Abstract
This article explores the theme of religious socialisation in an environment marked by modernity. It has three goals: first, to analyse the changes in religious socialisation across age groups; second, to analyse the impact of religious socialisation on religiosity; and third, to analyse the impact of religiosity on religious socialisation. Based on quantitative data about the Lisbon Metropolitan Area, results show the following: The level of religious socialisation increases with age, but young respondents have been significantly less socialised than older age groups; the level of religious education provided by respondents to their children increases with age; socialisation and religiosity are biunivocal, mutually reinforcing. In short, this article reinforces the theory of cohort replacement, which states that religiosity decreases in younger groups due to the decrease in religious socialisation over time.
Introduction
The last three decades of research have seen a strong interest in the problem of religious transmission. This process has been analysed mainly from the perspective of crisis or decline. Indeed, this erosion has been interpreted as part of the 1960s religious crisis in the North Atlantic region (McLeod, 2010). In the North Atlantic region, the social sciences have constructed their lexicon to identify these dynamics involving the family, the school, and religious institutions. In the case of the family, this new lexicon aimed to interpret the new generations’ situation. Although they had experienced a process of religious socialisation in a family environment, they were not now finding reasons and resources to continue this role of reproduction of religious identity. In this itinerary, the new lexicon seeks to disentangle itself from the idea of transmission based on the premises of reproduction and control. Studies on religious socialisation have developed models that are sensitive to multiple trajectories, avoiding inscribing the processes of religious socialisation into a linear teleology of decline. These multiple trajectories are part of the ‘situated pluralism’ that characterises the dynamics of religious identification (Voyé, 2003).
In any case, as observed in the first section of this article, studies tend to show that families play an essential role in shaping religious identities. Yet, studies focus on national data disregarding more diversified and complex scenarios such as metropolis with all the complex dynamics of late modernity involved. At the same time, there is a lack of recent studies in the European context that analyse religious socialisation, particularly in the metropolitan context. Therefore, to fill this gap, this study analysed data from Portugal, from its region most marked by the dynamics of social modernity: the Lisbon Metropolitan Area (LMA). In fact, Portugal is an interesting country to test the scope of secularisation, as it is between the process of modernisation and the strong influence of Catholicism for many centuries that still remains. Based on data from the project ‘Religious identities in the Area of Greater Lisbon’ and in the wake of other articles (e.g. Coutinho, 2022, 2023; Franca, 2022), this article analysed three aspects: the changes of religious socialisation across age groups, the impact of religious socialisation on religiosity, and the impact of religiosity on religious socialisation.
Religious transmission and socialisation: the lexicon of a crisis
The ‘crisis’ model, as a tool of analysis, was mainly developed by Hervieu-Léger. The sociologist proposed an interpretative model centred on the current crisis of the processes of transmission of belief, that is, the erosion of the devices of production of a belief lineage. This interpretative path aimed to interpret the phenomenon of ‘amnesia’ characteristic of modernity and understanding emerging religious recompositions (Hervieu-Léger, 1998, 1999). This section proposes mapping the concepts and interpretative models accompanying this phenomenon. In a way, it is an overview of the crisis lexicon.
In some national contexts, study results suggest that only in about half of the cases is parental religiosity effectively transmitted. However, the absence of religion is almost always transmitted (Voas and Crockett, 2005). Some macroscopic analyses attempt to demonstrate that this phenomenon is documentable in all industrialised societies (Bruce, 2011). However, it is essential to note that there are significant differences when comparing majority and minority identities (Kühle, 2012). Steve Bruce’s perspective continues the thesis of Bryan Wilson, for whom the modern process of ‘societalisation’ significantly impacts religion. On the one hand, the emancipation of individual trajectories from the contexts of social reproduction causes an erosion of the belief heritage. On the other hand, diminishing individuals’ contact with religious institutions, many of their operations in the daily life of individuals are replaced by different forms of secular agency (Wilson, 2007).
This is precisely the kind of treatment that religious socialisation gets within the framework of secularisation theories. Karel Dobbelaere (2011), for example, pays particular attention to the processes of ‘individual secularisation’. In this framework, individuals position themselves idiosyncratically in the private sphere, fabricating their religiosities, in the framework of the production of their subjectivity, to the detriment of adherence to communal forms. Moreover, this subjectivisation is intensified by the fact that individuals experience constraints in applying their religious beliefs to other subsystems that are part of daily life (Campiche, 2010; Day, 2016; Monnot, 2013).
However, the broadening of the contexts observed and the diversification of the analysis techniques have allowed for the identification of less linear social dynamics. By comparing the data from the ‘European Values Study’ of 1981, 1990, and 1999, Pierre Bréchon (2004) looked for evidence of the persistence of Christian heritage in Europe among younger generations. By compiling six indicators, he found that only 15% of respondents aged 18–29 were included in the group of those whose religious representations strongly influenced their world views and lifestyles (as opposed to 38% in the 60 + age group). Within the same analytical framework, it was observed that the structuring of religious identity was very dependent on family socialisation. Of those with a solid religious identity, 90% had participated in worship at least once a month when they were 12 (compared with 40% of those with a weak religious identity). Overall, 52% of the 18–29 age group said they had participated in worship once a month at age 12, compared to 75% of the older generation.
In this interpretative context, the very concept of transmission was relativised and often replaced by others. Collet-Sabe (2007) proposed replacing the transmission model with the ‘complex socialisation model’. The model is based on four observations: the processes of primary and secondary socialisation are now developed in a framework of various referential worlds – ‘All the worlds share contingency, relativity and uncertainty’; the construction of subjectivity is part of the social frameworks of modern reflexivity; affective socialisation favours a regime of validation between peers; the new place of socialisation is the ‘world of daily life’, marked by the dynamics of globalisation. In this research context, the study of beliefs has undergone a shift in perspective: The individual and their group of beliefs are taken as a reference rather than an objective orthodoxy. Its relationship with the belief heritage is somewhat indeterminate. As some studies suggest, many believe ‘vaguely’ (Hunt, 2015).
In the transmission model, beliefs have received precedence. In the complex socialisation model, beliefs are (re)constructed in the reflective and affective activity framework. Therefore, this literature highlights that adolescents and young people have become more active in creating their universe of religious references (Arweck and Jackson, 2014; Bengtson, 2013). Some qualitative studies have shown that younger people are distanced from propositional belief systems and are more interested in performance of religiosity (Vincett et al., 2012). Experience is more important than content, as has already been identified in the Baby Boomer generation. The socialisation of experiences materialises in constructing a cognitive and affective framework that guides the subject in their ‘navigating in life and the social network’ (Hervieu-Léger, 1999).
In this line of research, one tends to value the pluriform and distributed dimensions of religious socialisation processes. Admittedly, this process may occur, most prominently, between childhood and the onset of young adulthood. Nevertheless, it may intensify in other stages of life. This perspective on extended religious socialisation places the individual and their context at the centre rather than as external agents influencing the process (Klingenberg and Sjö, 2019).
Considering the diversity of contexts, interpretative models have become more available for thinking about the complexity inherent in the processes of transmission and socialisation. On the one hand, these processes cannot be restricted to the development period of the young adult since they continue throughout their whole life. On the other hand, in the case of parental socialisation, it is highlighted that the processes may have different directions (Sherkat, 2003; Thomson and Davignon, 2017). The most recent literature has identified three distinct morphologies: modelling, guiding, and relating (McPhail, 2019): (a) In the first, parents model religiosity that their children imitate; (b) in the second, parents guide their children towards religious institutions, environments, and social networks where religion is reinforced; (c) in the third, positive affective relationships between parents and children facilitate religious inheritance.
In the case of the ‘guidance’ model – in which parental responsibility values the referral of children to social environments and institutions that reinforce the logic of family religious socialisation – it is necessary to bear in mind that such orientation is strongly conditioned by processes of ‘cultural socialisation’ (McAndrew and Richards, 2020). Even in the case of parents who are active in the religious socialisation of their children, a significant part of this group values autonomy over obedience to the received tradition. This is demonstrated by research conducted in strongly secularised national contexts such as the Netherlands (Sieben and Halman, 2014).
Over the last 10 years of research, one can observe a renewed interest in the different family roles in the framework of religious socialisation. Take, for example, the study carried out in the United States (‘Longitudinal Study of Generations’), involving over 3500 respondents from 357 families, from three and four generations (Bengtson, 2013). In this trajectory of change, parents were expected to play an increasingly irrelevant role in their children’s religious identification process. However, the data did not confirm this linearity. Despite the many changes that have affected the American society, the results showed that a child was likelier to remain within the religious belonging in which they had been socialised than to abandon it. Moreover, this trend was also present among those with none. The researchers were surprised, for example, by the unexpected importance of the father’s role. The effect of interfaith marriage and divorce on children’s religious identity is the same. In mono-faith marriages, children were more likely to continue their religious heritage. Discontinuity was more likely in the case of mixed-faith marriages or in families that experienced parental divorce. The authors also point out that grandparents continued to play a significant role in religious transmission, sometimes bypassing the parents themselves (e.g. Bengtson, 2013; Voas and Storm, 2012). The research included the construction of a typology of trajectories concerning the results of religious socialisation: (a) the ‘rebels’, who refuse the religion of their parents; (b) the ‘zealots’, who are more devout than their parents; and (c) the ‘prodigals’, who grow up in the faith of their parents, abandon it (converting to another religion or becoming non-religious), and then return to the faith in which they were socialised.
Despite the multifactorial nature of religious socialisation processes, the importance of family cultures and practices remains evident (Guest, 2011; Sherkat, 2003; Smith and Adamczyk, 2021). Studies continue to show that parents’ religiosity influences their children’s religiosity (Denton and Flory, 2020; Smith and Adamczyk, 2021). However, fathers and mothers may have different roles and importance in socialisation. Women typically have more responsibility for childcare and tend to play a more active role in religious socialisation (Gutierrez et al., 2014). Previous European research has studied the differences between the Evangelical, Reformed, and Roman Catholic milieu. It was found that the greater effectiveness of evangelicals regarding the transmission of family religious identity depended on parental commitment, with a particular emphasis on mothers (Stolz and Favre, 2005). Even so, as has already been emphasised, the role of fathers should not be disregarded (Baker-Sperry, 2001; Smith and Adamczyk, 2021). Studies point to fathers having more influence on formal religious behaviour, such as participation in worship, than on the day-to-day experience of religion, where mothers tend to be more influential (Bader and Desmond, 2006). The likelihood of children’s religious disaffiliation increases when one parent is less religious than the other, when they communicate discordant religious messages, or when what is done does not match what is said (Bader and Desmond, 2006; Clements and Bullivant, 2022; Smith, 2014). Conversely, when parents share religious preference and commitment (‘religious homophily’), young people are likely to be more religious (Clements and Bullivant, 2022; McPhail, 2019; Voas and Crockett, 2005; Voas and Doebler, 2011; Voas and Storm, 2012).
The impact of religious socialisation in European societies is closely linked to the secularisation paradigm. Steve Bruce’s (2011) reaffirmation of secularisation theory emphasises, on the one hand, the widespread impact of factors such as pluralism, egalitarianism, and individualism on social and religious change and, on the other, notes that new forms of religiosity or spirituality cannot indeed be considered as a social substitute for previous modes of religious identification. The theory of cohort replacement clearly argues that generations are increasingly less religious due to the gradual decrease on socialisation (e.g. Stolz et al., 2021; Voas and Doebler, 2011). However, studies have focused on countries, which contain a lot of internal diversity, and not on specific areas where the degree of secularisation is much more noticeable. In fact, this is one of the main contributions of the present article, the focus on a metropolitan area – LMA – where its socially dynamic context is a laboratory for observing the changes taking place in Portuguese society and by extension in European and Western societies. Modernisation has been more accelerated in this region than in other Portuguese regions as the highest levels of population density, tertiary population, purchasing power, higher education, female workforce, and households with Internet, among other indicators, may prove (Instituto Nacional de Estatística (INE), 2022).
This regional differentiation was demonstrated by the study on Religious Identities in Portugal (Teixeira, 2013). Specifically, LMA showed a considerable erosion in the relative weight of Catholics, 67.6% – considering that the national figure was close to 80%. This region also contributed to the representative sample of the continental population, with 55.2% of the people not believing, 62.2% of respondents belonging to a Protestant denomination, 51% of Jehovah’s Witnesses, and 61.5% belonging to other religions. This metropolitan region also found the highest percentage of other Christians (47.2%) and believers without religion (43.5%). Despite its higher religious pluralism, LMA is more secularised than Portugal in general, which may raise the question of the higher importance of families to shape children’s religiosity (Kelley and Graaf, 1997).
The same study (Teixeira, 2013) also examined the ‘size of the locality’ variable (urban, semi-urban, and rural). It concluded that Catholic identity was the only one documenting a preponderance of the ‘rural’ typology. While Catholics accounted for around 80% of the total population surveyed, their relative position changed substantially when the universe was limited to respondents living in urban areas. There, their relative weight rose to 66.6%. The population belonging to other religious denominations is mainly urban and semi-urban. The same goes for those who belong to no religion at all. The urbanisation of lifestyles and the dynamics of the metropolisation of the territory can be identified as factors with a high impact on the erosion of traditional Portuguese religiosity, facilitating both the itineraries of religious disaffiliation and the emergence of a plural religious landscape. As a global phenomenon, within the scope of the so-called ‘Spatial turn’ (Lussault, 2013), it can have a particular impact on societies that experience a later and more accelerated change in how social space is practised.
Socialisation and religious transmission no longer benefit from a compact relationship with the territory. The consequences of the processes of de-territorialisation of people and communities favour the emergence of hybrid cultures and topoligamic identities, a context in which territories of belonging are multiple on different scales, requiring the mobilisation of adaptive skills (Haesbaert, 2004). The contexts of multi-territoriality and territory-networks do not make religious identification impossible. Still, they do imply relationships of loyalty and belonging that are less hierarchical or centralised and more adapted to the consequences of cosmopolitanism (Franca, 2022). Turner (2011) particularly highlighted the impact of modern cosmopolitanism on religious identities. His proposal takes on a normative character by stating that cosmopolitan values are the antidote to fundamentalism and nationalism in the context of ‘de-traditionalisation’. These values could guarantee new forms of integration, replacing ethnic and ‘familialist’ forms of socialisation.
The recent currents in urban cultures weaken transmission logic based on inheritance logic. The religious landscape in Portugal was particularly affected by the waves of urbanisation during the twentieth century, especially in metropolitan areas (Lopes, 2017; Rodrigues, 2010).
These data show that LMA is a unique region in Portugal. Although other studies, mainly in the US context, have been studying religious socialisation and religiosity, there are practically no studies focused on metropolitan areas, even less in European countries. This specific socio-cultural context of LMA is a good starting point for other studies on metropolitan areas about the relationship of socialisation and religiosity. It is possible to conjecture that in the more secularised LMA, compared to Portugal in general, despite the larger religious diversity, families are more decisive to shape children’s religiosity (Kelley and Graaf, 1997). So, based on the premise that we are studying a specific case, with its idiosyncrasies comparable to other European and Western metropoles, and on the literature review, we propose a set of hypotheses for a double characterisation: parents’ religiosity during childhood and the current religiosity of those surveyed (which includes decisions regarding the socialisation of their children). The respondents find themselves in a situation where they move from a place where they are mainly the ‘object’ of socialisation to a position where they are the protagonists, from ‘socialised’ to ‘socialising’. Thus, the three hypotheses are:
Hypothesis 1 (H1). Religious socialisation increases with age.
Hypothesis 2 (H2). More religious socialisation induces more religiosity.
Hypothesis 3 (H3). More religiosity induces more religious socialisation.
Method
This part presents methodological choices: data sources, population under study, variables, age groups, and statistical analyses. The data are from the project ‘Religious identities and social dynamics in Lisbon Metropolitan Area’ (Teixeira, 2019) applied by the Centre for Studies and Opinion Polls (CESOP), the survey centre of the Universidade Católica Portuguesa, with a random sampling method (stratified with random route), during June and July of 2018. The sample had 1180 respondents from LMA who were at least 15 years old and had a margin of error of 2.9% (95% confidence level). 1
ISSP 2018 is the only database with socialisation variables. There are ten European countries with a Catholic tradition: six ex-communist and four Western. Of the Western countries (Austria, France, Italy, and Spain), which are geographically, culturally, and historically closer to Portugal, mainly the last two, NUTS III Madrid has a sample of 241. In contrast, NUTS II Lazio has a sample of 102. This last one does not correspond to NUTS III, making comparison impossible. So, the most comparable metropolitan area is Madrid, but it has a much smaller sample than the LMA project. NUTS III Paris has only 29 cases, while NUTS III Vienna has 254 cases. The mean of the 10 countries and the most prominent sample are 190 and 267 (NUTS III Vilnius), respectively. In short, all the samples are much smaller than the LMA project, and Portugal is not in the ISSP.
The population under study included Catholics and non-religious people, the two main groups, while religious minorities were not included (see the top of Table 1). This group is small and heterogeneous, so religious minorities are too small to produce reliable analyses, and they could not be analysed together. Table 1 (bottom) shows that in LMA, the only critical religious changes occurred towards disaffiliation and/or diminishing religiosity (practice and/or belief). Conversions are shallow: becoming Catholic (0.85%) or converting from Catholicism to another religion (3.0%). This confirms that the two main religious fields in LMA are Catholicism and non-religious people.
Religious affiliation and changes in LMA.
Source: LMA dataset.
Weighted data. First part is from Q13, while the second part is from Q31.
This study used two types of variables: socialisation and religiosity. Socialisation variables/questions in the LMA questionnaire are of two types:
(a) Upstream of respondents’ religiosity. This type includes variables of the respondent’s parents and their religiosity when the respondent was a child (10 years old). They are important indicators of respondents’ religious socialisation during childhood, so they are good predictors of respondents’ religiosity. They are two types: • Father/Mother’s religious affiliation when the respondent was a child: Q35.
2
For this analysis, a dummy variable was created for both where ‘Catholic’ is 1, and the others are 0. • Father/Mother’s religious practices when the respondent was a child: Q36.
3
This group includes four variables for both. All variables are the dummy (1: yes; 0: no).
(b) Downstream of respondents’ religiosity. This type includes variables of respondents’ children and respective religious education. They are good indicators of respondents’ influence on their offspring, so they are good expressions of respondents’ religiosity. They are two types: • Children’s religious education: Q38.
4
This group includes five variables for both. All variables are the dummy (1: yes; 0: no). These variables represent different types of religious education which ultimately depend on respondents’ religiosity. • Number of children: Q37.
5
Although this variable is not a type of religious socialisation, it expresses respondents’ religiosity since larger families correlate with higher levels of religiosity.
Based on pioneering but still critical studies (Fichter, 1951; Glock and Stark, 1965), religiosity has a few dimensions: four for the first and five for the second. Usually, surveys have variables for just four dimensions: community (belonging), ritualistic (practices), ideological (beliefs), and consequential (attitudes). Still, researchers, including Stark and Glock (1968: 16), have long debated whether the consequential dimension is part of religiosity or derives from it. So, in this study, only the first three dimensions were considered. For each dimension, only two variables that best characterise it were selected among the available variables to simplify analyses. Each variable was recoded as a dummy to simplify and normalise the analyses where ‘1’ is always the category that expresses more religiosity. For each dimension, the respective questions/variables are:
(a) Belonging (Religious affiliation/Movement affiliation): Q13/Q26. 6 As already explained, only Catholics and non-religious people were considered in this article, so 1 corresponds to Catholics and 0 to non-religious people. Religious affiliation is one of the most important parameters to characterise religiosity and the variable par excellence of belonging, so it is regularly used. From the available variables for this dimension, the second variable is more interesting than another one (Q27) on recurring roles within the church. Therefore, 1 corresponds to ‘yes’ and 0 to ‘no’. Practice (Frequency of attendance at religious services /prayer): Q12a/Q9. 7 Analyses focused on regular practice, which is the sum of the categories ‘more than once a week’ and ‘once a week’ in the first variable and ‘every day’ and ‘a few times a week’ in the second variable. 1 corresponds to the sum of these categories, and 0 to the sum of the other categories in each variable. Like the first variable of belonging, attendance is a mandatory variable used in all religious studies. It is the best expression of regular practice, as attending Mass is the only of the Catholic Church’s five precepts and seven sacraments in which Catholics should participate weekly. Like the previous variable, the second variable is also a usual variable in this type of study though not as commonly used as the first.
(b) Belief (Beliefs in revealed God/meeting God after death): Q7b/g. 8 Analyses focused on the sum of categories ‘partially agree’ and ‘totally agree’ in both variables. 1 corresponds to the sum of these categories and 0 to the sum of the other categories in each variable. Belief in God is probably the most used belief variable, although ‘belief in revealed God’ is relatively different from strict ‘belief in God’. The second variable relates to life after death and Heaven, two variables most used in surveys.
When choosing the number of age groups, the fewer the groups, the simpler the analysis. For this article, the best option should allow distinguishing young people from older people, where the middle age group is for confirming trends. Therefore, three age groups was determined to be the most straightforward option. Thirty-four is an increasingly used upper limit for youth (e.g. Sagnier and Morell, 2021), mainly due to increasing delays in adulthood. Sixty-five is the minimum limit for older people, as it is the retirement age in the EU (66 in Portugal). 9 In the middle are the middle-aged, although the limits of this group are also debatable, but they are roughly between 40 and 60 years of age. 10 So, the age groups are 15–34 (young), 35–64 (middle age), and 65+ (older).
Based on SPSS, the following analyses regarding each hypothesis were applied:
(a) The first hypothesis (religious socialisation increases with age): comparison by age group of all socialisation variables and their indexes using ANOVA, since all variables are dummy or quantitative. Father/mother’s indexes of socialisation with five variables include all upstream variables, while with three variables include only religious affiliation and the two most important variables of practice (attendance and prayer). Parents’ indexes with 10 and 6 variables are the same for both parents. The quality of the indexes was measured through Cronbach’s alpha (CA): 11 The first four vary between 0.577 and 0.627, while the last two are practically the same (0.756/0.757). Data are from the LMA dataset (2018), weighted. ANOVA can only be applied rigorously when two conditions are met: normality, verified by the Kolmogorov-Smirnov test with Lillefors correction or by the Shapiro-Wilk test (n < 30), and homogeneity of variances, verified by Levene’s test (Maroco, 2010: 133–136). When homogeneity of variances does not exist, Fw of Welch can be used; when normality or both conditions are violated, Kruskal–Wallis’ test can be used 12 ; although for samples larger than 30, normality is accepted according to the central limit theorem (Maroco, 2010: 59, 227). Significant differences are verified through post-hoc tests. 13
(b) The second hypothesis (more religious socialisation induces more religiosity): application of multiple linear regression (MLR) of the indexes of socialisation on the indexes of religiosity with the moderator ‘age’. Two indexes of religiosity were used: one with the six variables of religiosity and another with the three most representative variables of each dimension. Both are theoretically consistent, although the first has a CA of 0.599, and the second has a CA of 0.676. 14 With the moderator ‘age’, it was intended to check if the influence of religious socialisation on religiosity depends on age. Age was considered instead of age groups because age groups are an arrangement of age, while what matters here is to check the effect of age and not of each age group per se, and age is a continuous variable. Since independent variables (age and indexes of socialization) are continuous variables, they were mean-centred. To apply MLR, six premises must be fulfilled (Maroco, 2010: 563; Pestana and Gageiro, 2000: 474): linearity between dependent and independent variables, 15 absence of multicollinearity between independent variables, 16 residuals with four traits (null mean, 17 homogeneous variance, 18 independent, 19 and normal distribution 20 ).
(c) Third hypothesis (more religiosity induces more religious socialisation): application of MLR of the two indexes of religiosity on the second type of socialisation variables with the moderator ‘age’. The same aspects considered in the previous analysis were applied here.
Results
As studies have shown, family or parents are the principal socialising agent, meaning that more religious parents have more religious children. So, the father’s and mother’s religiosity during the respondent’s childhood are good indicators of how he or she was socialised religiously. The first set of variables concerns the respondent’s religious socialisation carried out by their parents during his or her childhood (10 years old). Only these variables are available in the project about parents’ religiosity during the respondent’s childhood. Table 2 shows that, while Catholic affiliation does not have significant differences between age groups, the oldest group has significantly higher percentages in the two most important practices. Table 3 shows that the youngest age group has significantly lower percentages for the same three variables than the other age groups. Comparing these tables, mothers have higher percentages in all variables, demonstrating that they are more religious than fathers in all age groups without exceptions.
Father’s variables of religious socialisation by age group (%).
Source: LMA dataset.
Significant differences exist when the letters are different—a, b, c.
Tests: FCA – Fw (2) = 1.554, p = 0.212. FWA – Fw (2) = 4.104, p = 0.017. FDP – Fw (2) = 6.315, p = 0.002. FBG – Fw (2) = 1.954, p = 0.143. FPD – Fw (2) = 3.384, p = 0.035. In all tests, normality assumed (n > 30) but homogeneity of variances not assumed (Levene’s test: p < 0.05); DK/NA were omitted.
Mother’s variables of religious socialisation by age group (%).
Source: LMA dataset.
Significant differences exist when the letters are different—a, b, c.
Tests: MCA – Fw (2) = 9.914, p < 0.001. MWA – Fw (2) = 12.748, p < 0.001. MDP – Fw (2) = 10.257, p < 0.001. MBG – Fw (2) = 1.122, p = 0.326. MPD – F (2) = 0.120, p = 0.887. In all tests (except the last in homonogeneity of variances), normality assumed (n > 30) but homogeneity of variances not assumed (Levene’s test: p < 0.05); DK/NA were omitted.
The combination of these variables formed two types of variables, each with two indexes: one with the father’s and mother’s variables separately and another with their variables combined (Table 4). In the first type, the subtype with five variables was based on all the aforementioned variables, while the subtype with three variables was based on the most important variables. The second type is a combination of the subtypes of the first type. For all indexes, values increase with age, without exceptions, although significant differences are not the same. Even so, they have in common that young people have significantly lower values than just the older group, in the father’s indexes, or the other age groups, in the other indexes. Besides being theoretically more consistent by using variables from both parents, parents’ indexes have higher Cronbach’s alphas, above 0.7, strengthening their reliability. In all types of indexes, the use of five/three or ten/six does not influence the quality of the results, which may indicate that they are all valid.
Indexes of religious socialisation by age group.
Source: LMA dataset.
Significant differences exist when the letters are different—a, b, c. Cronbach’s Alpha–FIS5 = 0.627 (only if the first was deleted the value would be higher–0.628); FIS3 = 0.626 (only if the first was deleted the value would be higher–0.681); MIS5 = 0.598 (only if the first was deleted the value would be higher–0.600); MIS3 = 0.577 (only if the first was deleted the value would be higher–0.635); PIS 10 = 0.757 (no item deleted would increase value); PIS6 = 0.756 (no item deleted would increase value).
Tests: FIS5 – Fw (2) = 4.420, p = 0.012. FIS3 – Fw (2) = 4.710, p = 0.009. MIS5 – F (2) = 13.527, p < 0.001. MIS3 – F (2) = 16.975, p < 0.001. PIS10 – F (2) = 10.179, p < 0.001. PIS6 – Fw (2) = 11.830, p < 0.001. In all tests, normality assumed (n > 30) but homogeneity of variances not assumed (Levene’s test: p < 0.05) in FIS5, FIS3, and PIS6.
Table 5 summarises the previous three tables. Data clearly show that parents’ religiosity (affiliation, attendance, and prayer) increases with age, and so the likelihood of being religiously socialised by the parents increases with age. In short, youth have less religious parents mainly comparing with older people. The age group 35–64 years are in the middle, closer to youth concerning father’s religiosity and to older people concerning mother’s religiosity.
Summary of the variables and indexes of religious socialisation by age group.
Source: LMA dataset.
The next set of variables concerns the downstream variables: religious education and the number of children (Table 6). Again, all percentages increase with age, and in general, they differ significantly between all age groups: The only exceptions are in religious education by parents, where the youngest group has a significantly lower percentage than the other age groups, and by ‘other family’, with no significant differences between age groups.
Variables of religious education and number of children by age group.
Source: LMA dataset.
Significant differences exist when the letters are different—a, b, c. ‘Yes’ was originally ‘No’ which was inverted.
Tests: Yes − Fw (2) = 18.049, p < 0.001. Parents − Fw (2) = 9.476, p < 0.001. Other family – Fw (2) = 2.021, p = 0.136. Church – Fw (2) = 36.403, p < 0.001. School – Fw (2) = 11.612, p < 0.001. NC (Number of children) – Fw (2) = 199.879, p < 0.001. In all tests, normality assumed (n > 30) but homogeneity of variances not assumed (Levene’s test: p < 0.05); DK/NA were omitted.
Table 7 shows two indexes of religiosity: one with all six variables (two per dimension) – IR6 – and another with three variables (one per dimension) – IR3. Both show that religiosity increases significantly with age, which may indicate that they are both valid for measuring this concept.
Indexes of religiosity by age group.
Source: LMA dataset.
Significant differences exist when the letters are different—a, b, c.
Tests: IR6 – F (2) = 22.122, p < 0.001. IR3 – F (2) = 28.858, p < 0.001. In both tests, normality (n > 30) and homogeneity of variances (Levene’s test: p > 0.05) assumed.
Table 8 shows the effects of the two parents’ indexes of socialisation on the two indexes of religiosity based on simple linear regression. R squares and standardised betas 21 are very close or equal, so they are all valid. In a second phase, a moderation/interaction analysis based on age was produced with MLR, where all standardised betas are very low, and p was higher than 0.05, meaning that the effect of socialisation on religiosity does not depend on age.
Effect of socialisation on religiosity.
Source: LMA dataset.
The six premises of MLR were fulfilled in all analyses.
Interaction age*IV: IR6-PIS10 – SB = 0.014, p = 0.611. IR6-PIS6 – SB = 0.005, p = 0.857. IR3-PIS10 – SB = 0.020, p = 0.464. IR3-PIS6 – SB = 0.010, p = 0.704.
Table 9 shows the effects of the two indexes of religiosity on the variables of religious education and the number of children based on simple linear regression. While the other indexes are upstream of religiosity, these are downstream. The existence of religious education reveals at least the will to transmit parents’ religious convictions. The number of children can be a good indicator of religiosity since more religious families usually have more offspring, so it can also be seen as a proxy for socialisation. Data show that, while religiosity influences socialisation, with close standardised betas and R squares in both indexes, the number of children is not so dependent on it, with very low R squares and low standardised betas. Applying moderation/interaction analysis, although for education, p values are lower than 0.05, standardised betas are low. Besides, normality is not so clear in both analyses of education. As in the previous table, age seems to be unimportant in influencing socialisation and religiosity.
Effect of religiosity on socialisation.
Source: LMA dataset.
The six premises of MLR were generally fulfilled in all analyses.
Interaction age*IV: IR6-Yes – SB = −0.098, p = 0.007. IR6-NC – SB = −0.026, p = 0.337. IR3-Yes – SB = −0.083, p = 0.023. IR3-NC – SB = −0.030, p = 0.265.
Socialisation and religiosity: continuities and discontinuities
In discussing the results, returning to the previously stated hypotheses is essential. The first hypothesis, that religious socialisation increases with age, was confirmed in general by the data and all the indexes. The only exceptions occurred in the father’s Catholic affiliation and the two minor parents’ practices. Still, there is no single variable or index where the three age groups differ significantly from each other. The only common feature across all variables and indexes is that the youngest group always has the lowest values. Generally, our data confirm the theory of cohort replacement (e.g. Molteni and Biolcati, 2018; Stolz et al., 2021; Voas and Crockett, 2005; Voas and Doebler, 2011). In fact, data show that cohorts are increasingly socialised by less religious parents. The absence of significant differences between 35–64 and 65+ in some variables, and almost all indexes may come from at least two reasons. First, it is more difficult for the older group to remember what their parents did when they were children. Second, there is a generation gap between the youth and older generations. The first explanation needs further research to check its validity for this specific issue, but its theoretical ground is clear since time erases or changes memory, according to psychological knowledge. The second has been analysed by many authors, and in fact, young people (generations Y and Z) distinguishes from previous generations (e.g. Wilkins-Laflamme, 2022).
The second hypothesis, that more religious socialisation induces more religiosity, was confirmed by the data. This is in line with previous studies (e.g. Denton and Flory, 2020; Smith and Adamczyk, 2021). Lineages of belief are preserved mainly when the parents converge in their attitudes towards religion and propose the same religious commitment to their children. When there are disagreements in terms of affiliation and/or on the levels of parental religiosity, the likelihood to diverge from a specific lineage increases (e.g. Conway et al., 2023; Voas and Storm, 2012). Voas and Storm (2012) show in a study comparing England and Australia that when parents attend a religious community together, religious practice is mutually encouraged. The impact of this homogeneity on the socialisation of children tends to be more decisive in the transmission of religious identity than in situations where only one parent attends the religious community. Also, the positive atmosphere infused with good emotions where the religious transmission occurs helps to keep children inside the religious canopy. Studies agree on this (e.g. Bader and Desmond, 2006; Clements and Bullivant, 2022; Smith, 2014). Like the previous hypothesis, this confirms the theory of cohort replacement but in the opposite direction, that more produces more, instead of less producing less, as the theory argues. Data for this hypothesis show that the effect of socialisation on religiosity does not depend on age. It means that a younger person may be just as influenced by their parents to become religious as an older person. So, the differences of religiosity by age group (Table 7) or the gradual disappearance of the lineages of belief are caused by the individualisation process (e.g. Collins-Mayo and Dandelion, 2010), mainly in more urbanised places, such as LMA (e.g. Coutinho, 2023), and other factors, such as the predominance of forms of ‘cultural religion’, the majority or minority situation of the religious group, the type of implantation of religious institutions in the territory, family cultures, or the dynamics of lifestyle urbanisation (Teixeira, 2023). In sum, individualisation and other factors affect socialisation because they decrease the likelihood of being religious, which in turn affects the religiosity of their children. In the end, socialisation is an intermediate process, not the cause. So, these data may help to doubt about the capacity of socialisation as cause when crossing with age.
The third hypothesis, that more religiosity induces more religious socialisation, was also confirmed by the data. This is a logical consequence of receiving religiosity from their parents: Who receives more, gives more; who receives nothing, can offer nothing. And once again, this may be linked to the theory of cohort replacement in reverse. The idea of the snowball in reverse expresses well what this theory argues, meaning that the absence of religiosity and socialisation reinforce each other. As with the previous hypothesis, the data show that the effect of religiosity on socialisation does not depend on age. The same lines of explanation as the previous hypothesis may be used here.
The data analysis allows us to reaffirm that the contexts of primary religious transmission and socialisation in the family environment – especially in the context of parenthood – continue to have significant weight in the development of religious identities. However, the results of the analysis also show that there are fewer and fewer families able to be active in the processes of religious transmission and socialisation. These two research paths should continue with complementary studies that can, in any case, reach a more detailed understanding of this dual phenomenon.
Thus, a comprehensive model is needed to accompany these two checks. On the one hand, it is undeniable that religious identity weakens from the older to younger generations, becoming more fluid and composite. However, it is also undisputable that religious socialisation in childhood and adolescence is a determining factor for the construction of solid religious identities. This observation also accompanies the findings of the ‘non-religious studies’. Comparing the ‘nones’ who have not had any religious socialisation and the ‘nones’ who have experienced a process of religious disaffiliation throughout their lives, it can be observed that in the latter, a crucial religious residue is found (Beider, 2022). The possibility of a ‘religious residue’ is reduced in contexts where children are mostly socialised as nones (Stolz et al., 2020). It is essential to underline that religious practice is usually the most eroded indicator in longitudinal studies. For example, a marked decrease in church attendance was observed in a survey in the Netherlands, interviewing the same population in 1983 and 2007. However, in that study, religious education in the family did not produce different results in the Catholic or Protestant camp (Vermeer et al., 2011).
On the other hand, social contexts need to be considered. Studies that consider the diversity of religious groups and their minority or majority position have shown that local contexts matter in the study of this problem (Lim and De Graaf, 2021). The research results over the last decade on Spanish society – comparable to Portugal in terms of the prominent presence of a ‘cultural Catholicism’ (Herranz de Rafael and Fernández-Prados, 2022) – have shown that more religious Christians predominate in the older age groups. Conversely, those with no religion have a younger age profile. However, the results showed that the portrait of the younger generations is influenced by the fact that religiosity favours fertility. In other words, when characterising the phenomenon of socialisation, it is necessary to consider that more religious tend to have more children (Stonawski et al., 2015). Thus, even in a social context where the non-religious population continues to grow (Pérez-Agote, 2010), the religiosity of parents, within the framework of socialisation processes, continues to be an essential factor in the religious characterisation of young people. Kelley and Graaf’s study of data from the 1991 ISSP showed that people living in ‘religious nations’ adhere more to orthodox beliefs than those living in ‘secular nations’. In secular-trending nations, family religiosity powerfully shapes children’s beliefs, and the influence of the national religious context is small. In religiously inclined countries, family religiosity affects children’s beliefs less than the national context. These patterns maintain consistency regardless of the nations’ economic level, whether or not they have a communist past or an established democratic history. This trend persists among younger and older people, men and women, all levels of schooling, and Catholics and Protestants (Kelley and Graaf, 1997).
However, it must be borne in mind that this type of family agency is reconfiguring within the framework of models gradually guided by deinstitutionalisation (fewer marriages and more children out of wedlock), separation (more divorces), infertility (fewer children), and postponement (later marriages and children). These more discontinuous models opposed to a more traditional family model can be found in the younger generations because of individualisation, giving space to a more complex socialisation model, as defined by authors as Collet-Sabe (2007). Since family is key in religious socialisation (e.g. Smith and Adamczyk, 2021), changes in its models cause changes in its impact. Moreover, as children grow up, parents increasingly compete with secular alternatives to religion, functioning as alternative resources for guidance, potentially weakening the internalisation of religious content (Baker-Sperry, 2001; Kelley and Graaf, 1997; Stolz et al., 2016). The theory of cohort replacement points in this direction, arguing that younger generations (less religious) are replacing older generations (more religious) (Molteni and Biolcati, 2018; Stolz et al., 2021; Voas and Doebler, 2011).
Finally, the study of Coutinho (2023), that used data from the same source of this article and which point in the same direction, confirms the importance of individualisation and socialisation to explain the decrease of religiosity, mostly in urban areas. Comparing LMA and Portugal, the author concluded that these processes are in the centre of the increasing distancing from institutional religiosity. In fact, LMA, the least ‘parochial’ Portuguese region and which is much less religious than Portugal as a whole, is the wealthiest and most educated Portuguese region. The gradual increase in the values of these indicators in the long term supports the argument of a notorious combination between these processes.
Conclusion
This article had three goals in analysing data on the LMA: first, to analyse the changes of religious socialisation across age groups; second, to analyse the impact of religious socialisation on religiosity; and third, to analyse the impact of religiosity on religious socialisation. In general, the three hypotheses associated with these goals were confirmed, meaning that the theory of cohort replacement is correct. In other words, across age groups, people are increasingly less religious, and so the likelihood of passing their religious convictions on to their offspring diminishes. At the same time, if people are less prone to be less religiously socialised, they are going to be less religious and so on, like a snowball: Less gives less, more gives more. This is the main idea that the data presented here shows. This article contributes to reinforce the theory of cohort replacement mostly in the metropolitan areas of southern European countries from Catholic tradition. At the same time, it found an interesting aspect related with socialisation, that it is more an intermediate process than a real cause of decreasing religiosity. In fact, socialisation is between individualisation (and other factors) and religiosity meaning that it is also a consequence of individualisation. Individualisation is the motor of religious changes, not socialisation.
This study has some limitations that ask for more research. First, due to its quantitative nature, it lacks a deeper understanding of reality, something that is achievable by qualitative analyses. Second, situational and/or longitudinal analyses could overcome the problems associated with cross-sectional analyses bringing long-term analyses where the same age group or generation is studied in different time periods. Third, comparisons with other metropolitan areas and Portugal could bring two benefits: first, to situate this specific region among other similar regions; second, to compare with the country in general, as Coutinho (2022, 2023) made in terms of religiosity.
Footnotes
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
Authors biographies
Address: Centro de Investigação em Teologia e Estudos de Religião, Universidade Católica Portuguesa, Palma de Cima, 1649-023 Lisboa, Portugal.
Email:
Address: Centro de Investigação em Teologia e Estudos de Religião, Universidade Católica Portuguesa, Palma de Cima, 1649-023 Lisboa, Portugal.
Email:
Address: Politécnico de Leiria–Escola Superior de Educação e Ciências Sociais, Campus 1 Rua Dr. João Soares Apartado 4045, 2411-901 Leiria, Portugal.
Email: