A census is presented as a count, yet the results are argued over long after publication. The disputes arise because nobody has ever succeeded in counting an entire population directly.
Some people are always missed
A census reaches households through addresses, and people without a stable address, those moving, and those in irregular housing are the hardest to reach.
Undercounting is therefore not random: it concentrates in particular groups, places and age ranges, and it repeats in the same populations at each count.
Because the shortfall is patterned rather than scattered, it changes the relative size of communities rather than reducing every figure equally.
Coverage surveys estimate the gap
Statistical offices follow the count with an independent survey of sampled areas, conducted separately and matched against the census returns.
Comparing the two shows how many people the census missed in each area and demographic group, and the results are used to adjust the published figures.
Those adjustments are where much of the dispute sits, because they are estimates rather than observations, and the method chosen affects the outcome.
The stakes make disputes inevitable
Census figures determine the allocation of political representation and of funding distributed on a per-head basis.
An area whose population is understated therefore receives less of both, and the effect persists until the next count is taken.
Any group that believes itself undercounted has a concrete and continuing interest in challenging the result, which is why litigation is common.
Question design changes the answers
Categories for ethnicity, household relationships and residence are decided before the count and constrain how people can describe themselves.
Changing a question improves accuracy in the present but breaks comparison with previous censuses, so trends become harder to establish across time.
Statistical offices consult extensively on these choices precisely because they determine which populations become visible in the data at all.
Administrative data is changing the approach
Several countries now build population statistics primarily from existing registers of residence, tax and health records rather than from a periodic count.
That approach produces continuous figures and avoids the cost of a full enumeration, but it inherits whatever errors and omissions the underlying registers contain.
Whether a country can adopt it depends on what registers exist and how they may lawfully be linked, which differs by jurisdiction and changes over time.