Referenced by 7 other explainers
Perturbation is a method used by the Australian Bureau of Statistics (ABS) to protect privacy in published Census data. Before data is released, the ABS changes small counts by a tiny random amount. This helps make sure no single person can be picked out from Census tables.
Because of perturbation, Census figures should be read as close estimates, not exact counts. Two cells that look like they should add up to a third may not, and the same group of people can show slightly different figures in two tables.
The ABS must protect the privacy of people and homes in Census data. Small counts can sometimes make a person easy to pick out, above all in small areas or detailed groups. Perturbation lowers this risk while still letting the ABS share detailed data.
Small random changes are made to Census cells and totals. For example, a real value of 4 could show as 3, 4, or 5 in a published table.
Because each value is changed on its own:
These gaps are normal and expected. A small count shown as 0 may not be a true zero, and small cells should not be read as exact counts. Patterns across many cells stay useful. Single small cells do not.
All ABS data used in Place Forecast has already been perturbed before it is received. Place Forecast does not remove perturbation and does not add its own changes. Because of this, small gaps between Place Forecast figures and ABS figures are expected and are not errors.
See the marker methodology explainer for more detail on how these figures are created and their limits.