Parental spiritual and religious beliefs and behaviour data collected from the Avon Longitudinal Study of Parents and Children, 2020 [version 1; peer review: 3 approved with reservations]

There are few studies that chart the ways in which the religious beliefs and practices of parents and their offspring vary over time. Even fewer can relate this to aspects of their physical and mental health or distinguish the different facets of the environment that may have influenced the development or loss of religious/spiritual belief and behaviours over time. This paper describes the recent data collection in the Avon Longitudinal Study of Parents and Children (ALSPAC) on the beliefs and behaviours of the study parents some 27-28 years after the first measures were collected. Questions that were previously administered to the mother and her partner on religion, spirituality, behaviours, and beliefs (RSBB) were repeated for the fourth time, together with enhanced data on RSBB. The new data are described and compared with previous responses. The most notable difference between the 9 year and the 2020 sweep was the increase of professed non-believers in both the mothers (17.5% vs 29.8%) and partners (31.9% vs. 45.3%). As expected, on each occasion study partners were less likely to acknowledge RSBB compared to the study mothers. In the latest sweep, respondents were less likely to be unsure if they believed and more likely to not believe. Responses to “Do you believe in God or a divine power?” in mothers ranged from 49.9% stating ‘yes’ antenatally to


Introduction
There is increasing evidence that Western populations are becoming increasingly secular with each new generation (e.g., Chaves, 2017;Iles-Caven et al., 2019;Office of National Statistics, 2012; The Pew Forum on Religion and Public Life, 2017). For example, the British Social Attitudes Survey (Curtice et al., 2019) has demonstrated a dramatic decline, between 1983 and 2018, in those with no religion (31% to 52%, respectively); identification with a Christian denomination (66% vs. 38%) and a steady increase in non-Christian beliefs (2% vs. 9%). Those who identified as being very or extremely religious remained similar (6% vs. 7%) but those stating they were extremely, or very non-religious increased from 14% in 1983 to 33% in 2018. Those who professed they had never believed in God rose from 13% to 26%. Among the younger generations particularly, there was an increased tolerance towards different religious or belief systems, including non-belief. At every age, more women than men were affiliated with a religion, believed in God, and attended religious services. These findings echo those found in the Avon Longitudinal Study of Parents and Children (ALSPAC) over time (Iles-Caven et al., 2019).
Evidence, mainly from cross-sectional studies, has shown associations between religious/spiritual belief and positive health outcomes. These are described in brief elsewhere (Iles-Caven et al., 2019). Nevertheless, definitive proof of causal consequences of belief is lacking. Any study designed to identify whether some aspect of religious/spiritual beliefs or behaviours (RSBB) affects physical or mental health must analyse data longitudinally (i.e., by identifying beliefs first and determining their relationships with subsequent aspects of health). In addition, it is also important to have information collected longitudinally on factors that may be confounders, moderators, or mediators. To this end, the Avon Longitudinal Study of Parents and Children (ALSPAC) received funding to enhance the data resource on the topic of RSBB in 2020, which will be analysed along with extensive additional data on potential mediators, moderators, confounders, and physical and mental health outcomes to be collected over the next two years.
This paper describes the RSBB data collected in 2020 from the parents enrolled ALSPAC. These questions were designed to enable linkage to other longitudinal data from the cohort on the environment, traumatic incidents, physical and mental health, and genetic background. It can be used for research into various aspects of the antecedents and consequences of RSBB, and changes over time. A separate paper will describe the RSBB data collected on the 28-29-year-old offspring in 2020.
Previous data on RSBB collected prenatally and on two subsequent occasions (at 5 and 9 years post-delivery) are described in detail elsewhere (Iles-Caven et al., 2019). The data showed strong sex differences (all P<0.001) regarding RSBB (e.g., 49.9% of women vs. 37% of men stated that they believed in God/a divine power; 28.6% of men vs 14.9% of women declared they were non-believers). Among the 6256 women and 2355 men who responded at all three time points, a slight increase over time in the proportion stating that they were non-believers and a small reduction in professed belief were apparent.

Materials and methods Participants
The ALSPAC survey was specifically designed to determine ways in which the individual's genotype combines with environmental pressures to influence health and development [Golding et al., 2001]. The study is geographically based in the south-west of England, centred around the city of Bristol and its surrounding rural and semi-urban areas, with a population of about 1 million. To capture as much valid information as possible, unbiased by knowledge of details of the characteristics of the baby, the study was designed to start as early in pregnancy as possible. All women resident in the area at the time they were pregnant were eligible, provided that their expected date of delivery lay between 1 st April 1991 and 31 st December 1992. In total, 14,541 pregnant mothers, resident in the area, were recruited into the ALSPAC study. From these pregnancies, there were a total of 14,676 fetuses and 14,062 live births. Of the children, 13,988 were still alive at 1 year of age. Mothers were considered enrolled if they had returned at least one questionnaire or attended a "Children in Focus" clinic by 19 th July 1999. At the age of 7 years, the study team reached out to eligible mothers who had not been included in the study previously and thus recruited additional families to boost the number of participants. As such, from the age of 7 the total sample number is 15,454 live births, resulting in 15,589 fetuses, of which 14,901 were alive at 1 year of age (Boyd et al., 2013;Fraser et al., 2013).
Following advice from the ALSPAC Ethics and Law Committee, partners were recruited into the study only if the mothers wished them to be included. Questionnaires were sent to the mother who then passed the questionnaire on to the partner with a separate pre-paid return envelope. This method meant that ALSPAC were unable to follow up or communicate directly with the partners (Birmingham, 2018;Fraser et al., 2013). Therefore, the numbers of partners' questionnaires returned were less than those received from the mothers. Around 75% of partners participated in the study. Partners were subsequentlyenrolled in their own right in 2010 (n=3000).
A detailed data dictionary on the study web pages, and a detailed proposal form for access to specified data are available.
Data have been collected from pregnancy onwards using a variety of methods: (a) self-completion questionnaires; (b) assays of biological samples; (c) hands-on examination of the subjects; (d) linkage to educational and health data on the individuals; (e) linkage of addresses to measures of geographic exposures; (f) information on schools attended with details of behaviour of the child and his/her parents completed by teachers and head teachers.
Since the offspring were aged 22 years, data have been collected and managed using REDCap electronic data capture tools hosted at the University of Bristol (Harris et al., 2009). REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies.
Previous data collection on religious/spiritual beliefs and behaviour (RSBB) The population of Avon comprises a predominantly protestant Christian population, with a lower rate of non-Christian religions (ALSPAC: 3.5% mothers; 3.5% partners) than the general UK population (6.2%). Whilst population-level religious affiliation data for England or Avon in 1991 are not available, we can illustrate the similarities between ALSPAC, the city of Bristol and England and Wales ten years later, when the national Census collected data on religion for the first time. Partners, along with the general Bristol population, were more likely to state they had no religion compared with the rest of the country ( Table 1).
The initial religious behaviour and belief questions used by ALSPAC were asked at three time points (antenatally, and at 5 and 9 years later) were identical and are described elsewhere (Iles-Caven et al., 2019). In brief, the items covered the following aspects of belief: (a) the participants' fundamental beliefs: 'Do you believe in God or some divine power?' which had three possible responses: 'yes; not sure; no'; (b) whether they felt that they had received or asked for assistance from such a power; (c) the type of religion the parent reported, almost all were Christians of various denominations -they were also asked how long they had had that particular faith; (d) the frequency with which they attended religious services; and (e) whether they had received help and/or support from members of their own and/or other religions. When the child was 9 years old, an additional question was asked of parents concerning whether they prayed 'even when not in trouble'. Enhanced data collection on RSBB The religion questions used at the three earlier time points were repeated in the 2020 sweep (see Table 3, questions C1-C6, C9, C10), and enhanced with additional questions (Table 3: C7, C8, C11-C24, highlighted in bold) to capture details of extrinsic/intrinsic behaviour, and interests in religious worship using the written word and/or radio, television or other electronic media. Most of these new questions comprised elements from well-validated, standardised scales recommended by an international workshop of RSBB/Health experts in August 2019 (see Acknowledgements) and are described below. In the 2020 sweep, both the parents and their offspring received identical RSBB questions.

The new measures
The Duke University Religion Scale (DUREL) (Koenig et al., 1997), a five-item measure of religious involvement was developed for use in large cross-sectional or longitudinal studies. It assesses organisational and non-organisational religious activity and intrinsic religiosity. The scale has high test-retest reliability (intra-class correlation = 0.91), high internal consistency (Cronbach's alpha's = 0.78-0.91); and high convergent validity with other religiosity measures (r's = 0.71-0.86). The DUREL has been used extensively (Koenig et al., 2011). The five questions that comprised this scale were split up in the questionnaire (see Table 3, questions C9, C11, C14-C16). As can be seen from Table 4, women were more likely than their partners to attend organised religious worship and to practice private worship (e.g., prayer). Questions C9 and C11 measure organisational activity; and questions C14-C16 measure intrinsic religious motivation when combined, and again the women scored higher (means 6.44 vs. 5.52).
Specific questions to elicit extrinsic and intrinsic religious motivation were included for the first time. Extrinsic individuals are more likely to exploit religion, e.g., to provide security and solace, for social reasons, status, and self-justification. Whereas intrinsic individuals aim to live their life according to the All non-Christian 3.5% 3.5% 4.5% 6.2% Table 2. Proportion (n) of enrolled parents who completed the religion questions in a) pregnancy (1991-2) and in b) 2020 by selected sociodemographic factors.

Partners 2020
Age of parents at the birth

C12. How often do you listen to/watch religious programming on the radio/ television/social media?
Daily/Several times/week/Several times/month/Occasionally/ Never/Please describe Y3110 4587 2144

C13. How often do you read religious related texts or publications (e.g. the Bible, the Koran, prayer book, Watchtower, The War Cry, The Friend, Spirituality & Health, Catholic Digest)
Daily/Several times/week/Several times/month/Occasionally/ Never/Please describe Y3120 4586 2147

C14. In my life, I experience the Presence of the Divine (e.g. God)
Definitely true of me/Tends to be true of me/Unsure/ Tends not to be true of me/Definitely not true of me/ Not applicable Y3130 4551 2137

C15. My religious beliefs are what really lie behind my whole approach to life
Definitely true of me/Tends to be true of me/Unsure/ Tends not to be true of me/Definitely not true of me/ Not applicable Y3140 4544 2135

C16. I try hard to carry my religion over into all other dealings in life.
Definitely true of me/Tends to be true of me/Unsure/ Tends not to be true of me/Definitely not true of me/ Not applicable Y3150 4530 2131 Table 4. Duke University Religion Scale (DUREL) derived variables.

Mothers Partners
Organised religion activity score   (1967), 14 questions had been selected and revised by Gorsuch & McPherson (1989) so that the questions could be answered by non-believers. Each question has a five-point scale ranging from 'strongly agree' to 'strongly disagree', with an additional 'not applicable' option. We used two of the extrinsically weighted items (see Table 3, questions C17 and C18).

C17. I attend a place of worship mainly because it helps me make friends:
Three questions (see Table 3, questions C19-C23) are from the well-validated Fetzer Brief Multi-Dimensional Measure of Religiosity/Spirituality for use in health research (BMMRS) (Fetzer Institute, 2003). The questions were chosen to enquire about religious/spiritual history: whether an individual has had a religious/spiritual experience that changed their life or experienced a significant gain or loss of faith and if so when. Note that the study team added a request for a free text description of the experience/gain/loss in faith after discussion with the workshop attendees. This request was worded in such a way that it was seen as optional by the respondents.
At the suggestion of Connie Svob (personal communication) question C24, "How important to you is religion or spirituality?" was included. This question has been shown to be highly predictive in a transgenerational longitudinal study of depressed and nondepressed probands and their offspring followed over 30 years (Anderson et al., 2021;Weissman et al., 2016). Future comparison with the ALSPAC data was thought to be of great value.
Questions C7 and C8 were devised by the study team and asked whether the participants were brought up in a particular faith and whether they had brought their own child(ren) up in a particular faith.
Questions C12 and C13 were devised by Golding & Iles-Caven for this sweep to elicit contemporary forms of private belief/worship such as radio, TV programmes and social media, and the reading of religious texts and periodicals.
In each of the questionnaires administered at the four time points (pregnancy; +5years; +9 years, and in 2020) the question (C6) concerning the duration of their current faith was included.
In pregnancy, the majority responded: 'all my life' and fewer than 5% responded <5 years. These responses were consistent over time and enable the study to identify a large proportion of the population for whom there are consistent responses throughout the time span (see below and Table 8a, Table 8b for examples). We believe that using this data to extrapolate backwards for this large group of the population is valid, especially when supported by earlier data. We can identify a large sub-group for whom data on RSBB will be able to be extrapolated throughout the life-course (Table 5). Table 6 shows the mothers' and partners' response rates to each question in the 2020 sweep. There were many differences evident between the sexes (at p<0.001 level). Women were more likely than men to believe in a divinity, participate in private and public worship, and to lead their lives according to their religious principles. Table 7a and Table 7b show the total numbers of each parent who answered the RSBB questions at any of the four time points of data collection, and Table 8a and Table 8b show only those parents who completed the questions at every sweep. Modern statistical techniques will allow imputation increasing statistical power for those for whom we have incomplete data.
In the 20 years since the questions were previously asked at 9 years on belief in God/a divine power, whether God/a divine power had helped them or if they would appeal for help from such powers, mothers seem to have become more certain of their beliefs, with the largest reduction in the 'not sure' categories. Very few stated they had only followed their current faith for less than five years (3% at 9y and 1.7% at 2020), but 16.2% (at 9y) vs 23.3% had followed their faith for more than 5 years (but not all their life) (Table 7a). Similar results were shown for their partners (Table 7b).
The most notable difference between the 9 year and the 2020 sweep was the increase of professed non-believers in both the mothers (17.5% vs 29.8%) (Table 7a) and partners (31.9% vs. 45.3%) (Table 7b).
For those mothers who responded to the same questions at all four time points, a dramatic increase in those professing non-belief can be noted (from 6.6% antenatally to 19.6% in 2020). However, at 5 and 9 years the corresponding figures were similar at 10.0% and 9.8% respectively. There was a steady decrease in the numbers of mothers stating that they would appeal for help when in trouble (Table 8a).
For those mothers answering the type of religious beliefs they had had at each time point, those stating 'none' were fairly consistent at the first three points (range 25.6% to 27.3%) and then rose to 40.4% in 2020. A corresponding decrease was especially notable in those who stated they belonged to the Church of England (54.3% antenatally to 41.2% in 2020) (Table 8a).

Strengths and limitations of the data
The strengths of these data include the large sample size, with almost 7000 participants having data available from the 2020 sweep. The participants are broadly representative of the general population in the area, at the time of recruitment, in terms of sex, ethnicity, and socio-economic status (Fraser et al., 2013).
The extensive data on mediators, moderators, confounders, and physical and mental health outcomes to be collected over the next two years will facilitate huge amounts of research.
A key limitation of the data is the lack of ethnic diversity. At the time of enrolment, the county of Avon was mainly Caucasian, therefore there were too few Black, Asian and Minority Ethnic (BAME) participants (<6% in total) to allow for detailed analysis by ethnic background. A further limitation is that, as with all longitudinal studies there is increasing attrition over time. For these study parents, the loss is due mainly to mortality, change of address, as well as of reluctance to stay involved in the study.

Data availability
ALSPAC data access is through a system of managed open access. The steps below highlight how to apply for access to the data included in this paper and all other ALSPAC data. Note that Table 3 in this paper gives the variable numbers for the religion data. 1. Please read the ALSPAC access policy which describes the process of accessing the data and biological samples in detail, and outlines the costs associated with doing so.
2. You may also find it useful to browse our fully searchable research proposals database, which lists all research projects that have been approved since April 2011.
3. Please submit your research proposal for consideration by the ALSPAC Executive Committee using the online process. You will receive a response within 10 working days to advise you whether your proposal has been approved.
If you have any questions about accessing data, please email: alspac-data@bristol.ac.uk (data) or bbl-info@bristol.ac.uk (samples).
The ALSPAC data management plan describes in detail the policy regarding data sharing, which is through a system of managed open access.

Ethical approval and consent
Prior to commencement of the study, approval was sought from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees (Birmingham, 2018). Informed consent for the use of data collected via questionnaires and clinics was obtained from participants following the recommendations of the ALSPAC Ethics and Law Committee at the time. Questionnaires were completed in the participants own home and return of the questionnaires was taken as continued consent for their data to be included in the study. Full details of the approvals obtained are available from the study (http://www. bristol.ac.uk/alspac/researchers/research-ethics/). Study members have the right to withdraw their consent for elements of the study or from the study entirely at any time.
Yes required to explore the extent of selection bias and whether this may impact the subsequent conclusions drawn (Griffith et al., 2020;Smith, 2020). Additionally, other aspects of RSBB-belief in God/a divine power and religious affiliation-were frequently associated with continued participation in unadjusted analyses, but these associations were often attenuated to null after adjusting for potential confounders. This suggests that these RSBB factors may be 'missing at random' (MAR, i.e., adjusting for confounders included here, religious belief and affiliation are not associated with selection). These RSBB factors may not therefore be MNAR, and so are at less risk of causing selection bias; however, to avoid these RSBB variables resulting in selection bias, these demographic and socioeconomic confounders would also need to be controlled for in subsequent analyses. Although some studies have suggested that association with participation would have to be fairly severe for the subsequent bias to cause concern (Cornish et al., 2021;Pizzi et al., 2011), this needs to be explored on a case-by-case basis. It is hoped that highlighting the potential for selection bias in ALSPAC's RSBB data will inform future users of the resource." We have therefore added a couple of sentences to this effect in our revised version of the Data Note (to the end of Strengths and Difficulties). Morgan et al., 2022 reference has been added to the reference list.
paper (Piedmont, Fox, and Toscano 2020) 1 is an example of a cross-sectional study, using SEM analyses to identify the causal precedence of spiritual variables on mental health functioning. While not a longitudinal study, it nonetheless is able to evaluate the relative merits of different causal models. Only a research design can examine causality, whether one is conducting a cross-sectional or longitudinal study. Without an experimental manipulation and random assignment of subjects, any study becomes correlational in nature. Such sweeping statements need to be avoided. This longitudinal study has nothing to contribute to the examination of any causal effects, it is merely descriptive in nature. I am referencing some papers that do examine causal issues using cross-sectional data.
I am completely underwhelmed by the selection of measures employed in this study. Many of the study's own measures are simply single item scales that require only a "yes/no" response. Such items are quite useless because they contain so much error. It is unclear what a "yes" or "no" answer really means because no normative data is given. If someone says that they are spiritual, how spiritual are they? Do we just lump all the "yes" responders together in a single group? Not very informative. Further, without controlling for potential mediators, in this case factors like economic status, ethnicity, and related constructs, it is not clear what is uniquely important about responses to these items. It was noted that gender was a factor influencing responses, but no effort was made at explaining why this is.
No effort was made to determine what influence gender had on score changes over time. This would be important to know. What else is going on in the lives of these people that may be influencing their responses. England has been going through lots of economic and political upheaval over the course of this study (e.g., Brexit, the new "leveling up" policies, etc.), all of which may impact these scores. Some control for potential mediators and moderators of responses needs to be given, especially when only single items are used. More on this below.

2.
The other scales used in this study are also quite problematic. The DUREL scale is widely used in medical research, but because all the items are positively phrased, the scale is contaminated by acquiescence bias. One reason why this scale correlates well with other religious/spiritual (R/S) scales is because many of those are similarly biased (like those on the BMMRS scale). This issue needs to be acknowledged as a relevant limitation to the study. The other scale used, the Fetzer BMMRS is noted by the authors as being, "from the well-validated Fetzer Brief Multi-Dimensional Measure of Religiosity/Spirituality." I am not aware that the BMMRS is well-validated. The measure is NOT a scale in any sense of the term. It is merely a collection of approximately 32 items taken across 12 different constructs. There is no scalar information about this instrument and is best considered an epidemiological screener. Without any normative information about responses to these items, it is difficult to interpret their meaning. However, this is a fixable issue whereas the manual to this instrument does give mean level data for each item. This information can be used as a comparison to the current data. Finally, it is not clear to my why such out of date scales are used, like the Intrinsic-Extrinsic Religiosity Scale? This is a very old scale with problems to it. There are so many newer, and better scale available that can be more useful, like the Assessment of Spirituality and Religious Sentiments (ASPIRES) scale or the Numinous Motivation inventory (see www.centerforprofessionalstudies.com for information about both). One the one hand, continuity is important in a longitudinal study, but it would be helpful if some better instruments are also included.

3.
The data are just descriptive in nature, and clearly show a decrease in R/S salience for the sample over time. Why this is, is not discussed or examined. Similar data trends have been found by other organizations, like the Pew Foundation in the US which has documented the rise of the non-affiliated. It would be helpful conceptually for readers if connections with these other studies are made. Where data analyses are conducted no real statistical information is given. First, providing p-values as way of demonstrating effect is problematic; p-values are NOT measures of effect size and should never be used as a means of presenting results. I would like to know, what exact statistical tests were conducted in comparing mothers and their partners and what the observed values for those tests were. In this manner, readers can get a better sense of the magnitude of effects being presented.
Given the very large sample sizes noted here, it is completely expected that any statistical test will generate a very small p-value. Even small effects can generate small p-values. As we know, when a two-tailed test is used, as sample size increases the probability of rejecting the null moves to 1.0. The authors also need to comment on the actual effect sizes noted in their analyses. How much of a change is really going on here? I would suggest that the authors drop the p-values and give appropriate effect size estimates.
cross-cultural research.
I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Author Response 23 Aug 2022
Yasmin Iles-Caven, University of Bristol, Bristol, UK We thank you for your helpful and insightful comments on this paper. Your point 1. Firstly, this is purely a descriptive paper (data note) of the data available on RSBB collected in 2020 and the ability to link this to a rich phenotypic dataset collected over a 30-year period including RSBB data collected antenatally and at 5 and 9 years postpartum. It is not a research article.
We apologise for our sweeping statement concerning longitudinal studies. We agree crosssectional studies have their value, but in order to examine trajectories of poor mental and/or physical health and whether RSBB has any influence or causality on the (non-) development of disease, longitudinal studies are required (e.g. non-genetic inheritance via grandmaternal smoking in pregnancy and RSBB in granddaughters (Golding et al., 2022); or RSBB influences on the initiation or giving up of habits such as alcohol, smoking and drug use as well as aspects of diet, such factors may be on the causal pathway between RSBB and aspects of mental or physical health. Such features are most easily interpreted when it is clear as to the sequence of exposures, mediators and outcomes.

Selection of measures.
The RSBB data collected within ALSPAC questionnaires only comprises one section among many others including socioeconomic status, health, measures of anxiety and depression, personality, social interactions, locus of control, and traumatic experiences. We are therefore not limited in our exploration of RSBB in itself.
To repeat, this is a description of RSBB data collected in this wave. Data Notes, as defined by Wellcome Open Research do not involve detailed analyses. They are meant to describe the relevant data available and the ways in which the interested researcher may access it.
Your point 2. We have omitted the phrase 'well-validated' from the paper. We are sorry you have been less than enthusiastic about some of the measures used. We relied on recommendations from experts in the field. This paper merely describes what is available to the interested researcher.

Your point 3 Statistical analyses.
We have compared proportions using chi-squared, and quoted the corresponding P value. We do not claim that this is equivalent to an effect size. Rather it is there as an indicator of a difference that may not be explained by chance. The data provided in the tables is such that the interested researcher can calculate unadjusted effect sizes if required. Again, we reiterate this is purely a descriptive paper. ALSPAC welcomes collaboration (see data availability section).

Your point 4.
There are a number of researchers working with these data, but more are always welcomed. Because of the broad wealth of data collected by ALSPAC, there are a large number of questions that we hope that interested researchers may address.
the study. This is really a huge amount of loss, and it calls for some procedures, perhaps Heckman correction models or other ways of attending to the ways in which the sample is different from the population from which it came. Without this, and a much more major acknowledgement of the magnitude of the attrition, the implications of the language in the text and abstract, of a change over time in RSBB is quite misleading.