Diversified Agriculture and Water Management 2017-2019, Independent Performance Evaluation

BASELINE DATA COLLECTED BY THE PREVIOUS EVALUATORS

Di PAP baseline survey data: The Di PAP baseline survey is a retrospective baseline survey. The survey was administered by BERD to a representative sample of 500 PAPs (out of roughly 1500) in October 2013 as some PAPs began receiving their new plots on the perimeter. A total of 388 PAPs out of the selected 500 PAPs completed the survey. The survey collected information on household demographics, production and land use outside the perimeter, perspectives on compensation, anticipated land use in the perimeter, household assets, access to credit, revenue, and training received. The survey data were received and are thus being delivered by Mathematica in 16 separate files. The files map to specific sections of the baseline survey with each file at the observational unit level of the information collected. Mathematica used these data for the baseline report.

Di Lottery baseline survey data: The Di Lottery baseline survey was administered in late 2013 by CERFODES to all 2,178 Di lottery applicants who met the lottery eligibility criteria (though not necessarily admitted to the lottery). A total of 2,128 applicants completed the survey. The survey collected data on demographic characteristics, socioeconomic status, agricultural experience, and other background characteristics relevant to the criteria for admission to the Di lottery. These survey data are delivered in a single data file in which the unit of observation is the lottery applicant. Mathematica used these data for the baseline report..

Farmer training baseline household survey data: NORC and CERFODES designed and administered the farmer training baseline household survey which was to provide a baseline for the evaluation of a range of ADP activities. The baseline data were collected in two rounds in parallel with the two agricultural seasons in Burkina Faso. The first round of the baseline data was collected immediately after the 2011 dry season, and the second round was collected immediately after the 2011 rainy season. The survey's targeted sample comprised 1,082 matched pairs of farming households with each pair containing one household from the project's treatment area and the other from the comparison area. The lengthy baseline survey is comprised of seven modules focusing on the following content areas: household, agriculture, animal husbandry, forestry, consumption and credit, food security, and health. The data collected via each module was delivered to Mathematica in multiple data files given that the data collected under each module could have been collected at multiple levels (e.g. household, individual, plot, and crop levels). As such, we are delivering each data file as a unique restricted-use file (roughly 50 files per survey round/season) with each file at the observational unit level of the information collected. Mathematica used a subsample of these data for the baseline report.

Barymetric survey data: The barymetric survey was administered by CERFODES to a subsample of farmer training households in two rounds: baseline in mid-2012 and one-year follow-up in mid-2013. In total, 153 households completed the baseline survey of which 146 completed the follow-up survey. In each round, the survey obtained data on cattle herd size, health, weight, milk production, and other related bovine information from each sampled household. The data for each round were received, and are thus delivered, in three separate files mapping to the three distinct sections of the survey: i) cattle herd characteristics at the household level; ii) cattle weight and other bovine characteristics at the cattle level; and iii) milk production at the cattle level. Mathematica did not use these data in the baseline report.

Supplemental household survey data: Implemented by CERFODES in late-2013, the farmer training supplemental household survey primarily collected information on the training and support farmers in the project's treatment area received from AD10. A short additional section of the survey also collected estimates of 2013 household agricultural production. Of the 1,082 farmer training households in the treatment area, 949 completed the survey. The data were received, and are thus delivered, in two separate files which map to the two sections of the survey: the first section covering training and support at the household level, and the second section covering household production at the crop level. Mathematica did not use these data for the baseline report.

FOLLOW-UP/ENDLINE DATA TO BE COLLECTED BY MATHEMATICA

Di perimeter households: We interviewed a sample of households whose land was expropriated to construct the Di Perimeter and who received land in compensation. These are called PAP household, with PAP standing for persons affected by the project. The sample comprises all households who received only rice plots in compensation, households who received rice and polyculture plots and households with only a female PAP head. For households who received only polyculture plots, we draw a sample proportional to size. In addition we draw samples from non-APs from neighboring communities. Women and youth who received small plots of land as parts of women and youth groups are sampled in a two stage sampling process, whereby first groups are chosen, then individuals in these groups. Note that because the sampling for the interim and final evaluation differs from the baseline it is not possible to link interim and baseline samples.

Di Lottery households: We include all Di lottery households for whom baseline information is available in our survey.

Farmer training households: We include all farmer training households who are part of the CERFODES-NORC baseline survey and who participated in farmer training activities (per the AD10 trainee identification survey).

Qualitative Evaluations:
We identified our criteria for selecting participants before fielding the study. Certain key informants were selected purposively, based on their role or experience. For example, we interviewed the staff member who was most knowledgeable regarding each aspect of the implementation, while striving to avoid burdening any one agency. For farmer training participants, we used selection criteria to ensure balance and variation based on factors such as geography, demographic characteristics, and so on. For members of PAP households, we will use our criteria to identify participants through contacts and chose them purposively. The composition of the focus groups took a number of elements into consideration, including people's demographics, experiences with the project, geographic characteristics, and plot size. The local data collection firm handled participant selection, in conjunction with Mathematica.