Three weeks into our household survey in Mogadishu, the district commissioner changed. Within forty-eight hours, our local enumerators received threats. Not because of the research topic — we were surveying market access — but because the previous DC had approved the study, and the new one had not.
This is the reality of political economy research in fragile states. The barriers to good data are rarely methodological. They are political.
The Invisible Architecture of Fieldwork
When I design a research project in Somalia, I spend as much time mapping power relationships as I do designing questionnaires. Who controls which neighborhood? Which elders need to be consulted before a female enumerator can speak to a male household head? What is the current status of the district's relationship with the federal ministry that approved our project?
These questions do not appear in any research methods textbook. But they determine whether your data is representative or fabricated, whether your enumerators are safe or at risk, and whether your findings will ever see the light of day.
Gender as a Methodological Variable
Perhaps the most underappreciated dimension of fieldwork is gender. In our 2023 study on financial inclusion, we ran parallel survey teams — one all-female, one mixed. The all-female team reported household incomes 34% lower than the mixed team.
The reason? Men were underreporting women's informal economic activities to male enumerators. The women were simply not mentioning their side businesses to men they did not know. This was not response bias in the statistical sense. It was a privacy boundary enforced by social norms.
We redesigned the survey instrument to include sex-matched enumerator pairs for household interviews. The result was more accurate data, but also a 20% increase in survey costs and a three-week delay to recruit and train female enumerators in a region where few women have prior survey experience.
The Cost of Rigorous Research
All of this adds time and money. A household survey that might cost $80 per interview in Kenya costs $140 in Somalia, not because of security — though that is a factor — but because of the layers of negotiation, relationship-building, and adaptive sampling required to produce valid data.
Donors often balk at these costs. They compare our budgets to standard World Bank survey instruments and ask why we need more. The answer is simple: standard instruments assume a state. When the state is fragmentary, contested, or absent, the research itself must do the work of building the institutional context in which valid data can exist.
Why It Matters
I write this not to complain, but to argue. Too much of what passes for "evidence" in fragile states is produced under conditions that would be considered unacceptable in stable contexts. Convenience samples are treated as representative. Male-only enumerator teams are assumed gender-neutral. Districts that are too dangerous to enter are simply dropped from the sample, with no adjustment for the bias this introduces.
The result is policy advice built on foundations of sand. If we are serious about evidence-based policy in fragile states, we need to get serious about funding the real cost of producing evidence. That means longer timelines, higher budgets, and research designs that treat politics not as a confounding variable, but as the primary object of study.
Dr. Mariam Ali is a PhD student in Political Economy and a research fellow at the Institute for Development Economics. She has led field studies in Somalia, Kenya, and Uganda since 2019.