Choosing a dwelling occupancy rate in a fast-growing district

The housing unit method is simple on paper: count dwellings, multiply by persons per occupied dwelling, add institutional population, done. The arithmetic isn't where districts go wrong. The occupancy rate is.

Most statistics offices pull that rate straight from the last census tabulation for the district or its nearest comparable ward. In a stable area that's defensible. In a district that's added two or three wards' worth of new construction since the census, the rate you're carrying forward describes a place that no longer exists.

Where the occupancy rate usually comes from, and why it stops fitting

The standard sources are the last census, a recent household survey if one covers the area, or an administrative register of dwellings and residents. Each of these has a lag built into it, and growth makes the lag matter.

Census-derived occupancy rates reflect the household composition at the time of enumeration. If the district's growth is driven by young in-migrant workers filling new apartment blocks near an industrial zone, persons per occupied dwelling will often run higher than the district-wide census figure, because new arrivals double up before they split into separate households. If growth is driven by formal housing development aimed at middle-income buyers, occupancy per unit tends to run lower than the older housing stock, smaller households, more owner-occupiers living alone or in pairs.

Pick the comparator that matches the growth type you're actually seeing on the ground, then check it against a subsample. A field team visiting forty or fifty addresses in the newest construction tells you more than another round of the district average.

Adjusting vacancy for structure counts

This is where the structure count itself needs scrutiny before you apply any occupancy rate to it. A satellite-derived structure count doesn't know which buildings are occupied, vacant, under construction, or non-residential. That distinction has to come from somewhere else.

Three categories inflate a raw structure count if left unadjusted:

Buildings under construction. A fast-growing district will have a visible share of roofed structures with no occupants yet. Counting these as occupied dwellings overstates population immediately.

Non-residential structures. Warehouses, workshops, and school buildings read the same as houses in a structure count unless the building-height and footprint data are cross-checked against land use, or unless an analyst flags them by hand.

Speculative or seasonal vacancy. In districts attracting investment housing, a meaningful share of completed units may sit empty for a year or more. This differs from informal settlement growth, where vacancy is usually close to zero because every new structure gets occupied as soon as the roof goes up.

A reasonable starting adjustment is to apply a vacancy rate pulled from the closest available source, often a recent housing survey or utility connection data, rather than assuming the census-year vacancy rate still holds. If utility connection records are available for the district, the ratio of connected to counted structures is a useful cross-check, though it will undercount informal connections.

Setting a defensible range instead of a single number

Given the lag in every input, the more honest output is a range rather than a point estimate. Run the calculation twice, once with the pre-growth occupancy rate and once with a rate adjusted for the dominant construction type in the new area, and report both. A statistics office publishing an intercensal figure for a fast-growing district should be able to state which assumption drives the number and how sensitive the total is to it.

Where the input layers come from matters as much as the arithmetic. A built-up area, a structure count by type, and a building-height layer drawn from recent high-resolution imagery give you the denominator for this calculation without waiting for a field census you don't have the budget to run, which is the starting point Population Estimation is built around.

If your district's last census is old enough that the occupancy rate no longer fits what's being built, that's worth a closer look.