By: Yuliya Dudaronak, Research Director, Andy Fallon, Senior Researcher, and Alexis Eck, Researcher 1

Photo 1: ORB interview in Nigeria

Since 2019 ORB International has been using Gridded Population Sampling to conduct rigorous research in conflict and fragile environments. These locations often lack proper or up-to-date census data and have experienced large-scale population displacement. Because of this, Census-based sampling can be an ineffective method to achieve a truly representative sample. Gridded Population Sampling, based on internationally recognized population estimate datasets, can resolve this issue by providing a more representative sample frame.

ORB has completed 43 projects in 13 countries using Gridded Population Sampling, collecting about 86,000 interviews. Throughout the implementation of these research studies, we have focused on making the process increasingly more efficient and reliable. One of our best practices is what we call ‘viability review’ – a process by which ORB staff determines whether a particular location is feasible for fieldwork. Here is how we do it and why it matters!

ORB’s Viability Review Process

We start with creating Gridded Master Sampling Frames.  They use the UN-constrained WorldPop dataset with other layers marking population density (i.e. roads, lights) to create the shapes of Primary Sampling Units (PSU). PSUs are the exact locations where ORB will conduct surveys. These PSUs are based on 1) a target population of 1,200; 2) that don’t exceed a certain size (5 sq km); and 3) that don’t cross over administrative districts or urbanity designations. This means that PSUs in dense urban centers are smaller and PSUs in very rural areas are larger, with upper limits of 5 sq KM. This process creates what we call a Gridded Master Sampling Frame from which we draw the required sample size.

Figure 1: Gridded Master Sampling Frame Example

However, the creation of the Gridded Sample Frame is only the first step. While the algorithm seeks to get as close to 1,200 estimated population as possible within the constraint of 5 KM, you will have many PSUs in rural areas that are much smaller than 1,200 people, sometimes only a fraction of a person. Furthermore, these are modeled estimates that can result in overestimates of the population numbers. Deploying field teams to a PSU that turns out to be a non-residential area can waste valuable resources and delay data collection; therefore, trained ORB staff review recent satellite imagery of each selected PSU to determine if the desired number of respondents could live in that area. If any PSU appears to be non-residential (i.e. desert area, factory, airport, etc) or does not have enough visible residential properties, ORB will mark them as non-viable and exclude them from the sample draw.

Figure 2: GooglePro PSU Review Example

During the review we ensure that (1) the PSU is within the country and at least 5km away from the border, (2) there are enough residential properties, (3) the centroid (coordinates of the mathematical center of the PSU that enumerators use to navigate to the PSU and start their random walk) is located within the PSU boundary, and (4) the centroid is in the reasonable place within the PSU to start a random walk.

In the last year, we used this process in ten countries, reviewing close to 16,000 PSUs and excluding 2,188 total – an average of 13 percent. Most PSUs were excluded because not enough people live there. Our internal rule is we need at least 30 compounds in a PSU that we can verify.  Just because a modeled estimate says there are a certain amount of people there, doesn’t mean it’s precise in all cases.

Figure 3: PSU Review Results in 10 Countries

Conclusion

ORB’s viability review process is very labor intensive but vital to the successful fieldwork using GriddedPop methodology because:

  • It ultimately saves sample preparation time by reducing the time our local team needs to review all the locations and plan fieldwork logistics;
  • Guarantees smoother fieldwork with less in-field replacements
  • Improves future work in that country as any PSU we review won’t have to be reviewed again for the same sample frame
  • And, most importantly, it improves local partner relations since it shows we value their time, energy, and safety; enumerators won’t waste time and money travelling just to find out they can’t work there. This is particularly crucial given that we work in environments with poor infrastructure and ongoing conflict, where travel is often the most expensive and dangerous part of the fieldwork.

Viability review is a crucial part of using gridded population sampling in fragile and desert environments – a starting point in the process that allows us to implement rigorous methodologies in some of the most complex environments in the world.

Menu