Secularism and the Census: How States Count Religion

Whether a country counts its religious population says something about how it understands the relationship between the state and belief. The decision to ask, and equally the decision not to ask, is itself a policy choice with real consequences for secular governance.

Why states ask about religion at all

Governments collect demographic data to allocate resources, monitor discrimination, and track social change. Religion fits the same logic. A state that funds faith schools, grants religious bodies charitable status, or exempts certain groups from general laws needs to know the rough composition of its population. Without that data, resource allocation becomes guesswork, and legal protections for minorities are harder to calibrate.

The practical argument is stronger than it first appears. Countries with established churches — the United Kingdom, Denmark, Norway — have historically used census data to test whether institutional arrangements still reflect the population they govern. When the 2021 England and Wales census recorded that fewer than half of respondents identified as Christian for the first time, the result fed directly into public debate about whether the Church of England's twenty-six bishops in the House of Lords remain democratically defensible. Data, in that sense, is not merely descriptive; it is politically generative.

The design problem

How a census question is worded determines what it measures, and religious questions are unusually sensitive to wording. The two dominant approaches are identity questions ("What is your religion?") and practice questions ("How often do you attend religious services?"). These do not measure the same thing, and conflating them produces badly misleading conclusions.

In Ireland's 2022 census, respondents were asked their religion; 69 percent returned a Catholic answer, down from 78 percent in 2016 but still a large majority. Yet Mass attendance data collected separately suggests that fewer than a third of Irish people attend weekly. The identity figure captures cultural and ethnic self-understanding as much as theological commitment. Policymakers who treat the identity figure as a measure of active religiosity will systematically overestimate the social reach of the institutional church.

A further problem is the non-religion framing. When "No religion" appears as one option among many named faiths, it tends to attract fewer responses than when the question first asks whether the respondent has a religion at all, with a yes/no gate before the denomination list. The British Social Attitudes Survey, which uses a different methodology from the census, consistently returns higher "no religion" figures than the census for the same population. Neither figure is wrong; they are answering different questions. But only one of them is widely reported as if it settled the matter.

When states refuse to ask

Some secular democracies have concluded that the state should not collect religious data at all. France treats religion as a private matter incompatible with the republican principle of laïcité: official statistics do not record faith, and the prohibition extends to ethnicity in most public contexts. The argument is that a truly neutral state has no business categorising citizens by their inner commitments.

The counterargument is that refusing to count produces its own distortions. Without data on the religious composition of the prison population, the judiciary, or the civil service, it becomes very difficult to identify structural discrimination — which is real, documented in other ways, and affects both minority religious groups and the nonreligious. French authorities have responded by commissioning private surveys under legal exemptions, which produces a situation where the data exist but carry no official status and feed only informally into policy. Principled neutrality here generates practical opacity.

The German model sits between the French and British approaches. Germany does not ask about religion in its general census, but it collects church tax affiliation data through the tax system, because members of the major denominations opt in to a payroll levy that funds the churches. This produces very accurate figures for formal Lutheran and Catholic membership, but tells you nothing about Muslims, Buddhists, or the nonreligious except by subtraction. The data reflects the state's financial entanglement with specific institutions rather than the actual religious landscape.

What secular governance actually needs from this data

A secular state, properly understood, does not privilege any religion — but it still has to govern a population that includes people whose religious identity shapes their legal needs, their interactions with public institutions, and their vulnerability to discrimination. The data question is therefore not a concession to religion but a tool of fair governance.

What that requires is precision about what is being measured. Affiliation, practice, belief, and salience (how much religion matters to a person's decisions) are distinct variables that move in different directions across populations. A census that collapses them into a single tick-box generates a single number that satisfies no one's analytical needs.

The most honest approach is probably also the most uncomfortable one for survey designers: acknowledge that no single question captures religious life, publish multiple measures with explicit methodological caveats, and resist the temptation to produce a headline figure that appears to settle what proportion of the country is or is not religious. Governing well in a religiously diverse society means tolerating that complexity rather than designing it away.