A census attempts to count everyone in a country on a single reference date. The count itself is straightforward for most households, and almost all of the effort goes into the remainder.
The reference date fixes a moving target
Populations change continuously through births, deaths, arrivals and departures, so any count needs a fixed moment to be coherent.
A census defines that moment and asks where each person was on it, which is why forms specify a particular night rather than asking where someone usually lives.
The rule creates its own difficulties for people who were travelling, working away, in hospital or between homes, and detailed guidance exists for each case.
Non-response is concentrated, not random
Most households return a form with modest prompting, and the returns arrive from broadly the population that is easiest to reach.
Those missed are not spread evenly. Young adults, people who move frequently, those in shared or informal housing and recent arrivals are all under-counted.
Because the gap is concentrated in identifiable groups, it distorts the picture of particular places far more than a uniform shortfall would.
Field follow-up is the expensive part
Statistical agencies send staff to addresses that have not responded, sometimes repeatedly, and this is the single largest cost in the operation.
Building an accurate address list beforehand is a project in itself, since new dwellings, subdivided properties and informal accommodation are all easy to miss.
An address absent from the list is never followed up, so errors there propagate silently into the final count.
Coverage is measured after the fact
Agencies run a separate survey after the census, re-interviewing a sample of areas independently and comparing the results.
The comparison estimates how many people were missed and who they were, and the published totals are adjusted accordingly.
This is why headline figures are released some time after collection, and why they can differ from the number of forms actually returned.
The results are used for a decade
Census figures underpin the allocation of funding, the drawing of electoral boundaries and the planning of schools, transport and health services.
Those uses give an under-count real consequences for the areas affected, since a place counted short receives resources sized for a smaller population.
Administrative data now supplements the exercise in many countries, and some have moved towards continuous estimates rather than a single decennial count, though the methods and timings vary by jurisdiction.