Population volatility
How much a tract's year-to-year population change has varied over a five-year span.
How steady a tract's population has been. Two tracts can arrive at the same five-year growth figure by very different routes: one adding people at a similar pace every year, the other swinging between sharp gains and sharp losses. This family measures the second thing — the dispersion of a tract's year-to-year population change over a five-year span, expressed as a standard deviation of those annual percentage changes.
It is a stability measure rather than a growth measure. It carries no sign: a tract that lost people steadily scores as stable, and a tract that gained them erratically scores as volatile. What the level of population is doing is covered separately by population.
The underlying figure is the American Community Survey's estimate of the resident population, including people in group quarters such as dormitories, nursing homes and correctional facilities (ACS Subject Definitions).
Metrics in this family#
Five-year population volatility
stabilityThe variability of annual population change across a five-year span, from the most recent ACS 5-year vintage backwards. A tract is only given a value where the full span of annual changes is available; a shorter run of years produces no value rather than a partial one wearing the five-year label.
Considerations#
This measure is more sensitive to the properties of the ACS than most. Three of them matter.
First, ACS 5-year estimates describe a five-year period, and consecutive vintages share four of those five years. The Bureau's guidance is "do not compare overlapping datasets" (Comparing ACS Data). That matters more here than for most measures, because smoothing is precisely what this metric is trying to see through: consecutive vintages are rolling estimates rather than independent annual counts, so genuine year-to-year turbulence is damped and the figure understates real instability. It is usable as a relative signal — a tract measuring as more volatile than another almost certainly is — but not as a measurement of how much a tract actually moved.
Second, part of what moves between vintages is sampling rather than people. Every ACS estimate carries a margin of error, and in small tracts that error is a meaningful share of the population; noise of that kind adds to measured volatility.
Third, the Bureau controls the group quarters sample at the state level only, and warns that "users may observe greater fluctuations in year-to-year ACS estimates of the GQ population at sub-state levels than at state levels," typically where a facility has closed or its population differs from what was expected. A tract containing a prison, a large dormitory or a nursing home can register as volatile for that reason alone.
Census tract boundaries were redrawn for the 2020 census. A boundary change is not a population change, but it looks like one in a raw series. Estimates published on earlier boundaries are carried onto current ones before use — see geography — which removes most of that artefact, though tracts whose boundaries moved substantially will retain some residual noise.
Coverage runs through the most recent ACS 5-year release. Publication lags the present, and 5-year estimates are the least current of the ACS products by design.
References#
- 1ACS Subject Definitionshttps://www2.census.gov/programs-surveys/acs/tech_docs/subject_definitions/2024_ACSSubjectDefinitions.pdf
- 2Comparing ACS Datahttps://www.census.gov/programs-surveys/acs/guidance/comparing-acs-data.html