\ln \left( {\frac{{\mathcal{P}}}{{1 - {\mathcal{P}}}}} \right) = \beta_{0} + \mathop \sum \limits_{i = 1}^{n} \beta_{i} X_{i} + \varepsilon_{i} { }$$
(4)
where n is the total number of variables, \(\beta_{i}\) is the regression constant, \(X_{i}\) is the logit coefficient for the variable and \(\varepsilon\) is the error term. From Eq. 4, the quantity \(p/\left( {1 - p} \right)\) is the odds ratio expressed as a linear function of the independent factors. Data were analysed using Statistical Social Sciences (SPSS) version 16 (SPSS Inc., Chicago, IL, USA). In the MNL analyses, several models were tested with a forward entry method. At each step, the term whose addition caused the largest statistically significant change in − 2 Log Likelihood was added to the model. The final model included important predictors only.
Food availability
Food availability involved assessment of wild foods and food obtained through crop cultivation and/or purchase. As for the wild foods, a composite index for availability of the resources was constructed as follows:
$$I = { }\mathop \sum \limits_{i = 1}^{N} X_{i} .W_{i} /N$$
where, I = composite index of a particular wild food, \(X_{i}\)= individual wild food availability, Wi = respective weight for a given wild food (very low = 0.25, low = 0.5, high = 0.75, very high = 1), and N = total number of responses. Indices show the degree of availability of important wild foods. Food obtained through crop cultivation, collection from the wild or purchase was assessed through MAHFP and compared between households mainly subsisting on foraging, beekeeping and farming.
Food access
In calculating HFIAS score, the responses on frequency-of-occurrence were coded as 0 = never, 1 = rarely, 2 = sometimes, or 3 = often. A HFIAS score variable was calculated for each household by summing the codes for each frequency-of-occurrence question. All cases where the answer was “no” were coded 0. Thus, a possible maximum score for the household was 27 while the minimum score was 0 for households that responded “no” to all the occurrence questions. Average HFIAS score was computed as the sum of HFIAS scores in the sample/number of HFIAS scores (i.e. households in the sample). The higher the score, the more food insecurity (access) the household experienced. The lower the score, the less the food insecurity.
Further, socio-economic and demographic factors were used to predict household food insecurity. A total of nine variables were included in the regression analysis to identify the key predictors of household food insecurity. These variables were age of the respondent, sex of the household head, highest education level of the household head and household size. Others were the dependency ratio, farm size, possession of agricultural tools, involvement in wage labour and household main activity. Dependency ratio was defined as the number of dependent children < 18 years of age plus the number of dependent elderly over 65 years of age relative to the number of working aged adults in the household [36]. Diagnostic analyses were carried out to assess the potential for collinearity among the independent variables by assessing Variance Inflation Factors (VIFs). The observed VIFs ranged from 1.3 to 2.2 which is below the cut-offs above which collinearity may be considered a problem [37]. Thus, inclusion of the separate predictors in the model is statistically valid. Dependent variable had a binary outcome: food secure household (HFIAS ≤ 17) or food insecure household (HFIAS > 17) as described in the FAO report [38]. Analysis of food access was also performed by assessing the relationship between food insecurity conditions and predominant household activity (foraging, beekeeping or farming) using a Chi-square test or analysis of variance. In these analyses, the significance level was set at p < 0.05.
Dietary diversity
Dietary diversity scores (DDS) were calculated from 24 h recall data. These dietary diversity scores were defined as the sum of food groups (0–12) from the chosen food items. DDS was used as a response variable against independent socio-demographic factors such as age of the respondent, sex of the household head and highest education level of the household head, family size and dependency ratio. Means and standard deviation differences for DDS in households with foraging, beekeeping and farming were compared using a one way analysis of variance (ANOVA) at 95% level of confidence. Because the responses on household’s main activities were ubalanced, HFIAS scores were subjected to Welch F-test which is more conservative than the regular ANOVA F-test. This test statistic reduces the odds of commiting type I error which could result from unbalanced sample size or unequal variance.