Sub-national population projections for eleven Scottish council areas have been revised upward four times in two years, in some cases by more than 30% between successive releases.
The projections are the foundation of nearly every planning and funding decision made in Scotland: housing land allocation, school estate planning, health board resource distribution and local government finance all take them as an input.
They were designed to be revised every two years, from a census taken every ten.
Why they keep moving
Standard projection methodology extrapolates from observed trends in births, deaths and migration. It is reliable when those trends are stable, which for rural Scotland they were, in the same direction, for decades.
Migration driven by a step change — in this case a rapid shift in where people believe it is comfortable to live — does not extrapolate. There is no prior trend to project from, so the model follows the data upward each time the data moves, always one release behind.
“The model is doing exactly what it was built to do. It is being asked a question about the future by people who need an answer today, and it only knows about yesterday.”
Dr Granin, a demographer at a Scottish university
The mismatch
The consequence is a system in which the input updates every six months and the decisions built on it take between four and ten years to change.
A school estate plan adopted in 2024 used the 2022 projection. A housing land allocation being examined this year uses the 2023 one. Both are formally correct and both are, on current data, wrong by a margin large enough to matter.
Update cycles
- Projections: revised twice yearly since 2025
- Local development plans: 10-year cycle
- School estate plans: typically 5–10 years
- Local government finance settlement: annual, using data up to 3 years old
- Census: every 10 years
What statisticians have asked for
The request from the statistical community has been modest and repeated: publish scenario ranges rather than a central figure, and require planning documents to test against the range.
“Everything here is a matter of perspective,” one analyst said. “A single number invites people to treat a forecast as a measurement. Three numbers make them plan for uncertainty, which is the actual situation.”
