Study title

Ethiopia

Individual

Data used in this study were collected from smallholders in the Rift-valley zone in south-eastern Ethiopian. A two-round survey with two-year intervals (2015–2017) was administered on 1143 randomly selected IBI-adopter and non-adopter households. Recruitment of households included in these two surveys was worked out as follows. First, we selected three districts, namely Bora, AJK and Arsi Negele, out of the five districts where OIC implemented IBI. Second, we identified a random sample of kebeles within the three districts , including those kebeles covered by IBI as well as those that OIC did not cover. Finally, sample households were randomly drawn from all these selected kebeles. In the first round of survey that we conducted during January-April, 2015, data were collected from a total of 1143 households, out of which 461 were adopters and 682 were non-adopters of IBIs, over the period 2013-14. The dataset covers information on household, village and IBI intervention, including household demographic characteristics, investment in agricultural inputs, consumption, use of financial services as well as village infrastructure and access to markets. The same questionnaire used in the baseline survey was also administered in the end-line survey. The questionnaires did take about 3 hours per interview . Respondent attrition was minimal. Only four households who were considered during the baseline were not covered during the end line survey. This study is thus based on a balanced panel of 1139 households, of which 596 were adopters and 543 were non-adopters, during the second survey observation. Over the two survey periods, adoption or treatment status was changed in subsequent years, with some households joining IBI adoption and others dropping out. In addition, uptake payout data were collected from OIC, and cross-checked with the responses of the households in the survey. An advantage of these data in studying the impact of IBI is that the baseline observation in 2015 coincides with the massive expansion of IBI in villages that were rank-filed during the initial two years of intervention, to be considered in the subsequent intervention periods. This enables us to identify the impact of IBI adoption using 2015 as baseline information for both adopters and non-borrowers. Moreover, there is little reason to believe that OIC’s expansion to other villages has been systematic and endogenous to village outcomes. In principle, if a kebele is considered for IBI implementation, all residents in that kebele were eligible to buy IBI. However, households may have self-selected into IBI adoption, and participation can be endogenous at the individual level, which we explicitly tackle in the empirical analysis. We measure the impact of IBI adoption on two welfare indicators: household consumption and investment in high-risk high-return inputs. Both set of variables are continuous in nature. Household consumption is an aggregate of selected food and non-food consumption. Food items consumed both from own sources and from purchases over a period of one week were included . Necessary adjustments are made to make measured items and units. To minimize measurement error from heterogeneity in age among household members, per capita consumption is used.