This study aimed to identify the determinants of household DD using household panel data collected at four time points in rural Ethiopia. A mixed effect and zero truncated Poisson regression model was fitted. The findings indicated that the variety of crops produced by the households, the income generated from animal products, the harvesting season, the years of education of the head of the household, and the women’s decision-making power about agricultural issues in the household were positively associated with HDDS while being a female headed household was negatively associated with HDDS. No association was found between time taken to reach a market and HDDS.
Results of this study indicate that producing one additional crop variety contributes for about 3.4% relative changes to HDDS. The positive association between HDDS and food crop production variety found in the current study is consistent with many other studies conducted in African and Asian countries (29, 36, 49–50). In some other studies, the association of crop variety with HDDS varies under different conditions like harvesting seasons and the number of production diversity (30, 37, 50) but that was not the case in the current study. In this paper, we quantified the relative change to HDDS when one additional crop variety is added to the same household production (3.4%). This result is similar with a study done in India where one additional crop produced by the household contributed for about 4% of the HDDS (48).
A second variable found associated with HDDS was the income generated from animal products, defined as total value in birr obtained from the selling of animal products owned by the household. We estimated that a one birr increases in income generated from livestock products in a month contributed 2.6% relative change of HDDS of the household. This finding was in line with previously conducted studies, including some from Ethiopia, which have identified that either the availability of livestock, the number of owned livestock or the income obtained from owned livestock was significantly associated with HDDS (11, 30, 37, 49–51).
Being in the post-harvesting seasons was found to be positively associated with HDDS. The relative change in HDDS due to the change in harvesting season from pre to post harvesting was 6.4%. This finding is concordant with studies done in Nigeria, Ghana, Ethiopia and Tanzania (37, 38, 52, 53). Additional studies have also indicated that seasonality can play an important role on factors affecting HDDS, including production and access to markets (37, 54, 55). Contrary to our findings, HDDS was significantly higher during the lean season and the harvesting season as compared to the post-harvest season in a study conducted in Burkina Faso (51). The difference between the current study and the one conducted in Burkina Faso could, at least in part, be due to difference in the types of crops commonly produced in these two settings.
If a household has a female head, the relative change of HDDS was a reduction of 7.8% compared to a household with a male head. This finding is consistent with the findings reported by two studies conducted in Ethiopia (27, 56), while it contradicts the findings of studies conducted in three other African countries (Tanzania, Burkina Faso and Malawi) (19, 28, 49). The difference between the Ethiopian findings and those of the other African countries is probably attributable to differences in gender roles in controlling over the resources of the household used to improve DD.
It is expected that better education of the head of the household, has a positive effect on the DD of the household, as education is likely to be related to a higher knowledge of the benefits of eating diversified food. The findings of the current study confirm this expectation, as a unit increase in years of education of the household head had a 1.2% contribution on relative change of HDDS. Different studies conducted in Africa and South America have found similar associations, with years of education or better schooling of the head of the household positively contributing to diversified food for the household (11, 13, 19, 28, 53). Moreover, literacy of the head of the household has impact on having diversified food according to studies conducted in Ethiopia and India (27, 48).
The other important finding was the association of the decision-making power of the women regarding agriculture with HDDS. A one unit increase in the women empowerment index for agriculture was associated with a 3.8% increase of the HDDS. There was scarcity of study that explored the direct association of women’s decision making about agriculture and HDDS. However, studies exploring some indirect but relevant associations that mainly focuses on the association of women’s empowerment and either the maternal and child DD, the women’s DD or individual DD were identified. The better the decision-making autonomy of the women in the household, the better the DD of both the women and children. Positive associations of women’s decision-making and maternal and child’s Dietary Diversity Score (DDS) and many of the nutritional outcome measurement like body mass index (BMI), Weight for Height Z score (WHZ), and Weight for Age Z Score (WAZ) were documented in some studies (35, 41, 57, 58). Limited studies have assessed the direct association of women’s empowerment and HDDS and reported that female participation in decision-making is associated with a higher diversity both in terms of production and consumption and concluded that the intra-household decision-making is an important covariate for both production diversity and dietary quality (59). This assocition could be related to the fact that as the women get a relatevely better autonomy over the household’s resources, especially over production types and income generated from their own production, they are more likely to invest over the houshold food consumption than men.
Time taken to local markets was not found to be associated with HDDS. This was in contradiction with the findings of other studies. In most of the studies exploring determinants of DD, time taken to reach local markets or the distance to reach the nearest local markets were significantly associated with HDDS (37, 48, 55). Access to market was found to be associated to HDDS in further studies conducted in India and Ethiopia (48, 54). However, a study from Tanzania did not find this association, in line with the findings of our study (53). These inconsistencies in the literature could be due to differences in the season of data collection, as well as the methodological approaches used. Local markets are common in Ethiopia, and are widely spread, so that distance to them might not make any difference.
A different interesting way to quantify and compare the effect of the various determinants on the outcome is to estimate their effect on HDDS when they vary from their minimum to their maximum values. For example, a household producing the maximum number of different crops compared to a household who produces the minimum number of crops, is expected to have a HDDS 31% higher. A household where the women empowerment score is 4 (maximum) compared to 1 (minimum) will have an estimated 11.84% difference in HDDS. The longest education (15 years) compared to absence of education gives an estimated difference of HDDS of 19.5%. These quantifications can be used when planning interventions, each coming with a cost and then an estimated predicted benefit for HDDS. Our quantification of the relative change in HDDS when each determinant is increased by one unit, has the same scope. One can argue, for example that working on increasing the women empowerment by one unit (3.8%) can have an effect which is comparable to the increase of one additional crop type (3.4%). Each year of education is expected to contribute with a 1.2% relative change.