Religiosity, Secular Participation, and Cultural Socialization: A Case Study of the 1933–1942 Urban English Cohort

The nature of secularization is of enduring interest in the social science of religion. Numerous recent papers have established downward cohort trends as characterizing religious change. We examine potential mechanisms by assessing cultural participation and secular engagement during the formative period of one cohort. We provide estimates of active and nominal religiosity, nonreligion and religious belief for those born between 1933 and 1942, using multiple surveys ﬁelded between 1957 and 2018. We model the association between religiosity and secular cultural and social participation for this cohort in 1957, then examine how cultural socialization in childhood relates to religiosity in their later adulthood using surveys ﬁelded between 2005 and 2007. Increased secular competition is found to be associated with less active religiosity. These trends were underpinned by an ethic of increasing autonomy for the young. We conclude by afﬁrming the link between increasing secular competition, long-run modernization, and changing cultural socialization.


INTRODUCTION
What are the mechanisms of secularization? Modernization, rationalization, urbanization, the rise of industrial society, and declining existential insecurity have been posited as distal causes by those working within a secularization paradigm (Bruce 2011;Norris and Inglehart 2004). From socioeconomic perspectives, secularization has been linked with the rise of consumer society (Hirschle 2014), increased secular competition (Gruber and Hungerman 2008;Stolz 2010), and weak religious competition linked to state regulation (Stark and Iannaccone 1994).

JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION
Secularization has also been linked to the moral acceptability of fertility control and ideational change (Lesthaeghe and Surkyn 1988). Each of these accounts is separately compelling, so that the causes of secularization appear overdetermined. Scholars such as Warner (1993) and Stolz (2009) have accordingly argued that accounts of religious change require a broad paradigm capable of encompassing methodological individualist and sociological models and perspectives.
Latterly, a considerable body of work has established downward cohort trends as characterizing generational religious change, consistent with demographic and sociological models (Brauer 2018;Crockett and Voas 2006;King-Hele 2011;Voas and Chaves 2016;Voas and Crockett 2005;Wolf 2008). Modernization of the wider cultural environment, and its effects via religious and broader cultural socialization, has been posited as one crucial mechanism. This interpretation has however proved difficult to test directly using quantitative data. Ideally, researchers would have recourse to longitudinal studies of individuals over a range of contexts, both religious and secular. Repeated cross-sectional studies such as the General Social Survey and International Social Survey Programme (ISSP) have provided considerable insight, but have proved most valuable in permitting description of religious change rather than identification of its drivers, a point noted by Stolz with regard to the secularization paradigm more broadly (2010:256).
In this paper, we combine a range of data sources to examine the sources of religious change for one particular cohort coming to adulthood at a time of profound cultural change. We estimate aggregate rates of active and nominal religiosity, nonreligion, and adherence to religious beliefs from adolescence to late adulthood, using 38 surveys fielded over a 60-year period. In particular, we exploit a new data set to test the relationship between religiosity and cultural participation within local context. This combination allows a more robust test than hitherto possible of whether cultural socialization is a proximate source of secularization in England. We identify the timing of shifts within the cohort from active to nominal religiosity, and nominal religiosity to nonreligion. We find evidence that less active religiosity was associated with increased secular competition, and socialization into secular culture with nonreligion in adulthood. We conclude by affirming links between increasing secular competition, long-run modernization, and changing cultural socialization.

BACKGROUND
A considerable body of quantitative work has demonstrated cohort replacement as a key mechanism of secularization. U.S. debate has tended to focus on American exceptionalism and institutional vitality (Putnam and Campbell 2010;Warner 1993), but recent work corroborates that downward cohort trends are taking effect nevertheless (Brauer 2018;Voas and Chaves 2016). Voas and Chaves remark that the study of causal mechanisms underlying cohort differences should now be prioritized: "[w]hat social and cultural changes make each generation slightly less religious than the previous one? What is the relative importance of changes in geographical mobility, family structure, education, technology, economic conditions, and other factors? " (2016:1548).
The field has also devoted attention to religious socialization in youth, and its relation to religiosity in adulthood. Religious socialization is the "interactive process whereby social agents influence individuals' religious beliefs and understandings" (Sherkat 2003:151). Religious groups socialize adherents partly through providing nonreligious social goods: "access to mating markets, contacts for business, friendship networks for children, social status in the community" (Sherkat 2003:154). Inglehart (1977) centered socialization in his work on value change, theorizing that values are formed in childhood, conditioned by the social environment then growing up, and thereafter demonstrate relative stability. Accordingly, adolescence is a critical period for value formation and ensuing social change, although it is plausible that the effect of adolescence on values depends on being young in a particular period rather than "generation" being a purely additive effect. Indeed, Luo and Hodges (2019) have recently provided a careful account of cohort 249 effects as the interaction of age and period effects. With relevance to religious socialization, Regnerus, Smith, and Smith have used AddHealth, the U.S. National Longitudinal Study of Adolescent to Adult Health first fielded in 1994-1995 to those in grades 7-12. Their study examined adolescents' social context, identifying parents as the primary influence on young people's attendance, but friends, school, and the local community as also significant: "social relationships matter for the development of religiosity in adolescence" (2004:34-35). Their results confirm the importance of an ecological approach to understanding youth religiosity.
Scholarly attention to religious socialization does not necessarily imply adherence to a secularization paradigm. Nevertheless, findings of downward cohort change do imply that parents are unsuccessful in transmitting their level of religiosity to their children, and this raises the question of how religious socialization takes effect. Using cross-national data from the 1991 wave of the ISSP, Kelley and De Graaf (1997) argue that parental religious socialization is more intensive in secular societies; however, this was thoroughly rebutted by Voas and Storm (2019) with a larger sample of countries, using the 2008 waves of the ISSP and European Values Study. Rather, they found that religious contexts reinforce religious values and behavior, including parental prioritization of religious transmission.
Social structure and organizations (family, school, community, voluntary groups among others) affect human behavior at all life stages. Through secular socialization, young people acquire different values compared with religious socialization; and with social change, religious socialization in childhood tends to become less relevant to adult experience so that agents of socialization change in response (Ryder 1965). Cohort effects in religiosity are the result, due to particular events and social changes affecting a given cohort during their formative ages. King-Hele accordingly concluded of her study of five predominantly English-speaking societies that "the period of childhood or adolescent socialization seems to be at the heart of the dynamics" (King-Hele 2011:244). In finding primarily cohort-driven change, she suggests that increasing time in education, increased geographic mobility, media massification, and increased social diversity all encourage the value shifts leading to religious decline. She also considers changing social norms among the young and changing parental expectations. She concludes that "[g]iven the broadly similar patterns in all results for these countries, it seems unlikely that individual country characteristics are as important as some overarching factor such as modernization" (2011:291).
Earlier work has also identified youth socialization as critical for adult religiosity. Uecker, Regnerus, and Valler (2007) use Waves I (1994-1995, aged 11-17) and III (2001-2002 of AddHealth to identify the social sources of diminished religious attendance, religious salience, and disaffiliation in early adulthood. They found that not participating in higher education was associated with greater religious decline, as were cohabitation, nonmarital sexual activity, and alcohol and cannabis consumption. They identified religious decline as primarily passive: religious participation is "simply and subtly" crowded out so that it is "not a priority" (Uecker, Regnerus, andValler 2007:1686). They suggested that weak religious socialization "set in motion during adolescence" (1686) might account for it.
High-quality large-scale surveys of youth religiosity using population samples tend to be rare, however, including in England, our case of interest. Conclusions regarding the effect of cultural socialization are largely inferred from the differences evident between cohorts via pooled cross-sectional studies rather than investigated directly. They also rely on surveys established from the 1980s, well after the seismic social and legal changes of the 1960s (Hall 2013). However, it is plausible that religious socialization in contemporary secular and religiously-diverse contexts works differently compared with when religious organizations were mainstream and most young people had a religious identity, motivating a turn to earlier sources. Useful contributions include Wadsworth and Freeman's study of a birth cohort born in Britain in 1946, where religious beliefs were captured in 1972. They found 67 percent qualified as "believers" compared with 87 percent during upbringing (1983:430), with higher education identified as an important driver of decline in belief. They also noted that those who saw their parents less than monthly compared with weekly were more likely to report change in religious belief (Wadsworth and Freeman 1983:432). Tilley analyzed the British Election Study (BES) cross-sectional and panel studies from 1964 to 1992, finding period decline in attendance at a place of worship between the early 1960s and 1980s, and apparently trendless fluctuation thereafter. Otherwise, he found declining attendance to be primarily cohort-driven (Tilley 2003:277).

DATA AND METHODS
We seek a more definitive account of the causal mechanisms driving retreat from religious identification, belief, and practice. We approach this by focusing on a particular cohort experiencing adolescence during a time of increasing secular competition, moving in three distinct steps. We first explore change over time in religious practice, affiliation, and belief for the 1933-1942 birth cohort-close to the zenith of fuzzy fidelity (Voas 2009) in England, and falling immediately before the baby boom cohort-using a number of survey sources. While an individual longitudinal study would be preferable, in its absence we deploy techniques common in quantitative history (for example, Field 2017). We combine cross-sectional sources to examine religious change for this cohort, including the little-known Youth Research Council (YRC) survey of 1957, a large survey of English youth hailing from the earliest years of British survey research. Similar items were also fielded on NatCen Social Research's British Social Attitudes (BSA) survey in the 1990s and 2000s, allowing comparison of rates in young and later adulthood. Next, to probe potential mechanisms through which downward cohort change worked, we draw on rich data on sociocultural engagement in the YRC and its stratified design to assess the relationship between religious and secular engagement, and how it varied across local contexts for this cohort in adolescence and early adulthood. Finally, to examine the effect of cultural socialization further, we turn to the Taking Part 2005Part -2006Part /2006Part -2007 to test for associations between cultural socialization in early youth and religiosity in later adulthood for this cohort.
Accordingly, we examine a distinct cohort over a long period of time to establish how modernization has affected religiosity via cultural socialization and participation (King-Hele 2011:292). The YRC and Taking Part surveys were fielded in England only, but there is good reason to take England as the unit of analysis given differences between England, Wales, Scotland, and Northern Ireland in their religious, political, and administrative histories. Our approach accordingly responds to Voas and Chaves' call for new directions in the study of cohort-driven religious change, including focus on subgroups and regions (2016:1547).

THE 1933-1942 COHORT AS A CASE STUDY
Our analysis is centered on a novel data source, gathered at the beginning of the age of rock and roll. Affluence, mass consumption, and distinct youth cultures heralded significant cultural changes of concern to academics and public bodies as they related to working-class youth. Accounts of the 1950s have focused on the emergence of affluence, youth subcultures, and teenagerhood as a social and marketing category (Hebdige 1979). Todd describes nothing short of cultural revolution: "Being working class became fashionable. Teenagers could take some of the credit. Although 'affluence' remained limited, young wage-earners were increasingly significant consumers" (Todd 2014:236). By contrast, sociologists of generations distinguish a "Silent Generation" coming to adulthood in the 1950s from the more liberated 1960s Baby Boomers (Inglehart 1977). Biographies and social histories suggest that traditional activities continued to be widespread, with religious organizations providing social life in contexts where consumer society was less prevalent. Brown has described a time of religious retrenchment in the 1950s, ending "really quite suddenly in 1963" (Brown 2001:1).
1950s youth were beginning to accommodate affluence and social change in ways that older generations, committed to particular habits and lifestyles, could not (Ryder 1965). Moreover, from 1957, those born from 1939 were not subject to compulsory National Service in the armed forces. The 1944 Education Act established free secondary and tertiary education, although the majority left school before 16 as late as 1973 (Bolton 2012:10). This cohort also had relative employment security through postwar reconstruction, and was in any case a significantly smaller cohort compared with the middle-aged. Journalist Tosco Fyvel also noted the opening of public social spaces to affluent youth as a new phenomenon: "the plush cinemas, the modernized dancehalls, the pubs with a singing trade, the late-night cafés with juke-boxes blaring and girls to be talked to at the tables . . . [by] the type of lad of 17 to 19 who, after what he pays for his board at home, has about four pounds a week [$11.20 in 1957; $97.50 in 2017]" (1966:15, 22). This was also highlighted by Abrams' report The Teenage Consumer drawing attention to the surge of fashion demand among 15-25-year-olds (Abrams 1959). American rock and roll was becoming accessible via records played at home and on jukeboxes, and the skiffle craze began in 1957, the year that Lennon and McCartney formed The Quarrymen. Memoirist Michèle Hanson attended concerts in London at this time: "Buddy [Holly] did his strange and thrilling dance . . . [to] the huge waves of screaming from hundreds of girls . . . overcome with the excitement, rock and roll, noise and heat and sudden and entirely novel screaming opportunities" (Hanson 2012:207).
Equally, memoirs concerning the time describe the traditional as persisting: "[e]verything we wanted, even if it was legitimate, we were made ashamed of wanting" (Bradbury 2011(Bradbury [1962:5). Historian Ross McKibbin characterized it as "a golden age of youth discipline and respect" (2005:519). Music writer Jacky Hyams (b.1944) described "the rituals of going steady followed by an engagement ring. East-End girls my age still either lived at home or moved out to marry" (Hyams 2011:233). She also noted that 1950s Sundays were "another planet away from the Sundays we now take for granted. Silent streets; virtually everything closed. Pubs open briefly at lunchtime and for a couple of hours at night . . . People visited each other. Or they stayed indoors" (Hyams 2011:43). Professional football was still a Saturday game, with the Lord's Day Observance Society monitoring infringements closely (Abrams and Brown 2010:166, 169). Onethird of cinemas were closed in England and Wales on Sundays in 1957, and cinema proprietors had to pay a charity levy to comply with the Sunday Entertainments Act 1932 (HL Deb 20 February 1957. Indeed, a significant economics literature suggests that such restrictions may well have shored up youth religiosity (Gruber and Hungerman 2008).
Neither were many young people so affluent: wages were controlled for apprentices, with 35 percent of male school leavers going into 5-year apprenticeships between 1945 and 1975, paid at about one-fifth of the tradesman's rate (Vickerstaff 2003:284). Clothing was still expensive (Majima 2008), underpinning a culture of mending, knitting, and dressmaking, which half of female respondents to a 1948 Mass Observation survey cited as their chief leisure activity (Langhamer 2000:41). Only in 1957 did television overtake radio as the primary source of family entertainment (Kynaston 2013:14). Accordingly, youth social life continued in religious contexts, for rather longer than standard accounts of 1950s youth culture imply. Brown considers the 1950s "the last Victorian decade": "[r]eligion mattered and mattered deeply in British society as a whole in the 1950s. But it started to stop mattering in the 1960s" (Brown 2001:7).

THE YRC STUDY
Our core source, digitized from paper returns stored in a private archive, was fielded just after the Suez Crisis of October-November 1956. High-quality British social surveys are generally from later decades, with the Medical Research Council National Survey of Health and Development (or 1946 Birth Cohort Study), and Glass mobility study of 1949, noted exceptions. The headquarters of the Young Christian Workers (YCW) association had requested cross-national data from national sections for their 1957 international congress. The English section tasked the Newman Demographic Survey (NDS), a private research organization consulting primarily to Catholic organizations, to conduct a survey on scientific principles. Its head, religious sociologist Tony Spencer (b.1928), advised that a random sample survey could be fielded by local YCW volunteers: "this opportunity to obtain really valuable information about an age group which has been causing great concern was unlikely to recur for many years and should be embraced" (Spencer 1958:5). The YRC was accordingly formed as a joint venture between the NDS and YCW to field the survey in January/February 1957, and 1,500 young people recruited to field it-mostly volunteers in full-time work (Spencer 1958:6).
With investigators aiming for a sample of at least 1,000 young Catholics across all English dioceses, the target sample was 8,196, of which 5,834 returns have survived. A short version of the questionnaire was presented to 70 percent of Anglican respondents, which formed the religious majority, to ensure that sample sizes for Catholics (and, concomitantly, for Nonconformists) were reasonably large without exhausting interviewers. Surviving correspondence evidences how a random probability sample was drawn, while the survey instrument consisted primarily of closed-form items piloted in Highgate, Manchester, and Gateshead, designed in collaboration with researchers at Mass Observation, Gallup, and the official National Food Survey. Volunteer interviewers received training to ensure consistency in extracting responses (see study documentation). Multistage stratified sampling was employed: 58 local authorities were selected to reflect the geographic diversity of urban England, with data from 39 surviving. Addresses were sampled from the electoral roll with young people aged 15-24 then selected using the 'last birthday' method. Bristol, Rugby, and Walsall returns have been lost to damp, and those from 14 of the 28 London boroughs to causes unknown. We assume that missingness occurred completely at random and that the proportionate stratified sample retains external validity.

CROSS-TEMPORAL COMPARISONS OF THE RELIGIOSITY OF THE 1933-1942 COHORT
We begin by comparing the affiliation, attendance, and belief rates reported in the YRC 1957 survey and the BSA 1983-2018, categorizing respondents as actively-religious (attending a place of worship at least monthly), nominally-religious in reporting a religious affiliation but less than monthly attendance, or nonreligious, following Wilkins-Laflamme (2014). We also provide data points for active and nominal religiosity and nonreligion from the 1963, 1964, and 1966 waves of the BES, bearing in mind that its sample of electors relates to a slightly different population than that for the BSA, which samples those living in private households in Britain. Butler and Stokes (1970) provide detailed information on the BES study design; the 1963 sample is selfweighting while cross-sectional weights are available for the 1964 and 1966 waves. 1 Since the YRC survey was fielded in urban England at a time before large-scale immigration from the New Commonwealth, we restrict our analyses where possible to White respondents in England to reduce confounding by immigration from more religious contexts when making comparisons over time. 2 On the YRC survey, respondents were asked:

What is your religion? Church of England, Nonconformist, Roman Catholic, Other, None, Other Answer (write in).
A subsample (30 percent) of Anglicans, and respondents who were not Anglican, were offered the full version of the questionnaire, including a question on attendance: Would you mind telling me about how often nowadays do you go to Church or Chapel, apart from Weddings, Christenings and Funerals? Once a week or oftener, once in 2 weeks, once a month, four times a year, twice a year, once a year or less often, Christmas, Easter, never, other answer (write in)? This served to ensure good sample sizes for Catholics (N = 1,168) and Nonconformists (N = 568); poststratification weights were calculated to adjust for undersampling of Anglicans among other variables (see online data appendix). However, the full sample was asked their religious beliefs, including the following: Do you believe in . . . God? Heaven? Hell? with options for each comprising yes, no, don't know, and other answer (write in). We compared those indicating belief with those who did not, did not know, or provided another answer spontaneously.
Comparably, the BES asked respondents their religious affiliation and attendance at a place of worship in the years it was fielded between 1963 and 1970, and again from 1983 (from this point we use the BSA instead as a more frequently-recurring source). The question wording is reasonably close to that in the YRC (see online data appendix for full question wording). Serendipitously, the 1933-1942 cohort overlaps closely with the youngest age group in the 1963 BES, ranging in age from 21 to 30 at that point. Respondents were asked their religious affiliation (Church of England/Episcopalian; Church of Scotland/Presbyterian; Methodist; Baptist; other Nonconformist; Roman Catholic; Jewish; none); then: How often do you attend church (synagogue)? Several times a week, once a week, several times a month, once a month, several times a year, once a year, less than once a year, or never?
We estimate religious type for our cohort of interest: whether they were nonreligious (rejecting a religious affiliation), nominally-religious, or actively-religious in reporting at least monthly attendance. "Race" is available as a variable for 1963 and 1966; we estimate rates for White English respondents for those years. 3 Urban-rural status is available for 1964 only; however, the urban English in 1964 appear only very slightly more likely to be actively-religious or nonreligious than otherwise, and so any biases for the 1963 and 1966 appear small. Nevertheless, slight differences in underlying populations should be borne in mind.
For the BSA, the following items form part of its core questionnaire, fielded face-to-face every year since 1983 except 1988 and 1992: Do you regard yourself as belonging to any particular religion? Which? Apart from such special occasions as weddings, funerals and baptisms, how often nowadays do you attend services or meetings connected with your religion?  This does differ from the YRC and BES in not assuming that the respondent has a religion, which may encourage the unreligious to identify as having no affiliation compared with the 'what is your religion?' wording. For current purposes, we use having an affiliation versus no affiliation as a minimal indicator of religiosity; for attendance, we categorize respondents by whether they attend at least monthly or not (see online appendix for full response options). The BSA has also hosted the ISSP surveys of religion in 1991, 1998, 2008, and 2018, when additional items were Percentage actively-religious, nominally-religious, and nonreligious, 1933-1942 cohort (%). 1957, 1963-1964, 1966, 1983-2018  For comparability with the YRC, we dichotomized responses: those who indicated belief in God at least some of the time were scored 1, while those who did not believe in a personal God scored 0. Those who indicated they were absolutely or somewhat sure of their belief in God scored 1; other responses were scored 0. Where fielded separately on split samples, or where respondents provided responses to only one, we combined these variables to create a single indicator of belief in God. For the heaven and hell items, those responding "yes, definitely" or "yes, probably" were scored 1, otherwise 0. A potential criticism is that some religions do not have a clear concept of heaven or hell. However, the major world religions accounting for the vast majority of our religious adherents in this cohort in fact do, so we judge these items as relevant indicators of religiosity. A variable is also available to indicate whether respondents were White or identified with another ethnic category. Urban-rural status is only available consistently from 1985 to 1991 or 1993 to 2018 (using different items) and so we report rates for the White English sample. Changes in weighting from 2005 mean that we use the pre-2006 weights provided by the data publisher to allow cross-wave comparisons. Descriptive statistics are available in the online supplementary data appendix.
In graphing our indicators in Figure 1 we find that in 1957, 35 percent reported active religiosity (attending a place of worship at least monthly and a religious affiliation), and 61 percent reported nominal religiosity (religious affiliation only). Four percent reported that they were of no religion. Between then and the mid-1960s, there was a substantial fall in the percentage reporting active religiosity, from 35 percent to 17-20 percent. The nonreligious proportion appeared to stay stable; most of the reduction in active religiosity resulted in an increase reporting nominal  1933-1942 cohort, 1957- 1963, 1964, 1966andBritish Social Attitudes surveys 1983-2018. religiosity. By 1983, the proportion reporting that they were of no religion had increased to 20 percent, while active religiosity remained stable at around 20 percent, which appears to have persisted over the entire 1983-2018 period. 4 The increase appears to have been drawn from the hitherto nominally-religious, which retreated to about 50 percent by 1983 and then exhibited trendless fluctuation. Clearly, there was a sizeable shift from active to nominal religiosity over the course of the late 1950s and early 1960s, and a shift from nominal religiosity to nonreligion over the later 1960s and 1970s. Figure 2 summarizes data on reported attendance: a sizeable increase in the proportion reporting that they never attend occurred between the 1950s/1960s and 1980s, while weekly attendance notably decreased. The differences in percentages reporting that they never attend in 1957 and then 1963-1965 are suggestive of potential survey house effects. It is plausible that the rate for 1957 reflects the stratified design for which weights have not entirely adjusted, and those in the BES are affected by the civic focus of the questionnaires encouraging overreporting of generalized engagement. Until further surveys of the period are digitized it is impossible to tell the comparative quality of the surveys. Nevertheless, we retain them for completeness; and taken together, the 1957-1966 estimates indicate shifts in frequency of attendance and nonaffiliation taking place for this cohort before the 1980s.
As Figure 3 indicates ninety percent of the cohort indicated belief in God in 1957, 87 percent in heaven, and 55 percent in hell. Rates are considerably lower in 1991, 1998, and 2008, and by 2018 had reached 56 percent for at least some belief in God, 26 percent for heaven, and 10 percent for hell. Rates for 1998 do look higher for each item than in 1991, perhaps because of survey wave or period effects. However, rates for 2018 look somewhat lower than in 2008, suggesting  1957 1991 1998 2008 2018 1957 1991 1998 2008 2018 1957 1991 1998 2008 2018  that belief was not stable for this cohort even in later adulthood, and has retreated either with age or due to downward period effects. While belief in God looks moderately high, certainly compared with attendance at a place of worship for this cohort (between 56 and 77 percent 1991-2018 for belief, vs. 16-26 percent for at least monthly attendance), contemporary rates are much lower for more demanding beliefs. A relationship with a personal God is perhaps easier to retain compared with belief in an eternal punitive hell. While many respondents may have been offering socially-desirable responses, overreporting belief in 1957 and perhaps underreporting it in the 1990s and 2000s, it is nevertheless hard to argue that this cohort's levels of belief in youth were retained in later adulthood. The most conservative interpretation is that belief in God was essentially stable for this cohort from the 1990s, while decline in belief in heaven and hell continued during the 1990s and 2000s.
These trends provide an interesting contrast with the apparent cohort stability in affiliation and at least monthly attendance from 1983 as presented in Figure 1, and remind us that religiosity is multidimensional. They also suggest that the rate of secularization may vary across different dimensions. This variation in belief decline is also interesting given previous focus on belief in God as an indicator of religiosity, one exhibiting cohort stability in the pooled BSA (Crockett and Voas 2006;King-Hele 2011;Voas and Chaves 2016). It is plausible that weakening religious socialization via less frequent church attendance in both youth and adulthood may play a role, and that the external environment may exert pressure on particular aspects of religiosity throughout the life course, even if at the cohort level identification and attendance have been essentially stable since the 1980s. Furthermore, the religious behavior-belief link is clearly worthy of renewed attention. Given the nature of the BSA survey design, such relationships are difficult to test directly. In their absence, we turn to the plentiful measures of social participation in the YRC data set for what they can inform us about the youth sociality-religiosity relationship.

RELIGIOSITY AND SECULAR SOCIAL ENGAGEMENT IN 1957
The YRC survey provides detail on how young people spent a substantial proportion of their free time through questions on how they spent the previous Sunday. Spencer sought unprompted RELIGIOSITY, SECULAR PARTICIPATION, AND CULTURAL SOCIALIZATION 257 report of church attendance to compare with a later explicit question, aware of the concerns regarding social desirability bias later confirmed by Hadaway and Marler (2005). Accordingly, following a screening question to identify age, respondents were asked:

Did you do any of the following last Sunday: read books, go to the cinema, read newspapers, listen to the radio, go for a walk or cycling, play indoor games or cards, watch television? Was there anything else you did last Sunday?
Responses were coded for report in the table of descriptive statistics (see online data appendix for a complete set). Interviewers' record sheets have been lost so that we cannot estimate response rate, but survey correspondence suggests it was not of concern, at a time when 80 percent was "thought normal" (Savage and Burrows 2007:890). Sixty-eight percent read a newspaper, 61 percent listened to the radio, 52 percent watched TV, 30 percent went to the cinema, 27 percent went for a walk, and 14 percent to church (poststratification weights applied; see online supplementary data for details). Drawing on the written-in responses, 6.7 percent report engagement in activities we could describe as "youth culture," combining written-in responses of having listened to records at home or with a friend, having gone dancing, having gone out drinking, or having visited a café, nightclub, jazz, or rock concert. Nine percent reported doing house or garden work, craft, or home improvements; 4 percent went visiting or entertained; 2 percent did homework (about a quarter of those who were students or still in school). With these answers unprompted, categories are accordingly sparse, but it is also plausible that young people indeed had relatively few options. A total of 1.2 percent of respondents did not select any activity or provide a written-in response (see online supplementary data).
We next examined the effect of religiosity on secular cultural engagement, making the assumption that religious identification, beliefs, and behavior are often laid down in childhood and can be treated as causally-prior to young adult sociality. We use the following as dependent variables: -whether the respondent reported involvement with "youth culture", defined as having listened to records, going dancing, drinking, or visiting a cafe, nightclub, jazz or rock concert; -the extent of the respondent's secular engagement the previous Sunday, using a count of secular activities (reading books, reading newspapers, going to the cinema, listening to the radio, going for a walk, playing indoor games, watching TV, doing domestic work, doing paid work, homework, going visiting, playing sport, mending, doing craft, engagement in youth culture, practicing music, and going motoring or mending a car).
In the YRC data set, individuals are nested within towns. For the models of youth cultural engagement, however, likelihood ratio tests of random intercept versus single-level models justified a single-level (rather than multilevel) specification at the 5 percent level of significance. We therefore use nested logistic regression models, with results presented in Table 1, to examine the additional impact of potential confounders, calculating robust standard errors to take account of clustering within towns. 5 In the first model, we include age, gender, marital status, occupational status, religious affiliation, religious attendance, and belief as explanatory variables. Religious affiliation is captured by a variable for Anglican, Nonconformist, Catholic, other, or no reported adherence. For attendance, we use the measure of how often respondents reported attending church apart from weddings, christenings, and funerals as a rule. Respondents are "frequent attenders" if reporting attending at least weekly, which 29 percent do. We also use a measure of belief in God. In the second, we add a variable for whether the respondent still lives at home to account for direct parental influence. In the third, we add a term for whether the respondent reports attending church less often than when they were about 12. Church attendance predicts being significantly less likely to participate in youth culture (although in the final model, where addition of the change in church attendance variable reduces sample size significantly, p = .058). Women are less likely to take part. Regarding occupation, the unskilled are considerably more likely to participate (albeit at only the 10 percent level of significance in model 3), while students are considerably and significantly less likely to do so. These differences appear consistent with Todd's account of "working-class culture," although these were not young labor aristocrats but, as written-in data indicate, primarily laborers, porters, cleaners, and warehouse workers. Catholics and members of other religious minorities are more likely to take part, although differences are only significant at the 10 percent level for members of other religions in models 1 and 2. Catholics were considered socially-different in the 1950s, often of recent Irish origin, and were not generally Sabbatarian (see, for example, Field 2014). Of the "other" group, about a quarter were Jewish, the rest reporting "don't know," with very small numbers of members of other faith communities. It is plausible that these "other" adherents mostly comprised immigrants for whom religion and "going out" were important for social integration (the questionnaire did not record country of birth). Otherwise, Nonconformists and Nones are not significantly different from Anglicans. Belief in God has no effect on youth cultural participation across these specifications.
We extended the first basic model to account for living at home in model 2. A large negative effect of living with parents was found, suggestive of an individualization effect on youth leisure, while the negative effect of church attendance remains significant. In the final model, those who reported currently attending church less often than when aged 12 were more likely to participate in youth culture, with a coefficient larger than that for any other variable except for being married. Engagement in youth culture may well be subject to particular reporting biases whereby those engaged in it retrospectively judge that they also chose to reject churchgoing. It is also plausible that a latent "modern values" orientation drove both rebellion against church attendance and cultivation of new taste. Nevertheless, this association between reporting attending church less often compared with late childhood, and being more likely to participate in youth culture, is a novel finding. We argue that it captures a direct substitution effect between the church and milk bar or dance hall. 6 While the third model performs less well in terms of classification, the AIC and BIC are lower than for models 1 and 2 and serves as our preferred model of youth cultural participation.
We then investigated whether the strength of the effect of church attendance on secular engagement depends on local context, making use of the fortuitous survey design whereby respondents were nested within towns. We created an additive scale of Sunday engagement by scoring respondents from 0 to 8 depending on their number of secular activities-a nine-point scale we treat as continuous. We compared a null single-level model with a null two-level model (not reported), rejecting the hypothesis that the data had a single-level structure at the 1 percent level. Model 2.1 accordingly reports results of a random intercept model. We see no differences in secular engagement by religious affiliation, but there is a significant negative effect of frequent church attendance at the 1 percent level of significance. This is perhaps due to church attendance crowding out secular time use, notably contrasting with Katz-Gerro and Jaeger's (2012) important finding using a more contemporary source in the shape of the 2007 ISSP, that attendance and voracity of cultural participation are complementary.
We then specify random slopes models, with results presented in Table 2, to test whether the effect of frequent church attendance differs depending on context. The covariance between intercept and slope for frequent church attendance is negative and significant at the 1 percent level: the higher the random effect for the intercept (indicating town engagement is generally higher), the larger the (negative) effect of church attendance. In other words, religious attendance appears to have a larger negative effect in areas where there is more to do, and a smaller negative effect where there is less. This result might well be picking up a social capital or class spillover effect ascribable to the socioeconomic composition and "class cultures" of those areas rather than a religious context effect; or it could be that this effect works through religiosity more broadly rather than attendance specifically. For robustness, we accordingly allow the coefficients for the socioeconomic variables to vary over towns via a further set of random slopes models run separately (models fail to converge when attempting to fit too many random slopes simultaneously).
Results for models 3-5 show that the effect of religious belief does not vary over towns; neither does that for professional/intermediate status; nor that for student status. This suggests    To summarize, in the 1957 survey, there is evidence that attendance at a place of worship and secular social and cultural engagement are negatively associated. A strong negative relationship is also found between change in attendance and engagement in youth culture. For secular engagement, the association between reporting frequent attendance and secular engagement is generally negative. Moreover, the negative effect of reporting frequent church attendance is significantly larger where overall engagement is higher, indicating greater tension between religious involvement and secular involvement where more secular activities are available. This supports the prediction that secular opportunities were in competition with religious involvement, and that as they expanded, it proved more difficult to maintain church attendance. It also supports the interpretation of modernization as the mechanism of secularization, specifically via secular competition (Gruber andHungerman 2008, Stolz 2010). For a stronger test of the relationship between secular cultural socialization and religiosity in adulthood, we turn to a more recent source.

CULTURAL SOCIALIZATION IN EARLY ADOLESCENCE AND ADULT RELIGIOSITY FOR THE 1933-1942 COHORT
Secondary data sources including measures of both religious and cultural socialization are relatively rare. The BSA and ISSP are comprehensive regarding the former, but less so the latter. Accordingly, in our third analytic stage, we investigate the association between cultural socialization in adolescence and religiosity in adulthood using the large sample of the 1933-1942 generation available in Taking Part: The National Survey of Culture, Leisure and Sport, 2005-2006and 2006 Part is commissioned by the UK government's Department for Digital, Culture, Media, and Sport to generate cultural participation indicators relevant to subsidized cultural institutions. Surveying adults aged 16-plus in England, it has run annually since [2005][2006]. Primarily concerned with cultural participation, it also includes measures of religious affiliation and practice. Respondents are regularly asked their religious affiliation along the lines of the YRC and BES question wording (what is your religion?), with response options 262 JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION comprising no religion; Christian (including Church of England, Catholic, Protestant, and all other Christian denominations); Buddhist; Hindu; Judaism; Muslim; Sikh; and other. (For our analyses, we screen out those responding "don't know.") Those reporting a religious affiliation are then asked,

Are you currently practising this religion?
Response options comprise yes and no. (Those reporting no religious affiliation, and those who do not know if they have an affiliation, were not asked their current practice.) In the first three waves, questions relating to cultural socialization were put to half of each sample: When you were growing up, how often did your parent(s) or other adult(s) take you to . . . museums or art galleries? theatre, dance or classical music performances? sites of historic interest? libraries?
Response options comprised never, less often than once a year, one or two times a year, less often than monthly but at least three or four times a year, at least monthly, and don't know. Scherger and Savage (2010) describe these measures as tapping "high" cultural capital and engagement with "legitimate culture." In the absence of items capturing everyday cultural participation, we use these as proxies for secular cultural engagement. Respondents were also asked: How much did they encourage you [in] . . . reading books that were not required for school or religious studies? drawing or doing painting, writing stories, poems, plays or music? taking part in sport? playing musical instrument(s), acting, dancing or singing?
Response options comprised "encouraged you a lot," "encouraged you a little," "didn't encourage you at all," and "don't know." Scherger and Savage (2010) note that it is not clear how far respondents were thinking of parents rather than other influential adults, but nevertheless judged it fair to assume so. While these items refer to more "everyday" activities conducted at home (reading, art, practicing music) or school (sport, drama, singing, and so on), the wording does mean these items relate more closely to respondents' parents' behavior rather than their own. It is also plausible that people who value their own autonomy more remember their parents' encouragement as less than it actually was. This is not necessarily a problem, so long as we bear in mind that the item may capture individualist attitudes as much as an objective assessment of parental behavior. A further concern is that the interviewer clarified "[b]y 'growing up' I mean when you were aged around 12 to 15" in 2005-2006, but "around 11 to 15" in 2006-2007. We assume that respondents interpreted this period to refer to early adolescence in general rather than a precise age interval, and judged that the benefit in terms of sample size justified pooling both waves. 8 While Scherger and Savage combine the items into a single index to measure cultural socialization (the Cronbach alpha for waves 1 and 2 is .76), we treat them separately, to capture youth cultural participation and encouragement as different but related phenomena. The Cronbach alpha for the participation in secular "legitimate culture" items is .70; that for parental encouragement in everyday cultural participation .67. We sum and scale the items to create cultural participation and parental encouragement scales; the pairwise correlation is .53.
We examined the relationship between cultural participation and religious type for the 1933-1942 cohort, differentiating those who actively practice, those who report an affiliation but do not currently practice (nominally-religious), and those who do not identify with a religion (the nonreligious). This allows testing of the hypothesis that cultural socialization is associated with religiosity in adulthood. The measures of cultural socialization are limited to "improving" forms, with no measures of listening to the radio or records, watching TV or going to the movies, reading comic books and so on; and our measure of religiosity is thin. Nevertheless, this source offers potential for understanding how sociocultural life when growing up related to later religiosity for this cohort.
For comparability with the YRC sample, we restrict analysis to White respondents (in the absence of suitable migration measures) living in English urban areas who were born between 1933 and 1942. Including urban-rural classification loses us the third survey wave; nevertheless, the sample reaches 1894, of which 11 percent is categorized as nonreligious, 55 percent nominallyreligious, and 34 percent actively-religious.
We then modeled religious type to examine the association between cultural socialization and adult religiosity, using a multinomial logit specification with the nominally-religious the reference category. Results are presented in Table 3. In this case, it is more plausible that religiosity is a choice variable in later life, with respondents having less control over their cultural socialization in early adolescence, though we remain circumspect regarding the direction of causality. We also control for age, gender, whether the respondent were partnered rather than single, highest education level (degree level/more advanced or some qualifications vs. no qualifications), housing tenure, occupational status (professional/managerial, intermediate, or never worked vs. routine), whether the respondent reports a limiting illness, and year of interview.
Results generally support established findings. Women are more likely to be actively-religious rather than nominally-religious, and less likely to be nonreligious (Walter and Davie 1998). Those with some qualifications rather than none are more likely to be actively-rather than nominallyreligious. Having a degree-level education is associated both with active rather than nominal religiosity, and nonreligion rather than nominal religiosity. Those of professional or managerial occupational status rather than routine status are more likely to be nonreligious than nominallyreligious.
Turning to our key explanatory variables, cultural participation in youth has no association with being actively-rather than nominally-religious, while parental encouragement of everyday cultural activity is positively associated with active rather than nominal religiosity. By contrast, cultural participation is positively associated with identifying as nonreligious rather than nominally-religious, while reported parental encouragement is negatively associated with the same. This suggests that more extensive secular cultural participation in early adolescence was associated with being more likely to adhere to a secular identity in later life. Given the very low rates of nonreligion in 1957, this supports an interpretation of secular cultural participation being associated with a shift from at least nominal religiosity in young adulthood to nonreligion in later life, although unfortunately we do not observe religiosity type when growing up in this survey. Nevertheless, this finding and interpretation is consistent with what we observe in the 1957 survey: secular social and cultural opportunities are associated with lower religiosity or religious practice.
After calculating marginal effects, as summarized in Figure 4, we find the predicted probability of being categorized as nonreligious rather than nominally-religious ranges from 9 percent for those with a score of 1 (equating to never attending any of the listed sites) on the cultural participation scale to 19 percent for those with a score of 5 (attending each at least monthly), with all other variables held at their means. Regarding parental encouragement, as summarized in Figure 5, the probability of being nonreligious in later adulthood rather than religious is predicted to be 18 percent for those with an average score of 0 and 6 percent for those with an average score of 2, where zero equates to not being encouraged at all to read, draw, play sport, or practice music, and 2 equates to being "encouraged a lot" for each of the four activities. Notably, these differences are similar in size to those associated with education: 9 percent of those with no qualifications are predicted to be categorized as nonreligious rather than nominally-religious, compared with 11 percent of those with some qualifications, and 19 percent for bachelor's degree-holders. These results indicate that cultural socialization in early adolescence has substantive significance for adult religious identity and practice. Since the midcentury, successive generations have stayed in school longer, and been given greater access to higher education, more opportunity to engage in secular cultural life with economic growth; and experienced less authoritarian parenting. These features have combined to encourage retreat from active religiosity. Moreover, the nature of cultural socialization appears to have differential effects on adult religiosity: one  channel via the child's active participation, and a second via parental encouragement. This is notable given that our indicators of cultural socialization are positively correlated. It is plausible that the "encouragement" variable inversely captures prioritization of self-reliance, or an ethic of self-direction, as much as it relates to cultural engagement. Furthermore, those perceiving themselves to have been independent relatively young may be relatively more inclined to report that parents were not involved with their social and cultural interests. Alternatively, parental encouragement of cultural attendance perhaps tended to be more participatory for those who grew up to be nonreligious, who in turn cultivated a variety of secular interests in childhood; and more regulatory, involving concerted inculcation of everyday cultural behavior, for those religious in later life. The former channel relates primarily to activity on the part of the child, albeit accompanied by adults; the latter to activity on the part of the parent.

DISCUSSION
Social scientists of religion tend to seek parsimonious accounts of religiosity and religious change applicable across contexts. To explicate mechanisms for downward cohort change, we examined a particular cohort in detail, using cross-sectional surveys fielded in their adolescence and early youth, and then in later life. This allowed testing of relationships between religiosity and cultural participation and cultural socialization, and interpretation of cohort effects as arising at least partly due to growing up during a particular time. It also enabled focus on specific mechanisms through which secularization occurs. We identified secular social and cultural participation and parental socialization strategy as relevant mechanisms. We additionally provide evidence for downward period effects on active religiosity associated with the 1960s and perhaps the 1970s, corroborating Brown (2001) and Tilley (2003), and contrasting with more recently established wisdom that secularization is generated primarily by cohort replacement. The shift away from frequent attendance for our cohort likely reflected a combination of expanding alternative opportunities, change in cultural socialization reflecting their parents' own preferences, and broader normative change reducing the cultural status of religion (Bruce 2011).
Looking beyond belief in God, trends in belief in heaven and hell demonstrate downward period effects for this cohort, likely a product of declining identification with the institutions cueing religious doctrine. Organizations are closely involved with the production of belief and identity, and without active ties to places of worship, religion becomes less salient and deprioritized. While reprioritization is unremarkable to those to whom it occurs, over several decades the result is a substantial reconfiguring of worldviews and of social life.
These findings also offer a further perspective on cultural reproduction, suggesting that basic values should be incorporated into relevant models. A body of evidence documents how socially advantaged parents "concertedly cultivate" their children's educational and social lives to cultivate their cognitive skills and social networks (Jaeger and Breen 2016;Lareau 2003). Cultural reproduction is also subject to parental preferences for children developing the capacity to choose well for themselves, so that they acquire cultural capital relevant to the contemporary world. Accordingly, we should rethink how parents transmit religiosity to their children less in terms of transmission failure and more in terms of what parents positively prioritize.
A potential response is that this merely pushes the explanandum back one stage. If the young were less likely to retain religiosity because they were increasingly given more freedom, why were their parents giving them a free choice, or the church community not inclined to make religious participation more compelling in the face of this freedom? Social class may account for important variation in deprioritization on the part of parents just as it accounts for variation in youth cultural engagement. Davie (2014) has discussed class percipiently in her expansive survey of religion in contemporary Britain, and it is plausible that weakening of religious transmission began earlier among the working classes. However, we also raise the possibility that religious decline may have been partly unanticipated: once independence was prioritized, enforcement became impossible. This also raises the question of the conditions under which religious involvement proves attractive given the emergence of youth culture, more permissive parenting, and secular opportunities. Were relevant items available on the large cohort studies, we could examine how both religious and cultural socialization are associated with adult religiosity using individual longitudinal data for this and later cohorts. If we find that relationships change for later cohorts-perhaps due to the religious environment becoming more secular and more diverse-that will allow a more precise understanding of the nature of cohort change qua pure generational change versus age-by-period effects.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article. Appendix S1.Religiosity, Secular Participation and Cultural Socialization: Online Supplementary Data Appendix.