2011 Census – Searching for Explanations

The general direction of travel revealed by the 2011 census results for religion in England and Wales, published on 11 December 2012, in relation to those for 2001 (when the question was first asked) came as no surprise.

However, many commentators have been surprised by the speed of change over the past decade, not least in terms of the reduction in the number of professing Christians and the increase in those declaring no religion.

The headline figures for England and Wales are summarized in the table below:

From this it will be seen that Christians were the only major religious group to have lost ground between the two censuses, in terms of absolute numbers. All other main groups expanded and, apart from the Jews, did so at a much faster rate than the growth in population. The ‘nones’ increased most of all, more than ten times greater than the population as a whole and even faster than the Muslims.

We are not yet in possession of cross-tabulations of the religion data, especially by demographics, and will have to wait well into next year for them to become available. In particular, these will be key to understanding regional and local, as well as national, variations.

Nevertheless, given the scale of Christian ‘losses’ and of ‘gains’ by the ‘nones’, it is not too early to start seeking explanations and to begin to map out an agenda for future research into the 2011 religion census which might begin to suggest explanations.

This post, therefore, is a preliminary attempt to identify some of the factors which would appear to merit further consideration, to assess how they may have impacted upon the responses which were given to the religion question.

Under coverage of the census

As with all censuses, the 2011 census did not initially reach 100% of English and Welsh residents through completion of the household schedule. This will have arisen through gaps in administrative knowledge and some measure of non-compliance. Therefore, the final population figure incorporates a degree of estimation. In the first statistical bulletin to contain headline results from the census, issued on 16 July 2012, the Office for National Statistics (ONS) explained:

‘The 2011 Census achieved its overall target response rate of 94 per cent of the usually resident population of England and Wales, and over 80 per cent in all local and unitary authorities … A good response was achieved to the 2011 Census but inevitably some people were missed. The issue of under coverage in a census is one that affects census takers everywhere and ONS designed methods and processes to address this. A Census Coverage Survey was carried out to measure under coverage in a sample of areas and, based on this and rigorous estimation methods, the census population estimates represent 100 per cent of the usually resident population in all areas. The estimation methods were subject to an independent peer review … All census estimates were quality assured extensively, using other national and local sources of information for comparison and review by a series of quality assurance panels. An extensive range of quality assurance, evaluation and methodology papers are being published alongside this release.’

Non-response to the religion question

Over and above the issue of ‘under coverage’, noted above, the religion question was affected by additional non-response. This arose from the fact that it was the only voluntary question to be asked in the whole census, being clearly marked as such, in reflection of the sensitivities which continue to exist around ‘the state’ investigating a topic which is often regarded as a personal and private matter.

In the event, only 7.2% chose not to answer the religion question, 0.5% fewer than in 2001 (albeit the absolute number was up). The slightly improved relative response may have reflected the efforts of ONS and faith communities to explain the rationale of the religion question among ethnic minorities. At present, we have no accurate means of knowing whether the religious profile of the ‘religion not stated’ category matched that of the 92.8% who did state their religion, or whether it was skewed in some way. A post-census survey which enquired into people’s motivations for declining to state their religion might have explicitly or implicitly shed some light on their hidden religious profession, but it is doubtless too late now to contemplate such a study. Meanwhile, it might be potentially misleading to reallocate this 7.2% as though they did match the 92.8%. It would naturally be very dangerous to assume that ‘religion not stated’ can be equated with ‘no religion’. But the fact that the number of ‘religion not stated’ was broadly similar at the two censuses perhaps enables us to discount it as a driver or explanation of religious change.

Question-wording

Previous research has indicated that variations in question-wording can produce significantly different results, and this is especially true of investigations into religion. There are several formulations which have been used to assess religious identity over the years, and these have often produced a wide gap between maxima and minima, not least for numbers of Christians and those professing no religion.

But it seems unlikely that question-wording can explain much of the change in the religious landscape of England and Wales evident from a comparison of the 2001 and 2011 censuses. The core question was unchanged: ‘What is your religion?’ While the question is viewed by some as imperfect, in implying that a religious identity is expected, this factor would only be relevant if there was a greater degree of such expectation in the minds of respondents in 2001 than there was in 2011. Given other religious changes during the decade, it might be possible to suggest that the ‘prestige’ effect of religion had lessened, and that some individuals felt it less necessary to declare a faith in 2011 than in 2001. However, other questions, such as frequency of attendance at religious services, still seem to engage the ‘prestige’ effect, leading to aspirational if not exaggerated answers.

Not only was the core question on religion the same, but the reply options were identical, with one variation. In 2001 ‘none’ was used and in 2011 ‘no religion’. ONS explained that this change was ‘for clarity’ and consistency with other questions. It would be possible to run a test, for example using a split sample, to see whether ‘no religion’ produces more affirmative replies than ‘none’, and thus whether the adoption of the code ‘no religion’ contributed in any way to the growth in the number selecting that option. Instinctively, however, this does not seem a plausible explanation for the growth in ‘nones’, certainly not in any large measure.

The running order of the reply options was likewise unaltered: no religion, Christian, Buddhist, Hindu, Jewish, Muslim, Sikh, any other religion. In the case of any other religion, the instruction in 2001 was ‘please write in’, in 2011 the more peremptory ‘write in’.

Reallocation of write-in replies

The write-in option was designed to be used, both in 2001 and 2011, by those who ticked the ‘any other religion’ box. In practice, some ticked the box but wrote nothing alongside (as yet an unknown number in 2011, but 19,000 in 2001). Others used the write-in option to qualify an answer which ONS judged would better fit another category, particularly Christian or no religion.

In fact, in 2001 only 17.2% of the 878,000 write-ins for any other religion were actually ultimately credited to any other religion. A further 33.3% were a Christian denomination and reassigned by ONS to the Christian category and 49.6% were judged by ONS to sit within the no religion rather than any other religion category, by far the biggest element being 390,000 Jedi Knights. 

For 2011 ONS has not yet divulged how many write-ins were reclassified as Christians, but it has revealed that 260,000 were reassigned from any other religion to no religion, of whom the Jedi Knights, at 177,000, were still the single biggest component, albeit a shadow of their 2001 selves. Others moved to no religion included 32,000 agnostics, 29,000 atheists, and 15,000 humanists.

All in all, it seems unlikely that the processing by ONS of write-ins will have materially affected the results of the 2011 census or help to explain the major changes which have taken place since the 2001 census. There may have been some minor alterations in coding and treatment, but these will have had little impact on the big picture; these have been explored further in the recent post by David Voas about write-ins at: 

http://www.brin.ac.uk/news/2012/census-2011-any-other-religion/

Head of household factor

Unlike in most sample surveys, the replies to questions in a census may not be given by the individual concerned but by proxy. The census is based on a household schedule, which will generally be completed by the head of the household or an equivalent senior figure. It is therefore possible that, in answering the question on religion, the head of household may impute to other members of the household his/her own religious views or record what he/she believes to be their affiliation but without ever asking them outright.

In their article on the 2001 religion census, published in Journal of Contemporary Religion (Vol. 19, 2004, pp. 23-8), David Voas and Steve Bruce hypothesized that as heads of households ‘are likely to be older than other members and religious affiliation is strongly correlated with age, one might thus expect the census figures to be biased upwards’. This might have accounted, in part, for the 72% of the population professing Christianity in 2001, which was the phenomenon that Voas and Bruce were concerned to explain at that time.

In fact, their brief analysis of the religion data from 2001 for both heads of households and all adults in the household did not suggest any striking differences between the two groups. One of the variations, albeit small, was that heads of households were 0.5% more likely to subscribe to no religion than adults as a whole. This will almost certainly have reflected the fact that heads of households were disproportionately men, and that males tend to declare no religion more than women.

Nevertheless, this head of household factor should not be lightly discounted. It could in principle be tested through a sample survey which asked heads of households to give the religious affiliation of household members while simultaneously asking the question directly of household members themselves. Given that men are disproportionately heads of households and disproportionately of no religion (one-fifth more likely than women in a 64,000 person survey by YouGov in 2011), there is at least a potential (if small) explanation for the big increase in ‘nones’ between the 2001 and 2011 censuses.  

Cohort replacement effects

It is unnecessary to say much here about the importance of cohort replacement effects since they have already been explored in some detail by David Voas in his recent post at:

http://www.brin.ac.uk/news/2012/religious-census-2011-what-happened-to-the-christians/

Suffice it to say that sample surveys, such as the British Social Attitudes Surveys, have mostly tended to show that religious affiliation is fairly stable across a person’s lifetime but declines over time in the country as a whole as a result of the death of older cohorts with a strong religious (especially Christian) identification and their replacement by young people with much lower levels of religious allegiance.

However, the scale of the fall in the Christian population between 2001 and 2011 is far more than can be explained by cohort replacement alone. According to the modelling undertaken by Voas in his post, based upon deaths during the decade, he projects that the number of Christians would only have fallen by 1.5 million, or from 72% to 69% of the population, as a direct consequence of cohort replacement. 

Disaffiliation

We therefore have to face up to the possibility that a lot of people who gave their religion as Christian in 2001 recorded themselves as of no religion in 2011.

A superficially attractive explanation for this might be sought in the British Humanist Association (BHA)’s Census Campaign, under the somewhat tongue-in-cheek slogan of ‘If you’re not religious, for God’s sake say so’.  Is it possible that this campaign tipped the balance by persuading many to abandon their ‘cultural Christianity’ and to embrace no faith? At present, thinking in the BRIN team is that this was probably a highly marginal influence, for the following reasons:

• The campaign was launched as early as 27 October 2010, and it is very hard to maintain momentum for such an initiative over such a long period as five months (census day was 27 March 2011)
• The BHA is a fairly small organization in terms of its paid-up membership and immediate circle of influence, and it therefore had to rely upon advertising, publicity in the commercial media, and social media (such as Facebook and Twitter) in order to reach the general public
• The BHA never succeeded in realizing its advertising fund-raising target for the campaign
• The BHA was informed that its slogan was likely to cause widespread and serious offence and had to be modified for advertising purposes (a previous BHA campaign in 2009, under the banner of ‘There is probably no god, now stop worrying and enjoy your life’, had already proved controversial)
• Companies owning advertising space in railway stations refused to display three different BHA census posters, thereby depriving BHA of a major promotional opportunity
• BHA census posters on 200 buses in London and six other cities had to be rephrased to read (less strikingly): ‘Not religious? In this year’s census, say so’
• To the best of our knowledge, the BHA has published no independent market research to testify to the visibility and impact of the campaign
• The only other major new religion-related census campaign in 2011, to promote Heavy Metal as a religion, was a singular flop, attracting a mere 6,242 write-ins

Therefore, defection or disaffiliation of Christians since 2001 is a probable major cause of the decline of Christian allegiance over the decade. Even though it is not the complete answer (after all, the net decline in Christians constitutes no more than 64.1% of the net growth of ‘nones’), it should undoubtedly be a primary focus of research effort. Notwithstanding the census in England and Wales did not distinguish between different types of Christians (it only did so in Scotland and Northern Ireland), such research needs to be undertaken at a denominational level and should particularly concentrate on affiliation to the Church of England, which has long been known to be the weakest and most nominal of all religious groups. Besides, we know from sample survey evidence that affiliation to the Roman Catholic Church has remained relatively stable and that, while support for the historic Free Churches (such as the Methodists and United Reform) has waned, there has been balancing growth at the Pentecostal and charismatic end of the spectrum, especially of black minority churches, which represent a new form of ‘Free Churchism’.

The Church of England undoubtedly seems to be in a fairly bad way in terms of the decline of popular identification with it. Fifty years ago, in 1963, when Gallup asked the question ‘what is your religious denomination?’ of 20 samples aggregating to more than 21,000 adults in Great Britain, 61% replied Church of England (presumably including its sister Churches in Wales and Scotland). By 2011, when YouGov asked 64,000 adult Britons ‘what is your religion?’ (the identical question to the census), professing Anglicans had reduced to 31%, although 43% had been brought up as an Anglican as a child.   

The rites of passage, another bulwark of residual Anglicanism, tell a similar story. Back in 1900 Anglican baptisms represented 65% of live births, but by 2000 this was down to 20% and by 2010 to 12%. In 1900 the Church of England conducted 65% of all marriages, but only 24% in 2000 and 2010. Even funerals, over which the Churches in general and the Church of England in particular held a near monopoly until well after the Second World War, have seen a collapse. In 2000 46% of deaths were still followed by an Anglican funeral, but only 37% in 2010.

It is not enough for Anglican apologists to counter these facts, as they often do, by reference to pockets of growth in the Church of England (Fresh Expressions and cathedrals are most often cited). Growth and decline have coexisted in organized religion for a very long time, so green shoots in some places do not contradict the overall downward trajectory in Anglican fortunes. In statistical terms, the Church of England can no longer count even on the nominal support of the majority of the nation.

So, if large-scale disaffiliation of Christians has occurred between the 2001 and 2011 censuses, it is probably in the Church of England that it needs to be sought. The motivations for it also require to be explored, in particular, the extent to which defections are a manifestation of growing dissatisfaction with the Church and thus a kind of ‘protest vote’. Three opinion polls in 2012 (by YouGov in January and November and ComRes in July) have shown that a clear majority of the nation feels that the Church of England is out-of-touch with modern society, 76% saying so in the most recent survey, following the General Synod’s rejection of women bishops. Indeed, at a time of growing liberalism in public attitudes to diversity issues, the Church’s continuing difficulties over sexuality and gender do present it with something of a public relations mountain to climb.       

International migrants

One of the key demographic changes to have occurred between the 2001 and 2011 censuses was the dramatic increase in the number of international migrants, people born outside the UK but who were usually resident in England and Wales at the time of the census. In 2001 there had been 4.6 million answering this description (8.8% of the population); by 2011 there were 7.5 million (13.4%, roughly one-third from the European Union and two-thirds from the rest of the world). Of these 7.5 million, over half (3.8 million) had arrived in the UK between 2001 and 2011. Regionally, the number of residents born outside the UK was highest in London (36.7%). Summary details are provided in the ONS statistical bulletin on International Migrants in England and Wales, 2011, published on 11 December 2012 and available at:

http://www.ons.gov.uk/ons/dcp171776_290335.pdf

It is naturally important to understand how immigration may or may not have affected the religious landscape of England and Wales. At present, we have no cross-tabulations from the census for religion by country of birth, although such data should become available in due course and will provide conclusive proof of the relationship between the two variables. All that can be done for the moment is to look at the religious profile of the countries from which these migrants have come and assume that the religion of the migrants broadly matches that national religious profile.

To take a very crude example, the top ten countries for non-UK born residents in England and Wales at the 2011 census are set out below, together with an indication of the religious profile of those countries (the latter information being taken from the very recent report on The Global Religious Landscape by the Pew Research Center). 

The superficial inferences we can draw from this table are that immigration a) may have contributed to the growth of the non-Christian population of England and Wales between 2001 and 2011 but b) seems unlikely to explain the big increase in the number professing no religion. In these ten countries the proportion with no religious affiliation was highest in Germany (24.7%), followed by Jamaica (17.2%), the United States (16.4%), and South Africa(14.9%); it fell to less than 0.1% in Bangladesh, India, and Pakistan. Moreover, the collapse in the proportion of Christians might almost certainly have been worse but for international migrants, who often came from countries where Christianity was the dominant faith. For example, there was a ninefold growth in the number of Poles (who are preponderantly Catholic) resident in England and Wales between 2001 and 2011, following Poland’s accession to the European Union in 2004.

Natural growth

Natural growth is probably irrelevant as an explanation for the increase in the ‘nones’ but almost certainly accounts for much of the rise in the non-Christian population between 2001 and 2011, especially of Muslims. The 2001 census had demonstrated that Muslims had the youngest age profile of all religious groups, as well as the largest households. A similar picture is eventually likely to be revealed by the 2011 census and merits detailed investigation once the data are available to do so. Meanwhile, other evidence also testifies to the fact that Muslims are disproportionately young (and thus more likely than average to be in the child-rearing phase) and inclined to have more children than the norm. The phenomenon is explored on a global basis by Eric Kaufmann in his book Shall the Religious Inherit the Earth? (Profile Books, 2010), which suggests that a fecundity-driven growth of ‘fundamentalist populations’ (Christian, Jewish or Muslim) will reverse the march of secularism before 2050. 

Needless to say, all the foregoing are very much preliminary observations and subject to revision in the light of further reflection and the availability of more data. At present, our ability to explore the 2011 census results for religion is very constrained by the limited amount of information released by ONS thus far.