A national count is treated as a fixed number, yet every census produces an estimate with known error. The difficulty is structural, not a failure of effort or budget.
The count is anchored to a single day
A census defines a reference date and asks where each person was living on it, which gives a common rule for a population that moves continuously.
Collection itself runs for weeks or months, so responses arrive from people who have already moved, and enumerators visit addresses whose occupants have changed since the reference date.
Reconciling those responses to one day is a reconstruction exercise, and it is the first place where judgement enters what looks like a simple headcount.
Some households are structurally hard to reach
Counting depends on being able to associate people with addresses. That association weakens for anyone in temporary accommodation, moving frequently, or sharing a dwelling informally.
Households with several families behind one door, young adults between residences and people without fixed housing are all systematically harder to enumerate than a stable single-family address.
Because the difficulty attaches to circumstance rather than chance, the same groups are undercounted repeatedly, which turns a random-looking error into a persistent one.
Overcounting happens as well
People with more than one residence can be recorded twice, and students, children in shared custody arrangements and seasonal residents are all common sources of duplication.
Statistical agencies run matching procedures to identify duplicate records, but matching relies on names and dates being written consistently, which they often are not.
Undercount and overcount therefore coexist in the same national figure, and they do not cancel neatly, because they fall on different groups and different places.
Imputation fills what enumeration cannot
Where a household cannot be contacted after repeated attempts, its occupancy is estimated from administrative records and from the characteristics of the surrounding area.
The methods are published and audited, but they are inference rather than observation, and their share of the total is larger in exactly the places already hardest to count.
Follow-up surveys conducted after collection measure how far the count strayed, which is how the size of the error becomes known at all.
The results are used long after collection
Census figures underpin the allocation of representation and funding, and they serve as the denominator for other statistics for years until the next count replaces them.
An area undercounted at the reference date carries that deficit forward across the whole intercensal period, which is why the stakes of small errors are disproportionate.
Agencies publish estimated coverage alongside the headline number for that reason, though the headline is what circulates and the coverage note rarely travels with it.