Smoking and Religion: Untangling Associations Using English Survey Data

This paper explores the association of religion with smoking. In recent years, there has been increased interest in this topic (Chitwood et al. 2008; Anthony et al. 2013; Ford and Hill 2012; Garrusi and Nakhaee 2012; Karlsen et al. 2012; Karlsen and Nazroo 2010). Patterns of smoking are known to vary significantly by religion but less is known about how this association is affected by other factors or how, if at all, it differs between younger and older people. We address this gap in knowledge through a focussed case study of England, where recent falls in smoking prevalence have taken place alongside significant changes in religious affiliation but where the association between religion and smoking has received little attention. Understanding more about this association is potentially significant for the design of effective tobacco control interventions that take account of the specific needs and characteristics of religious groups while also reflecting the distinctiveness of populations of different ages.

Our motivation for focussing on the association between religion and smoking in England is twofold. First, as noted, recent years have seen the size of different religious groups in the England change markedly (ONS 2012). Between 2001 and 2011, the number of people identifying as Christian decreased by 13% (from 72 to 59%), while those who reported having no religion increased by 10% (from 15% in 2001 to 25% in 2011). Among the other main religions, the population of Muslims increased the most, from 3% in 2001 to 4.8% in 2011. To a significant extent, these changes reflect underlying demographics, hence our interest in comparing adult and youth populations in terms of smoking prevalence.

A second motivation relates to more theoretical and theological concerns. The major world religions have positions that are largely opposed to smoking (Khayat 2000; Garrusi and Nakhaee 2012). For example, within Christianity, Biblical interpretations condemn smoking as bodily pollution and an unnatural vice that runs counter to Christian values of temperance and moderation. Equally, Muslim perspectives are marked by leading clerics urging abstinence and pronouncing a fatwa against tobacco on the grounds of its potential to cause ill-health and offend Koranic injunctions to ensure personal health and the health of others. We ask whether such positions are evidenced in differential smoking prevalences between religious groups in the predominantly secular context of contemporary England.

Past studies of religion and smoking have generally focused on measures of religiosity, that is the depth or extent of religious belief. This body of work has very clearly pointed to higher levels of smoking among people who do not profess any religion and conversely lower smoking prevalences among religious people. Such associations have been found across much of the world, implicating Christian denominations, different forms of Islam and Eastern faiths. Research in the USA, for example, has linked greater religiosity with lower levels of smoking among both adults (Whooley et al. 2002; Garcia et al. 2013; Hayward et al. 2016; Bowie et al. 2017) and younger people (Alexander et al. 2016; Nonnemaker et al. 2006; Amey et al. 1996; Wallace and Forman 1998). Elsewhere similar conclusions have been drawn for young people in Central America and the Dominican Republic (Chen et al. 2004), Hungary (Kovacs et al. 2011), Switzerland (Becker et al. 2015), Iran (Ameri et al. 2016) and Jordan (Alzyoud et al. 2015), for adults in Brazil (Martinez et al. 2017), mainland China (Wang et al. 2015; Wang and Jang 2016), Zambia and Malawi (Pampel 2005) and South Africa (Prinsloo et al. 2008), and for pregnant women in San Luis, Brazil (Barbosa et al. 2015). Analogously, adults in South Korea have been found to be more likely to quit smoking if they are religious (Myung et al. 2012).

In terms of identification or affiliation with particular religions, a US study has suggested that tobacco use among Muslim college students is lower than that for non-Muslims (Ahmed et al. 2014). This finding is sustained for adult populations in the former Soviet Union (Pomerleau et al. 2004 and for pregnant Muslim women in Thailand who smoke less than pregnant Buddhist women (Assanangkornchai et al. 2017). Wang et al. (2015) link lower levels of smoking in China to more religiously observant Muslims. Ghouri et al. (2006), in contrast, link the Muslim religion to high and rising rates of smoking through a focus on national levels of smoking in ‘predominantly Muslim’ countries. Lakew and Haile (2015) find that Muslims (and Catholics) in Ethiopia smoke more than the people from the dominant Coptic Orthodox community. Chen (2014), in a Taiwanese study focussed on Eastern Religions, has made the important point that links between religious affiliation and smoking may not be robust to confounding.

Evidence focussed on the association between smoking and religion in England is sparse. It has drawn substantially on localised survey research in the West of Scotland highlighting the interplay of ethnicity, religion and life stage (Williams et al. 1994; Williams and Shams 1998; Bradby and Williams 2006). Youthful abstinent behaviour erodes earlier among non-Muslims and a higher prevalence of ever-smoking is evident among young Christians and ‘Others’ and a lower prevalence among young Muslims. This assessment broadly tallies with the conclusions of Anthony et al. (2013), who also used local survey data, showing lower ever-smoking and current-smoking prevalences in Leicester, England, among Muslims as compared to Christians and (more so) those who reported no religion. These differences in prevalence may reflect underlying beliefs: Francis (2008) suggests that, in England, 34% of young people not professing a religion believe that it is wrong to smoke compared to 39% of Christians and 54% of Muslims.

In the light of this current literature, we identify the need for national scale research that considers the impact of religion on smoking behaviour, contrasting youth and adult populations and controlling for potential confounding variables, particularly ethnicity.

We used a secondary analysis approach with a cross-sectional research design contrasting data on youth and adult smoking behaviour drawn from the same source over a common time period.

In contrast to previous UK studies that have focussed on specific locations (the West of Scotland or Leicester), the present study has provided national evidence for England comparing three measures of smoking behaviour between youths and adults and highlighting the extent of association with religion while controlling for other relevant factors. Our findings respond to concerns about confounding articulated by Chen (2014) in a very different national context and develop and enhance suggestions by Bradby and Williams (2006) and Anthony et al. (2013) about the interplay of ethnicity, religion and socio-economic status in understanding smoking behaviour.

Initial indications from simple cross-tabulations suggested that Muslim youth are far less likely to be current smokers than their Christian or no-religion counterparts. This confirms evidence from the US, China and the former Soviet Union (Ahmed et al. 2014; Pomerleau et al. 2004; Wang et al. 2015). Moreover, Muslim youth are less likely to be current smokers in comparison with Muslim adults; this discrepancy is not evident for Christians, sustaining Frances’ (2008) argument that Muslim youth are particularly likely to deem smoking to be wrong. Muslim adults and youth also stand out as being less likely to have ever smoked. These simple associations suggest that the research in the West of Scotland pointing to abstinence persisting longer among Muslim youth (Bradby and Williams 2006) may have wider relevance to England.

Across both adult and youth groups, simple descriptive analyses pointed to smoking (both ever and current) being highest among people professing no religion. This confirms that the widely held global equation of lower religiosity with higher levels of smoking applies to England and adds to knowledge by demonstrating that this finding is relevant beyond adolescent English populations (Francis 2008) and the City of Leicester (Anthony et al. 2013). This position is sustained our simplest models, indicating that it is not an artefact of age or sex. Both Christians and Muslims appear to be less likely to smoke than people with no religion with Muslims generally being particularly averse. This initial finding gives strength to suggestions that religion may somehow protect against smoking, perhaps by binding its adherents in social communities with shared norms of abstinence and obedience to recommendations by leaders, as well as scope for mutual support (Gryczynski and Ward 2011; Mason et al. 2012). Wray-Lake et al. (2012) in national repeated cross-sectional study of US adolescents has shown how such social capital constructs have independent negative associations with smoking.

Our analysis of quitting challenges this conclusion. If religion points towards a lower smoking prevalence, we would expect that it might also point to higher levels of smoking cessation. While this is the case with Christianity, it is not evident with the Muslim religion. In a simple cross-tabulation, Muslim adults are less likely to quit smoking than adults declaring that they do not identify with any religion. It is well established that smoking cessation and continued smoking are distinct processes (Hyland et al. 2006) so it would be entirely possible for religion to simultaneously assist individuals in stopping smoking initiation while also hindering quitting. Why it might work differentially for Christians and Muslims is unclear. Croucher and Choudhury (2007) offer potential insights with their suggestion, based on qualitative work, that continued smoking among Muslims reflects anxieties about harassment, low-status employment, and the long shadow of migration experiences. Though these factors are undoubtedly significant for Muslims, they are not, however, exclusive to Muslims. Potentially more pertinent is the possibility that smoking provides a counter to the stresses and strains of being a minority religion. Padela and Curlin (2013) have developed this argument in the US context in relation to a range of health conditions and it draws strength from established theories about relative inequality and health behaviour (Jen et al. 2009). To unpack these possibilities, we need to turn to our modelling analyses.

Our models add to knowledge by demonstrating that, in England, our initial finding of an association between smoking and the presence of a religious affiliation is generally robust to confounders and moderators only in the case of Christianity. This conclusion suggests that the hypotheses linking religious social capital to smoking cited above may be relevant in England within a Christian context. The association with Christianity applies to current-smoking by both adults and youth and to adult ever-smoking. With our youth sample we were, however, unable to demonstrate a statistically significant association between ever-smoking and a Christian affiliation. In contrast, the initial associations linking the Muslim religion to low levels of smoking and also paradoxically to low levels of quitting are not robust to the impact of other relevant variables. We are thus unable to sustain the relative inequality/minority religion hypothesis. Ethnicity and, particularly, socio-economic factors trump the effect of religious affiliation on smoking prevalence for Muslims in England. Socio-economic status also over-rides any suggestion that Muslims are less likely to quit smoking. It is also clear that religious social capital is, at least in England, not a significant factor in smoking cessation, either for Muslims or Christians. This conclusion echoes that found in the very different context of Thailand by Yong et al. (Yong et al. 2009, 2013) who have emphasised that religion and religious authority are both potentially important in driving smoking cessation but neither ensure success, particularly in secular societies.

Our study has strengths and limitations. We present evidence from linked runs of a well-found long-established routine national survey using appropriate statistical methods and standard measures of smoking behaviour. However, despite merging 5 years of data, our sample size remained relatively small and led us to employ broad and potentially confusing ethnic categorisations. ‘South Asian’ and ‘Black’, for example, cover very diverse communities and there is no clear congruency between our ethnic and religious categorisations. Equally, we were unable to separate out different forms of Christianity or Islam. Small numbers are also evident in our youth samples though potential shortcomings have been addressed through a bootstrapped analysis. A further limitation is, of course, the cross-sectional design of our study. As a consequence, we do not seek to draw conclusions regarding the causal nature of the association between religion and smoking.

The potential implications of our study concern both future research and the practice of tobacco control. An enhanced qualitative component to future research will be essential if we are to explore more fully the relationship between religion and smoking. In-depth information drawing on interviews, ethnographic observation and the voices of different religious groups (and the non-religious) will be needed to draw out the extent to which people understand the impact of religion on smoking initiation, cessation and maintenance, and its interaction with other factors. Equally quantitative longitudinal studies are also needed to trace the interplay of religion, smoking and other confounding and moderating factors over time.

In terms of tobacco control, our results raise issues for faith-based health interventions. Evidence primarily from the USA but also from the Far East and Muslim countries has been hopeful but equivocal about the effectiveness of such measures (Campbell et al. 2007; Schoenberg et al. 2016; Ismail et al. 2016; Byron et al. 2015; Elkalmi et al. 2016). Our research points to the need for faith-based interventions to move beyond baseline prevalences to understand how religion interacts with other factors that may be more important in driving smoking behaviour, notably socio-economic disadvantage and ethnicity. We also underline the importance of targeting those without a religious faith and recognising that the association between smoking and religion is not uniform across all faiths. The potential for effective faith-based interventions in England would appear to be greatest for interventions based around Christian congregations drawing strength from the independent association of Christianity with lower smoking prevalences. There is, however, potential for all faiths provided it is recognised that religion is both more complex in terms of its role as an epidemiological construct (Levin 1996) and more complex than is commonly understood in health promotion (Liu et al. 2016). As Ward et al. (2014) note the link between religion and smoking can vary significantly across different religious communities and must (Schoenberg et al. 2015) be deployed with careful attention to community norms if it is to be effective.