Measuring religious affiliation in Great Britain: the 2011 census in historical and methodological context

Abstract

The British religious census of 2011 is located in its broader historical and methodological context. The principal developments in the measurement of religious affiliation (proxy-assigned or self-assigned) in Britain are traced from the Reformation to the present day, charting the relative contribution of the Churches, the State and empirical social science. The key statistics which have emerged from their respective efforts are summarised, with nominal religious affiliation universal until the time of the French Revolution and preponderant until as late as the 1980s. For recent decades, when the profession of faith has been rejected by large numbers of Britons, particular attention is paid to the variant results from different question-wording. Depending upon what is asked, the proportion of the population currently making sense of their lives without asserting a confessional religious identity ranges from one-quarter to one-half. The difficulties of trying to construct a religious barometer through a single, unitary indicator are thus illuminated.

Key Words:

• Great Britain
• religious affiliation
• religious census
• religious statistics
• sample surveys

Introduction

At the individual level, religion is often said to divide into three discrete but overlapping facets: affiliation (alternatively expressed as profession, allegiance, self-identification or belonging); beliefs; and practices (private and public). Of the three, affiliation tends to be regarded as part of the socio-public domain by virtue of helping to define people's identity, and thereby to predict their social attitudes and behaviours (Fane 1999a, 122). This contrasts with religious beliefs and practices which are said to be positioned within the private/personal domain (Francis 2008, 149). Conceptually, affiliation is a relatively weak and generalised indicator of religiosity, not necessarily implying great commitment and often merely denoting ancestry or empathy with a particular religious community, albeit some scholars (Purdam et al. 2007, 153–154) see it as additionally embracing a sub-category of ‘membership’, which others would link more with practice.

Interest in the weaker notion of affiliation, and especially how to measure it, has been rekindled since publication of results from the religion question in the 2011 census of population for England and Wales. Various practitioner groups – academics, statisticians, policy-makers, faith leaders and media – have sought to understand what these findings mean for the religious landscape of contemporary Britain. This is particularly so given the pace of religious change since the 2001 census was faster in England and Wales (less so in Scotland) than some commentators anticipated, not least in terms of the decline in profession of Christianity and the rise in those proclaiming ‘no religion’.

Rather than discussing the 2011 census results for religion in isolation, the present paper seeks to locate religious affiliation in its broader historical and methodological context. It traces the principal developments in the measurement of religious affiliation in Britain since the 16th-century Reformation, charting the relative contribution of Churches, State and empirical social science (especially sample surveys). Key statistics emerging from their respective efforts are summarised, covering either the population as a whole or adults; sources focusing solely on children or adolescents are excluded. The different methodologies employed to record current religious affiliation are discussed (religion of upbringing is omitted, partly for space reasons and partly because it has been less commonly studied), with special reference to question-wording and the effects different formulations can have on outcomes. Comparison of the 2011 census with large-scale contemporaneous social surveys highlights the maxima and minima of religious affiliation, and demonstrates the difficulties of constructing a religious barometer through a single, unitary indicator.

Churches

Following the Christianisation of Britain during the Anglo-Saxon period, from the very late 6th century, and until the Protestant Reformation of the 16th century, both Church and State could assume the population of Britain was Roman Catholic, albeit there would have been variable degrees of Mass-going and occasional manifestations of ‘heresy’. Once the Reformation commenced under Henry VIII, and was consolidated under Elizabeth I, it became obvious that Britain was no longer religiously monochrome, with the new Anglican via media challenged by Catholic sympathisers at one end of the spectrum and radical Protestant modernisers at the other. Uniformity legislation was introduced to enforce adherence to the Church of England, including compulsory attendance at parish churches, a provision which – amazingly – was on the statute books for almost the entire period 1552–1969 (Field 2008). The nascent Tudor Protestant state felt particularly concerned about perceived threats from Catholics, whose allegiance was to an extra-national temporal power (the Papacy) and who were also often thought to be in league with England's foreign enemies (notably France and Spain). However, Protestant sectaries, from whom Nonconformity grew after the Restoration, were likewise viewed as a potentially destabilising force, for their socio-political opinions as well as their theological and ecclesiological tenets.

Under these circumstances, it became important for Church and State to gauge the extent of religious ‘deviance’, and thus the first general religious censuses of England and Wales took place in 1603 and 1676, the extant documents for which have recently become available in scholarly editions (Dyer and Palliser 2005; Whiteman 1986). Because, through the parish system, the local administrative machinery of the Church was more effective than that of the State at the time, information was gathered through individual clergy. The former enquiry sought a return of communicants, recusants and non-communicants, although in practice it concentrated on communicants and recusants (equivalent to a mere 0.4 percent of communicants). The latter investigation (Compton Census) enumerated conformists, papists and nonconformists. Both surveys suffered from an imprecise and inconsistent application of these categories, non-response and underestimation.

A heavily qualified freedom of religion was introduced for Trinitarian Protestant Nonconformists by the Toleration Act 1689, but Roman Catholics remained suspect, at least before the Roman Catholic Relief Act 1791. Indeed, there were no fewer than four occasions in the 18th century when the House of Lords called on Anglican bishops to enumerate English and Welsh recusants: 1705, 1706, 1767 and 1780. The 1767 investigation is the most detailed and complete (Worrall 1980–89). In Scotland it was also deemed essential to know the respective strengths of Protestants and Catholics, and thus a census of religious profession was organised in 1755, on the initiative of Alexander Webster, Moderator of the Church of Scotland, and again employing parish ministers as informants (Kyd 1952); it showed one person in 75 was a papist. The situation in Ireland (then an integral part of Britain) was thought to be graver still, for here Catholics formed the overwhelming majority, not the tiny minority on the mainland. Accordingly, in 1732–33 a census of Irish Protestant and Roman Catholic families was taken in connection with returns to the Hearth-Money Office (Bindon 1736), revealing that 72 percent were Catholic. In 1764–66 the House of Lords ordered a fresh enumeration of Irish Protestants and Catholics, the returns to which were mostly lost when the Public Record Office of Ireland was destroyed in 1921.

Besides these nationwide investigations, local data were gathered in England and Wales, on a diocesan basis, through parochial clergy. With the progressive weakening (after 1689) of the system whereby churchwardens presented laity for religious and moral failings at episcopal visitation, for subsequent trial before ecclesiastical courts, bishops turned to incumbents to ascertain the true state of their dioceses. Commencing with the Diocese of Lincoln in 1706 (Broad 2012), pre-visitation questionnaires were issued to clergy. By the 1760s the practice had become the norm, and by the turn of the 19th century questionnaires increasingly had to be completed and sent to the bishop prior to, rather than (as previously) delivered at, visitation. The latter development meant that, from the 1820s, some rudimentary aggregation of clergy answers appeared in published episcopal visitation charges. The fullest record of visitation returns is in manuscript, through original questionnaires and/or digests (specula). Critical editions of both classes of document are being brought out by record societies (Field 2010: appendix 4), and there has been some secondary analysis, for instance examination of the religious composition of the Diocese of Salisbury in 1783 (Field 2013a). Information about Nonconformists and Catholics was routinely sought as part of these clergy returns, albeit quantitative estimates were not always offered, notably in populous settlements. Assessment was also made of the conformity of the nominally Anglican population, but, despite ministerial railing against ‘practical atheism’ (subsuming indifferent churchgoing, Sabbath breaking, intemperance and ‘loose morals’), there was little avowed unbelief before the 1790s. The ‘infidel tradition’ in Britain is usually traced to the French Revolution.

Beyond the Church of England, other faith groups attempted to calculate the sizes of their constituencies during the ‘long 18th century’. Putting these Anglican and non-Anglican primary sources together, ironing out inconsistencies and adding moderate estimates to fill gaps, preliminary figures for the religious distribution of the English and Welsh population can be computed for various dates between 1660 and 1840 (Field 2012a; Table 1). Although these are proxy statistics, in that they were obtained not by asking people directly to nominate their religion but by third parties (usually religious officials) making a judgment about the religious persuasion of their flocks/neighbours, they demonstrate unmistakable trends. In particular, the Church of England lost almost one-fifth of its market share between 1680 and 1840, from 94 to 77 percent, mostly after 1800, with an ever-increasing number of nominal Anglicans also ceasing to practise. Nonconformity more than quadrupled, mainly from 1760 and especially post-1800, representing 20 percent of the population by 1840. Roman Catholicism kept pace with demographic growth, but, even reinforced by Irish immigration from the 1780s, remained a limited force in 1840 (less than 3 percent). Judaism and irreligion were negligible.

Table 1. Estimated religious profile of adult/whole population of England and Wales/Great Britain, 1680–1939 (percentages).

Date	Area	Population	Church of England	Free Church/Presbyterian	Catholic	Non-Christian	No religion
1680	EW	All	94	4	1	*	–
1720	EW	All	92	7	1	*	–
1760	EW	All	94	5	1	*	–
1800	EW	All	88	10	1	*	*
1840	EW	All	77	20	3	*	*
1914	GB	Adults	61	28	9	1	1
1939	GB	Adults	55	29	11	1	4

Notes: * denotes a figure of less than 0.5 percent. EW denotes England and Wales, GB Great Britain.

Sources: Figures for 1680–1840 from Field (2012a, 711); for 1914 from Field (2013b, 62) (adjusted to include an estimate for nominal Catholics not known to priests); for 1939 from Field (2013c, 91).

This national picture naturally conceals many local and regional variations, glimpses of which can sometimes be caught in topographical sources, as well as visitation returns. There was no official census of population until 1801, but there were unofficial counts in various places, some recording religious profession, as in Hertford (1747), Stockport (1754) and Woodbridge (1770, 1777) (Field 2010, 13). These were isolated initiatives at first, but, by the second quarter of the 19th century, it became common for whole neighbourhoods to be surveyed on a house-by-house basis by social investigators (many Anglican clergy) enthused by the early Victorian ‘statistical movement’ or by city missionaries. A notable Anglican pioneer was Abraham Hume, active in Liverpool from the 1840s to the 1880s (Field 2010: appendix 5; Lesourd 1974; Pickering 1971). Particularly impressive was Hume's census of religious profession of the Anglican Diocese of Liverpool in 1881–82, revealing a population 57 percent Anglican, 24 percent Catholic and 18 percent other denominations and faiths (Hume 1882). Nevertheless, the scale of this undertaking was not entirely innovative, for, in Scotland during the 1830s, Church of Scotland and other clergy had also gathered affiliation data house to house as part of their evidence to the Royal Commission on Religious Instruction, Scotland (1837, 6–11), including two competing religious profiles of Edinburgh. Other Scottish parish-level data, of variable quality, can be found in The Statistical Account of Scotland (Sinclair 1791–99) and The New Statistical Account of Scotland (Society for the Benefit of the Sons and Daughters of the Clergy 1834–45).

During the second half of the 19th century most Churches started to collect statistics about their following, with Methodists leading the way in 1766. The drivers for this were partly internal, to facilitate denominational resource planning, and partly external, to demonstrate success in relation to other denominations. In particular, rivalry between Church and Nonconformity, when disestablishment and voluntaryism respectively were live political issues, often transformed religious statistics into a battleground between warring parties, as will be seen from the population census. In Wales, where disestablishment of the Church in Wales was not achieved until after the First World War, this quantitative strife reached fever pitch over the accuracy and interpretation of Anglican and Nonconformist statistics for 1905 collated by the Royal Commission on the Church of England in Wales and Monmouthshire (1910–11).

Data collected by individual Churches were frequently methodologically inadequate and conceptually incompatible, Anglicans and Presbyterians counting communicants, Nonconformists members, and Catholics population (including children) known to parish priests. So, simple aggregation of these disparate indicators does not paint a picture of overall religious affiliation. In the Church of England, for instance, very many identified as Anglican even though they were non-communicants and attenders only at the more communal services (such as harvest festivals or familial rites of passage). Before the First World War, the majority of Nonconformist adults were adherents, not members, but not necessarily any less committed in terms of chapel-going and financial support, yet adherents were rarely counted. Similarly, there were many ancestral and non-observant Catholics, especially of Irish background, not known to their priests, subsequently confirmed by the mismatch between the Catholic Church's official figures and national sample surveys (Spencer 1982, 228). Nevertheless, through careful interpretation, and factoring in other primary sources, we can reconstruct what the religious profession of the country might have looked like. Thus, in 1914 it has been conjectured the adult population of Britain comprised 61 percent Anglicans, 9 percent Catholics, 28 percent Nonconformists and Presbyterians, 1 percent other religions (mainly Jews) and 1 percent no religion. By 1939 the Anglican share had fallen to 55 percent and the ‘nones’ increased to 4 percent (Table 1).

After the Second World War some Christian denominations began to take a still more serious interest in religious statistics, inspired by the French school of sociologie religieuse. In particular, there were important initiatives (national and local) in the Anglican, Catholic and Methodist Churches during the 1950s and 1960s, while, in concert with the Church of Scotland, John Highet made a notable contribution in Scotland from 1947 to 1965, albeit he dismissed as ‘singularly worthless’ the notion of a census question on religious profession (Field 2010: appendix 8). Slightly later, there were fledgling ecumenical schemes such as, in the 1970s, ‘Stand Up and Be Counted’ from the British Council of Churches (1972), and the Inter-Churches Research Group, which in 1978–80 sought to establish a programme of religious research to parallel the 1981 census. The only one of these endeavours to bear real fruit was the Nationwide Initiative in Evangelism, which, in 1979, conducted a census of churchgoing and membership in England. The driving force behind this was Peter Brierley, ex-government statistician but then Bible Society programme director. Successively through MARC Europe, Christian Research and Brierley Consultancy, he undertook further censuses of churchgoing (in England, Wales and Scotland) and compiled multiple editions of the UK Christian Handbook and its supplement Religious Trends, the latter doing much to quantify Christian congregations, members and ministers but shedding little light on non-Christian religions and on religious affiliation generally (reflecting Brierley's dependence on church sources). Another survey of churchgoers was organised by Churches Information for Mission on 29 April 2001, coinciding with the civil census; however, it was confined to congregations of five major Protestant traditions in England.

State

As noted, the State had motivated the Church of England's efforts to gather information about nonconformity, and by the early 19th century it had begun to develop the administrative and statistical capacity to gather data in its own right, both about the Church (whose unreformed condition was causing concern in an era of political reform) and nonconformity. In the latter field an early enterprise in England and Wales was the Home Office's 1829 return of places of worship not of the Church of England, and of their adherents. Unfortunately, apart from an inaccurate edition for Lancashire, this was never printed, and the central record went up in flames with the Palace of Westminster in 1834; however, original local replies often survive in county record offices. In Ireland there had been some aspiration that the first population census of 1813–15 might encompass religion, but this was not so, and the census was never completed in any case, arising from opposition by the Catholic majority. By the 1831 Irish census the climate seemed more propitious, with Daniel O'Connell, advocate of Irish Catholic interests, pressing for a census of religious profession (Levitan 2011, 152). It did not happen in that year but in 1834, under the auspices of the Commissioners of Public Instruction, Ireland (1835). This was on the basis of information supplied by local ministers of all denominations and enumerators of the 1831 Irish census of population. Four-fifths of Ireland's inhabitants were found to be Catholic, the remainder sub-divided between Anglicans, Presbyterians and other Protestants.

In Britain a population census was taken decennially from 1801, but its scope was limited until 1841, when it was widened considerably. The newly founded London (later Royal) Statistical Society unsuccessfully petitioned for several additions to the household schedule in 1841, including religious affiliation (which it also urged at subsequent 19th-century censuses), yet there was no strong official support for tackling religion until ten years later. Sir George Lewis, member of the 1830s Irish Commission, was Under-Secretary at the Home Office by the 1851 British census and keen to see religion included, as was Registrar General George Graham, and head of his Statistical Department, William Farr. But an enquiry into religion had not been authorised under the Census Act, and objections were raised, inside and outside Parliament, to the inquisitorial nature of government asking about it in any form. The notion of investigating religious affiliation was quickly abandoned, and even the alternative (census of accommodation and attendance at places of worship, recorded by their ministers/officials) proved controversial, so eventually the government agreed it should not be completed under penalty. There is an enormous literature on the 1851 religious census, but Thompson (1978) provides a useful overview of methodology and results, while Field (1999, 2010: appendix 2) offers a bibliography of all but most recent publications.

The controversy reignited when the headlines of the 1851 religious census were published, in England and Wales (1853) and Scotland (1854). In particular, the count of sittings and attendances seemingly confirmed the relative success of voluntaryism and failures of state Churches, fuelling the disestablishment campaign. Polarised denominational positions blighted attempts to include religion in later 19th-century censuses. Thus, for 1861, the government proposed (in 1860) a census of religious profession, as implemented in several European countries. The suggestion appealed to the Church of England and Church of Scotland because it seemed likely to maximise their adherence. But it was fiercely resisted by Nonconformists; they felt either there should be no religious census at all (for reasons of religious freedom and privacy) or it should be a count of churchgoers, as in 1851, which would naturally put Nonconformists in the best statistical light. In the end, Lord Palmerston's government, commanding a small Parliamentary majority, abandoned a religious census altogether (Drake 1972, 17–18; Levitan 2011, 36–37, 88–94; Snell and Ell 2000, 449–452). Successor administrations had no appetite to resurrect the matter at subsequent British censuses but came under pressure to do so.

Prior to the 1871 census, the General Assemblies of the Church of Scotland and Free Church of Scotland memorialised the Home Office for a religious census, but the government was soon ‘flooded’ with counter-memorials from Free Church and United Presbyterian congregations. The government decided there was insufficient consensus to ask about religious affiliation, but an amendment to the Census Bill to include religious profession was tabled by John Ball at committee stage. It was voted down in the House of Commons, reinstated by the House of Lords in committee (notwithstanding government opposition), and objected to again by the Commons, whereupon the Lords resolved not to pursue the issue (Hansard, Commons, 22 July 1870: cols. 805–818; Commons, 26 July 1870: cols. 1003–1006; Lords, 2 August 1870: cols. 1399–1406; Lords, 9 August 1870: col. 1730). Similar amendments, again contrary to government wishes, were laid during Parliamentary scrutiny of Bills for the 1881 and 1891 censuses and also lost on division (Hansard, Commons, 2 September 1880: cols. 1063–1086; Lords, 16 June 1890: cols. 950–960; Commons, 21 July 1890: cols. 399–421; Commons, 22 July 1890: cols. 517–534; Drake 1972, 19). Lord Hugh Cecil's amendment to include religious affiliation in the 1901 census was likewise lost in the Commons after a brief debate (Hansard, Commons, 15 March 1900: cols. 1002–1013).

Religious profession was more exhaustively discussed in connection with the Census Bill 1910, largely because of agitation for Welsh disestablishment. The government's line was that any attempt to extend the census to religion would be of ‘a most fiercely controversial nature’. While it saw off John Rawlinson's amendment in the Commons on 21 June, Lord Newton's amendment was carried by a narrow margin in the Lords at committee on 12 July and reaffirmed in a second division on 19 July. The amendment was overturned by the Commons on 27 July, and a government motion in the Lords on 1 August not to press for the amendment was accepted without division, thereby averting a constitutional crisis. However, the issue did not disappear since the Anglican Sir Arthur Griffith-Boscawen introduced two bills in 1912 and 1914 to take a census of religious profession in Wales and Monmouthshire. Both were resisted by the government and Nonconformists and did not advance beyond second reading (Field 2013b, 59–60).

With Welsh disestablishment achieved, there was no longer a political imperative for a religious census. Besides, opportunities for detailed Parliamentary debates about each census had been minimised by the Census Act 1920, a perpetual piece of legislation (still in force) providing for taking each British census under authority of Order in Council. Religion was not one of the census topics specifically mentioned under the Act's Schedule, although discretion was given to include ‘any other matters with respect to which it is desirable to obtain statistical information with a view to ascertaining the social or civil condition of the population’. So far as can be seen, there was no Parliamentary discussion of religion in the 1921 census, although some statisticians were in favour (Bisset-Smith 1921, 10–21, 99–101, 210–213).

Nevertheless, while reluctant to countenance religious censuses in Britain, the government was not averse in principle to collecting data about religious affiliation. This was a standard (and, until 1971, compulsory) feature of the Irish decennial census from 1861 (and, following partition, in Northern Ireland from 1926). Religious profession likewise featured in censuses throughout the British Empire, in 28 colonies and dependencies in 1901 (Census Office 1906). Also, at home, the government enquired into the religious affiliation of particular groups, notably members of the armed services, commencing with the Army (1860) and extending to the Royal Navy (1939) and Royal Air Force (1963), as well as prisoners, intermittently at first, and then annually from 1962. It should be noted that, although religion of patients became routinely recorded by NHS hospitals, data have never been collated and reported nationally.

After 1911, the possibility of a population census on mainland Britain investigating religion seems not to have been discussed in Parliament for half a century, until it briefly reared its head with the 1961 census. John Parker, Dagenham's Labour MP, moved to include religion when the draft Order in Council for the census was debated in the Commons in 1960, following his earlier correspondence with the Home Office on the subject. Parker's rationale was public expenditure on Church-run schools and the need for its evidential basis. Responding for the government, Niall Macpherson (Joint Under-Secretary of State for Scotland) dismissed the request, commenting: ‘There is considerable resistance towards giving this kind of information and we are advised that the questions would probably be widely resented and in consequence we would not be likely to get the sort of truthful answers which we seek to obtain in the census’ (Hansard, Commons, 4 May 1960: cols. 1193–1196).

The advice Macpherson received was from the Registrar Generals for England and Wales and Scotland, and it long remained the case that the key government department – the Office for Population Censuses and Surveys (OPCS, merged into the Office for National Statistics, ONS, in 1996) – appeared slow to engage with a religious census. Doubtless, anxieties about public resentment and lack of veracity expressed in 1960 continued to inform official thinking. Perhaps, too, what has been labelled the ‘religious crisis’ of the 1960s contributed to a feeling that religion was becoming less significant socially, and thus did not need enumeration as an aid to government policy. On the other hand, large-scale immigration after the Second World War made parts of Britain increasingly multicultural and ethnically diverse. Through the strong link between ethnicity and religion, not least with immigrants from the Indian sub-continent, it was also becoming more religiously pluralistic, suggesting the requirement for better government intelligence.

However, it was Christian rather than non-Christian advocates who sowed the seeds for an extra-Parliamentary campaign to include religious affiliation in the census. The first tentative suggestion appears in connection with the 1971 census, when Wallis Taylor, University of Manchester statistician and Methodist, raised the topic at the official census advisory committee, of which he was a member. The matter does not seem to have progressed beyond the committee, which, on that occasion, prioritised a question on language. Both ethnicity and religion were candidates for inclusion in 1981, but neither was adopted, seemingly for cost and fear of exciting public controversy, the religious cause also suffering from late intervention (at the end of 1978) and a narrow basis of faith support (primarily Taylor, Brierley and Tom Houston, Bible Society executive director). Although the British Council of Churches' Division of Ecumenical Affairs, at the behest of the Inter-Churches Research Group, initially favoured approaching the OPCS about including religion in the census, the Council's executive committee, in November 1978, declined to recommend to senior Church leaders that they should lobby the OPCS, raising ‘rather far reaching questions about the usefulness of such a question to the community as a whole’ (letter from Martin Conway of the Council to Brierley, 1 December 1978). An ethnicity question did finally make it to the census in 1991, but only after prolonged debate about benefits and pitfalls. The potential relevance of religion was recognised in the OPCS census test of October 1985, in which persons of Indian, Pakistani, Bangladeshi or Sri Lankan origin were asked about their religion in a supplementary to the primary question about race and ethnicity, but the idea had been dropped by the time the 1991 census was taken.

Not until the 2001 census did religion finally appear in the household schedule. The story of how this happened has been told by several participants (Aspinall 2000; Dixie 1998; Fane 1999b; Francis 2003; Kosmin 1998, 1999; Sherif 2011; Southworth 1998, 2001, 2005; Weller 2004; Weller and Andrews 1998). By the mid-1990s there was a developing groundswell of support for including religion in the census. Paul Weller (University of Derby) had launched a Religions and Statistics Research Project under whose auspices was convened a seminar of faith and academic representatives in May 1994, from which sprang a pilot survey in 1994–95, testing the interest of faith and secular stakeholders in covering religion at the next census. Brierley raised the possibility of a census question at a conference organised by Churches Together in England in October 1994, from which emerged a working party on a ‘Religious Question for the 2001 Census’ chaired by Leslie Francis (then of the University of Wales, Lampeter), which reported in 1996. The Board of Deputies of British Jews, which had opposed a census of religion in 1991, favoured one in 2001 and raised it with the OPCS. Two Jewish organisations, the Parkes Centre (University of Southampton) and Wiener Library (London), arranged a symposium in July 1997 to consider ‘The Ethnic and Religious Questions in the British Census’, whose proceedings were published in the April 1998 issue of Patterns of Prejudice. The Muslim community, feeling beleaguered and unrecognised, and – like most non-Christian faiths other than Jews – lacking its own statistics, latched on to a religion question in the 2001 census as providing official affirmation of Muslims' collective identity (Sherif 2011). Inter-faith organisations, such as the Inter Faith Network for the United Kingdom and Inner Cities Religious Council, added their weight. There was also interest from several government departments, the Commission for Racial Equality and some academics, but there were a few outspoken academic critics, notably Graham Zellick (1999), human-rights lawyer and University of London Vice-Chancellor, who denounced a religious census as a breach of the European Convention on Human Rights.

OPCS consultations on the 2001 census began in 1995, with the establishment of a content working group. The initial departmental view, articulated by John Dixie (manager of the census Data Needs Project), was sceptical about (if not – in the minds of some – antipathetic to) covering religion in 2001. Another senior ONS colleague is recalled by David Voas (writing to the author on 30 March 2013) as being unenthusiastic, arguing the census should be reserved for measuring ‘objective’ and not ‘subjective’ characteristics. As Francis (2003, 46) recollected: ‘The campaign to have such a question in the national census was initially met by incomprehension and disbelief by the civil servants within the Office for National Statistics.’ Nevertheless, Dixie led an initial meeting with faith bodies in March 1996, and, following intervention by the Department of Health, Social Security and Education, the ONS agreed to constitute a Religious Affiliation Sub-Group. This commenced in August 1996, convened by Francis, with a membership of faith leaders and academics. It was disbanded by the ONS in June 1998 but continued independently as the Census 2001 Religious Affiliation Group, again led by Francis. The sub-group's task was to develop an indicative business case for including religion in the census and thereafter to advise on a possible question; it was clear that the focus should be on religious identification, rather than belief or practice. The business case was ready by May–June 1997, and in June, following a small-scale test in March–April, a religious question was included in a test involving 97 000 households. The response rate was good, suggesting public willingness to be asked about religion, but, according to qualitative follow-up, the wording used (‘Do you consider you belong to a religious group?’) led to confusion as to whether affiliation, belief or practice was being probed. On the Sub-Group's advice, the question was changed to ‘What is your religion?’ and subject to a small-scale test in December 1997.

However, successful testing did not imply religion a guaranteed place in the 2001 census. While not without support, it was in contention with other new topics. On this relative scale of priorities, it was clear by 1998 that religion had come fairly well down the list, not simply in a trawl of government departments but even among academics, judging from a survey conducted by the Economic and Social Research Council and Joint Information Systems Committee. Notwithstanding, the White Paper issued the following March (HM Treasury 1999: paragraphs 28–29, 64–67, 176) recommended a religion question in England and Wales, albeit not Scotland (on the grounds that the business case was weaker there, a position the Scottish Parliament ultimately reversed, under concerted pressure from Scottish sociologists and historians). But the rationale given for the question (‘What is your religion?’) was that it ‘would help provide information which would supplement the output from the ethnicity question by identifying ethnic minority sub-groups, particularly those originating from the Indian sub-continent, in terms of their religion’. In other words, religion was not viewed by the government as a critical variable in its own right but as an adjunct to ethnicity. This helps explain the decision to locate the question on the 2001 census form for England and Wales after the question on ethnicity, which some commentators (Day 2011, 65, 184; Voas and Bruce 2004, 27) believe contributed to an inflated number of Christians (in Scotland, by contrast, religion appeared before ethnicity). Another manifestation of this essentially ethnic motivation was the government's refusal to differentiate within Christians in England and Wales, thereby preventing quantification of allegiance to individual denominations (unlike in Scotland, where the Scottish Executive relented, or in Northern Ireland).

There was another problem, for the White Paper contained an important caveat, noting that, on advice from government lawyers, primary legislation would need to be changed (an amendment to the Schedule of the Census Act 1920) to authorise a question about religion, and that, before it proceeded to move in this direction, the government ‘would want to be satisfied that the inclusion of such a question in a census commanded the necessary support of the general public’. The successful census rehearsal in April 1999 helped provide the requisite assurances on the latter point, and a Census (Amendment) Bill was introduced into the House of Lords by Lord Weatherill as a Private Member's Bill in December 1999 (and sponsored, on the same basis, by Jonathan Sayeed in the House of Commons). It eventually received Royal Assent in July 2000, despite predictions it would run out of Parliamentary time, not least in the House of Commons, where it got bogged down, despite an early concession (Lords Second Reading) that the religion question would be voluntary and not answerable under penalty, thereby ensuring it would not be perceived as infringing civil liberties (Hansard, Lords, 27 January 2000: cols. 1709–1722; Lords, 3 February 2000: cols. 363–368; Commons, 20 June 2000: cols. 266–311; Commons, 26 June 2000: cols. 670–673; Commons, 26 July 2000: cols. 1133–1176). Similar legislation was enacted in Scotland in April 2000.

A religion question thus appeared in the 2001 census, notwithstanding government ambivalence. As Southworth (2005, 75, 85) argued, the campaign proved ‘a difficult process’ which ‘almost failed’, its eventual success, she suggested, lying disproportionately with ‘the Muslim population’. There can certainly be no denying that Muslims consciously used the issue to assert their collective identity and political clout (Sherif 2011). The first release of results from the religion question occurred in February 2003, separately for England and Wales and Scotland, with much more information subsequently available online; there were also two printed reports (Office of the Chief Statistician 2005; ONS 2006). In the event, religion was not stated for 8 percent of the population of Britain, 72 percent being Christian, 5 percent non-Christian and 15 percent of no religion. Despite initial reservations, the government increasingly recognised the value of the religious data and, as we shall see, incorporated a census-style question on religion, albeit with variant wording, into its large-scale sample surveys, notably the Labour Force Survey in Britain (partially from 2002 and fully from 2004, prior to which religion had only been covered in Northern Ireland).

By the time it came to planning the 2011 census, the government needed no persuading that religion should again be included. When the ONS launched a public consultation in May 2005 about the proposed content for the next census, religion was assigned the highest (‘category 1’) priority. The consultation, which ran until August, attracted just under 50 responses focusing on religion, the scoring of which in March 2006 confirmed ‘category 1’ status. A more detailed public consultation on user needs for religion commenced in December 2006, extending to March 2007. Responses to this, analysed in October 2007, revealed four-fifths of English and Welsh organisations required information on religion from the 2011 census and that, of those with such a need, three-quarters felt it would be met by the question and categories proposed (the same as in 2001) for the May 2007 census test. When the White Paper on the census in England and Wales was published in December 2008 (Cabinet Office 2008: paragraphs 3.59–3.64), a replication of the 2001 (voluntary) question was suggested, and three requested user enhancements were rejected. These were: differentiation within the Christian and Muslim categories; inclusion of codes for additional religions and non-religious beliefs; and incorporation of a measure of religious practice.

Nevertheless, the ONS was at least partially sensitive to criticisms of its approach in 2001, among them the ‘positive presumption’ or ‘affirmative grammatical form’ of the question, implying respondents would or should have a religion (Day 2011, 33, 64, 66; Voas and Bruce 2004, 26). The ONS accepted the 2001 question was ‘potentially leading’. Therefore, the programme of qualitative and quantitative testing (February 2005–July 2009) investigated four alternatives for asking about religious affiliation, besides ‘What is your religion?’ These were quantitatively evaluated through the ONS monthly omnibus: ‘What is your religion or belief?’ (September, November–December 2007); ‘What is your religion, even if not currently practising?’ (May–June 2008, April–July 2009); ‘Do you regard yourself as belonging to a religion?’ (June 2008); and ‘Which of these best describes you?’ (April–July 2009). As well as these variations in wording the principal question, there was experimentation with the order and wording of response codes. ONS (2009) provides a detailed report on this testing programme, while outcomes of the omnibus research are summarised in Table 2. Following post-test consultation with key stakeholders, the majority of whom apparently preferred to keep the 2001 question-wording (to ensure comparability between 2001 and 2011), the ONS resolved on this final recommendation, albeit the ‘none’ response code was changed to ‘no religion’. The ONS further determined to continue to locate the religion question in England and Wales after the one on ethnicity, although in 2011 two (in England) or three (in Wales) questions on language separated them.

Table 2. Religious profile of adult population of Great Britain according to ONS omnibus testing of five questions about religious affiliation, 2007–9 (percentages).

Question	Date	Christian	Non-Christian	No religion
What is your religion or belief?	Sep. 2007	64	12	25
	Nov. 2007	67	8	25
	Dec. 2007	66	9	25
	Mean	66	10	25
What is your religion?	May 2008	76	8	17
	Mean	76	8	17
What is your religion, even if not currently practising?	May 2008	74	6	18
	June 2008	77	9	15
	April 2009	74	8	18
	May 2009	73	5	21
	June 2009	69	9	22
	July 2009	77	5	18
	Mean	74	7	19
Do you regard yourself as belonging to a religion?	June 2008	65	6	29
	Mean	65	6	29
Which of these best describes you?	April 2009	73	7	19
	May 2009	68	5	26
	June 2009	68	9	23
	July 2009	71	6	23
	Mean	70	7	23

Note: Approximately 1000 interviews were conducted each month.

Source: ONS (2009): annexe A.

Some understanding of how the 2011 religion census was operationalised, by its respondents rather than the ONS, can be gained from a survey by Ipsos MORI (2012) for the Richard Dawkins Foundation for Reason and Science (UK). The sponsor's intention was demonstrating how relatively feebly Christian those who professed to be Christian in the census actually were, in terms of beliefs, practices and attitudes, but this paper's ends are served by the first of the screening questions, put to 2107 adults in Great Britain and Northern Ireland in face-to-face interviews on 1–7 April 2011, just days after the census form had to be completed (27 March 2011). In 69 percent of cases people claimed to have answered the religion question themselves against 18 percent who had it answered by somebody else on their behalf (one-third of whom had not bothered to check with the informant first). The remaining 13 percent either had not replied to the religion question, had not completed the census at all, could not remember who had answered the question about their religion, or preferred not to say. This highlights a methodological drawback of censuses, that data are often collected by proxy (typically via head of household), rather than supplied directly by the respondent. It partly explains why, in 2001, all people in a household were classified to the same religion in 85 percent of households in England and Wales, with a further 14 percent of households recording one religion apart from no religion and/or religion not stated; that left just 2 percent of households as religiously heterogeneous (ONS 2006, 13). Since heads of households are disproportionately male and older, demographics known to align with religion (negatively in one case, positively in the other), there is some potential for the proxy factor to skew census results (Voas and Bruce 2004, 24–26). Proxy responses are presumably very common for children under 16 years, who may often not be consulted about their religious identity. On the other hand, it is notable that, at the 2011 census for England and Wales, children were six points more likely to be recorded as without religion as adults (30 versus 24 percent) and one point more likely to be entered as religion not stated (8 versus 7 percent).

The first release of 2011 religion census data for England and Wales was on 11 December 2012, with further releases on 30 January and 16 May 2013, the last accompanied by a substantive paper on religion (ONS 2013), and 31 July 2013. The first release of Scottish data did not come until 26 September 2013. In England and Wales, there was a decline of 11 percent between 2001 and 2011 in the number professing to be Christians and an increase of 83 percent in those declaring no religion, with 57 percent more non-Christians. The scale of these changes is greater than some commentators anticipated, and the search quickly began for possible explanations (Field 2012b). There seems little doubt that natural growth and immigration account for the progress of non-Christian faiths, but methodological issues (under coverage, non-response, amended question-wording, reallocation of write-in replies, head of household factor) do not seem to explain adequately inter-censal movements in Christian and no religion categories. The debate here is whether cohort replacement effects (i.e., death of elderly Christians) or disaffiliation from Christianity has been the major force for change (Brierley 2013a, 2013b; Voas 2012, 2013). If the latter, what difference might the British Humanist Association's 2011 census campaign, under the slogan of ‘If you're not religious, for God's sake say so’, have made?

Sample surveys

Advances in probability theory, sampling techniques and invention of the Hollerith tabulating machine provided the impetus for the third source of data on religious affiliation: national sample surveys. Even then, it was left to the requirement for better predictions of outcomes of American presidential elections to trigger the birth, through George Gallup in 1935, of the public-opinion polling industry, based on interviews with representative adult samples. The movement spread to Britain in 1937, with the foundation of the British Institute of Public Opinion (BIPO), later Social Surveys (Gallup Poll). Many other agencies have since entered the field, and some have left it.

Although Gallup's first full-scale study on religion in Britain was not undertaken until 1957, individual questions were included in earlier polls, commencing with professed church membership and regular churchgoing in its November 1937 omnibus. This was despite the alleged conviction of Henry Durant, BIPO's owner, that ‘religion was no longer an important factor shaping public opinion’ (Roodhouse 2013, 237). The first known occurrences of a Gallup question on religious affiliation were in August and December 1943, when the religious profile of Britons aged 21 and over was: 47 percent Anglican; 25 percent Free Church and Presbyterian; 10 percent Roman Catholic; 14 percent another religion; with 4 percent not answering. It was a pity that Gallup initially failed to differentiate between other religion and no religion, which did not happen until July 1946, and even then the no religion category remained somewhat ‘contaminated’ by the inclusion of non-respondents. Since the 1940s, there have naturally been countless national sample surveys measuring religious affiliation, albeit it has never been a question routinely asked in its own right, generally inserted only when there has been a specific requirement to analyse by religion replies on some other topic (often morality-related). Space limitations preclude a comprehensive catalogue of surveys here. However, a reasonably complete listing until 1982 can be found in Field (1987, 365–387) and a more selective record thereafter reconstructed through the source database on the British Religion in Numbers website (by searching under ‘religious affiliation’).

As noted elsewhere (Field forthcoming), sample surveys on religion present sundry methodological and interpretative challenges. In the case of religious affiliation, there has been only limited standardisation of question-wording, and still less of the categorisation and running order of response codes. This can make comparison between surveys problematical, as illustrated by the 1970 British Cohort Study, which has already used many different measures of religion across five waves, revealing that ‘apparently small differences in question wording can lead to dramatic differences in responses’ (Sullivan, Voas, and Brown 2012, 19). To mitigate these problems, our emphasis is on a limited number of time series where the agency, methodology and question have been held constant, thereby enhancing comparability. The focus is generally also on surveys which employed larger than average samples or on chronologically adjacent surveys whose results can be aggregated, ironing out the inevitable volatility from one poll to the next. For this reason, several well-known series such as the European Values Surveys (Williams, Francis, and Village 2009), which use fairly small samples and pose consistency issues across the constituent waves, have been ignored. Even so, apart from the two largest communions (Anglicanism and Catholicism), individual national surveys cannot be guaranteed to provide a totally reliable measurement of specific Christian denominations, although pooling data from multiple studies, as Field (2009) has done for Methodists, can sometimes lead to significant discoveries. Neither do samples constructed to be representative of the entire adult population of Britain tend to be fully accurate for non-Christian religions, on account of their relatively recent growth and spatial concentration. The incorporation of booster samples of ethnic minorities naturally improves reliability, but this practice has typically only been followed in government-sponsored and a few academic investigations.

For most of the second half of the 20th century, the best evidence derives from cumulated Gallup Polls (Table 3), through face-to-face interviews of adults aged 16 and over. It will be seen that between the 1940s and 1970s the self-assigned religious profile of Britons was relatively stable, other than a marked decline in the Free Church constituency, reflecting the progressive disappearance of ‘adherents’. With that exception, there was remarkably little change, and certainly no sign of the ‘religious crisis’ of the 1960s which scholars such as Callum Brown (2012) have written about. Indeed, Table 3 suggests the 1980s may have been the key secularising decade, in terms of numbers professing no religion. Of course, the question posed by Gallup – ‘What is your religious denomination?’ – will be seen as somewhat leading, implying that respondents not merely professed a religion but could also denominationalise it, the effect being reinforced by ‘no religion’ appearing last in the response codes. This factor doubtless helped trigger the phenomenon we now know as ‘Christian nominalism’ (Day 2011) or ‘cultural Christianity’ (where religion, tradition and nationality converge), with four-fifths of Britons still self-identifying as Christian as late as the 1990s. Gallup's principal rival, National Opinion Polls (NOP), used a very similar face-to-face question during the 1960s, and it is noteworthy that, when it switched to the milder formulation of ‘Which religious group would you say you come into in terms of your beliefs?’, the Church of England's share declined by several points and the number selecting no religion rose. As with Gallup, NOP data reveal that the 1960s did not see a particular flight from religion, which became pronounced from the late 1970s (Table 4).

Table 3. Religious profile of adult population of Great Britain according to cumulated Gallup Polls, 1947–96 (percentages).

Date	N =	Church ofEngland	Church ofScotland	FreeChurch	Catholic	Otherreligion	Noreligion
1947–49	?7000	51	8	15	11	6	9
1957	2261	55	7	15	9	5	9
1963	21 495	61	8	11	10	4	6
1964	10 708	62	8	11	10	4	6
1974	17 968	61	7	7	11	6	8
1978	11 061	60	7	7	12	6	8
1979	11 097	61	7	7	12	5	8
1982	5800	58	8	6	14	6	8
1987	3918	58	7	4	13	7	12
1992	4064	55	6	5	14	5	15
1996	4666	53	5	5	12	8	16

Notes: The question asked is ‘What is your religious denomination?’ Church of England and Church of Scotland figures in 1992 are estimated since only a total for both denominations combined was coded. Other religion represents a combination of Christians other than Anglican, Catholic or Free Church and of non-Christians, the latter mostly not being separately identified by Gallup.

Sources: Statistics for 1963 from Brothers (1971, 12); for 1964 from Gallup Election Handbook, March 1966; and for 1974 and 1979 from Gallup Political Index. All other figures from unpublished data held by the author.

Table 4. Religious profile of adult population of Great Britain according to National Opinion Polls, 1965–93 (percentages).

Date and question	N =	Church of England	Church of Scotland/Presbyterian	Catholic	Other religion	Atheist/agnostic	No religion/don't know/refused
What is your religion, if any?
March 1965	2160	63	9	10	15	1	2
Feb. 1967	1899	63	10	10	15	2	0
Sep. 1967	1792	64	8	11	13	3	1
Aug. 1968	1218	64	9	10	13	3	1
Jan. 1970	1705	65	8	10	13	3	1
Which religious group would you say you come into in terms of your beliefs?
June 1970	1396	60	8	10	17	4	1
April 1973	1974	57	7	9	16	5	6
July 1978	1985	59	8	9	14	6	3
Aug. 1978	1852	58	7	11	13	6	4
Sep. 1981	1991	57	8	11	14	5	6
April 1982	1104	60	7	12	13	5	4
Oct. 1984	1992	59	6	9	13	6	7
Feb. 1985	1709	54	5	12	15	10	4
Jan.–Feb. 1986	1652	53	7	11	15	6	8
April–May 1989	1960	54	7	11	16	9	4
Regardless of your religious upbringing, would you tell me what your religion is now?
Sep. 1976	2125	59	6	10	13	7	4
Nov. 1978	1952	59	7	9	13	7	5
March–April 1993	2017	52	6	10	15	11	6

Sources: Published and unpublished reports in author's possession.

It is instructive to compare Gallup and NOP figures with those from BSA Surveys, undertaken by NatCen annually since 1983 (except 1988 and 1992). Interviews are conducted face-to-face with adults aged 18 and over. The filter question asked has been: ‘Do you regard yourself as belonging to any particular religion?’ An affirmative answer then prompts the follow-up question of ‘which?’ The filter wording might be seen as more neutral than Gallup's, in carrying no implication that respondents would or should have a religion. On the other hand, the BSA question introduces the concept of ‘belonging’, which many people might well interpret to mean religious membership or some other formal connection with a faith body. The wording thus tends to maximise the proportion self-declaring without any religion, which has grown steadily from one-third in the early 1980s to almost one-half today (Table 5, where data are aggregated quinquennially, to generate larger samples). There has been a corresponding reduction of two-fifths (from 61 percent to 36 percent) among denominational Christians, with the Church of England losing almost one-half its market share over the lifespan of BSA, and other denominations contracting less strongly. Also notable is the trebling in those opting to describe themselves as undenominational Christians and a doubling of non-Christians (principally Muslims). These trends are brought into sharper relief by analyses of BSA data by birth cohort (1900–04 to 1985–89) as opposed to fieldwork date (Lee 2012; McAndrew 2011), which reveal ‘an increasing proportion of the younger birth cohorts are none, other religion or non-denominational Christian’. In England, for example, the generation born in the 1980s was almost four times as likely as that born in the 1900s to profess no religion at the time of interview. Such findings support a generational replacement explanation of religious change, with each generation less likely than its predecessor to have a religious upbringing. This, coupled with BSA evidence that people do not generally ‘convert’ from no religion in later life, seemingly points to the inexorability of on-going decline in religious affiliation.

Table 5. Religious profile of adult population of Great Britain according to British Social Attitudes Surveys, 1983–2010 (percentages).

Date	N =	Church ofEngland	Catholic	Other denom.Christian	Undenom.Christian	Non-Christian	Noreligion
1983–85	5115	38	11	12	3	3	33
1986–90	11 438	37	10	12	3	3	35
1991–95	12 842	33	10	11	5	3	38
1996–2000	14 570	29	9	10	6	4	43
2001–05	18 525	28	9	9	7	5	42
2006–10	19 521	21	9	7	10	6	47
2011–12	6520	21	9	6	10	7	47

Note: The question asked is ‘Do you regard yourself as belonging to any particular religion?’

Source: Calculated from http://www.britsocat.com.

The British Election Studies (BES) had also become relevant by this point. Although they commenced in 1964, the BES question on religious affiliation was not standardised until 1983, when the BSA formulation of ‘Do you regard yourself as belonging to any particular religion?’ was adopted (Table 6). Despite the identical question, sample profile and interviewing method (face-to-face), and a similar long-term trend, BES has tended to report more professing Anglicans and fewer ‘nones’ than BSA, particularly during the 1980s. There are likewise some discrepancies between relevant years of BSA and BES on the one hand and wave 1 (1991–92) of the British Household Panel Survey (BHPS) on the other, which periodically asked ‘Do you regard yourself as belonging to any particular religion?’ By restricting analysis to surviving members of the original panel (Table 7), we can measure changes in religious affiliation over time for the same individuals. It will be seen that, within less than two decades, between wave 1 (1991–92) and wave 18 (2008–09), the proportion professing no religion grew by 6 percent, with a 4 percent drop in Anglicans and smaller decreases for other Christians. Slightly different again is the European Social Survey (ESS), which, despite using virtually identical wording to BSA and the same face-to-face data collection, has obtained rather higher figures (51 percent in 2012) for ‘nones’ (Table 8).

Table 6. Religious profile of adult population of Great Britain according to British Election Studies, 1974–2010 (percentages).

Date	N =	Church of England	Catholic	Other religion	No religion
1974	2349	42	9	16	34
1979	1854	31	10	17	42
1983	3944	45	11	18	26
1987	3813	41	10	17	32
1992	3521	33	11	26	30
1997	3591	32	11	25	32
2001	3857	29	11	19	41
2005	1201	31	10	15	44
2010	1576	30	8	15	48

Note: The question asked has varied: ‘Do you belong to any religious denomination?’ (1974); ‘Do you belong to any church or religious group?’ (1979); and ‘Do you regard yourself as belonging to any particular religion?’ (1983–2010).

Source: Calculated from http://www.besis.org.

Table 7. Religious profile of adult population of Great Britain according to British Household Panel Survey, 1991–2009 (percentages).

Date	N =	Church of England	Catholic	Other Christian	Non-Christian	No religion
1991–92	9912	36	9	14	3	38
1997–98	6900	35	9	12	3	41
1999–2000	6538	35	9	12	3	41
2004–05	5440	32	8	12	4	44
2008–09	4683	32	8	12	4	44

Notes: The question is ‘Do you regard yourself as belonging to any particular religion?’ The analysis is restricted to surviving members of the original 1991–92 panel.

Source: Supplied by Professor David Voas.

Table 8. Religious profile of adult population of the United Kingdom according to European Social Survey, 2002–12 (percentages).

Date	N =	Catholic	Other Christian	Non-Christian	Any religion	No religion
2002	2052	8	35	5	49	51
2004	1892	–	–	–	51	49
2006	2389	–	–	–	47	53
2008	2351	10	32	5	47	53
2010	2422	8	29	7	45	55
2012	2277	11	31	7	49	51

Note: The question is ‘Do you consider yourself as belonging to any particular religion or denomination?’

Source: Calculated from http://nesstar.ess.nsd.uib.no

And so, we reach 2001, when the census of population first covered religion. The census differed from sample surveys in four main respects: data were gathered by means of self-completion schedules rather than interviews; many responses were given by proxy and not obtained directly from each individual; the whole population was investigated and not just adults; and the census was voluntary, with many consequently not stating their religion, whereas in sample surveys remarkably few choose not to answer the question on religious affiliation. If the 2001 census results are rebased to exclude non-responses and children, then the religious profile of Britain in that year, as determined by the question (in England and Wales) of ‘What is your religion?’, was: 78 percent Christian, 5 percent non-Christian and 16 percent no religion. This distribution is remarkably similar to the 1996 Gallup Poll data, which is unsurprising, given that Gallup posed a similar question (‘What is your religious denomination?’) to the census. Perhaps equally predictably, the rebased census continues to differ markedly from BSA, whose actual result for 2001 alone was: 54 percent Christian, 4 percent non-Christian and 41 percent no religion. BSA's much stricter concept of religious identity, wrapped up with the notion of ‘belonging’, inflated the no religion category by a factor of two and a half, relative to the census. For ESS in 2002 (Table 8) there was a threefold differential.

The decade following the 2001 census witnessed several new government sample surveys, to parallel the less official BSA, BES and BHPS studies, which continued to enquire into religious affiliation (Tables 5–7), as did ESS from 2002 (Table 8). These government newcomers included: the Annual Population Survey (APS)/Integrated Household Survey (IHS) from 2004, sponsored by the ONS, covering Great Britain (Table 9); the Citizenship Survey from 2001, sponsored by the Home Office and subsequently Department for Communities and Local Government, covering England and Wales (Table 10); and the Taking Part Survey from 2005, sponsored by the Department for Culture, Media and Sport, covering England (Table 11). All three used face-to-face interviewing of adults, as did BSA, BES, BHPS and ESS. The seven time series are consistent in confirming the overall direction of travel during the course of the noughties, specifically weakening allegiance to Christianity and growing willingness to declare no religion. But they disagree about the scale of these changes, not least in reflection of the different questions asked. Especially interesting is Table 9, reporting on APS/IHS, partly because it is the largest pool of social data after the population census and partly because the question was altered in January 2011, with dramatic results, as Hawkins (2012, vi–vii) observed. Until that time respondents were asked ‘What is your religion, even if you are not currently practising?’ and the no religion option appeared last in the list; thereafter, the question was truncated to ‘What is your religion?’ and no religion was moved to the head of the list, in both respects for consistency with the 2011 census. Comparing quarters 1–3 (April–December) and quarter 4 (January–March) for 2010–11, these adjustments had the immediate effect of shifting a net 5 percent of the population from the Christian to no religion category. This can only mean that some people understand ‘What is your religion?’ to imply the practice of religion, rather than just loose attachment. The qualifying statement ‘even if you are not currently practising?’ had formerly made them comfortable to tick the Christian box and to feel nominalism was entirely acceptable; once this crutch is kicked away, a significant number are driven to confront reality and concede they are not meaningfully attached to Christianity. ‘What is your religion, even if you are not currently practising?’ has also been used in the British Crime Survey from 2005, replacing the question introduced in 2003 of ‘Which of these religious groups do you belong to, if any?’

Table 9. Religious profile of adult population of Great Britain according to Annual Population Survey/Integrated Household Survey, 2004–12 (percentages).

Date	N =	Christian	Non-Christian	No religion
2004–05	518 743	78	6	16
2005–06	365 016	76	7	17
2006–07	353 914	75	7	18
2007–08	348 699	74	8	19
2008–09	339 805	72	8	20
2009–10	442 266	73	7	20
2010–11, Q1–3	311 793	70	8	22
2010–11, Q4	102 039	64	9	27
2011–12	342 506	62	8	29
2012	338 174	61	9	30

Notes: Until 2010 the question asked was ‘what is your religion, even if you are not currently practising?’ From 2011 it changed to ‘what is your religion?’ The base excludes Northern Ireland.

Sources: Figures to 2008–09 from Perfect (2011: 5); later ones calculated from http://nesstar.esds.ac.uk

Table 10. Religious profile of adult population of England and Wales according to Citizenship Surveys, 2005–11 (percentages).

Date	N =	Christian	Non-Christian	No religion
2005	9654	77	8	15
2007–08	9325	74	10	16
2008–09	9317	72	10	18
2009–10	9291	70	8	21
2010–11	10 296	68	10	22

Notes: The question asked is ‘What is your religion even if you are not currently practising?’ The base is the core sample, excluding ethnic minority boosts.

Source: Calculated from http://nesstar.esds.ac.uk

Table 11. Religious profile of adult population of England according to Taking Part surveys, 2005–12 (percentages).

Date	N =	Christian	Non-Christian	No religion
2005–06	21 058	72	8	20
2006–07	23 547	71	7	22
2007–08	25 070	70	7	24
2008–09	14 117	70	6	24
2009–10	5954	70	6	24
2010–11	13 712	67	6	26
2011–12	8941	69	6	26

Note: The question asked is ‘What is your religion?’

Source: Calculated from: https://www.kmrsoftware.net/netquestdcms/Default.aspx

Table 12 brings our discussion full circle by juxtaposing results of the 2011 census of religion with contemporaneous sample surveys (to ensure like-for-like comparison, census figures have been rebased to omit non-respondents and children). It is superficially reassuring that the census and IHS for 2011–12, both asking ‘What is your religion?’, are in substantial agreement. At the same time, when YouGov posed exactly the same question weeks after the census, they found 11 percent fewer Christians and 14 percent more ‘nones’. Might the variance be explained by the fact that their sample was drawn from a self-selecting panel and they only poll online, with a risk of under-representing older people, who are at once more likely to profess religion than the young and to be less likely to use the Internet? YouGov's other survey in Table 12, the BES Internet Panel in 2010, asked ‘Do you regard yourself as belonging to any particular religion?’, as did NatCen for Understanding Society in 2009–10 and for BSA in 2010–11. This is a narrow definition of religious affiliation, predicated on the notion of belonging, and implying formal links with a faith body, and all three surveys therefore produced a high number (around half the adult population) declaring no religion. A similar concept of ‘belonging’ (Opinion Research Business 2010) or ‘membership’ (Populus 2011) probably helped drive down Christian adherence and drive up those without religion in these two surveys. The question wording most calculated to maximise Christians and minimise ‘nones’ seems to be ‘What is your religion even if you are not currently practising?’, used in APS/IHS before 2011 (Table 9), as well as in the Citizenship Survey (Table 10). Interestingly, differences in question-wording do not appear to have much impact on non-Christians, whose proportion remains remarkably constant; their effect is confined to the extensive borderland between the most nominal Christians and people of no faith. How one frames a question on religious affiliation has the potential to shift large numbers from one camp to the other.

Table 12. Religious profile of adult population of Great Britain according to various surveys, 2010–12 (percentages).

Question	Date	Agency	Method	N =	Christian	Non-Christian	Noreligion
SAMPLE SURVEYS
Which religion, if any, do you regard yourself as belonging to?	2010	Opinion ResearchBusiness	online	4030	53	8	39
Which of the following religious groups do you consider yourself to be a member of?	2011	Populus	online	44 551	58	8	35
Do you regard yourself as belonging to any particular religion?	2010	YouGov	online	16 816	41	6	53
What is your religion?	2011	YouGov	online	64 301	55	5	40
Do you regard yourself as belonging to any particular religion?	2009–10	NatCen	face-to-face	45 741	44	7	49
Do you regard yourself as belonging to any particular religion?	2010–11	NatCen	face-to-face	6560	45	7	48
What is your religion even if you are not currently practising?	2010	ONS	face-to-face	311 793	72	8	21
What is your religion?	2011–12	ONS	face-to-face	342 506	64	8	28
CENSUS
What is your religion?	2011				66	8	27

Sources: Opinion Research Business (2010) from data tables in author's possession; Populus (2011) from 16 polls at http://www.populus.co.uk and http://lordashcroftpolls.com; YouGov (2010) supplied by Dr Ben Clements; YouGov (2011) from http://www.yougov.polis.cam.ac.uk; NatCen (2009–10) supplied by Professor David Voas; NatCen 2010–11 from http://www.britsocat.com; ONS (2010) and ONS (2011–12) from http://nesstar.esds.ac.uk

Summation

Does all this matter? To the extent that we are unable to provide a definitive answer to what is ostensibly a simple question – ‘What is the religious profile of the population?’ – it surely does. It appears to be a failure of professional social science and statistics that there are multiple and incompatible answers when it comes to measuring religious affiliation. And yet, in some senses, this should not surprise us. In the same way that we understand religious affiliation, religious beliefs and religious practice are conceptually distinct entities, so each of these three domains encapsulates a spectrum of attachment running from weak to strong. Nor are people necessarily consistent, logical and rational in operating across these domains. The beliefs and practices of some Christians may not live up to their faith, as the Richard Dawkins Foundation sought to prove (Ipsos MORI 2012), yet equally many allegedly faithless manifest signs of faith, as Theos contended (Spencer and Weldin 2012). Religious identity is understood in different ways by different people, ranging from a secularised ethno-culturalism to some form of religious membership, and the variant language used in the questions gives expression to this mix of understandings and misunderstandings. It is not that one form of question is right and another wrong; all are equally valid in demonstrating the complexity of the popular religious landscape.

Of course, some argue that the taxonomic framework within which we are trying to operate is fundamentally outmoded. Linda Woodhead (2012), for example, commenting on the 2011 religion census, criticised reliance on simple, unitary and confessional categories of religion, which have increasingly broken down and ignore realities of religious heterogeneity on the ground. Indeed, for many, she argues, the very word religion has become toxic. It is undoubtedly the case that some of the old denominational Christian labels have worn thin; self-identifying as Anglican is no longer the default option, and, in the most recent BSA surveys (Table 5), one-tenth classify themselves as undenominational Christians. In the YouGov poll which she commissioned for the 2013 Westminster Faith Debates (Religion and Society Programme 2013), Woodhead posed a number of more experimental questions probing religious identity. Although one-half of her sample continued to regard themselves as Christians, far fewer, 39 percent, said they were currently influenced by Christianity, the gulf being explained by the residual effect of a Christian influence in earlier life (reported by 59 percent). Likewise, while 37 percent professed no religion, as many as one-half reported no religious or spiritual influence on their lives or merely that of humanism or secularism.

In sum, when it comes to religious affiliation, while the master narrative may be clear enough – we have moved from a society before the French Revolution where there was precious little renunciation of all forms of religion to one where between one-quarter and one-half the population now make sense of their lives without a religious identity – the detail remains uncertain and ambiguous. Religious affiliation has often been seen as an easy mechanism for differentiating the religious from the irreligious. But such a differentiation is both conceptually and methodologically fraught and, even if the goal can eventually be reached, it surely cannot be attained on the basis of a unitary measure of religious affiliation. The quantification of religion requires multiple indicators to be able to capture its diversity. A single question in a decennial census of population provides a useful snapshot of one dimension, but we should not fool ourselves into thinking that it depicts the entire religious landscape.

Acknowledgements

The author is indebted to Professor David Voas (University of Essex) and Dr Ben Clements (University of Leicester) for commenting on a draft, and providing some of the sample survey data.

Funding

Preparation of this paper was facilitated by a grant (to British Religion in Numbers) from the RCUK Global Uncertainties Impact Support Fund, 2012/13.

Additional information

Clive D. Field is Honorary Research Fellow, School of History and Cultures, University of Birmingham, UK and Institute for Social Change, University of Manchester, UK. He has researched and written extensively on the social history of religion in Britain from the 18th century to the present day, with special reference to statistical sources. He is co-director of British Religion in Numbers at the University of Manchester.