Leaving My Religion: Understanding the Relationship Between Religious Disaffiliation, Health, and Well-Being

Abstract

Religious disaffiliation—leaving the religious tradition in which one was raised for no religious affiliation in adulthood—has become more common in recent years, though few studies have examined its consequences for the health and well-being of individuals. We use an innovative approach, comparing the health and subjective well-being of religious disaffiliates to those who remain affiliated using pooled General Social Survey samples from 1973 through 2012. We find that religious disaffiliates experience poorer health and lower well-being than those consistently affiliated and those who are consistently unaffiliated. We also demonstrate that the disadvantage for those who leave religious traditions is completely mediated by the frequency of church attendance, as disaffiliates attend church less often. Our results point to the importance of the social processes surrounding religious disaffiliation and emphasize the role of dynamics in the relationship between religious affiliation and health.

Introduction

Research on the sociology of religion has focused on how modes of religious affiliation and participation intersect with other dimensions of social life. Sociologists have noted that religious individuals exhibit better health and well-being outcomes, have lower death rates in old age, and report higher subjective well-being (“happiness”) than the less religious or non-religious (George, et al., 2002, Idler, et al., 2003, Sullivan, 2010). Since 1990, there has been a rapid rise in the fraction of Americans claiming no religious affiliation, and a substantial portion of this growth reflects an increase in disaffiliation among Mainline Protestants and white Catholics (Putnam and Campbell, 2010). According to the General Social Survey in 2012, 20% of all American adults claimed no religious affiliation, and 75% of those with no religion report being raised in a religious tradition. Religious disaffiliation has become increasingly common in the United States, and the processes surrounding its causes and consequences have received more attention recently (Schwadel, 2014, Vargas, 2012). Religious disaffiliation represents a particularly important process in religious change since it may be associated with significant changes in the social relationships surrounding religious affiliation, practice, and behavior (Uecker, et al., 2007). In turn, these changes may have implications for health and well-being and can improve our understanding of the relationship between religious participation and health.

Little research has focused on the individual implications of leaving a religious tradition, and the few existing studies attempt to link life course events to changes in religious affiliation, identifying the social experiences that lead individuals to leave religious congregations. The primary existing study that has examined the relationship between religious switching and health is the work of Scheitle and Adamczyk (2010). They demonstrate that switching from high-cost religions—those that induce a greater commitment—is associated with poorer reported health. Our study extends this consideration to a greater examination of religious disaffiliation itself, and examines subjective well-being in addition to health. We use a nationally-representative survey to examine the association between religious disaffiliation and health. Instead of simply comparing the religious and non-religious, we investigate the health and well-being of individuals who leave the religions in which they were raised, examining all religious traditions together as well as focusing on specific religious denominations in the United States. We demonstrate that religious disaffiliates experience poor health and lower subjective well-being than their counterparts who remain affiliated. We find that this effect is strongest for Evangelical Protestants and Catholics and does not exist for Mainline Protestants. We suggest that these disadvantages largely reflect the loss of the social benefits of religion following disaffiliation, but also find evidence for a role of the spiritual aspects among Evangelical Protestants. Like Scheitle and Adamczyk (2010), we find that disaffiliates from high-cost religions report poorer health. However, they do not report lower subjective well-being, as disaffiliates from other religious groups do.

Background

Religious Affiliation, Health, and Well-Being

The religious tend to have significantly better health than the nonreligious; Jews, Mainline Protestants and Catholics tend to live longer than those affiliated with no religion (Sullivan, 2010). The relationship between religion and health holds across many measures of religious participation, but the most consistent findings refer to active religious engagement (Dupre, et al., 2006). Religious individuals also report that they are happier than the nonreligious. Although this relationship reflects a similar process to the relationship between religion and health, well-being is likely to emphasize different dimensions of religious involvement (Childs, 2010, Ellison, 1991). This is an important area of study for sociologists since the most promising explanations for the health and well-being advantage associated with religion are social in nature (Shor and Roelfs, 2013).

First, it is possible that the relationship between religion and health is not causal, and instead reflects observed or unobserved differences between religious and non-religious individuals. Happier and healthier people may be more likely to participate in religion and may reap greater benefits from the social aspects of religious involvement (Miller and Thoresen, 2003). While negative health events may increase religiosity (McFarland, et al., 2013), they may also serve as barriers to religious participation, particularly if functional disability limits the ability to attend religious services (Benjamins, et al., 2003). Previous research has shown that reverse causation accounts for only a small portion of the religion-health relationship (Oman, et al., 2002).

Second, the relationship may be causal, with religion directly leading to better health and well-being. Religious participation may bring positive emotional benefits as a means to cope with stressful or adverse life events (Nooney and Woodrum, 2002, Pollner, 1989). There is some evidence that individuals use prayer as a means of coping with stress (Pargament, et al., 2004), although these benefits differ considerably by race and socioeconomic circumstances (Krause, 2003). Overall, the health or psychological well-being benefits of prayer depend considerably on images of God as either personal or remote (Bradshaw, et al., 2008, Bradshaw, et al., 2010).

Finally, the relationship may be causal but the effect of religious participation may be indirect. For instance, religion may provide strict behavioral directives regarding lifestyle factors that have effects on physical health. In addition to exhorting members to practice behavioral temperance, church congregations may also monitor members’ behavior in which religious social networks influence healthy behavior change (Scheitle and Adamczyk, 2010). Social networks may have generalized health and well-being benefits as well. Those individuals with strong networks of social ties and support feel a stronger sense of integration into social life, avoiding the negative emotional and socio-psychological effects of social group isolation (Seeman, et al., 2004). Social relationships have a notable positive impact on physical and mental health, and a great deal of research has been devoted to the underlying mechanisms (Umberson and Montez, 2010). Here, sociologists of religion have an important role in linking religious involvement to health and well-being, since religion plays an important role in individuals’ social embeddedness and social integration, providing for many a crucial source of both strong and weak social ties (McClure, 2013).

Religious attendance is often used as a proxy for the social characteristics of religious involvement (Dupre, et al., 2006, Sullivan, 2010). Attendance at religious services is inherently social, and qualitative research suggests that many individuals place a high value on the social relationships developed through church-based social networks (Banerjee, et al., 2014, Holt and McClure, 2006). The social benefits of attendance are the primary determinants of improvements of well-being for those who attend frequently (Childs, 2010), but these benefits may not occur specifically during attendance (Idler, et al., 2009). Individuals do not receive most of their social ties benefits from the actual act of attending, but instead from the friendship and acquaintance networks that derive therefrom (Rote, et al., 2013), as well as from secular activities such as volunteering and civic engagement (Lewis, et al., 2013). However, as a proxy of the social aspects of religion, attendance is one of the primary reasons for better health and well-being among the religious (Strawbridge, et al., 2001).

Religious Disaffiliation

Sociology has traditionally assumed that religious institutions are losing members over time as western societies secularize in late modernity (Berger, 1967). More recently, scholars have argued that religious pluralism does not lead to secularization but rather to greater religious participation due to competition (Finke and Stark, 2005). Although the majority of American adults remain affiliated with the religion in which they were raised, a significant minority “switch” to other religions (Schwadel, 2010). A unique, and increasingly common, form of religious switching is disaffiliation—switching from religious to non-religious. Since the early 1990s, there has been an unprecedented growth in the fraction of individuals claiming no religious affiliation in the United States. This growth during the 21^st century has increasingly been driven by disaffiliation (Schwadel, 2010). To the extent that religious participation confers a health and well-being benefit, disaffiliates represent a theoretically and empirically significant group.

Explaining why an increasing number of Americans choose to leave the religions in which they were raised has generated increased attention in the sociology of religion (Schwadel, 2014, Vargas, 2012). First, some may leave for religious differences, such as having lost a belief in the spiritual or dogmatic aspects of the church. Theological dissatisfaction and changes in religiosity are important predictors of leaving a religious faith (Hoge, et al., 1995, Sands, et al., 2006). However, research suggests that those with no religious affiliation are not necessarily atheist or agnostic; self-identified atheists make up a very small percentage of the US population (Hout and Fischer, 2002). Individuals may also disaffiliate for political reasons. Hout and Fischer (2002) indicate that the recent increase in religious disaffiliation in the United States is related to the strengthened political alignment of many religious organizations with the political right. Wanting to avoid political connections to the Republican party, many younger individuals choose to identify with no religious affiliation, although these effects appear to be driven by a few salient political issues (Vargas, 2012). Alternatively, dissatisfaction with the social aspects of the religious congregation may encourage individuals to disaffiliate. For instance, some research indicates that religious switching correlates with familial instability at younger ages, and the effect appears to be stronger for disaffiliation (Sandomirsky and Wilson, 1990). Likewise, strong religious-based social networks may be an important “holding factor” keeping individuals affiliated with the church (Wilson and Sandomirsky, 1991).

Disaffiliation, Health, and Well-being

Although there is little evidence regarding the health effects of disaffiliation (Scheitle and Adamczyk, 2010 are an exception), there is a larger literature related to the loss of a source of support more broadly and its effects on health and well-being. This process is most well-developed with respect to marital status, and the negative health and well-being implications of divorce and spousal death (Liu and Umberson, 2008). Individuals who experience marital dissolution through divorce or widowhood report poorer health and lower happiness not just compared to the married, but also compared to those who never marry (Hughes and Waite, 2009). In this case, it is the loss of a source of support that is related to health and well-being rather than simply its presence or absence. This may occur because those who lose a source of social or financial support may have a more difficult time generating suitable substitutes (Sullivan and Fenelon, 2014), while those who simply never had that source of support were able to adapt to its absence with alternative sources of support. Viewed from the perspective of loss, we might expect those who lose the benefits of religious involvement to experience poorer health and well-being than both those who are raised and remain either affiliated or unaffiliated.

Specific denominations differ in the characteristics of religious involvement and thus are likely to vary in the benefits provided to affiliates and to present differing risks for disaffiliates. We consider Evangelical Protestants, Catholics, Mainline Protestants, and high-cost religions (Mormons, Jehovah’s Witnesses and Seventh Day Adventists). Differences that exist in the social, spiritual, and institutional characteristics of these denominations have implications for our expectations surrounding the mechanisms that link disaffiliation to health and well-being.

One pathway linking the loss of support to health and well-being is reduced resources for emotional coping. To the extent that the spiritual aspects of religion provide a source of emotional strength for an individual (Pollner, 1989), disaffiliation presents a challenging scenario in which individuals must compensate for the loss of this source. Individuals may use prayer as a means to reduce experiences of stress during particularly stressful periods of life (Pargament, 2001), which may be more difficult following disaffiliation. Alternatively disaffiliates may experience spiritual struggles that impact both the likelihood of disaffiliation as well as perceived well-being (Ellison, et al., 2013). Disaffiliation may make it more difficult to manage negative or confusing emotions through God, and can exacerbate guilt or negative feelings toward the church (Sharp, 2010). Spiritual effects would be more likely to pertain to disaffiliates from strict denominations, such as Evangelical Protestants and high-cost religious groups. These effects should also have a stronger effect on well-being than physical health, and to remain even after accounting for the social benefits of attendance (Childs, 2010).

Some religious denominations provide strict directives regarding health-related behaviors and lifestyles. Iannaccone (1994) notes that strict churches are distinguished by behavioral control in addition to theological beliefs, and that behavioral directives serve improve health for affiliates. This effect should be stronger for some groups than for others, and particularly high-cost religions (Scheitle and Adamczyk, 2010). Mormons have well-known restrictions on alcohol use and cigarette smoking, as well as other beverages such as coffee and tea (Lyon, 1992), while Seventh Day Adventists also typically avoid soft drinks and red meat, and the church argues for a well-balanced vegetarian diet. Jehovah’s Witnesses do not have explicit restrictions, but the governing body discourages use of alcohol, tobacco, and other drugs, and emphasizes moderation with respect to diet (Penton, 2002). Disaffiliation from these groups is associated with poorer health (Scheitle and Adamczyk, 2010), which may reflect the uptake of unhealthy behaviors. It is unclear whether this would be expected to impact well-being.

Disaffiliation may also impact health and well-being through the loss of social support and social relationships developed through the church. While strong social networks can discourage disaffiliation (Ellison and Sherkat, 1995), disaffiliation may induce significant changes in an individual’s social experience, particularly social ties and relationships (Uecker, et al., 2007). “Affiliation must be understood as a complex of ties, some of them religious, some of them familial. Severing those ties means that the individual must give up not only secondary associations but also primary group commitments” (Sandomirsky and Wilson 1990, p. 1225). Being unaffiliated does not necessarily indicate that an individual is hostile to religion, and many with no religious affiliation continue to attend religious services (Petts, 2009), although they are considerably less likely to attend than Evangelicals, Mainline Protestants, and Catholics. According to the 1998 and 2000 General Social Survey, 63% of those with no religious affiliation never attend church services, compared with just 13% of the affiliated. But there is considerable variation among those with no religious affiliation, with 5–10% reporting that they attend at least somewhat often (Hout and Fischer, 2002).

Although disaffiliation does not preclude an individual from retaining religiously-based social networks (Lim, et al., 2010), certain denominations make that more difficult than others. Supply side theories of religious change predict that stricter churches induce a greater commitment from affiliates (Iannaccone, 1994, Perl and Olson, 2000), which leads to greater social embeddedness of affiliates. Although most social networks are dominated by individuals with similar religious affiliation, this is particularly true for Evangelical Protestants; Evangelicals tend to establish a strong boundary between themselves and non-Evangelicals and are more likely to believe that their religion is the only path to eternal life (Smith 1998). This may foster stronger social ties among Evangelicals through social network closure, orienting affiliates toward church-centered networks (Iannaccone, 1994, Stroope, 2011).

In contrast to Evangelical churches, Catholics do not appear to exhibit stronger commitment as a function of the religious social network (Stroope, 2011). However, Catholic churches may foster strong commitment through familial identity. Catholic identity is also quasi-ethnic (Sandomirsky and Wilson, 1990), and attendance serves to strengthen family ties as well as friendship ties among Catholics, although Catholic ethnic identity may have weakened over time (Sherkat, 2001). Mainline Protestants report having the least religious homogeneity in their family, friend, and neighborhood networks, and are less likely to rely on church ties (Putnam and Campbell, 2010). Mainline Protestants are also more likely to be involved in secular volunteer organizations than other groups (Beyerlein and Hipp, 2006). As a result, they may receive fewer social benefits of religious participation (Wuthnow, 2004), but also experience fewer disadvantages associated with disaffiliation.

Hypotheses

1. We expect disaffiliates to report poorer health and lower subjective well-being than the consistently affiliated

   1. If this effect is driven by a loss of social support, we expect this disadvantage to be mediated by the frequency of church attendance. We also expect strong effects for Evangelical disaffiliates and weak effects for Mainline Protestant disaffiliates.
   2. If this effect reflects health-related behaviors, we expect disaffiliates from high-cost religious groups to report poorer health, but not necessarily poorer well-being.
   3. If this effect is partially driven by the spiritual/emotional characteristics of religion, we expect poorer well-being (but not health) among disaffiliates to remain after controlling for attendance.

2. If disaffiliation is more important for health and well-being than merely having no religious affiliation, disaffiliates should report poorer health and well-being than both the consistently affiliated and consistently unaffiliated.

Data and Methods

Sample

We use data pooled from General Social Survey (GSS) covering the period between 1973 and 2012. GSS is collected and maintained by the National Opinion Research Center at the University of Chicago, and contains detailed information on lifestyle, religion, politics, and opinion over time. The pooled GSS forms a series of repeated cross-sections of nationally-representative population-based samples that are representative of the US non-institutionalized population in the year of the survey. During this period, GSS contains detailed information on religious affiliation, attendance, and observance that allows individuals to be categorized based on standard denominational religion measures. Pooling data across many waves of the survey provides a large sample size allowing us to consider finer religious switching behaviors than is possible with typical smaller samples. 1,464 individuals are removed from the sample because they did not provide sufficient information to classify their religious denomination.¹ The total sample includes 34,565 individuals aged 18 or above pooled across the 28 sample years.

Dependent Variables

A large subset of GSS respondents are asked to rate their general health status, taking everything into account. Respondents report their health as one of four categories: excellent, good, fair, or poor health. Self-reported health is an incredibly widely-used measure of general health. It has the primary benefits of being relatively easy to collect, since it involves individual general assessment of health, all things considered (Jylhä, 2009). While self-reported health indicates perceptions of health rather than objective or clinical measures, it has been shown to be among the best known predictors of subsequent mortality, because individuals take into account many dimensions of health that would not be visible to physicians in a single observation (Schnittker, 2005). We employ the frequently-used dichotomized version of health status into fair/poor versus excellent/good (Idler and Benyamini, 1997).² This dichotomy has been shown to be appropriate with respect to respondent interpretations of the question as well as within-individual differences in reporting (Zajacova and Dowd, 2011).

We also consider a measure of subjective well-being in which respondents report whether they are “very happy”, "pretty happy”, or “not too happy”. Research demonstrates that happiness is a measure of affective well-being which is contrasted with the more global measure of “life satisfaction”. Health and well-being are both positively related to religious participation and have implications for quality of life throughout the life course (Green and Elliott, 2010). However, whereas health is more likely to reflect the indirect behavioral benefits of religious participation (George, et al., 2002), well-being may also respond to the strength of individual religious faith (Ellison and Levin, 1998). Well-being also responds to short-term shocks and represents an accurate assessment of well-being in terms of material resources (Margolis and Myrskylä 2011). We model well-being as an ordinal variable, measuring the odds of moving from very happy to pretty happy or from pretty happy to not too happy.

Independent Variables of Interest

One of the primary benefits of the GSS is that it contains detailed information on religious affiliation, beliefs, and attendance that allows researchers to categorize individuals into sociologically meaningful religious traditions. We use the Steensland et al. (2000) RELTRAD scheme to sort Protestant denominations and identify meaningful religious groups. This approach uses histories of religious traditions to sort by affiliation rather than ideology. This scheme categorizes religious groups into six categories: Mainline Protestant, Evangelical Protestant, Black Protestant,³ Catholic, Jewish, and other (which includes Muslims, Unitarians, and other smaller groups). Respondents can also identify as having no religious affiliation or no religious preference. To date this is the best scheme to classify religious affiliation using the GSS (Woodberry, et al., 2012). Following Scheitle and Adamczyk (2010), we also consider those who affiliate with strict, high-cost religious groups: Mormons, Jehovah’s Witnesses, and Seventh Day Adventists.

In addition to current religious affiliation, the GSS collects valuable data about respondents’ experiences with religion during childhood and adolescence, and respondents report the religious denomination in which they were raised. This follows the suggestion of Hadaway and Marler (1993) that researchers should rely on religious self-identification variables rather than parent’s religion. We construct a variable of religious disaffiliation that takes into account the religion in which a respondent was raised and current religious affiliation. We use four categories: (1) Raised and remain religious (consistently affiliated)⁴ (2) raised religious and switch to no religious affiliation (disaffiliates); (3) raised with no religious affiliation and switch to religious (converters); and (4) raised and remain with no religious affiliation (consistently unaffiliated). In addition to broad categories of religious disaffiliation, we consider disaffiliation from specific denominations, focusing on Evangelical Protestants,⁵ Mainline Protestants, Catholics, and high-cost groups.

Religious attendance is intended to measure the strength of an individual’s relationship to the church-based social network, even if the individual does not strongly affiliate with the spiritual or dogmatic elements of the church (Dupre, et al., 2006). There is some evidence that individuals over-report their frequency of attendance, and this effect is likely to be stronger for stricter groups (Hadaway and Marler, 2005). Attendance is measured using a nine-category scale ranging from never to more than once a week.

Controls

We include controls for socioeconomic and demographic variables that may be correlated with both religious participation and health/well-being.⁶ Demographic covariates are age, gender, geographic region, marital status (married, divorced/separated, widowed, never married), number of children, and race (white, black, other). Socioeconomic variables are educational attainment (less than high school, high school, some college, college graduate), family income quartile (including missing), parents’ education (highest degree earned by respondent’s parent), geographic mobility since childhood (lives in same city, different city in same state, lives in different state), and childhood family experience (lived with both parents, parents divorced, parent died, did not live with both parents for other reason). Family experience is intended to account for the possibility that greater familial instability may be associated with higher likelihood of disaffiliation. All models also control for survey year, a common practice in analyses of pooled cross sectional data.

Analytical Strategy

We model the association between disaffiliation and self-reported health using binary logistic regression predicting the odds of reporting fair or poor health. We model well-being using ordered logistic regression, predicting the odds of moving one category on the scale of well-being towards being “not too happy”. The baseline model includes the measure of disaffiliation and a vector of socioeconomic and demographic controls. The second model for each outcome includes the frequency of religious attendance, intended to reflect whether individuals associate with church congregations even in the absence of religious affiliation. The third model includes an interaction between year and disaffiliation status, in order to examine whether the relationship between disaffiliation and health and well-being has changed over time. The first set of models combines all disaffiliates, and the second set stratifies by specific denomination (Evangelical Protestants, Mainline Protestants, Catholics, and high-cost groups).

Results

Religious Affiliation and Disaffiliation

Although the majority of respondents in the sample were raised and remain religious, it is quite common to switch religious affiliation. Table 1 documents changes in religion from childhood to adulthood. Of 34,565 respondents, 30,230 (87%) were raised and remain religious. Of those raised religious, 2,696 (8%) disaffiliate, reporting no religious affiliation in adulthood. Those raised with no religious affiliation are particularly likely to join a religious tradition: 45% report an affiliation in adulthood. Consistent with the rising fraction of Americans having no religious affiliation, the fraction of individuals disaffiliating has also increased. Between 1973 and 1980, 6% of those raised religious disaffiliated. During the period 2005–2012, 14% did.

Table 1.

Religious Disaffiliation: Religion at age 16 versus adult religion: GSS 1973–2012

Religious Affiliation at age 16	Religious Affiliation
	Affiliated	No Affiliation	Total
Affiliated	30,230 (87%)	2,696 (8%)	32,926 (95%)
No Affiliation	731 (2%)	908 (3%)	1,639 (5%)
Total	30,961 (90%)	3,604 (10%)	34,565 (100%)

Notes: Religious affiliation determined at age 16 and at the time of interview. Percentages of total sample shown in parentheses.

Source: General Social Survey Samples 1973 – 2012.

Consistent with previous research, we find large socioeconomic and demographic differences between religious affiliates and disaffiliates. Table 2 provides a detailed sociodemographic description of individuals by religious disaffiliation behavior. Compared to those who are raised and remain religious, disaffiliates are younger, more likely to be male, less likely to be married, and are more socioeconomically advantaged. They have higher educational attainment, higher incomes, are more geographically mobile, and are more likely to come from well-educated parents. On the other hand, disaffiliates are more likely to have experienced family instability in childhood.

Table 2.

Descriptive Statistics of General Social Survey Sample 1973 – 2012 (N=34,565)

	Total	Affiliation/Disaffiliation
		Consistently Affiliated	Religious -> None	None -> Religious	Consistently Unaffiliated
N	34,565	30,230 (87%)	2,696 (8%)	731 (2%)	908 (3%)
Mean Age	44.5	45.3	39.1	44.0	35.5
Female (%)	56.0	57.5	40.3	59.0	48.4
Race
White (%)	81.6	81.4	86.0	89.1	78.6
Black	13.8	14.6	8.5	6.3	11.7
Other	4.6	4.4	5.5	4.7	9.7
Marital Status
Married (%)	54.8	56.1	41.3a	57.5	36.7
Divorced/Separated	16.0	15.0	19.5a	16.4	16.4
Widowed	8.5	10.4	3.5a	8.6	3.1
Never Married	20.8	18.4	35.5a	17.5	43.8
Mean Number of Children	1.9	2.0	1.2a	2.1	1.1
Education
Less Than High School (%)	21.6	24.0	16.2a	23.5	23.1
High School	30.9	31.7	22.9a	32.4	28.2
Some College	24.1	22.9	28.4a	22.4	24.9
College Degree	23.4	21.2	32.4a	21.6	23.7
Income Quartile
Bottom (%)	24.2	24.2	21.7a	27.8	31.4
Bottom-Middle	25.5	25.5	25.6	24.5	26.3
Upper-Middle	25.0	25.0	25.5	25.6	21.0
Upper	25.3	25.3	27.3a	22.1	21.2
Parental Education
Less Than High School (%)	38.6	41.8	23.9a	38.5	22.7
High School	42.3	40.7	48.2a	44.0	49.8
Some College	3.2	2.9	4.3a	3.1	3.6
College Degree	16.0	14.7	23.7a	14.4	23.9
Childhood Family Situation
Lived with both parents (%)	74.2	75.2	69.4a	66.0	62.4
Parents divorced	12.4	11.3	18.3a	18.6	24.2
Parent died/other	13.4	13.5	12.2	15.4	13.4
Mobility since age 16
Same City (%)	40.9	41.7	35.6a	34.5	43.5
Difference City, Same State	25.9	25.4	26.0	29.0	26.0
Different State	33.2	32.9	38.3a	36.5	30.5
Self-Reported Healthb
Fair/Poor (%)	22.6	23.9	23.4	23.9	20.8
Good/Excellent	77.4	76.1	76.6	76.1	79.2
Subjective Well-Beingb
Very Happy (%)	31.9	33.0	25.2a	29.2	26.3
Pretty Happy	55.9	55.0	59.7a	58.2	58.3
Not too Happy	12.2	12.0	15.1a	12.7	15.3
Attendance
Never (%)	15.6	9.9	60.7a	14.6	71.2
At Least Once a Year	33.5	33.8	34.2a	29.4	24.2
At Least Once a Month	22.3	24.6	3.4a	20.1	3.0
At Least Once a Week	28.7	31.8	1.7a	35.8	1.7

Notes: Reflects demographic and socioeconomic differences among religious affiliation groups (columns).

^aSignificantly different from those who remain religions at the p<.05 level.

^bSelf-reported health and well-being percentages are age-standardized.

Without adjusting for controls, disaffiliates report similar health and lower subjective well-being than the consistently affiliated. 23.9% of the consistently affiliated report fair or poor health, compared to 23.4% of disaffiliates. The lack of a large difference here primarily reflects the fact that disaffiliates are more likely to be male and have higher socioeconomic status than the consistently affiliated. 15% of disaffiliates and those who are raised and remain with no religion report that they are “not too happy” compared with only 12% of the consistently affiliated. The frequency of religious attendance varies considerably across groups. 32% of the consistently affiliated attend at least once a week, and 56% attend at least once a month. Disaffiliates and consistently unaffiliated attend rarely, with only 5% of each group attending at least once a month. However, disaffiliates are more likely than those raised none to attend a few times a year, with 40% reporting that they attend church at least sometimes.

Health and Well-Being

Disaffiliates tend to report worse health and well-being than those who remain in the religion in which they were raised, net of sociodemographic controls. The models in Table 3 predict fair/poor health and reporting lower well-being in adulthood as a function of disaffiliation.⁷ Model 1 predicts health and includes demographic characteristics and socioeconomic background variables. Compared to those who are raised and remain religious, disaffiliates have 21% higher odds of perceived fair/poor health. The effect size is similar to that of being never married (OR 1.18) or having a parent with a college education (OR 0.74). Those who are raised and remain with no religion exhibit no health disadvantage compared to the consistently affiliated, and considerably better health than disaffiliates. The same is true of those who are raised with no religion and switch to being religious in adulthood (converters).

Table 3.

Odds ratios predicting fair/poor health and subjective well-being as a function of religious disaffiliation using General Social Survey 1973 – 2012

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Fair/poor Health	Fair/poor Health	Fair/poor Health	Well-Being	Well-Being	Well-Being
Marital Status
Married (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Divorced/Separated	1.17**	1.16*	1.16**	2.78**	2.65**	2.64**
Widowed	1.01	0.93	0.92	2.76**	2.74**	2.74**
Never Married	1.18**	1.16**	1.10*	1.97**	1.92**	1.92**
Number of Children	1.02**	1.03**	1.04**	1.00	1.01	1.01
Education
Less Than High School (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
High School	0.56**	0.58**	0.58**	0.89**	0.92**	0.92**
Some College	0.48**	0.50**	0.51**	0.81**	0.84**	0.84**
College Degree	0.35**	0.37**	0.38**	0.74**	0.80**	0.80**
Income Quartile
Bottom (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Bottom-Middle	0.62**	0.62**	0.63**	0.80**	0.80**	0.80**
Upper Middle	0.48**	0.49**	0.51**	0.68**	0.68**	0.68**
Upper	0.35**	0.36**	0.37**	0.53**	0.53**	0.53**
Missing	0.67**	0.67**	0.67**	0.69**	0.69**	0.69**
Parental Education
Less Than High School (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
High School	0.82**	0.81**	0.80**	0.94*	0.93**	0.93**
Some College	0.85	0.86	0.84	0.93*	0.93	1.94
College Degree	0.74**	0.74**	0.72**	0.83**	0.82**	0.82**
Childhood Family Situation
Lived with both parents (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Parents divorced	1.06	1.03	1.02	1.15**	1.12**	1.12**
Parent died/other	1.05	1.04	1.04	1.06	1.05	2.05
Mobility since age 16
Same City (%)	1.00	1.00	1.00	1.00	1.00	1.00
Difference City, Same State	0.98	0.97	0.98	0.99	0.98	0.98
Different State	0.95	0.95	0.96	1.05	1.04	1.04
Affiliation/Disaffiliation
Consistently Affiliated (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Religious -> None (Disaffiliates)	1.21**	1.00	0.97	1.30**	1.05	1.25*
None -> Religious	0.99	1.00	1.01	1.15	1.20*	1.12
Consistently Unaffiliated	0.97	0.80*	0.77	1.13	0.93	1.20
Attendance
Never (ref.)		1.00	1.00		1.00	1.00
< Once a Year		0.87*	0.87*		1.04	1.04
Once a Year		0.87*	0.87*		0.94	0.94
Several Times a Year		0.82**	0.82**		0.85**	0.85**
Once a Month		0.74**	0.74**		0.78**	0.78**
2–3 Times a Month		0.78**	0.78**		0.77**	0.77**
Nearly Every Week		0.67**	0.67**		0.69**	0.69**
Every Week		0.59**	0.59**		0.62**	0.62**
More Than Once a Week		0.62**	0.62**		0.44**	0.44**
Disaffiliation-Year Interaction
Disaffiliate × Year			1.00			0.99*
No. of Observations	34,565	34,565	34,565	34,565	34,565	34,565

Notes: Models 1–3 predict the odds fair/poor health. Models 4–6 predict lower well-being as an ordered logit. Models 2 and 5 add frequency of church attendance. Models 3 and 6 add an interaction between disaffiliation and year. All models also control for age, sex, race, geographic region, and year of interview.

^*p <0.05

^**p <0.01.

Model 4 predicts subjective well-being as a function of disaffiliation using an ordered logit.⁸ Disaffiliates are considerably more likely to report lower subjective well-being than those who remain religious (OR 1.30, p<0.01). As with self-reported health, a well-being disadvantage exists only for disaffiliates. There are no statistically significant well-being disadvantages for the consistently unaffiliated and those who are raised unaffiliated but affiliate with a religious tradition in adulthood.

Attendance

Differences in the frequency of religious attendance across affiliation/disaffiliation groups mediate differences in health and well-being. Models 2 and 5 in Table 3 add the frequency of church attendance to Models 1 and 4, respectively. The health and well-being disadvantages of disaffiliates are entirely mediated by the frequency of religious attendance. After the inclusion of this variable, health and well-being differences between the consistently affiliated and disaffiliates are substantially attenuated and lose statistical significance.

Time Trends

Models 3 and 6 add an interaction between disaffiliation and year in order to assess change in the disaffiliate disadvantage over time. While there is no significant change in the health disadvantage, there is a statistically significant decline in the well-being disadvantage of disaffiliates since 1973.

Disaffiliation from Specific Denominations

Considering specific denominations can shed light on some of the underlying mechanisms responsible for the disaffiliate disadvantage. Table 4 shows the frequency of individuals who disaffiliate from major religious traditions in the United States: Evangelical Protestants, Mainline Protestants, Catholics, and high-cost groups. A similar fraction of those raised in each major religious tradition disaffiliate (around 7–10%), while a greater fraction of high-cost groups disaffiliate (17%). 76% of those raised Evangelical Protestant remain affiliated in adulthood, compared to 78% of those raised catholic, and just 70% of those raised Mainline Protestant, and 70% of those raised in a high-cost group.

Table 4.

Disaffiliation from Specific Denominations: GSS 1973–2012

Religious Affiliation at age 16	Affiliation/Disaffiliation
	Consistently Affiliated	Other Religion in Adulthood	Disaffiliate	Total
Evangelical Protestant	6,606 (76%)	1,490 (17%)	633 (7.2%)	8,729
Catholic	7,937 (78%)	1,349 (13%)	953 (9.3%)	10,239
Mainline Protestant	5,851 (70%)	1,770 (21%)	691 (8.3%)	8,312
High-Cost Groupa	322 (70%)	74 (16%)	61 (13%)	457

Notes: Religious affiliation determined by the scheme developed by Steensland et al. (2000). Percent of those raised in each religion in parentheses. Other religions refer to all other religious traditions besides the focal tradition and no religion.

^aIncludes Mormons, Jehovah’s Witnesses, and Seventh Day Adventists

Source: General Social Survey Samples 1973 – 2012.

The results in Table 5 examine differences in health and well-being between individuals who disaffiliate from these traditions and those who are consistently affiliated. We find health disadvantages for those who disaffiliate from Evangelical (33% higher odds of fair/poor health) and Catholic congregations (27%). High-cost group disaffiliates experience the greatest disadvantage in terms of health, having 2.7 times the odds of fair/poor health compared to those remaining affiliated. Mainline Protestant disaffiliates do not report significantly poorer health. The health disadvantages of Evangelical Protestant and Catholic disaffiliates are entirely mediated by the frequency of religious attendance, and each coefficient loses statistical in Model 2. Model 3 examines whether these disadvantages have changed over time. None of these interactions is statistically significant.

Table 5.

Odds Ratios of Fair/Poor Health and Well-being as a Function of Religious Disaffiliation from Specific Denominations

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Affiliation/Disaffiliation	Fair/Poor Health			Well-beinga
Remain Evangelical (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Evangelical -> None	1.33**	1.09	1.01	1.54**	1.21*	1.05
Year × Disaffiliate			1.00			1.02
Remain Catholic (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Catholic -> None	1.27**	1.04	1.17	1.19*	0.97	1.42*
Year × Disaffiliate			0.99			0.98**
Remain Mainline (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Mainline -> None	1.05	0.90	0.88	1.27*	1.10	1.22
Year × Disaffiliate			1.00			1.00
Remain high-costb (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
High-cost -> None	2.75**	2.18*	4.40	1.22	0.90	0.44
Year × Disaffiliate			0.97			1.04

Notes: All models control for sociodemographic characteristics and year of interview. Models 2 and 5 add frequency of church attendance. Models 3 and 6 add an interaction between disaffiliation and year.

^aModeled as an ordered logit predicting the odds of moving one category towards lower well-being

^bIncludes Mormons, Jehovah’s Witnesses, and Seventh Day Adventists

^*p <0.05

^**p <0.01.

Model 4 demonstrates that Evangelical disaffiliates are more likely to report lower subjective well-being than those who remain Evangelical (OR 1.54), while Catholic and Mainline Protestant disaffiliates experience modest disadvantages (OR 1.23 and 1.27 respectively). Disaffiliates from high-cost groups show no significant disadvantage in well-being. The frequency of attendance in Model 5 completely mediates the disadvantage for Catholic and Mainline Protestant disaffiliates, and the coefficients lose statistical significance. Although attendance substantially mediates the disadvantage for Evangelical disaffiliates, it remains statistically significant (OR 1.22, p=0.02). Model 6 demonstrates that the decline over time in the well-being disadvantage of disaffiliates is driven by the experience of Catholics.

Discussion

Sociologists of religion have focused on the social implications of religious participation and institutions, and have been particularly adept at identifying the social factors contributing to the relationship between religion and health and well-being. The implications of strong social ties for meaningful outcomes of health and well-being have been central to much sociological research on health since the early 20^th century (Berkman, et al., 2000). Recent increases in the fraction of those claiming no religious affiliation has brought renewed focus on the relationship between religious involvement, social relationships, and health. Despite growing importance of understanding those who leave religious traditions, previous research has largely focused on the institutional processes surrounding religious switching or the social determinants of switching behavior. Research on the consequences of religious disaffiliation is more nascent (Schwadel, 2010), and the relationship between disaffiliation and individual well-being remains somewhat unclear. We contribute to this literature by identifying an adult health and well-being disadvantage associated with religious disaffiliation, which is largely mediated the social aspects of religious congregations through attendance, and varies by specific denomination of origin.

Our study joins the findings of Scheitle and Adamczyk (2010) that disaffiliation is related to self-reported health. We extend these results to demonstrate the association between disaffiliation and subjective well-being, as well as investigating denomination-specific differences in the relationship between disaffiliation and health, drawing additional insights from the context-dependent nature of the disaffiliation process. Beyond Scheitle and Adamczyk we identify important ways in which well-being differs from health with respect to the consequences of disaffiliation, demonstrating that high-cost religious disaffiliates experience no disadvantage in well-being.

Theoretical considerations surrounding these patterns are applicable to the expanding literature on religious disaffiliation and its consequences for the individual, particularly when disaffiliation is framed in terms of the loss of support provided by religious involvement (Hypothesis 2). Although our data are cross-sectional, our study design thus improves upon previous analyses that have relied on comparison of religious and non-religious individuals, and our results stand in contrast to expectations that having no religion by itself has unfavorable health implications (Ellison and Levin, 1998). We find evidence for the notion that health and well-being erodes only for those who disaffiliate, which has an interesting corollary with respect to other forms of social support, particularly spousal support. Most studies of marital status and health compare divorced and widowed individuals to the continuously married, showing that both types of spousal loss are associated with lasting declines in health, which are often attributed to reductions in social support (Sullivan and Fenelon, 2014). However, in more recent years divorced and widowed individuals often report poorer health than those who never marry (Liu and Umberson, 2008), indicating that there is something unique about the loss of social support that reduces health. This is an important theoretical contribution in the literature on religious affiliation and health, indicating that the social support deriving from religious involvement is important for health, and health may suffer if ties are severed (Ellison and George, 1994).

Although not the primary focus of the analysis, our results have implications for understanding the health of those raised with no affiliation but who affiliate with a religion in adulthood. Religious “converters” represent a theoretically important group, since they presumably have access to the social benefits of religion, but do not exhibit better health and well-being than the consistently unaffiliated. This finding suggests that gaining the social benefits of religion may not improve health and well-being relative to social ties developed among the consistently unaffiliated, further clarifying the importance of loss in the religion-health relationship. This may reflect the fact that social networks can directly induce religious conversion through changing preferences, or that individuals may convert if they are seeking new social ties (Loveland, 2003). Among unaffiliated individuals, those successful at developing secular social networks may be less likely to convert, and those who are satisfied with their social networks are not likely to be at any additional social disadvantage. The finding of similar health for both converters and the consistently unaffiliated further demonstrates the distinctiveness of disaffiliates and represents an important theoretical insight in research on religious affiliation and health.

Our results are consistent with research suggesting that the frequency of church attendance is an important mediator of the relationship between religion, health, and well-being (Dupre, et al., 2006, Sullivan, 2010). Although adults have many possible sources of social support, religion is for many the primary source of social ties outside the family (Ellison and George, 1994), and recent research indicates that frequent church attendance is associated with developing social networks that provide benefits in secular as well as religious contexts (Lewis, et al., 2013). Our study adds an additional layer to this literature, demonstrating that individuals who disaffiliate but continue to maintain regular attendance experience no health or well-being disadvantage (Hypothesis 1a). This is significant given that secular sources of social engagement and social ties do not fully compensate for the absence of church-based social networks (Acevedo, et al., 2014).

The results for Evangelical Protestants and Catholics largely support the role of social ties (Hypothesis 1a). Although some disaffiliates maintain their religiously-based social networks (Lim, et al., 2010), the social dynamics of certain denominations discourage this. Social networks among Evangelical Protestants are more likely revolve around the church (religious homophily), and they are less likely than Mainline Protestants to have extensive secular networks of social ties (Smith, 1998). The strong divide that Evangelical Protestants perceive between themselves and non-Evangelicals reinforces their network religious homophily (Putnam and Campbell, 2010), and suggests that more social ties are severed at disaffiliation (Merino, 2014). The same is true for Catholics, particularly Latinos, whose religious affiliation is more likely to reflect the unity of familial relationships and friendships (Calvillo and Bailey, 2015). Overall, Catholics are more likely than other denominations to indicate the importance of family relationships in continued association with the church (Smith and Sikkink, 2003). The importance of church-based social ties for Catholics and Evangelicals suggests a role for social integration. Classical sociological theory linked religion to lower risk of suicide via social integration (Durkheim, 1951 [1897]), which has had great impact on sociological understandings of religion and health. This insight remains relevant in the modern US context as well; Evangelical Protestants and Catholics enjoy more extensive and supportive social networks than Mainline Protestants, which is partially reflected in lower suicide rates (Pescosolido and Georgianna, 1989). Mainline Protestants are less likely to depend on the church for their social networks and may be more successful at maintaining social relationships and engagements in a secular context (Wuthnow, 2004). However, Mainline Protestant disaffiliates do report a well-being disadvantage, which may reflect the greater benefits or religious social support as compared to secular social support (Krause, 2006).

We also find some evidence of the importance of the spiritual aspects of religion, largely with respect to the well-being of Evangelical disaffiliates (Hypothesis 1c). The frequency of attendance attenuates the well-being disadvantage, but does not completely mediate it. We do not find a residual disadvantage for any other denomination after adjusting for attendance, perhaps reflecting that religious belief is more central to the spiritual lives of Evangelical Protestants (Kooistra and Pargament, 1999). Moreover, previous work suggests that the linkage between religious participation and health is stronger for Evangelical Protestant congregations than for other Protestants (Maselko and Kubzansky, 2006). As a result, disaffiliation may reduce resources for emotional coping, and may raise issues surrounding preoccupation with human sin, divine retribution, or guilt, which can impact cognitive well-being (Exline, 2002). Strong focus on human sin is also related to increased anxiety, which may worsen following disaffiliation (Ellison, et al., 2009). In any case, it appears that the well-being disadvantage for Evangelical disaffiliates goes beyond the social aspects of religion.

The health disadvantage of those who disaffiliate from high-cost groups may partially reflect the loss of behavioral regulation, a particularly important characteristic of Mormons, Jehovah’s Witnesses, and Seventh Day Adventists (Hypothesis 1b). Although this health disadvantage may have to do with the social benefits of religion (Scheitle and Adamczyk, 2010), it would not explain why this group does not experience a similar disadvantage in well-being. This is a surprising finding given that Evangelical and Catholic disaffiliates report disadvantages in both health and well-being. This may suggest that behavioral regulation has a protective effect on the health of those consistently affiliated with high-cost religions, along with social benefits (Sloan and Bagiella, 2002). There is considerable evidence of the salutary effects of healthy behaviors established within these groups, such as proscriptions against smoking or heavy drinking, which might be expected to have a stronger impact on health and less of an effect on well-being. The role of behavioral change in the health of disaffiliates is a promising topic for future research.

Finally we demonstrate that the well-being disadvantage associated with disaffiliation appears to be declining over time, and this is driven largely by the experience of Catholics. This suggests that since at least the early 1970s, the well-being disadvantage of Catholics who disaffiliate has been reduced. This pattern may reflect temporal changes in the strength of ethnic identity of Catholics. As many Catholic ethnic groups, such as Italians and Irish, are now considered white, their ethnic distinctiveness has declined, and the connection of Catholic identity to ethnic identity has weakened (Ignatiev, 2009). Alternatively, the rise of cultural Catholic identity means that Catholic disaffiliates in more recent years may be less likely to forfeit support from Catholic institutions. As more Catholics identify with Catholic ethnic identity and heritage, leaving the church may have fewer implications for loss of support.

Limitations

There are some limitations to our analysis that largely reflect issues of data availability and complexity. First, our data are cross-sectional and lack the ability to make explicitly causal statements surrounding the effects of disaffiliation. Our data use retrospective accounts of religious affiliation in childhood but we are unable to directly measure the reasons that individuals choose to disaffiliate. Although the GSS sample allows us to consider several measures of socioeconomic and family background, longitudinal data are necessary to document the process of disaffiliation and its specific consequences for social ties and well-being. As a result, there remains a possibility that unobserved differences between the consistently affiliated and disaffiliates can explain some of the health disadvantage of disaffiliates. For instance, individuals who are less happy and feel socially disconnected in their religious congregations may be more likely to leave these congregations (Krause and Ellison, 2009). It is possible that less healthy people cannot attend, and that is explaining some of the results. We would expect health to impact religious attendance (Musick, et al., 2004), which could lead to greater disconnectedness from the religious congregation and make disaffiliation more likely. Likewise, less happy people may be less likely to attend church regularly, which could increase the chance that they disaffiliate. As a result, poor health and well-being might be an underlying cause of disaffiliation, and could partially confound the relationship we observe here (Regnerus and Smith, 2005). However, it is less likely that this pattern accounts for denominational differences in the relationship between disaffiliation and health. Our results should not be interpreted as reflecting the causal effect of religious disaffiliation on health/well-being but instead as a window into the various mechanisms that might explain the association. Nevertheless, our study provides an important step towards understanding the role of changes in religious affiliation in reported health and well-being in adulthood, emphasizing the context-dependent nature of religious disaffiliation and the implications of a loss of social support for subsequent health and well-being. Future research should expand our understanding of the social processes surrounding disaffiliation, and how the quality of religious social networks may function as both a cause and a consequence of disaffiliation.

Conclusion

Our study contributes to knowledge about how religious participation is related to individual health by accounting for changes in religion from childhood to adulthood. We demonstrate that these changes are associated with adult health and well-being, which adds an additional layer to previous understandings of the mechanisms linking religious affiliation to health and well-being. To the extent that religion provides a major source of social support for many adults, changes in affiliation represent important processes in quality of life and well-being over time. As the population of individuals with no religious affiliation has grown, sociologists have placed greater emphasis on the individual, both in terms of the processes surrounding religious disaffiliation as well as the consequences of disaffiliation for individual well-being (Schwadel, 2010).

The rapid increase in religious “nones” since 1990 has generated some concern that Americans are becoming detached from the primary societal sources of social support and integration. Despite these worries, there had previously been no comprehensive studies of the health and well-being consequences of leaving one’s religion, a gap that our analysis seeks to fill. These worries may be somewhat overstated, as our study joins others that point to the overwhelming importance of social integration and social relationships, factors that appear to be driven more by religious attendance than by affiliation. Moreover, our finding that those raised without religion experience no disadvantage is important for demonstrating that simply having no religion does not imply poor health. Our study joins other research asserting that the social aspects of religion, as measured by frequent attendance at religious services, is the primary source of the health benefits for the religiously affiliated. Future work should increasingly focus on the distinct experiences in different denominational congregations, both for affiliates and disaffiliates, as these differences can highlight the importance of social ties across religious contexts.