Religion, Race/Ethnicity, and Norms of Intergenerational Assistance among Older Adults

Abstract

Using data on adults ages 55 and over from the second wave of the National Survey of Families and Households (NSFH-2), this study models the main and interactive effects of religious involvement and race/ethnicity on four items of attitudes towards intergenerational assistance. Results indicate that African Americans and Hispanics tend to express stronger support for intergenerational assistance than non-Hispanic Whites. Conservative Protestants, Mormons, and Catholics are more likely than others to believe that adult children should offer co-residence to their aging parents. In addition, theological conservatism is positively associated with support for each type of intergenerational aid, and the net effect of theological conservatism is stronger for African Americans than for non-Hispanic Whites. However, religious attendance is statistically unrelated to norms of intergenerational assistance. It is concluded that religious factors are important in shaping norms of intergenerational support, particularly within minority communities.

1. Introduction

In recent years, researchers have documented the growing size and racial/ethnic diversity of America’s aging population. According to projections by the U.S. Census Bureau, by 2050 there will be more than 88 million older adults (aged 65 and over), constituting around 20% of the U.S. population. Moreover, in 2010, minority group members comprised approximately 20% of the older adult population; this figure is projected to rise to 41.5% by 2050 [1,2]. Scholars and policy makers have expressed alarm over the possible fiscal implications of these demographic shifts. These concerns reflect growing recognition of the escalating costs of caring for the older adult population, alterations in the “social contract,” domestic social policies, and funding priorities, as well as the fact that older adults of color tend to have more health problems and fewer financial resources with which to deal with health care and other expenditures [1,3,4]. The available data indicate that older Americans from racial and ethnic minority backgrounds may be more likely to use certain formal supports, but are less likely to utilize institutions for long-term care, and rely much more on help from family members, as compared with their non-Hispanic white counterparts [1,4]. The topic of intergenerational assistance has taken on added importance due to recent economic recession and labor market shifts, which have increased pressure on older adults to provide financial and co-residence support to their adult children [5,6].

In response to the changing demographic portrait of the United States, numerous studies have explored the patterns and correlates of intergenerational aid, including financial assistance and co-residence, and the attitudes and values that may influence such familial support [7,8,9,10]. Prior theory and research have identified a number of influences on norms and attitudes regarding intergenerational exchange. A partial list of these factors includes emotional attachment, proximity, and frequent interaction among family members, as well as resource availability and perceptions that previous exchanges have been reciprocal [8,9,11,12]. Parental divorce and remarriage in late life also appear to influence norms and practices of intergenerational assistance [10].

Clearly, we need a better understanding of norms and practices of intergenerational aid within diverse segments of the population. To date, however, research on cultural diversity in this area has centered almost entirely on racial and ethnic differences [13,14]. This literature has tended to show that Asian and African Americans hold stronger norms of most types of support than non-Hispanic Whites, although findings involving Latinos have been somewhat inconsistent. Despite the resurgence of interest in linkages between religion and aging [15] and religion and family life [16,17], religious variations in norms of intergenerational assistance have received limited attention from researchers. In one notable exception to this general pattern of neglect, Goldscheider and Lawton reported that fundamentalist Protestants and Catholics were especially prone to agree that adult children should provide co-residence to parents who needed it, whereas Jewish Americans—and to a lesser extent fundamentalists and Catholics—were particularly supportive of parents providing co-residence to adult children experiencing difficulty [18]. In a subsequent study of patterns of intergenerational exchange, Myers found that: (1) religious congruence between parents and adult children was positively associated with intergenerational support; (2) congruence was more important than individual religiosity; and (3) the effects of religious congruence surfaced only among continuously married parents and adult children [19].

Our study contributes to the modest literature on religion and norms of intergenerational assistance in the following ways. First, we outline a series of arguments linking multiple dimensions of religious involvement—i.e., affiliation, practice, and belief—with norms of intergenerational assistance among older adults. We examine views on four issues: i.e., adult children’s duties to provide financial assistance and co-residence to aging parents, and aging parents’ duties to offer similar types of aid to adult children in need. Although the broader literature on norms and practices of intergenerational support has explored various domains of assistance, co-residence and financial help are among the most important and widely studied types of intergenerational aid. Next we discuss several possible reasons to anticipate racial/ethnic differences in the nature of these relationships, focusing on African Americans, Latinos, and non-Hispanic Whites. We then test relevant hypotheses using data on older adults drawn from a large nationwide U.S. survey. Key implications of these findings are discussed, and study limitations and promising directions for future investigation are identified.

2. Theoretical and Empirical Background

How and why might aspects of religious involvement be related to support for norms of intergenerational support among older Americans? Several generations of social scientists have recognized that religion is a multidimensional construct, and consequently that specific dimensions of individual religiousness may differ in their association with a common outcome of interest [20,21]. This study focuses on three dimensions that are commonly examined in the areas of aging and family studies: (a) organizational religious involvement, often gauged in terms of (self-reported) frequency of attendance at services; (b) theological conservatism, typically measured via support for conservative doctrines such as biblical inerrancy; and (c) religious affiliation, gauged via religious or denominational preference.

First, organizational religious involvement has been associated with stronger commitment to conventional family values and lifestyles in numerous previous analyses [22,23,24,25]. Religious institutions and communities generally support traditional beliefs and practices regarding family life [16,17,23]. This endorsement may take the form of formal messages, rhetoric from clergy, denominations, classes and seminars on family life, and other institutional mechanisms. In addition, informal interactions with coreligionists may strengthen norms regarding family solidarity and intergenerational bonds [24,25]. Moreover, most religious traditions and belief systems encourage members to focus on altruistic values, and they tend to promote models of family life that encourage individuals to sacrifice individual self-interest and gratification for the good of other family members, and for the good of the family unit as a whole [26]. Religious communities seek to foster a spiritual focus on emotions and qualities such as love, kindness, gratitude, and respect, along with forbearance and forgiveness (e.g., with regard to past conflicts, etc.), and reconciliation with others. Finally, religious institutions often provide a context in which informal exchanges of assistance take place, and offer formal programs, counseling resources, and other types of material, socio-emotional, and informational aid to persons and families in stress [27,28,29]. Taken together, these mechanisms seem likely to promote stronger norms regarding affection, harmony, and commitment within families, suggesting the following hypothesis:

H1: The frequency of religious attendance will be positively associated with acceptance of norms of intergenerational support.

Second, in addition to overall religiosity, theological conservatism—and particularly beliefs in the inerrancy and authority of the religious scripture—may also be linked with older adults’ norms regarding intergenerational assistance. Within the Christian tradition, convictions about the proper interpretation of scripture and the centrality of the Bible as a guide and yardstick for human affairs are hallmarks of contemporary evangelicalism [30]. Once associated primarily with a narrow range of conservative Protestant denominations, this theological tendency now has at least some presence in virtually all moderate and liberal (mainline) Protestant denominations, and even among certain segments of Catholicism [31]. These are precisely the elements that support renewed emphasis on “family values” and family-centered lifestyles, and the identification of this familial ethic as an essential part of Christian teaching. Conservative religious beliefs have been linked with an array of family-related outcomes, including sexuality and cohabitation, fertility, gender roles, and childrearing practices, among others (for review, see Ellison and Xu, 2014) [16].

Any approach to the interpretation of a complex, multivocal text (including the Bible) tends to highlight certain themes and passages over other possible points of emphasis. Contemporary evangelicals who embrace biblical inerrancy often emphasize conservative readings of scriptural passages on family-related matters such as sexuality, marriage, gender roles, and childrearing. However, numerous passages in the Old and New Testaments extol the virtues of loyalty, generosity, and care for family members, and older adults in particular (Ex. 20:12; Deut. 5:16; Eph. 6:1–3). Indeed, the Bible repeatedly instructs the faithful to honor and respect parents, and to repay their love and compassion when they become frail (Prov. 23:22; Matt. 15:3–9; Mark 7:9–13; Eph. 4:31–32; Tim. 5:4). Age is valorized in many biblical passages, and older adults are revered in the Bible as sources of worldly and spiritual wisdom. For their part, parents are enjoined to treat their children with love and compassion, to teach them well, and to avoid alienating them or provoking them to anger. Taken together, this line of argument suggests a second hypothesis:

H2: Theological conservatism—i.e., belief in the inerrancy and authority of scripture—will be positively related to support for norms of intergenerational aid.

Third, there are sound reasons to anticipate subcultural differences, i.e., distinctive patterns among specific denominations and religious traditions, even when religious commitment and doctrinal conservatism are taken into account. For example, several major evangelical Protestant denominations, such as the Southern Baptist Convention, and parachurch ministries, such as Focus on the Family, have published materials advising their members on options for appropriate care of aging parents [32]. Although they stop short of advocating co-residence or other specific strategies, many evangelicals endorse family-centered approaches to helping older adults as an alternative to expanded government involvement or more impersonal or bureaucratized forms of care. Other conservative and sectarian faiths may take even stronger positions on this issue. The Latter Day Saint (Mormon) tradition has long embraced the view that adult children should help aging parents in every way needed, while maintaining an ethic of self-reliance on the part of all able-bodied adults [33,34].

Catholic tradition and teaching have also stressed the importance of loving, generous familial relationships. Although earlier generations of U.S. Catholics tended to hold more traditional family norms than Protestants [35], ethnic assimilation and other factors have diminished the distinctiveness of non-Hispanic white Catholics relative to other religious groups [36,37]. Nevertheless, residual ethnic influences, combined with more recent immigration from Latin America and other parts of the world, could lead to elevated levels of Catholic support for intergenerational assistance and co-residence for family members.

Finally, the literature suggests that Jewish Americans may exhibit distinctive norms of intergenerational assistance. Although Jewish ethnic traditions valorize familial solidarity, on average, older Jewish Americans are less religious (by virtually any conventional indicator) than other older adults, and they also tend to be more affluent. According to Cantor and Brennan [38], older Jewish Americans also tend to value (a) their own independence, and seek to avoid burdening their children, and (b) the socioeconomic mobility of their children, and try to avoid interfering with the education, careers, or financial success of their children. Thus, they may be less likely than other older adults to require financial help, and they may tend to purchase caregiving aid (e.g., nursing, in-home care) and co-residence as needed [38]. Taken together, this brief discussion suggests the following broad hypotheses concerning religious subcultural patterns:

H3a: Conservative Protestants and members of sectarian groups (e.g., Mormons) will express comparatively strong support for norms of intergenerational assistance.

H3b: Catholics will express comparatively strong support for norms of intergenerational assistance.

H3c: Jewish respondents will express comparatively low levels of support for norms of intergenerational assistance.

3. Racial/Ethnic Differences in the Role of Religion

To this point we have discussed the links between multiple dimensions of religious involvement and norms of intergenerational aid. In addition to these hypothesized associations, there are also several reasons to expect them to vary according to race and ethnicity. Why might this be the case? Older adults tend to be more religious, in practice and belief, than younger persons, and persons from racial and ethnic minority backgrounds—especially African Americans—tend to exhibit particularly high levels of religious attendance, private religious devotion, religious identity, and doctrinal conservatism [29]. In addition, many scholars have argued that older Latinos tend to be at least as religious as their non-Hispanic white counterparts, and embrace conservative norms and teachings [39,40].

For African Americans, especially older adults, religious institutions and values have long been centers of collective life, and the Church was a key source of vitality for decades of civil rights activism [29,41]. Moreover, religious faith and practice have provided emotional strength and moral guidance for many older African Americans, assisting them via coping skills, formal and informal social support systems, and sustaining feelings of worth and dignity in the face of discrimination and economic marginality [27,28,42]. For generations, the church and the family were the only two institutions that were developed and operated by and for African Americans, largely beyond the scope of White influence. Thus, it is surprising that more empirical studies have not examined the influence of religious communities or values on the families of African Americans, especially in later life. However, a small but growing body of literature has linked religiousness and spirituality with various aspects of older African Americans’ family lives, including: (a) positive marital and relationship quality [43]; (b) more frequent and satisfying familial interactions [44]; and (c) closer relationships between aging parents and their adult children [45].

Among older Hispanics, religious ties and practices are also important, sustaining identity and ethnic community, and fostering emotional wellbeing and physical health [40,46,47]. Although a long tradition of theory has linked religion with family life among U.S. Latinos [39,48], only a few studies have investigated such connections empirically [49], and to our knowledge none of these focuses on older Latinos. However, the logic of the preceding arguments offers a basis for two final hypotheses:

H4a: The association between theological conservatism and norms of intergenerational aid will be stronger among African Americans and/or Latinos, as compared with non-Hispanic Whites.

H4b: The association between religious attendance and norms of intergenerational assistance will be stronger among African Americans and/or Latinos, as compared with non-Hispanic Whites.

4. Date and Measures

4.1. Data

The data for this study came from the second wave of the National Survey of Families and Households [50], a nationwide cross-sectional probability sample of 10,008 surviving members of the original sample (aged 18 and over) residing in the 48 contiguous United States. The survey was designed by the Center for Demography and Ecology at the University of Wisconsin-Madison, and field work for the second wave was conducted by the Institute for Survey Research at Temple University during 1992–1993. Although these data were not collected recently, we are aware of no other large-scale survey dataset that contains the information on norms of intergenerational assistance and multiple dimensions of religion, as well as the racial and ethnic diversity, needed to test our study hypotheses. In fact, researchers continue to use NSFH data and publish their findings in highly recognized social science journals [51]. In addition, the third wave of the National Survey of Families and Households (NSFH-3), conducted during the 2001–2003 period, focused heavily on childrearing and child outcomes, and that some of the key survey items needed for our study were unavailable in the NSFH-3. Therefore, despite its age, the NSFH-2 is the best available source of data with which to conduct a study of this topic because it has requisite items on the dependent variables as well as (a) multiple dimensions of religion and (b) sufficient racial and ethnic diversity to permit reliable estimation. For these reasons, and given the dearth of scholarly attention to the specific research questions considered here, we believe that analyses of the NSFH-2 data can still provide valuable new information.

Our analyses were restricted to a subsample of older adults, consisting of primary respondents who: (a) were 55 and over at the time of the wave two interview; (b) had at least one adult child; (c) were African American, Hispanic, or non-Hispanic White; and (d) were either religiously unaffiliated, Catholic, Jewish, Mormon, mainline Protestant, or conservative Protestant. Our effective sample size was 2195. To retain as many cases as possible, the small numbers of missing cases on several control variables (e.g., household income, and residence proximity to adult children) were replaced with valid sample means; ancillary analyses (not shown, but available upon request) indicate that this approach did not affect the key findings reported here. In addition, a sample of 2354 respondents with similar characteristics from the NSFH-1 was also analyzed and similar results were obtained. Since our selection models did not reveal any significant biases due to sample attrition across the two NSFH waves for this subsample (2354 − 2195 = 159), we decided to report our findings from the more recent NSFH-2 data.

4.2. Dependent Variables

Four items were used to measure norms of intergenerational support. Respondents were asked to indicate the degree of their (dis)agreement with each of the following statements: (1) “Parents ought to provide financial help to their adult children when the children are having financial difficulty”; (2) “Parents ought to let their adult children live with them when the children are having problems”; (3) “Children ought to provide financial help to aging parents when their parents are having financial difficulty”; (4) “Children ought to let aging parents live with them when the parents can no longer live by themselves.” For each item, response categories ranged from “strongly disagree” (1) to “strongly agree” (4). Responses were coded so that higher scores denoted greater agreement with the item.

4.3. Religious Variables

We considered three aspects of respondents’ religious involvement: affiliation, attendance, and theological conservatism. First, the original NSFH-2 project coded a broad array of religious denominations, and these responses were subsequently recoded into the following categories: (a) Unaffiliated; (b) Catholic; (c) Jewish; (d) Mormon (Latter Day Saints); (e) Mainline Protestant (e.g., Episcopal, Lutheran (except Missouri Synod and Wisconsin Synod), Methodist, Presbyterian (except Presbyterian Church in America), Congregationalist, etc.); and (f) Conservative Protestant (e.g., Baptist (except American or Northern Baptist), Church of Christ, Church of God, Assemblies of God, all Pentecostal and/or Holiness churches, Apostolic, Christian Alliance, and other fundamentalist and evangelical churches). Our classification scheme closely follows the one developed by Steensland and associates [52], except that we did not create a separate category for all Black Protestants; instead, African Americans who were Protestant were classified as belonging to either conservative or mainline denominations.

Second, the (self-reported) frequency of attendance at religious services was recorded in the NSFH-2 in terms of (a) number of times and (b) unit of time (e.g., month, week), as reported by the respondent. These responses were subsequently recoded into a seven-category ordinal measure, ranging from “never” (0) to “several times a week” (7). This resulting item closely parallels widely used measures from other major surveys, such as the NORC General Social Surveys.

Third, consistent with previous studies using NSFH data [53], theological conservatism was measured via the mean degree of respondent’s (dis)agreement with the following two items: (1) “The Bible is God’s Word and everything happened or will happen exactly as it says”; (2) “The Bible is the answer to all important human problems.” Responses to each item were first recoded, so that they ranged from “strongly disagree” (1) to “strongly agree” (5), and then averaged. Higher scores on this index (alpha = 0.81) indicated greater theological conservatism.

Finally, to test the intersection of religion and race/ethnicity in predicting norms of intergenerational assistance as specified in Hypotheses 4a and 4b, cross-product (interaction) terms between religious attendance, theological conservatism and race/ethnicity were created, respectively. Prior to constructing these interactions, component variables were zero-centered in order to reduce collinearity between raw and multiplicative terms.

4.4. Additional Variables

A number of additional variables were included in our statistical analyses. Race/ethnicity was dummy-coded as follows: 1 = African American, 1 = Latino, vs. 0 = non-Hispanic White. Following closely the previous work of Burr and Mutchler [13], our regression models included statistical adjustments for several characteristics of the respondent that could potentially confound the associations between religion, race/ethnicity, and norms of filial responsibility. Among these variables were: gender (1 = female, vs. 0 = male); marital status (1 = married, vs. 0 = other; different coding schemes for marital status were explored and no significant differences emerged); whether the respondent had at least one living parent (1 = yes, vs. 0 = no); region of residence (1 = Northeast, 1 = North Central, 1 = West, vs. 0 = South); age (measured in years); education (in years); respondent’s household income (in dollars, log transformed); number of adult children (actual number); and proximity to the nearest adult child (in miles, log transformed). Finally, each respondent’s physical impairment was measured via a short form of the Activities of Daily Living (ADL) scale, the items in which tapped the existence of physical or mental condition(s) that limited his/her ability to: (1) “care for personal needs, such as dressing, eating, or going to the bathroom”; (2) “move about inside the house”; (3) “do day-to-day household tasks”; (4) “climb a flight of stairs”; (5) “walk six blocks.” Responses to each item were dummy-coded (1 = yes, vs. 0 = no), and then summed, so that higher scores reflected greater levels of physical impairment (alpha = 0.87).

5. Results

Descriptive statistics on all variables used in our analyses presented in Table 1. Several points merit brief discussion. Not surprisingly, norms of intergenerational support were, for the most part, relatively popular among these older NSFH-2 respondents. This is particularly true of exchanges of financial aid between adult children and parents; approximately 77% of the respondents agreed that parents should provide financial assistance to offspring in need, and over 60% agreed that adult children should do the same for parents in difficult straits. Few respondents unequivocally disagreed with these views. Co-residence was less popular; approximately 40% of respondents agreed that adult children should provide co-residence for their aging parents, and a similar percentage of respondents agreed that parents should provide co-residence to their adult children. For each of these outcomes, more than one-third of respondents expressed ambivalence, while a solid minority—roughly one-quarter of respondents—rejected norms of co-residence.

Table 1. Sample Characteristics.

Variables	Minimum	Maximum	Count	Percent
Dependent Variables
Parents Ought to Provide Financial Aid
1. Strongly Disagree	-	-	35	2
2. Disagree	-	-	94	4
3. Neither Agree Nor Disagree	-	-	381	17
4. Agree	-	-	1267	58
5. Strongly Agree	-	-	418	19
Parents Ought to Provide Co-residence
1. Strongly Disagree	-	-	121	6
2. Disagree	-	-	447	20
3. Neither Agree Nor Disagree	-	-	752	34
4. Agree	-	-	737	34
5. Strongly Agree	-	-	138	6
Children Ought to Provide Financial Aid
1. Strongly Disagree	-	-	66	3
2. Disagree	-	-	153	7
3. Neither Agree Nor Disagree	-	-	614	28
4. Agree	-	-	1094	50
5. Strongly Agree	-	-	268	12
Children Ought to Provide Co-residence
1. Strongly Disagree	-	-	104	5
2. Disagree	-	-	364	17
3. Neither Agree Nor Disagree	-	-	816	37
4. Agree	-	-	705	32
5. Strongly Agree	-	-	206	9
Religion Variables
Affiliation
None	-	-	91	4
Catholic	-	-	487	22
Jewish	-	-	51	2
Mormon	-	-	65	3
Mainline Protestants	-	-	755	34
Conservative Protestant	-	-	746	34
Church Attendance	1	7	4.16	2.22
Theological Conservatism (Index)	1	5	3.67	1.08
Control Variables
Race/Ethnicity
White	-	-	1727	79
Black	-	-	355	16
Hispanic	-	-	113	5
Gender
Male	-	-	780	36
Female	-	-	1415	64
Respondent’s Age	55	96	67.80	8.91
Respondent’s Marital Status
Currently Married	-	-	1146	52
Other	-	-	1049	48
Respondent’s Education	0	20	11.47	3.25
Household Income (Log)	−2.30	12.90	9.54	2.39
Number of Adult children	1	14	3.45	2.09
Residence Proximity to Adult Children (Log)	−2.30	8.61	1.77	2.83
Respondent’s Physical Activity (Index)	1	15	6.56	2.45
Respondent Has Living Parent
Yes	-	-	390	18
No	-	-	1805	82
Region of Residence
Northeast	-	-	377	17
North Central	-	-	612	28
South	-	-	845	38
West	-	-	361	16
Total n (%)			2195	100

Notes: for continuous variables, means and standard deviations are displayed.

Table 1. Sample Characteristics.

Variables	Minimum	Maximum	Count	Percent
Dependent Variables
Parents Ought to Provide Financial Aid
1. Strongly Disagree	-	-	35	2
2. Disagree	-	-	94	4
3. Neither Agree Nor Disagree	-	-	381	17
4. Agree	-	-	1267	58
5. Strongly Agree	-	-	418	19
Parents Ought to Provide Co-residence
1. Strongly Disagree	-	-	121	6
2. Disagree	-	-	447	20
3. Neither Agree Nor Disagree	-	-	752	34
4. Agree	-	-	737	34
5. Strongly Agree	-	-	138	6
Children Ought to Provide Financial Aid
1. Strongly Disagree	-	-	66	3
2. Disagree	-	-	153	7
3. Neither Agree Nor Disagree	-	-	614	28
4. Agree	-	-	1094	50
5. Strongly Agree	-	-	268	12
Children Ought to Provide Co-residence
1. Strongly Disagree	-	-	104	5
2. Disagree	-	-	364	17
3. Neither Agree Nor Disagree	-	-	816	37
4. Agree	-	-	705	32
5. Strongly Agree	-	-	206	9
Religion Variables
Affiliation
None	-	-	91	4
Catholic	-	-	487	22
Jewish	-	-	51	2
Mormon	-	-	65	3
Mainline Protestants	-	-	755	34
Conservative Protestant	-	-	746	34
Church Attendance	1	7	4.16	2.22
Theological Conservatism (Index)	1	5	3.67	1.08
Control Variables
Race/Ethnicity
White	-	-	1727	79
Black	-	-	355	16
Hispanic	-	-	113	5
Gender
Male	-	-	780	36
Female	-	-	1415	64
Respondent’s Age	55	96	67.80	8.91
Respondent’s Marital Status
Currently Married	-	-	1146	52
Other	-	-	1049	48
Respondent’s Education	0	20	11.47	3.25
Household Income (Log)	−2.30	12.90	9.54	2.39
Number of Adult children	1	14	3.45	2.09
Residence Proximity to Adult Children (Log)	−2.30	8.61	1.77	2.83
Respondent’s Physical Activity (Index)	1	15	6.56	2.45
Respondent Has Living Parent
Yes	-	-	390	18
No	-	-	1805	82
Region of Residence
Northeast	-	-	377	17
North Central	-	-	612	28
South	-	-	845	38
West	-	-	361	16
Total n (%)			2195	100

Notes: for continuous variables, means and standard deviations are displayed.

The religious characteristics of this sample are also noteworthy. Very few (4%) of these older respondents reported having no religious affiliation; this figure is lower than comparable estimates for the overall U.S. population during the same time period. By contrast, the sample contained a higher proportion of Protestants and somewhat smaller percentage of Catholics than in the overall population. Approximately 70% of the respondents identified themselves as Protestants, and they were nearly evenly divided between mainline (i.e., moderate or liberal) and conservative (i.e., fundamentalist or evangelical) denominations. Nearly one-quarter (22%) of respondents were Catholic, with much smaller proportions of Mormon and Jewish respondents. The mean score on the theological conservatism index was 3.67, almost the equivalent of “agree” responses on the two items tapping biblical inerrancy and scriptural authority. Levels of attendance at religious services also appear relatively high; the mean of 4.16 was equivalent to roughly 1–2 times per month.

Although a large majority of respondents in this NSFH-2 subsample were non-Hispanic Whites, approximately 16% were African American and the remainder were Latino. The average respondent was female and married, and was nearly 68 years old, with roughly 11.5 years of formal education and a total household income of approximately $31,000. The average respondent reported approximately 3.45 adult children, at least some of whom resided within the same state. Nearly one in five respondents also had a living parent, and the average respondent reported a moderate level of physical impairment based on the ADL measure. Approximately two-thirds of the older NSFH-2 respondents resided in either the South or North-Central regions of the U.S.

Table 2 through Table 5 report the odds coefficients of a series of ordered logistic regression models, estimating the net effects of religious variables and covariates on rank-ordered agreement with each type of intergenerational assistance. Because there is an unobserved and continuously distributed variable that underlies the observed ordinal variables (i.e., the items tapping norms of intergenerational assistance), ordered logistic regression models were appropriate for our statistical analyses. Unless stated otherwise, all the ordered logistic regression models presented here were estimated using Stata software (release 10), with the maximum likelihood estimation method. It is worth noting that additional analyses were conducted to check the proportional odds assumption or parallel regression assumption and the unobserved heterogeneity. No problems surfaced.

Table 2. Odds Coefficients of Ordered Logistic Regressions to Predict Parents Ought to Provide Financial Aid (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as reference)
Non-affiliated			0.75						0.84		0.81
Catholic			1.08						1.07		1.06
Jewish			1.04						1.12		1.05
Mormon			1.06						0.99		1.01
Conservative Protestant			0.91						0.85		0.86
Church Attendance					1.03				1.02		1.02
Theological Conservatism							1.09	*	1.09		1.03
Black * Conservatism											1.59	***
Hispanic * Conservatism											1.23
Control Variables
Race (White as reference)
Black	1.49	***	1.56	***	1.43	**	1.43	**	1.49	**	1.21
Hispanic	1.58	*	1.44		1.52	*	1.57	*	1.39		1.37
Gender (Female as reference)	1.17		1.18		1.19		1.18		1.20	*	1.22	*
Respondent’s Age	1.03	***	1.03	***	1.03	***	1.03	***	1.03	***	1.03	***
Respondent Married	1.07		1.06		1.06		1.07		1.05		1.04
Respondent’s Education	1.08	***	1.08	***	1.08	***	1.09	***	1.09	***	1.09	***
Household Income	1.02		1.02		1.02		1.02		1.02		1.01
Number of Adult Child	0.94	**	0.94	**	0.94	**	0.94	**	0.94	**	0.94	**
Distance (log)	1.03		1.03		1.03		1.03		1.03		1.03
Respondent’s Physical Activity	0.97		0.97		0.97		0.97		0.97		0.97
Respondent’s Living Parent	1.10		1.10		1.09		1.10		1.09		1.09
Region (South as reference)
North East	0.75	*	0.71	**	0.77	*	0.79		0.74	*	0.75	*
North Central	0.53	***	0.52	***	0.54	***	0.54	***	0.53	***	0.53	***
West	0.58	***	0.56	***	0.59	***	0.60	***	0.59	***	0.58	***
−2 Log Likelihood	4870		4866		4867		4866		4861		4848
Model Chi-square	125.7	***	129.3	***	128.7	***	129.8	***	134.8	***	147.1	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		3.7(5)		2.9(1)		4.1(1)	*	9.1(7)		21.4(9)	*

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Table 2. Odds Coefficients of Ordered Logistic Regressions to Predict Parents Ought to Provide Financial Aid (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as reference)
Non-affiliated			0.75						0.84		0.81
Catholic			1.08						1.07		1.06
Jewish			1.04						1.12		1.05
Mormon			1.06						0.99		1.01
Conservative Protestant			0.91						0.85		0.86
Church Attendance					1.03				1.02		1.02
Theological Conservatism							1.09	*	1.09		1.03
Black * Conservatism											1.59	***
Hispanic * Conservatism											1.23
Control Variables
Race (White as reference)
Black	1.49	***	1.56	***	1.43	**	1.43	**	1.49	**	1.21
Hispanic	1.58	*	1.44		1.52	*	1.57	*	1.39		1.37
Gender (Female as reference)	1.17		1.18		1.19		1.18		1.20	*	1.22	*
Respondent’s Age	1.03	***	1.03	***	1.03	***	1.03	***	1.03	***	1.03	***
Respondent Married	1.07		1.06		1.06		1.07		1.05		1.04
Respondent’s Education	1.08	***	1.08	***	1.08	***	1.09	***	1.09	***	1.09	***
Household Income	1.02		1.02		1.02		1.02		1.02		1.01
Number of Adult Child	0.94	**	0.94	**	0.94	**	0.94	**	0.94	**	0.94	**
Distance (log)	1.03		1.03		1.03		1.03		1.03		1.03
Respondent’s Physical Activity	0.97		0.97		0.97		0.97		0.97		0.97
Respondent’s Living Parent	1.10		1.10		1.09		1.10		1.09		1.09
Region (South as reference)
North East	0.75	*	0.71	**	0.77	*	0.79		0.74	*	0.75	*
North Central	0.53	***	0.52	***	0.54	***	0.54	***	0.53	***	0.53	***
West	0.58	***	0.56	***	0.59	***	0.60	***	0.59	***	0.58	***
−2 Log Likelihood	4870		4866		4867		4866		4861		4848
Model Chi-square	125.7	***	129.3	***	128.7	***	129.8	***	134.8	***	147.1	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		3.7(5)		2.9(1)		4.1(1)	*	9.1(7)		21.4(9)	*

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

In each table, the sequence of regression models begins with a baseline model that included only socio-demographic predictors, similar to those employed by Burr and Mutchler [13]. In the next three models, religious affiliation, attendance, and theological conservatism were added to the baseline model individually. This is followed by a full main effects model, in which all religious variables and controls were included together. In the final model, we added cross-product (interaction) terms to test the hypothesis (H4a) that allows the links between theological conservatism and norms of intergenerational assistance to vary across racial/ethnic groups.

In ancillary models (not shown), we also examined the hypothesized interactions between race/ethnicity and religious attendance (H4b). However, none of these cross-product terms was significant, and to save space, those additional models are not tabulated here. In addition, in the interest of conserving space, our discussion of the multivariate results centers solely on the estimated religious effects that are of primary theoretical interest to our study.

Table 2 presents the ordered logistic regression models predicting agreement that parents should provide financial aid to adult children in need. No meaningful denominational differences surfaced in these models. Religious attendance exhibited a positive but insignificant association with support for this type of intergenerational assistance. Theological conservatism emerged as a positive and significant predictor of the belief that parents should give financial help to their adult children in Model 4. However, this relationship was clarified by the inclusion of interaction terms in Model 6. The significant link between theological conservatism and support for this type of intergenerational aid was stronger for older African Americans than for their non-Hispanic white counterparts; that is, within the African American subsample, theological conservatism was a strong and positive predictor of this norm of intergenerational support.

Findings regarding support for parental provision of co-residence are displayed in Table 3. Here again, there were no significant denominational differences. The frequency of religious attendance had no bearing on this attitude in Model 3. Once the other religious variables were included in subsequent models (Models 5 and 6), a slight suppressor pattern was revealed; religious attendance actually exhibited a significant inverse association with support for the parental provision of co-residence. Theological conservatism was a robust positive predictor of support for this co-residence norm, and there was no evidence of racial/ethnic variations in the magnitude of this relationship.

Table 3. Odds Coefficients of Ordered Logistic Regressions to Predict Parents Ought to Provide Residence (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as reference)
Non-affiliated			1.06						1.18		1.17
Catholics			1.17						1.24		1.24
Jewish			1.20						1.44		1.40
Mormon			0.98						0.91		0.92
Conservative Protestant			1.02						0.94		0.94
Attendance					0.99				0.96	*	0.96	*
Theological Conservatism							1.20	**	1.27	***	1.25	***
Black * Conservatism											1.11
Hispanic * Conservatism											1.28
Control Variables
Race (White as reference)
Black	1.55	***	1.58	***	1.57	***	1.43	***	1.58	***	1.53	**
Hispanic	2.70	***	2.45	***	2.72	***	2.69	***	2.40	***	2.35	***
Gender (Female as reference)	0.95		0.95		0.95		0.98		0.96		0.96
Respondent’s Age	1.01		1.01		1.01		1.01		1.01		1.01
Respondent Married	1.06		1.06		1.07		1.05		1.08		1.07
Respondent’s Education	0.97	*	0.97	*	0.97		0.99		0.99		0.99
Household Income	1.03		1.03		1.03		1.03	*	1.03		1.03
Number of Adult Child	0.95	*	0.95	*	0.95	*	0.95		0.95	*	0.95	*
Distance (log)	0.99		0.99		0.99		1.00		0.99		0.99
Respondent’s Physical Activity	1.02		1.02		1.02		1.02		1.01		1.01
Respondent’s Living Parent	1.11		1.11		1.12		1.10		1.11		1.11
Region (South as reference)
North East	1.00		0.95		1.00		1.10		1.00		1.00
North Central	0.71	***	0.71	***	0.71	***	0.74	**	0.72	***	0.72	***
West	0.91		0.91		0.90		0.97		0.94		0.93
−2 Log Likelihood	6029		6026		6028		6009		5997		5995
Model Chi-square	78.7	***	81.0	***	79.0	***	98.3	***	110.4	***	112.4	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		2.3 (5)		0.35 (1)		19.6 (1)	***	31.7 (7)	***	33.7 (9)	***

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Table 3. Odds Coefficients of Ordered Logistic Regressions to Predict Parents Ought to Provide Residence (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as reference)
Non-affiliated			1.06						1.18		1.17
Catholics			1.17						1.24		1.24
Jewish			1.20						1.44		1.40
Mormon			0.98						0.91		0.92
Conservative Protestant			1.02						0.94		0.94
Attendance					0.99				0.96	*	0.96	*
Theological Conservatism							1.20	**	1.27	***	1.25	***
Black * Conservatism											1.11
Hispanic * Conservatism											1.28
Control Variables
Race (White as reference)
Black	1.55	***	1.58	***	1.57	***	1.43	***	1.58	***	1.53	**
Hispanic	2.70	***	2.45	***	2.72	***	2.69	***	2.40	***	2.35	***
Gender (Female as reference)	0.95		0.95		0.95		0.98		0.96		0.96
Respondent’s Age	1.01		1.01		1.01		1.01		1.01		1.01
Respondent Married	1.06		1.06		1.07		1.05		1.08		1.07
Respondent’s Education	0.97	*	0.97	*	0.97		0.99		0.99		0.99
Household Income	1.03		1.03		1.03		1.03	*	1.03		1.03
Number of Adult Child	0.95	*	0.95	*	0.95	*	0.95		0.95	*	0.95	*
Distance (log)	0.99		0.99		0.99		1.00		0.99		0.99
Respondent’s Physical Activity	1.02		1.02		1.02		1.02		1.01		1.01
Respondent’s Living Parent	1.11		1.11		1.12		1.10		1.11		1.11
Region (South as reference)
North East	1.00		0.95		1.00		1.10		1.00		1.00
North Central	0.71	***	0.71	***	0.71	***	0.74	**	0.72	***	0.72	***
West	0.91		0.91		0.90		0.97		0.94		0.93
−2 Log Likelihood	6029		6026		6028		6009		5997		5995
Model Chi-square	78.7	***	81.0	***	79.0	***	98.3	***	110.4	***	112.4	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		2.3 (5)		0.35 (1)		19.6 (1)	***	31.7 (7)	***	33.7 (9)	***

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

The models predicting support for the belief that adult children should provide financial relief to their parents are presented in Table 4. As can be seen from the table, there were no significant denominational differences. Although religious attendance was positively associated with endorsement of this norm in Model 3, that association was also eliminated in subsequent models. Once again, theological conservatism was a strong predictor of the belief that adult children should give financial aid to their parents. The inclusion of interaction terms in Model 6 indicated that, whereas theological conservatism was a significant and positive predictor among all racial/ethnic groups, the magnitude of this estimated net effect was significantly stronger for older African Americans than for their non-Hispanic white counterparts.

Table 4. Odds Coefficients of Ordered Logistic Regressions to Predict Children Ought to Provide Financial Aid (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as Reference)
Non-affiliated			1.01						1.25		1.19
Catholics			1.11						1.11		1.10
Jewish			1.25						1.53		1.41
Mormon			1.27						1.10		1.13
Conservative Protestant			1.19						1.05		1.05
Attendance					1.04	*			1.01		1.01
Theological Conservatism							1.25	***	1.26	***	1.18	***
Black * Conservatism											1.65	***
Hispanic * Conservatism											1.40
Control Variables
Race (White as Reference)
Black	2.47	***	2.40	***	2.35	***	2.24	***	2.24	***	1.81	***
Hispanic	2.19	***	2.19	***	2.10	***	2.19	***	2.07	***	2.03	***
Gender (Female as Reference)	1.22	*	1.22	*	1.25	*	1.27	**	1.26	*	1.28	**
Respondent’s Age	1.02	**	1.02	**	1.01	**	1.01	*	1.01	*	1.01	*
Respondent Married	0.82	*	0.81	*	0.80	*	0.81	*	0.81	*	0.80	*
Respondent’s Education	1.00		1.01		1.00		1.02		1.02		1.02
Household Income	1.03	*	1.04	*	1.03		1.04	*	1.04	*	1.04	*
Number of Adult Child	0.96	*	0.96	*	0.96	*	0.96	*	0.96	*	0.96	*
Distance (Log)	1.00		1.00		1.00		1.00		1.00		1.00
Respondent’s Physical Activity	0.98		0.98		0.98		0.98		0.98		0.98
Respondent’s Living Parent	1.42	**	1.42	**	1.41	**	1.40	**	1.40	**	1.41	**
Region (South as Reference)	1.000		1.000		1.000		1		1		1.000
North East	0.59	***	0.60	***	0.60	***	0.65	***	0.64	***	0.65	***
North Central	0.56	***	0.58	***	0.57	***	0.59	***	0.59	***	0.59	***
West	0.70	**	0.70	**	0.73	**	0.76	*	0.75	*	0.74	*
−2 Log Likelihood	5345		5341		5340		5316		5313		5297
Model Chi-square	147.7	***	151.6	***	152.7	***	176.5	***	180.1	***	196.2	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		3.8(5)		4.9(1)	*	28.8(1)	***	32.4(7)	***	48.5(9)	***

Notes: * p < 0.05; ** p< 0.01; *** p < 0.001.

Table 4. Odds Coefficients of Ordered Logistic Regressions to Predict Children Ought to Provide Financial Aid (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as Reference)
Non-affiliated			1.01						1.25		1.19
Catholics			1.11						1.11		1.10
Jewish			1.25						1.53		1.41
Mormon			1.27						1.10		1.13
Conservative Protestant			1.19						1.05		1.05
Attendance					1.04	*			1.01		1.01
Theological Conservatism							1.25	***	1.26	***	1.18	***
Black * Conservatism											1.65	***
Hispanic * Conservatism											1.40
Control Variables
Race (White as Reference)
Black	2.47	***	2.40	***	2.35	***	2.24	***	2.24	***	1.81	***
Hispanic	2.19	***	2.19	***	2.10	***	2.19	***	2.07	***	2.03	***
Gender (Female as Reference)	1.22	*	1.22	*	1.25	*	1.27	**	1.26	*	1.28	**
Respondent’s Age	1.02	**	1.02	**	1.01	**	1.01	*	1.01	*	1.01	*
Respondent Married	0.82	*	0.81	*	0.80	*	0.81	*	0.81	*	0.80	*
Respondent’s Education	1.00		1.01		1.00		1.02		1.02		1.02
Household Income	1.03	*	1.04	*	1.03		1.04	*	1.04	*	1.04	*
Number of Adult Child	0.96	*	0.96	*	0.96	*	0.96	*	0.96	*	0.96	*
Distance (Log)	1.00		1.00		1.00		1.00		1.00		1.00
Respondent’s Physical Activity	0.98		0.98		0.98		0.98		0.98		0.98
Respondent’s Living Parent	1.42	**	1.42	**	1.41	**	1.40	**	1.40	**	1.41	**
Region (South as Reference)	1.000		1.000		1.000		1		1		1.000
North East	0.59	***	0.60	***	0.60	***	0.65	***	0.64	***	0.65	***
North Central	0.56	***	0.58	***	0.57	***	0.59	***	0.59	***	0.59	***
West	0.70	**	0.70	**	0.73	**	0.76	*	0.75	*	0.74	*
−2 Log Likelihood	5345		5341		5340		5316		5313		5297
Model Chi-square	147.7	***	151.6	***	152.7	***	176.5	***	180.1	***	196.2	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		3.8(5)		4.9(1)	*	28.8(1)	***	32.4(7)	***	48.5(9)	***

Notes: * p < 0.05; ** p< 0.01; *** p < 0.001.

Finally, Table 5 displays the results of similar analyses predicting support for the provision of co-residence by adult children. In marked contrast with the earlier analyses, these models revealed a number of notable denominational variations with regard to this outcome. Specifically, support for the provision of co-residence for aging parents was significantly stronger among conservative Protestants, Mormons, and Catholics as compared with mainline Protestants. These group differences persisted even with controls for other religious variables. Although religious attendance was positively associated with the provision of co-residence by adult children in Model 3, this association faded with the inclusion of other religious predictors. In a now-familiar pattern, theological conservatism emerged as a consistently strong predictor of support for this type of intergenerational assistance. Although this association surfaced for all racial/ethnic groups, the link between theological conservatism and support for the norm of co-residence for aging parents in need was even more potent within the African American subsample.

Table 5. Odds Coefficients of Ordered Logistic Regressions to Predict Children Ought to Provide Residence (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as Reference)
Non-affiliated			1.08						1.33		1.28
Catholics			1.29	*					1.32	*	1.30	*
Jewish			1.07						1.39		1.31
Mormon			2.01	**					1.79	*	1.82	*
Conservative Protestant			1.55	***					1.35	**	1.35	**
Attendance					1.04	*			0.99		0.99
Theological Conservatism							1.34	***	1.33	***	1.26	***
Black * Conservatism											1.51	***
Hispanic * Conservatism											1.11
Control Variables
Race (White as Reference)
Black	2.67	***	2.45	***	2.54	***	2.36	***	2.32	***	1.97	***
Hispanic	1.98	***	1.99	***	1.90	***	1.96	***	1.89	**	1.90	**
Gender (Female as Reference)	1.34	***	1.33	**	1.36	***	1.40	***	1.37	***	1.39	***
Respondent’s Age	1.01		1.01		1.01		1.00		1.01		1.01
Respondent Married	0.83	*	0.82	*	0.81	*	0.81	*	0.82	*	0.81	*
Respondent’s Education	0.94	***	0.95	***	0.94	***	0.96	**	0.96	**	0.96	**
Household Income	1.01		1.01		1.01		1.02		1.02		1.01
Number of Adult Child	0.97		0.97		0.97		0.97		0.97		0.97
Distance (Log)	1.00		1.00		1.00		1.01		1.01		1.01
Respondent’s Physical Activity	1.02		1.01		1.02		1.02		1.01		1.01
Respondent’s Living Parent	1.33	**	1.31	*	1.31	*	1.29	*	1.29	*	1.30	*
Region (South as Reference)
North East	0.68	***	0.73	**	0.70	**	0.79	*	0.79	*	0.80
North Central	0.65	***	0.70	***	0.66	***	0.69	***	0.71	***	0.72	**
West	0.81		0.80		0.85		0.90		0.85		0.84
−2 Log Likelihood	5930		5906		5925		5878		5864		5853
Model Chi-square	203.8	***	227.4	***	208.7	***	255.3	***	269.2	***	280.3	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		23.7(5)	***	4.9(1)	*	51.5(1)	***	65.5(7)	***	76.5(9)	***

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

Table 5. Odds Coefficients of Ordered Logistic Regressions to Predict Children Ought to Provide Residence (n = 2195).

Variables	Base Model		Model 2		Model 3		Model 4		Model 5		Model 6
Religion Variables
Religious Affiliation (Mainline Protestant as Reference)
Non-affiliated			1.08						1.33		1.28
Catholics			1.29	*					1.32	*	1.30	*
Jewish			1.07						1.39		1.31
Mormon			2.01	**					1.79	*	1.82	*
Conservative Protestant			1.55	***					1.35	**	1.35	**
Attendance					1.04	*			0.99		0.99
Theological Conservatism							1.34	***	1.33	***	1.26	***
Black * Conservatism											1.51	***
Hispanic * Conservatism											1.11
Control Variables
Race (White as Reference)
Black	2.67	***	2.45	***	2.54	***	2.36	***	2.32	***	1.97	***
Hispanic	1.98	***	1.99	***	1.90	***	1.96	***	1.89	**	1.90	**
Gender (Female as Reference)	1.34	***	1.33	**	1.36	***	1.40	***	1.37	***	1.39	***
Respondent’s Age	1.01		1.01		1.01		1.00		1.01		1.01
Respondent Married	0.83	*	0.82	*	0.81	*	0.81	*	0.82	*	0.81	*
Respondent’s Education	0.94	***	0.95	***	0.94	***	0.96	**	0.96	**	0.96	**
Household Income	1.01		1.01		1.01		1.02		1.02		1.01
Number of Adult Child	0.97		0.97		0.97		0.97		0.97		0.97
Distance (Log)	1.00		1.00		1.00		1.01		1.01		1.01
Respondent’s Physical Activity	1.02		1.01		1.02		1.02		1.01		1.01
Respondent’s Living Parent	1.33	**	1.31	*	1.31	*	1.29	*	1.29	*	1.30	*
Region (South as Reference)
North East	0.68	***	0.73	**	0.70	**	0.79	*	0.79	*	0.80
North Central	0.65	***	0.70	***	0.66	***	0.69	***	0.71	***	0.72	**
West	0.81		0.80		0.85		0.90		0.85		0.84
−2 Log Likelihood	5930		5906		5925		5878		5864		5853
Model Chi-square	203.8	***	227.4	***	208.7	***	255.3	***	269.2	***	280.3	***
DF	14		19		15		15		21		23
Chi-square (df) Changes	-		23.7(5)	***	4.9(1)	*	51.5(1)	***	65.5(7)	***	76.5(9)	***

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001.

6. Conclusions and Discussion

In recent years, investigators have demonstrated considerable interest in (1) the important and complex role of religion in late life; (2) racial/ethnic diversity in aging; and (3) multifaceted linkages between religion and family life. Building on the literature, this study has used data from the NSFH-2 to investigate religious variations in norms of intergenerational assistance, with particular attention to racial/ethnic variations in these associations. Specifically, we have examined the net effects of multiple dimensions of religious involvement, as well as racial/ethnic variations in these effects, on beliefs about the duties of adult children to provide co-residence and financial support to their parents in need and vice-versa. Viewed broadly, our results underscore the potential importance of religious factors in shaping orientations toward intergenerational assistance. Particular constellations of religious belief, e.g., regarding biblical inerrancy and scriptural authority, appear to be especially important in this regard. In some instances, our results also reveal evidence of distinctive religious subcultures.

The role of theological beliefs looms large in these data. As anticipated by H1, persons holding stronger views concerning the inerrancy and authority of the Bible were consistently more supportive of all four aspects of intergenerational assistance, even with controls for other aspects of religious involvement. Although theological conservatism was an important predictor of these attitudes for each of the three major racial/ethnic groups examined here (African Americans, Hispanics, and non-Hispanic Whites), for three of the four outcomes they were even stronger among African Americans, lending partial support to H4a. Taken together, these findings suggest that analyses of aging and family life within an increasingly diverse aging population will be incomplete without greater attention to the nexus of family and faith within these populations, and that religious beliefs, in particular, deserve more careful scrutiny in this literature.

In addition to these patterns involving theological conservatism, these analyses also revealed significant net denominational differences in one of the four attitudinal outcomes, the belief that adult children should provide co-residence for their parents. Although this is the only outcome for which these subgroup variations surface, the patterns are broadly consistent with our major expectations: Members of conservative and sectarian denominations (i.e., evangelical Protestant, LDS or Mormons) and Catholics tend to endorse this norm of intergenerational support, findings that were predicted by H3a and H3b, respectively. Particularly in the case of conservative Protestant and sectarian groups, these patterns reflect distinctive denominational belief systems regarding familial and intergenerational relations. For quite different theological reasons, these various religious subgroups are much more likely to embrace the provision of co-residence to parents than are older adults who are mainline Protestant or religiously unaffiliated. Contrary to H3c, we did not find lower levels of support for co-residence or other types of intergenerational assistance among Jewish respondents, perhaps due in part to the small number of older Jewish respondents in the NSFH-2 dataset.

Interestingly, while core theological convictions—and to a more limited extent, denominational subcultures—are clearly linked with norms of intergenerational aid, religious commitment per se (gauged by the self-reported frequency of attendance at services) had only negligible estimated effects, especially when other key religious variables are controlled. Here our findings run counter to H2, and they also contrast with the thrust of some other religion-family studies in recent years, which tend to emphasize the importance of more “universalistic” indicators of religiosity or religious commitment over more “particularistic” theological and denominational subcultural differences [36]. One possible reason for these modest—indeed, largely null—patterns may be the tendency of many persons to misreport the frequency with which they actually attend religious services [54]. Another reason may reflect the multiple motives and reasons that influence individuals’ decisions to attend (or not to attend) religious services. Regardless, studies that attempt to gauge the role of religion by considering only single items or single dimensions to tap the complex domain of religious life are likely to underestimate the role of religious factors in family relations, as with other outcomes of interest [20].

Given the findings reported here, a number of research directions merit close attention in future work. First, it would be useful to extend this line of inquiry by examining the total and indirect effects of religious affiliation, belief, and practice on the subsequent receipt of assistance and the residential choices of older adults [13,19]. Several studies have questioned the link between expectations and subsequent receipt of support within this population [55].

Second, comparing the role of religious factors in shaping norms of intergenerational support across cohorts would also be a profitable undertaking. Previous research has documented both age-related increases and cohort-based declines in some facets of religiousness [56,57], as well as (what appear to be) cohort-based declines in norms of one important type of intergenerational assistance, namely, co-residence [18].

Third, given the age of the NSFH-2 data, it will be important for future research to replicate and extend our findings using more recent data containing appropriate items on norms of intergenerational support, multiple facets of religion, and samples with sufficient racial and ethnic diversity. It is conceivable that cohort-based shifts in norms of co-residence and other forms of support—as well as religious belief, practice, and affiliation across various racial and ethnic subgroups—could have altered the patterns observed in this study.

Fourth, it is important to gain a greater understanding of the mechanisms that explain links between religious affiliation and belief and norms of intergenerational support. To what extent might the patterns observed here result from (a) denomination—and tradition—specific subcultures; (b) congregational socialization mechanisms (e.g., sermons, religious education programs, family ministries); or (c) spiritual modeling and reference group processes within religious groups? Do our findings partly reflect religious differences in the quality of relationships between aging parents and their adult children? In addition, although this study focused on adherents to Judeo-Christian faiths, due to the religious composition of the NSFH-2 sample, and on members of three racial/ethnic groups (African Americans, Latinos, and non-Hispanic Whites), clearly it will be important to explore these issues among members of other world religions (e.g., Islam, Buddhism, Hinduism, etc.) and other racial/ethnic minority groups in the U.S. (e.g., Asian Americans and Pacific Islanders, Native Americans). It will also be helpful to distinguish among specific Latino and Asian national origin groups in future research.

Fifth, previous research has pointed to the influence of parental divorce and remarriage in shaping differences in norms and practices of intergenerational assistance [10,14]. At least one study has reported that such relationship transitions among parents can complicate the association between religion and intergenerational exchange [19]. Given that racial/ethnic groups differ significantly by religion and family structures and transitions, it would be valuable to examine racial/ethnic variations in the influence of religion on norms of intergenerational aid when parents have divorced and remarried (or re-partnered).

Sixth, this study has built on a long tradition of work addressing norms of intergenerational support. However, the next logical step in this agenda will be to examine the role of religious factors in shaping actual patterns of intergenerational assistance, including the provision of co-residence to older adults, as well as the extent to which these dynamics may vary by race and ethnicity. For example, previous research has reported that racial and ethnic differences in the living arrangements of older adults may be due largely, if not entirely, to cultural variations in norms of intergenerational provision of co-residence [13]. It will be important to extend this line of inquiry to explore religious group differences in intergenerational assistance. Indeed, theological conservatism—perhaps the most significant religious predictor in our study—tends to be inversely associated with socioeconomic status (SES), which may bear upon the capacity of families to enact fully the supportive norms identified in our work.

Seventh, prior research has revealed that social support, including intergenerational assistance, is conducive to better health and perhaps longer life across most racial/ethnic subgroups [58]. Moreover, a growing body of research has also linked religion with salutary health outcomes, such as physical health, mental health, and psychological wellbeing [59,60]. Therefore, it would be fruitful to investigate if intergenerational aid helps to account for part of the religion-health connection and if so, whether this may vary by race/ethnicity in future research.

Finally, caregiving and intergenerational assistance can be significant sources of stress, impacting mental and physical health. In light of the role that religion and spirituality can play in coping with such stressors, especially within certain minority populations [61], additional research is needed to assess whether and how religious beliefs and practices can mitigate these deleterious effects, and the ways in which this may vary by race/ethnicity [62].

More generally, studies of racial/ethnic variations in linkages between religion and other family-related attitudes should be an urgent priority for future researchers. As we noted earlier, given the historical and contemporary importance of religion within racial/ethnic minority populations, it is surprising that there is still only a modest body of information about the religion-family connection among African Americans or Hispanics. In light of the increases in minority aging, the growing interest in multicultural family diversity research, and the pressing concerns about access to support and care for older racial and ethnic minority populations, as well as the salience of religion within these and other racial/ethnic subgroups, it is vital for scholars to address this major oversight in the empirical literature.