The 2021 census in England and Wales suggests that self-identified Christians are now less than half the population. Anyone following the British Social Attitudes (BSA) survey will regard this development as old news: it has put the Christian share below 50% in every year since 2009, with estimates as low as 38% in 2018 and 2019. The esteem accorded to official statistics, however, helps to explain headlines such as “Christians now a minority in England and Wales for first time” (Daily Telegraph, 29 November 2022).

 
It is worth remembering the features of each data source. To start with the census:

• It seeks to reach everyone in the population.
• It is conducted by the Office for National Statistics in England and Wales and by comparable public agencies in Scotland and Northern Ireland.
• Completion of the census form is a legal requirement, though the question on religion is voluntary.
• The question “What is your religion?” could be viewed as implying that everyone has one, though the first option listed is “No religion.”
• There are tick-boxes for world religions (Christian, Muslim, etc.) rather than specific denominations; there is also the option to choose ‘Other religion’ and write in a response.
• The question on religion directly followed the one on ethnicity in 2001 and 2021, which might imply that it is about family heritage rather than formal affiliation. In 2011 these topics were separated by questions on language.
• The form is often completed by one person in the household on behalf of some or all of the individuals in it.

As for the British Social Attitudes survey:

• It goes to a random sample of individuals aged 18 and above.
• It is carried out by NatCen, a leading survey company.
• The question is “Do you regard yourself as belonging to any particular religion?”
• The response options include denominational labels, for example Church of England, Roman Catholic, Methodist, Baptist, and Presbyterian, which might induce people who are not involved with a specific church to select ‘No religion’.

Regarding this last point, BSA respondents can identify themselves simply as ‘Christian – no denomination’, a category that probably includes committed members of evangelical and Pentecostal independent churches on the one hand and nominal Christians on the other. It seems likely that many non-churchgoers who in the past might have accepted ‘Church of England’ as a default label are now opting for the unspecified ‘Christian’ designation instead: the category has grown considerably in recent years, from 3% in 1983 to 13% in 2018. Unfortunately, NatCen now supplies only a summary variable for religion rather than the full breakdown shown in the questionnaire. Among Christians, the only groups identified are Church of England, Catholic and other. It is noteworthy, though, that in 2020 the ‘other Christians’ outnumbered Anglicans and Catholics combined. Once Presbyterians, Methodists and Baptists are subtracted, the ‘Christian – no denomination’ share is probably around 17%.

One plausible explanation for the differences in religious affiliation reported by the census and BSA survey is that the census question “What is your religion?” encourages people to choose one, in conjunction with the breadth and vagueness of the undifferentiated Christian category, which can easily be interpreted – particularly in this context – as a quasi-ethnic identity. By contrast, asking about belonging to any particular religion, where the usual answers include specific denominations, may nudge respondents without concrete connections to choose ‘no religion’ instead of a nominal Christian identity.

The fact that the Christian category on the census seems to be more inclusive than the same label used in surveys is only part of the mystery, however. The BSA survey has been running since 1983. Religious affiliation as measured by its question has declined steadily since that time, but the decline is wholly explained by cohort replacement. Elderly self-identified Christians die and are replaced in the population by young people who have no religion. Within each generation, average levels of affiliation are virtually unchanged during adulthood. Around 80% of people born around 1920 had a religion, and that percentage did not change significantly from one year to the next; by contrast, fewer than 30% of people born around 1990 regard themselves as belonging to a religion, and again that level has remained stable. Each successive birth cohort is less religious than the previous one, but the average within each cohort stays fairly constant over time.

Interestingly, the census measure of religious identity has been different. While generational differences are very apparent in the census figures from 2001 and 2011, the declines in Christian identification from one census to the next (2001 to 2011 and then 2011 to 2021) are much too large to be explained by cohort replacement alone. Millions of people who ticked the Christian box in one census chose ‘No religion’ ten years later.

The Christian share of the population in England and Wales is falling fast, as measured by the census: from 72% in 2001 to 59% in 2011 and then 46% in 2021. Using the detailed results on religion from 2011 together with the age distribution in 2021 (published earlier in the autumn), it is possible to estimate what the Christian share of the population would have been if religious identity was a stable characteristic. If change was purely the result of cohort replacement, 54% of the population would still be Christian.

Many people whose religious identity was nominal, weak and volatile were evidently classified as Christian in previous censuses. Quite a few still choose that affiliation – the Christian percentages remain higher than those from the BSA – but a substantial number have decided that ‘no religion’ is a more appropriate label. The term ‘disaffiliation’ might be applied to this shift, but it is hard not to suspect that most of the people concerned were scarcely affiliated in any real sense to begin with. When religion plays little part in one’s life, affiliation may amount to no more than a somewhat arbitrary decision about which box to tick.

Why has there been such stability in the BSA measure of affiliation and such instability in the census measure? The answer is presumably that the BSA question captures people whose religious identity is at least moderately strong and persistent, while the census picked up millions of nominal Christians for whom religion was not really part of their personal or social identity. Many have fallen off the fence into self-declared non-religion. There has been some convergence between the two data sources.

The generational contrasts in the census are now just as large as in the BSA. While we will have to wait for the release of more complete data to be sure, it is likely that around 80% of people aged 85+ will have called themselves Christian in 2021 but perhaps only 30% among those in their early 20s. Indeed, the Christian share will probably be less than 40% in all age groups below 45.

If what we know about religious change is correct, the slide in Christian affiliation will continue for decades into the future. In these circumstances, the role of the Church of England is naturally being questioned. Again, the BSA survey is arguably more helpful than the census, as it encourages respondents to identify with a particular Christian denomination. The age gradient for affiliation is especially steep: among the hundreds of respondents in the 18-24 age group, only three in 2018 and two in 2019 identified themselves as Anglicans. In order to obtain reliable estimates, we can pool the datasets from 2018, 2019 and 2020. One finds that just 4% of people in England under the age of 45 regard themselves as belonging to the Church of England.

As a final remark, our attention has been focused here on self-identification with a religion, and no single measure of religious involvement can tell the whole story of secularization. (Note, by contrast, Clive Field’s use of 21 key performance indicators in his recent book Counting Religion in Britain 1970–2020.) The problem is especially acute when affiliation is treated as binary: present or absent. In the United States, the General Social Survey (GSS) question on religious preference has since 1974 been followed by one that asks “Would you call yourself a strong X or a not very strong X?”, where X is the group chosen. About one in ten respondents volunteer the description “somewhat strong” and are so recorded. The GSS thereby discerns four levels of affiliation: strong, somewhat strong, not very strong, none.

This approach has important advantages. Renouncing any religious identity is a high bar, and relatively few people born before the end of the Second World War reached this threshold of secularity. There was little change before the Baby Boom generation in the proportion of Americans saying they have no religion. By contrast, strong / somewhat strong affiliation weakened for every successive generation from as far back as we can see, to people born more than a century ago. It seems likely that the same would be true of Britain.

If religious decline started earlier than some scholars suggest, it also seems to be continuing past the point when many expected to see a levelling out. (I was co-author of a book chapter published in 2010 entitled ‘The triumph of indifference’, but by the end of that decade I was commenting in British Social Attitudes: the 36th Report on the surprisingly assertive secularity of the unaffiliated.) The debate will continue, not least when more data from the census become available, but theories that link secularization to particular periods or generations seem hard to sustain.