Geospatial data encompass all information that is ‘geotagged’ to an exact geographical location on earth. This information can be remotely sensed from space—i.e. satellite imagery—but can also be collected from databases, surveys, project documents, and Monitoring & Evaluation (M&E) systems. The use of geospatial data on project variables has become an attractive solution to fill the void of field missions during the Covid-19 pandemic. Evaluators, however, were using geospatial data in evaluation even before travel was restricted. There is now an incredible opportunity for evaluators to use geospatial data more effectively and efficiently.

The last decade has seen rapid advances in all aspects of geospatial data, especially remote sensing data. First, satellite imagery has become more readily available, at lower (or zero) cost, and with higher quality. Terabytes of free and high-resolution raw data are created every single day. But more importantly, this raw satellite imagery is now rapidly processed into meaningful geospatial data by using machine learning. For example, raw images from the MODIS satellite are daily processed into geospatial data on land cover and forest fires that evaluators can use directly. Second, along with the revolution in big data, many data collection efforts—ranging from open-sourced platforms to household surveys—record the location of their observations. Similarly, more projects report on the geographical targeting of project investments. As a result, all of this geotagged information can be combined into one geospatial dataset. Finally, the analysis of geospatial data has become more efficient and user-friendly through open-source statistical programs. 

Innovative geospatial data and software provide evaluators at the Independent Evaluation Group (and other evaluation functions) with unique tools to better address evaluation questions around the relevance and effectiveness of World Bank Group interventions. To assess the relevance of development interventions, evaluators can compare the spatial variation in a project variable with the spatial targeting of development interventions. For example, in a recent Country Program Evaluation for Mexico, IEG assessed whether investments to reduce poverty were directed towards areas with the highest poverty levels. Using regression analysis, and controlling for relevant exogenous variation, the analysis showed that World Bank support at the state level is positively correlated with the presence of the poorest 40% and is fairly independent of national public spending. To assess the effectiveness of development interventions, evaluators can use geospatial time-series to proxy changes in the project outcome indicators and construct a spatial counterfactual. A previous blog elaborated on how IEG exploits the spatial and temporal aspects of geospatial data in a robust impact assessment of World Bank projects in Mozambique, India, Ethiopia, and Madagascar.

Geospatial data also help overcome some of the methodological challenges to rigorously assess the sustainability of project impacts. Remote sensing data (e.g. satellite imagery) can provide unbiased and objective information on project outcomes on a granular level in every part of the globe. The availability of such data over time enables us to understand the evolution of particular variables over the entire life span of an intervention (and many years after the intervention). For example, IEG’s ongoing evaluation of Bank Group support to Municipal Solid Waste Management is using geospatial data to assess the sustainability of such support regarding the intended and unintended environmental and land use impact around supported landfill sites long after the respective projects have ended. Until recently, such analyses typically used to be beyond the scope (and feasibility) of a conventional (project) evaluation.

The use of geospatial data is, however, no silver bullet. Whether evaluations can apply geospatial data depends on the nature of the evaluand (e.g. the sector and type of intervention to be evaluated) as well as the analytical skills of the evaluators. Moreover, the objective and rigorous assessment of effectiveness using geospatial analysis is not sufficient on its own to assess why interventions are effective (or not). For example, it remains difficult to proxy political economy and human behavior aspects from available geospatial data. Ideally, any geospatial analysis requires some type of verification and triangulation ‘on the ground’. One of the most challenging constraints regarding the use of geospatial data in many multilateral and bilateral international development agencies (as well as some other organizations) is the disconnect between operations (which focuses on design and implementation) and evaluation. Independent evaluation functions are not directly involved in the intervention cycle (especially project design and implementation). Evaluators, therefore, have relatively little influence on the M&E frameworks of the interventions financed by their organization. A well-known consequence is that public and private investments often lack granular information on project implementation which complicates the use of geospatial data in evaluation afterwards.

Going forward, how can evaluation functions like IEG enhance their use of geospatial data? The first step is to focus on some of the low-hanging fruits. In the examples mentioned above, IEG has applied geospatial analysis to a particular set of interventions with a clear temporal and spatial nature. The analyses have been facilitated by the availability of numerous data portals with open access and ready-to-use geospatial data on a wide range of economic, environmental, and agroecological indicators. In some cases, evaluations have benefited from collaborative efforts with research colleagues. These examples have not only generated interesting and useful findings, they also provide useful lessons on the potential feasibility and desirability for conducting geospatial analysis in the framework of an evaluation. To better understand which geospatial data are useful and for which purposes, piloting new methods in the framework of different evaluation modalities should be encouraged. Investments in staff capacity development, hiring external experts, computing capacity and specialized software should be weighed against the results of these pilots.

The next step constitutes an organizational dialogue on the integration of geospatial data (collection and analysis) in the design and implementation of interventions. Evaluators can help make a stronger case for informed investments in geospatial data collection and analysis, leveraging the support from like-minded champions in research and operations departments. For example, the World Bank has launched two mobile applications (the Geo-Enabling Initiative for Monitoring and Supervision, GEMS, and the Smart Supervision App, SSA) that precisely register project locations and collect information that feeds into remote and real-time M&E systems. The GEMS initiative and the Geospatial Operations Support Team (GOST) also provide trainings and advice to build the capacity of clients and World Bank staff for remote project monitoring and supervision. This fits in the World Bank’s broader strategy to support client countries in developing the infrastructure, legal framework, and human capacity needed for the management and utilization of geospatial data. After the necessary ‘proof of concept’ experiences, a concerted organizational effort is needed to unleash the potential of geospatial data for better intervention design, implementation, and M&E.