Plan Foncier Rural Impact Evaluation 2018

Baseline

The impact evaluation consists of gender and youth disaggregated data collection at base line, before the start of the intervention, in both the treatment and control villages. End line data will be collected at least 2 growing seasons after issuing of documentation to farmers.

The sample consisted of 2968 households, which were taken from 26 villages selected for the implementation of a Plan Foncier Rural (PFR), or rural landholding plans, these were the treatment villages and 27 control villages that did not benefit from a PFR.

The treatment villages were assigned by the ProPFR team in geographic clusters. The assignment of control villages followed this geographic clustering, also using further village level data with the aim of finding similar villages to maximize comparability.
These clusters were spread across the communes of Bembéréké, Sinendé and Kalalé in the north and Tchaourou in the south of the department of Borgou.

Villages were selected from 11 geographical clusters of villages facing similar issues, allowing easier logistical planning for the rollout of the PFRs.

Villages selected to be part of the programme had the following characteristics:
• Bordering/near to a classified national forest
• At high risk of land grabbing,
• The presence of another GIZ supported SEWOH project1
• Agropastoral areas (in particular the presence of transhumance –cattle driving - corridors)

But should not have the following:
• Villages bordering Nigeria, within the band of increased security
• MCA intervention with a PFR
• Suffered serious conflict which could block the realisation of a PFR, or where a PFR may reignite past conflicts.

These characteristics alongside the desire of the implementing team to select villages in clusters, for practical reasons presented the first challenge in selecting suitable comparison villages to measure the impact of the ProPFR programme. Clustering meant that villages selected for comparison should be near the clusters to be comparable, but given the typical geography of villages in northern Benin, in that most people live in the village centre rather than spread evenly with sufficient density at the village boundary, and the lack of clearly defined village boundaries, a geographic discontinuity could not be exploited.

The second challenge in selecting comparison villages arose due to a change in the village definitions in 2013, when Benin changed from 3758 to 5290 villages which is often referred to as the “nouveau découpage”. Some old villages were split but there are no clearly defined village boundaries for the new set of villages. ProPFR selected from among the new villages, so the control villages also needed to be selected from this list. Given that the last census was collected prior to this new definition of villages, no data about the villages existed that could easily be used in matching villages to those selected for the ProPFR.

Due to this lack of data on the characteristics of the people residing in the villages, Geographical Information Systems (GIS) data were used to match each of the treatment PFR villages to a control village. Villages which were previously included in the MCA’s wave of PFRs were excluded from our study due to the difficulty in separating the effects of the two programs (MCA vs ProPFR).
For each PFR village, a buffer of 20km was drawn and the union constructed for each cluster. Within this area, other villages were considered as a potential control village. Of the selection criteria, the only one applicable from GIS data is the proximity to a national forest. Where villages were close to a national forest, we attempted to match it with a control village also close to a national forest.
The additional criteria on which villages were matched were the proximity to a main road (as classified by the Open Street Map shapefiles for roads) and the number of buildings in the central agglomeration of a village. Main roads are used as a proxy for access to markets and thereby potentially income levels.

The size of a village and the amount of land which can be used around it will be influenced by the size of the population as well as the presence of national forests.
This strategy is similar to a Coarsened Exact Matching (CEM) strategy (see Blackwell et al, 2009), in which key characteristics are reduced (perhaps from continuous variables) to a small number of categories and matched with one another exactly.
In our selection of villages, one control village was selected for each treatment village based on the key characteristics, defined as proximity to national forests (5km) and main roads (1km), and having a similar number of buildings (within 1km of the central point).

For a small number of villages, we faced an issue of common support, meaning there were no exact matches on the key characteristics. In this case other nearby villages were selected which fulfilled as many of these characteristics as possible.
Data were collected on a wide range of variables following the theory of change, which states that the improvements in institutions and the PFRs may lead to improved perceived land tenure security and improved access to land for women and young men through the activities carried out by the ProPFR team.
This perceived land tenure security is often seen as key to agricultural investments and thereby food security in the long term, as it allows long-term planning. The issuing of official documentation provides collateral for a loan should households wish to borrow and invest in productive activities or smooth consumption.