The tension at the heart of it
Good sampling is about representativeness — every unit in the population having a known, non-zero chance of selection. Insecurity works directly against this, because the areas that are hardest to reach are often systematically different from those that are easy to reach. Ignore that, and your sample quietly over-represents the accessible and the safe.
The goal in these settings is not perfection — it is transparency about the trade-offs, and design choices that minimise and document bias rather than pretend it away.
Practical strategies
Several approaches help. Stratification by accessibility makes the constraint explicit and lets you report separately on reachable and hard-to-reach strata. Replacement protocols, agreed in advance, prevent enumerators from making ad-hoc substitutions that introduce hidden bias. And remote data collection — phone surveys, key-informant networks — can extend reach into areas fieldwork cannot safely enter, provided their limitations are acknowledged.
In insecure environments, the honest sample is the one whose gaps are documented, not the one that pretends it has none.
Reporting with integrity
Whatever the design, the credibility of the results rests on honest reporting of coverage. That means stating clearly which areas were excluded and why, how substitutions were handled, and what those choices mean for generalisability. Donors and decision-makers can work with acknowledged limitations; they cannot work with hidden ones.
PRIME's field teams are trained to maintain rigour and safety together — and to document every access constraint so the resulting evidence can be trusted for what it is.