Impact of Pedestrian Footbridges on Economic, Health and Educational Outcomes in Rural Communities

The study will be a matched-cohort study, in which twelve bridge sites will be matched to twelve comparison sites. The bridge sites have been defined and identified by B2P staff in conjunction with government officials through a systematic needs assessment carried out in 2014. Of the twelve bridge sites that have been confirmed for the study, seven are located in the Southern Province, and five are located in the Western Province. Comparison sites were identified from a 2018 needs assessment carried out by B2P. As with the bridge sites, seven of the comparison sites are in the Southern Province and five are in the Western Province. The table of bridge and control sites with their respective districts and provinces can be found in Appendix A.

The minimum detectable effect analysis for outcomes of interest determined that, with these 24 sites, the investigators should survey representatives from 100 households in each treatment site (those receiving a bridge) and comparison site (those that will not). For both treatment sites and comparison sites, B2P will identify the impacted side - that which their assessment indicates would benefit the most from the bridge. Investigators will then obtain the administrative list of households from the umudugudu chief (village leader) or cell secretary for the village that is closest to the bridge on the impacted side. If the list consists of at least 100 households, investigators will randomly select the 100 households to visit from this list. If the village has fewer than 100 households, enumerators will visit all households in that village and then, from the list for the next closest village, randomly select the remaining number required to reach 100 households.

As part of the pilot study, investigators are testing an experimental method to determine which households to sample within a community. This method involves tracing satellite imagery of the relevant communities to derive a vector dataset of all the households in the area, from which investigators randomly select certain ones based upon an algorithm. This approach is intended to make the sampling more robust by reducing the opportunity to introduce human bias. The tracing of homesteads will be done using freely available Bing aerial imagery and the data generated will be contributed to OpenStreetMap (OSM), in accordance with the Bing imagery license for OpenStreetMap. While all the households traced will be uploaded to OSM and be publicly available, all subsequent analysis of which households are to be surveyed and survey results themselves will be kept in private on password-protected devices and will only be used for the purposes of the study.

The study will utilize a difference-in-difference empirical design. That is, it will compare changes over time in communities that receive a bridge against those that do not. This is recovered from a linear regression:

y_vt = a + β*Bridge_vt + E_vt In this equation, y_vt is some outcome of interest in community v at time t and Bridge_vt is an indicator of community v has a bridge at time t. Under the identification procedure described above, the coefficient β estimated in the above regression identifies the causal effect of bridge construction.