De facto vs de jure population: why it changes a structure-based estimate

A quiet suburban street with parked cars, houses, and trees under a clear sky.

Pull two population tables for the same district and they can disagree by a wide margin, even when both claim to cover the same year. Nine times out of ten the gap comes down to definitions: one table counted usual residents, the other counted whoever was physically there on a given night.

Usual residence vs who's actually there

De jure population counts people where they usually live, the usual residence definition most national censuses use as their official standard: a household is attributed to its registered or habitual address even if a member is working elsewhere that week. De facto population counts people where they actually were at a fixed moment, typically census night. A seasonal worker sleeping in a dormitory three districts from home shows up in the de jure count back at the home address, and in the de facto count at the dormitory.

Most statistics offices report de jure figures as the headline number, because that's what budget formulas, electoral boundaries and service-planning norms are built around. De facto counts still earn their keep for specific planning questions. Tourism boards, disaster response teams and anyone sizing peak-load infrastructure, water, power, emergency shelter, need to know who's physically present on the ground, whatever address those people are registered at.

Where a structure-based estimate lands

A structure count, a building-footprint layer and an average household size combine into a population figure by counting occupied structures and estimating who's sleeping under them. That method has no way to ask whether a household considers itself "usually resident" somewhere else; it just sees an occupied roof. In practice, it produces a de facto, night-population estimate: a count of where people are housed right now.

For a stable district where in- and out-migration roughly cancel out, the gap between the two definitions is small, and you can treat the structure-based figure as a reasonable stand-in for the de jure total. For a fast-growing district, the gap can be large and systematic in one direction. Districts that pull in construction crews, factory labor or seasonal agricultural workers tend to run a de facto population well above their de jure population: dormitories and worker housing fill up with people whose usual residence, by the census definition, is a home village two provinces away. The structure count sees every occupied room. The civil registry doesn't.

The reverse also shows up. A district with a large commuter outflow, people who sleep in a cheaper peripheral district but work and are registered in the center, can show a lower de facto night population than its de jure total implies, because its own census definition counts residents who are frequently absent.

What this means for the number you publish

None of this is a reason to avoid structure-based estimates between censuses. It's a reason to label the estimate correctly. If your last census used the usual residence definition and your intercensal structure-based figure is de facto by construction, the two aren't directly comparable without a note explaining the difference. Anyone downstream, a finance ministry, a line ministry planning clinic staffing, will default to assuming de jure unless you tell them otherwise. A one-line footnote in the methodology, "this estimate reflects de facto (night) population derived from occupied-structure counts, not usual residence," saves a lot of correction emails later.

If you're already weighing day population against night population for service planning, a structure-based estimate answers the night-population side of that question better than it answers the registered-residence side, since it's counting roofs and rooms, not registration status.

Population Estimation turns a built-up area, structure count and building-height layer into an intercensal population figure for districts where the last census is getting old or never happened. Worth a look if you need a documented number for the years between counts, not a replacement for the census itself.

Start a pilot

Back to the blog