AidData at William & Mary is teaming up with the Swedish government’s Expert Group for Aid Studies (EBA) on a geospatial impact evaluation (GIE) to study the long-term impacts of a flagship rural roads project in Liberia.
A decades-long civil conflict in the West African country destroyed and isolated rural communities. Towns and villages were cut off, with crumbling roads impeding movement of goods, access to health and education services, and civic participation. That’s not all: “In interviews with engineers, we learned that because of the intense rainy seasons, roads in Liberia are some of the most dangerous in the world,” said AidData Chief Economist Ariel BenYishay, who is leading the GIE. “These downpours wash out roadways every few years, further isolating rural communities.”
Enter the Liberia Swedish Feeder Roads Project (LSFRP), an ambitious three-phase project of the Liberian government and the Swedish International Development Cooperation Agency (SIDA). Its aims were to improve rural connectivity and mobility and catalyze economic development and democratic participation. Implemented between 2009 and 2025, the LSFRP restored and maintained approximately 1,000 kilometers of roads—especially secondary “feeder” roads in rural counties—at an estimated cost of $60 million. The LSFRP also sought to build local capacity for ongoing road maintenance through skills training, institutional strengthening, and the adoption of new construction standards.
A previous EBA evaluation as well as Swedish Embassy reports have shown several positive outcomes thanks to the LSFRP, including new building standards, contractor capacity building, improved road quality and increased access to markets and services. But this new GIE will help with assessing longer-term impacts, such as whether the new or rehabilitated roads have affected incomes, jobs, and everyday finances; whether there were any unexpected effects, both positive and negative; and under what conditions road projects actually do the most good.
“Despite the scale and long duration of the LSFRP, no quantitative study has previously assessed its impacts using counterfactual evidence,” said Torbjörn Becker, EBA reference group chair. “By combining geospatial data with quasi-experimental methods, this GIE can identify the causal effects of the intervention. Such evidence is crucial for learning and improvement. It helps policymakers refine interventions, target investments more effectively, and ensure that development efforts generate meaningful and inclusive benefits, particularly for vulnerable populations.”
Counterfactual evidence compares what might have occurred in the absence of a project or program, versus the known and measurable impacts of the intervention. AidData has conducted a number of GIEs looking at road projects throughout the developing world, but in general ex post evaluations (which are conducted after a project is over) of transportation sector investments are rare.
The LSFRP targeted seven counties, beginning with Lofa and Bong in the northwest, expanding to Nimba, and eventually reaching the southeastern counties of Grand Gedeh, Grand Kru, Maryland, River Gee, and parts of Nimba once again. “Our advanced causal inference techniques allow us to dive deep and measure the causal effects of the project even after completion, exploiting the staggered roll-out of the LSFRP across communities in Liberia,” said AidData Research Scientist Rachel Sayers. “It allows us to extract credible and robust estimates of this intervention’s impact across space, time and several important social dimensions.”
AidData Data Analyst Pratap Khattri assembled the datasets that are the foundation of the GIE and is now conducting the analysis. Information sources gathered include nighttime lights data from satellites, high-resolution daytime satellite imagery (including European Space Agency images at the 10-meter scale), geocoded infrastructure data, and multiple rounds of the Demographic and Health Surveys (DHS) and Afrobarometer surveys. These are combined with big data processing capabilities using computing that is both cloud-based (i.e., Google Earth Engine) and on-site (William & Mary’s High Performance Computing clusters).
In rural Liberia, roads help farmers grow and get crops to market, where they can also buy fertilizer, pesticides and other commodities needed for the growing process. To understand these road networks and how they are used, the AidData team is examining transport connections to a subset of 26 major markets in Liberia. From this, they are constructing a Market Activity Index (MAI). The researchers are also looking at vegetation cover and indices over time, as a measure of crop production and farm health—which has knock-on impacts on nutrition outcomes for everyone living in rural communities.
“By integrating a wide array of satellite-based remote sensing data together with geolocated administrative and survey data, we can assess a range of highly relevant outcomes—such as regional economic, agricultural, and market activity, households’ material well-being, nutrition and food security, and democratic participation,” said Khattri.
A core dataset was a set of georeferenced shapefiles of road segments rehabilitated under the LSFRP, which were provided by the Embassy of Sweden in Monrovia and Liberia’s Ministry of Public Works. These shapefiles contain detailed metadata on the timing of road construction, phase-specific designations, and precise geographic alignments of completed segments. “This allows us to spatially map the rollout of the roads program and construct treatment indicators based on proximity to completed road segments over time. Specifically, this will permit us to construct buffers around each treated road segment to define zones of potential exposure,” said Khattri.
The research is not without its challenges. “We've been struggling to develop a measure of road quality,” said Khattri. “We know the location of the road improvements, but because roads wash out, it is hard to know their degree of accessibility and to have a consistent way to rate the roadways on quality.” The team is in discussion with satellite imagery providers who have already developed road detection and quality measures in other settings, to see whether these could be applied to Liberia.
“Evidence-based development cooperation is essential to ensure that limited resources generate the greatest possible impact,” said Becker. “Rigorous methods and robust analysis help us understand what works, for whom, and under what conditions, particularly for large-scale infrastructure investments such as rural roads, where effects may vary across locations and populations.”
BenYishay agreed: “In a tighter funding environment, impact evaluations become all the more important in informing policy.”
The EBA evaluation wraps up this summer. “This is part of the attractiveness of GIEs: we can scale them up as appropriate, and they are faster and cheaper than conventional impact evaluations,” added BenYishay. The GIE is expected to not only provide deep insights into the impacts of the LSFRP, but also generate important guidance for the design, implementation, and evaluation of similar development projects in the future. “We hope that the results, as well as the evaluation tools and technologies we employ to generate these results, will be useful in other countries too, especially where roads and infrastructure are critical for improving the lives and livelihoods of rural communities,” said BenYishay.
The LSFRP was part of Sweden’s long-standing investments in Liberia, through a partnership that has emphasized state-building, inclusive economic growth, and improved service delivery in a post-conflict setting. EBA—or Expertgruppen för biståndsanalys—is a government committee consisting of an Expert Group of ten members and a secretariat, mandated to evaluate and analyze the direction, governance, and implementation of Sweden’s official development assistance, with a specific focus on results and effectiveness. Stockholm-based EBA’s aim is to contribute to the efficient implementation of well-designed aid, focusing primarily on overarching topics and strategies of Swedish development assistance rather than individual projects.
