Do Religious Children Care More and Provide More Care for Older Parents? A Study of Filial Norms and Behaviors across Five Nations

INTRODUCTION

Much of the literature on the normative obligation to provide care to older parents centers on its role in motivating supportive behavior. However, most studies find that the relationship between norms and behavior in the domain of parent care is weak or holds true only under particular circumstances (Silverstein, Gans & Yang, 2006; Silverstein, Parrott, & Bengtson, 1995). While previous studies have noted the complexity of the relationship between filial norms and filial behavior, little is known about the societal and religious contexts within which normative obligations are enacted into actual support. This investigation examines the correspondence between filial norms and supportive behavior of adult children toward their older parents in four European countries (Norway, England, Spain, and Germany) and Israel, emphasizing the role of religiosity in the development and enactment of filial norms in those national contexts.

Despite recent gains in the longevity of older adults, many of the years added to life are accompanied by chronic conditions and experienced with frailty (Freedman, Schoeni, Martin, & Cornman, 2007). As a result, adult children are likely to spend prolonged periods of time assisting aging parents. However, increasing demand for eldercare is not matched by an increasing supply of potential family caregivers as a result of declining fertility, delayed parenthood, elevated rates of divorce and remarriage, and increased labor force participation of women (Easterlin, Schaeffer, & Macunovich, 1993). In such an environment, it is imperative to understand the ideological and structural conditions that promote or inhibit the development of normative obligations to support aging parents, as well as the conditions that encourage or impede the translation of such commitment into actual support.

RELIGIOSITY AND INTERGENERATIONAL SUPPORT

Despite growing evidence of the importance of religion in family life (Chatters& Taylor, 2005; Mahoney, Pargament, Tarakeshwar & Swank, 2001) and its obvious relevance as a normative factor in family decision-making, only few papers have addressed the role of religiosity in intergenerational relationships.

From the somewhat sparse literature on this topic, three general aspects of religion that are likely to shape intergenerational supportive relationships emerge: religious doctrines that prescribe appropriate behavior toward older parents; religious values emphasizing compassion and reinforcing helping behaviors; and institutional structures as well as religious rituals reinforcing strong intergenerational bonds and commitment to family members.

Doctrinal aspects of religion are best summarized by the dictate of the Old Testament to honor one’s father and mother, a filial prescription made in various forms by virtually all religions of the world. Individuals who are more involved in practicing religion through both the public sphere (participation in religious services) and the private sphere (family prayers, personal salience of religion) are more likely to be exposed to messages that promote strong family commitment (Pearce & Axinn, 1998). Moreover, co-religionists can reinforce positive adherence and performance of desired family norms as well as provide role models for preferred family behaviors and place informal sanctions on those deviating from desired norms (Chatters & Taylor, 2005; Ellison, 1997).

Religious teachings also inculcate collectivistic values that encourage service to the most vulnerable members of society (Ellison, 1997; Myers, 2004). Dollahite and Marks (2005) suggest that religion creates a sense of community supportiveness from which evolves “deep and abiding caring relationships” (p. 537). It is likely that individuals who are so influenced by their religious orientations are more inclined to provide support to needy parents. Additionally, some religious organizations provide formal education programs such as family life education, which explicitly promote and support desired family behaviors including care for frail parents (Chatters & Taylor, 2005). Such programs may indicate that religious commitment to helping other family members may be even stronger than the commitment to help other individuals.

Finally, most religious groups endorse strong family relationships (Myers, 2004). In the public sphere, religious institutions typically provide opportunities for family activities including family camps and retreats, which promote stronger intergenerational family bonds (Pearce & Axinn, 1998; Chatters & Taylor, 2005). In the private sphere, celebration of Holidays, as well as various religious rituals, provides opportunities for bringing families together, emphasizing the importance of families, and making individuals more aware of their role within the generational structure of the family (Chatters & Taylor, 2005). Pearce and Axinn (1998) demonstrated that religiosity had a significant positive effect on the quality of parent-child relationships by strengthening the affective bond between them. Given the strong relationship between affective ties and support provided to parents (Cicirelli, 1983; Rossi & Rossi, 1990), it is likely that religiosity indirectly enhances supportive behavior by strengthening intergenerational cohesion.

Taken together, it is clear that religiosity is a useful vantage point from which to study filial commitment and filial behavior.

In addition, religion and religious institutions are likely to attract adherents and congregants who already adhere to an ideology that emphasize selflessness and providing for others. Thus, religiosity may be selective with respect to characteristics that predispose individuals to help their aging parents.

A host of factors beyond religious values are important in motivating adult children to provide care for their aging parents or inhibiting them from doing so. Previous articles (see Gans & Silverstein, 2006; Silverstein, Gans & Yang, 2006; Silverstein, Conroy, & Gans, 2008) provided extensive literature reviews of these factors and we summarize them briefly here. These factors typically divide into three main groups: parental needs, intergenerational relationship quality, and adult child availability. Prime among the factors is parental need for support, typically manifest by physical and cognitive impairments, or widowhood (Silverstein, Gans & Yang, 2006). Another factor facilitating supportive behavior is having an emotionally close and engaged relationship with parents (Rossi & Rossi, 1990; Silverstein, Parrott, & Bengtson, 1995). Finally, competing demands such having minor children in the household may serve as a barrier to provision of support by imposing barriers on adult children’s availability and resources (e.g., time, money, energy). Normative orientations inconsistently predict whether supportive roles are enacted, but as discussed earlier, the expression of filial values by supportive behavior is highly contingent on these aforementioned factors. Other characteristics of children and parents may further be important. For example, Myers (2004) found that the relationship between religiosity and parental assistance was conditioned on gender of the child and the marital status of the parent, with more religious adult daughters providing higher levels of support to continuously married parents compared to provider sons and recipient divorced parents.

The political economy and family culture have increasingly been considered as important contextual factors in structuring how intergenerational relations of the elderly are maintained. The availability of internationally comparative data across European nations has been especially useful for studying how national context influences micro-family interactions (Glaser, Tomassini, & Grubndy, 2004; Hank, 2007; Lowenstein, Katz & Daatland, 2005). These investigations generally find that the salience of intergenerational relations—as measured by proximity, frequency of contact, and provisions of support—tends to be more common in nations with less generous social welfare regimes and more familistic cultures. For the most part, these differences follow a north-south divide. Collectivistic cultures tend to thrive in the south compared to individualist traditions of the north (Reher 1998), while the public sector is more developed in the social welfare regimes of the north compared to the more restricted system in the south (Epsing-Andersen 1990, 1999; Ferrera 1996).

The goals of the current study are to: (1) examine the degree of consistency between filial norms and support—the gap between what adult children believe they should do to support aging parents and their actual supportive actions; (2) identify how national context shapes the consistency between norms and behavior by examining the norms-support relationship across five countries with varying political economies, family cultures, and religious compositions; and (3) explore the role of religiosity in whether filial norms and support are consistent or inconsistent. It is hypothesized that more strongly religious adult children and those from more familistic national cultures will be more likely to be congruent in their norms and support. They will be more likely to express strong filial values toward older parents as well as more likely to act on those values.

METHODS

Sample

Data from the five-nation study known as OASIS (Old Age and Autonomy: The Role of Service Systems and Intergenerational Family Solidarity) are used. OASIS includes data collected in Norway, England, Germany, Spain, and Israel, each nation representing a unique welfare regime and familial culture. Data were collected using face-to-face structured interviews with random samples of about 1,200 adults aged 21 and older from each country, totaling 6,100 respondents (see Lowenstein and Ogg [2003] for a full description of the OASIS design and methodology). Over-samples were drawn of individuals aged 75 and older. As the data do not include weights, it was impossible to recalibrate the data for sampling distortions related to non-response, design effects, or disproportionate sampling. However, refusal rates were low in each nation (Lowenstein & Ogg, 2003) and focusing on the younger portion of the sample renders inconsequential the over-sampling of aged individuals. The sub-sample for this analysis is comprised of 2,327 adult children from the five countries, who have at least one surviving parent. Table 1 shows the descriptive characteristics of the subsample. The sample is slightly more educated than the general sample, likely due to the inclusion of respondents with living parents. It is possible that there is a selection effect of healthier families who are likely to be more educated and possibly wealthier.

Table l.

Descriptive characteristics of the sample (N=2,364)

	Norway N=546	England N=372	Germany N=417	Spain N=468	Israel N=561	Overall N=2,364
	M(SD)/%	M(SD)/%	M(SD)/%	M(SD)/%	M(SD)/%	M(SD)/%
Age	38.8 (10.1)	41.8 (10.2)	40.6 (9.9)	38.5 (10.3)	38.4 (10.9)	39.4 (10.4)
Female	55.7	62.2	49.5	52.2	63.5	56.8
Married	63.2	68.5	69.3	61.2	66.9	65.6
Has Minor children	51.2	66.0	51.5	43.9	59.1	53.9
College Education +	34.1	7.2	18.0	17.5	11.9	18.5
Very Religious	4.1	7.8	4.6	4.5	1.7	4.3
Not-At-All Religious	38.3	43.3	33.9	28.5	60.8	42.1
Age of Parent	67.9 (11.4)	69.8 (10.2)	67.6 (9.8)	68.9 (10.7)	67.5 (10.9)	68.3 (10.7)
Closeness with parents1	4.4 (.9)	4.5 (1.2)	4.3 (.8)	4.5 (.9)	4.7 (.9)	4.5 (.9)
Contact with parents2	3.9 (1.5)	4.2 (1.8)	3.5 (1.6)	4.5 (1.6)	5.1 (1.5)	4.3 (1.6)

¹Closeness to parents is measured on a scale of 1-6 with 1 indicating the most distant and 6 indicating the closest relationship.

²Contact with parents is measured on a scale from 0-6 with 0 indicating once per year or less and 6 indicating daily or more often contact.

The five countries considered in the current study can be ranked along several dimensions that roughly follow a north-south divide. Spain has weak welfare institutions, and along with Germany imposes legal obligations on adult children to support their older parents The mixed Israeli model is characterized by legal family obligations and strong collectivist orientation, but also has high service levels. Norway, and to a lesser degree England has strong social care policies, and no legal obligations between generations. Roughly we predict stronger congruence between elder-care norms and behavior in Spain, followed by Israel, Germany, England, and Norway. As religiosity is a key variable of interest, it is useful to note that in the five sub-samples Catholics form a plurality in Spain (59.7%) Jews are the large majority in Israel (98.5% given that only the Jewish population was surveyed), and Protestants are a plurality in Norway (73.3%). In England and Germany large portions of respondents reported no religious denomination (42.6% in England and 34.39% in Germany). The next largest faction was Protestant in England (28.3%) and Catholics (29.8%) in Germany.

Measures

Filial Norms and Filial Behavior

The dimensions used for classification are filial norms and filial behavior, each assessed with respect to two domains or content areas: geographic proximity and instrumental support. Parallel measures allow us to match the norm to the corresponding behavior in the two domains.

The first domain relates to geographic proximity. Filial norms in this domain were measured by level of agreement with the following statement: “Adult children should live close to their older parents so they can help them if needed”. Responses were dichotomized whereby “1” represents strong endorsement of this filial norm (agree or strongly agree), and “0” represents weak endorsement (neither disagree nor agree, disagree, or strongly disagree.) Filial behavior in the domain of proximity was measured using response to the following question: “how long does it take you to travel to parent(s)?” Responses were dichotomized whereby “1” represents a travel time of 29 minutes or less, and “0” indicates longer travel time.

The second domain relates to Instrumental support. Filial norms in this domain were measured using the level of agreement with the following statement: “Older parents should be able to depend on their adult children to help them do the things they need to do”. Responses were dichotomized whereby “1” represents strong endorsement of this norm (agree or strongly agree with the statement), and “0” represents weak endorsement (neither disagree nor agree, disagree, or strongly disagree). Filial behavior was measured using the response to following question: “During the last 12 months, have you provided any of the following kinds of help, assistance or support to any of your parents?” We considered only the following tasks that were deemed to be instrumental in nature: house repair or gardening; transportation or shopping, help with household chores, and personal care. Affirmative responses were dichotomized whereby “1” represents provision of at least one supportive instrumental task, and “0” indicates that no instrumental support was provided.

Predictor Variables

Subjective Religiosity

Respondents were presented with a single item question asking “Do you consider yourself to be not-at-all religious, somewhat religious, moderately religious, or very religious.” In this analysis we are especially interested in the extreme cases, those not-at-all religious and those who are very religious. Two dichotomous variables were created to represent the two extreme responses, leaving the middle group of those defining themselves as somewhat or moderately religious as the reference group.

Parental needs were measured using a dichotomous variable whereby “1” indicated parents older than 75 years of age and “0” indicated a younger parent. Advanced age is used as a proxy for parental health status. The data do not present with an objective health measure for the parents and about 24% of the subsample had missing data on the subjective measure of health (asking whether one’s parent needs help). Additionally, a variable measuring marital status of the parent was included whereby “1” represents widowed parent and “0” represents other marital status.

Relationship quality was measured using the mean score of responses to three questions measuring affectual solidarity—the emotional closeness between the child and the parent (Bengtson & Mangen, 1988). These items were: (1) How close do you feel to your parent?; (2) How good is your communication with your parent?; and (3) How do you and your parent get along? Responses to each item were scored on a six-point Likert scale, ranging from “not-at-all well” to “extremely well”. Frequency of contact with parents was measured by a continuous variable indicating the highest frequency of contact (via email, phone or in person) with a parent. This variable ranged from 0 to 6 where 1 = once a year or less to 6 = daily or more often.

Competing demands of adult children were measured using two variables: marital status (1 = married or living in partnership; 0 = any other marital status) and the presence of young children (1 = has at least one minor child aged 21 or younger; 0 = has no minor children). Other variables included in the model were gender (1=female; 0=male), age in years and a squared age term, and education level (1 = university degree and beyond; 0=lower levels of education).

Procedure

In the first stage of the analysis, a model that defines adult children based on the congruence between their filial norms and their filial behavior is developed. To do this, we use latent class analysis (LCA), an exploratory method for classifying individuals into unobserved or latent classes based on the association among observed variables (Vermunt & Magidson, 2000; Clogg, 1995). A key assumption of the technique is that the latent classes account for the true covariation among the measured variables. Model selection is based on progressively adding latent classes until a good fit to the observed data is achieved, using the following statistics to evaluate model fit: (1) chi-square likelihood test (L²), where a non-significant L² indicates that the predictions of the theoretical model fit the observed data, (2) Baysian Information Criterion (BIC) statistic indicates the relative fit of competing models, with lower values indicating a better fit (Raftery, 1986), and (3) classification error rate indicates the percent of misallocated cases in a specific model, where lower errors are preferable. Whereas the L² is prone to showing poor fit in large samples, BIC and classification error statistics are not sensitive to sample size.

Once a best-fitting model is identified and assignments to the classes are made, two sets of probabilities are examined to interpret the meaning and identify the prevalence of the classes: (1) Conditional latent class probabilities describe the distribution of observed indicators within each class and are used to label the classes. These are analogous to factor loadings and describe the measurement model. Nationality is included as a nominal covariate in the latent class analysis such that the measurement model is considered invariant across nations. (2) Latent class probabilities signify the distribution of sample members across classes, and describe the prevalence of each class in the population. These probabilities are free to vary across national contexts.

In the second stage of the analysis, we use logistic regression analysis to predict the likelihood of belonging to specific latent classes. The dependent variable at this point is nominal class membership based on assignments made by the LCA best-fitting model. Four dichotomous country indicators are included as predictors in the logistic regression with Israel (representing a unique mixed model) treated as the reference category.

RESULTS

Classification of Individuals Based on Filial Norms and Filial Behavior

The goodness-of-fit statistics for latent class models ranging from one to six classes is presented. The one-class model, which assumes that there are no relationships between the four measured items, fits poorly with the data. Successive addition of latent classes reveals that the L² of each is statistically significant and not until the six-class model does the model fit the observed data. In choosing between competing models we dismiss the L² as an absolute indicator of model fit due to the large sample size and rely on the other indicators. The BIC statistic reaches a generally similar minimum for the three-class, four-class, and five-class models. Among these, the three-class model has the lowest classification error at 17%, a rate somewhat higher than desired, but in keeping with those found in other applications of LCA (Vermunt & Magidson, 2003). Thus, we select the three-class model as the best-fitting model to our data.

Table 3 represents the latent class conditional probabilities associated with each class in the selected three-class model. We use these to profile the classes. Those assigned to the first class are defined by high probabilities of endorsing filial norms and engaging in filial behaviors in both domains, a group we label the committed supporters. Their actions are consistent with their normative commitment in both domains. The second group has low probabilities of endorsing filial norms and relatively low probabilities of engaging in filial behaviors of either type, a group we label as independent. Finally, the third group favors the domain of support over that of geographic proximity in both norms and behavior, and are labeled long-distance supporters. They tended to endorse the norm of being a dependable support provider and actually provide support, but tend not to endorse a norm of proximity, nor do they live close to their parents. In general, the types can be characterized as consistent in norms and behavior within each content domain.

Table 3.

Conditional Latent Class Probabilities for Constrained Three-Class Model.

	Conditional latent class probabilities
Measure	Committed Supporters	Independent	Long distance supporters
Filial Norms
Should live close to parent	.74	.01	.38
Should be dependable	.73	.19	.76
Filial Behavior
Lives close to parent	.70	.48	.40
Provides instrumental support	.53	.41	.67

Note: Probabilities greater than .5 are shown in bold.

In Table 4 the latent class distributions across the total sample as well as by country are presented. The class described as committed supporters represented slightly less than half (46%) of all respondents. The group termed independent was the next most frequently occurring class, describing about 31% of respondents. The remaining 23% of respondents fell into the class of long distance supporters. Striking differences emerge in the latent class distribution when examined by country. These figures are shown in Table 4 and shown graphically in Figure 1. Norway is unique in that almost three-quarters (72%) of its respondents in the long-distance supporters class and very few (2%) in the committed supporters class. Spain and Israel are the mirror image of Norway with very few (3% and 1% respectively) of its respondents in the long-distance supporters class and almost three-quarters (71% and 74% respectively) in the committed supporters class. Germany came closest to the mirroring the distribution for the total sample with about one-third of its respondents in each of the three classes. England was notable for its high representation (45%) in the independent class, more than half (53%) in the committed supporters class, and very few (1%) in the long distance supporters class, reflecting a polarization in the class assignments made in that nation.

Table 4.

Latent Class Probabilities for the Three-Class Model: Total and by Country

Country	Committed supporters	Independent	Long distance supporters
Norway	.02	.26	.72
Spain	.71	.26	.03
England	.53	.46	.01
Israel	.74	.25	.01
Germany	.34	.34	.32
Total	.46	.31	.23

Figure 1.

Predictors of Class Membership

As discussed, the distribution of classes varies markedly across the five nations considered. Because four out the five nations were sparsely represented in at least one class, multivariate estimation predicting membership across the three classes is not feasible with covariates indicating national membership. Therefore, we collapse the committed supporters and the long-distance supporters to form a single class that is differentiated from the independent class. Aside from being methodological convenience, this also provides a meaningful contrast of those who have weak normative commitment and are unlikely to be support providers with those who have moderate to strong normative commitment and tend to provide support at least in one of the two domains (instrumental support).

The results of the logistic regression are presented in Table 5. The analyses were performed using the combined committed supporters and long-distance supporters as the reference category. Odds ratios are reported. As expected, those defining themselves as not-at-all religious were 30% more likely (OR=1.3) to be in the independent group than in the other more committed and more involved group. Moreover, those defining themselves as very religious were about half as likely (OR=.47) to be independent.

Table 5.

Logistic Regression Analysis Predicting Independent Compared to Combined Committed Supporters and Long-distance Supporters^a.

Predictor	eb	SE
Religiosityb
Not-at-all religious	1.31**	.14
Very religious	.47**	.13
Parental need
Parent aged 75+	.98	.14
Widowed parent	1.00	.13
Relationship Quality
Closeness to parents	.81***	.04
Frequency of contact	.93*	.03
Competing Demands
Married	1.07	.12
Has minor children	.85	.10
Age	1.04	.04
Age squared	.99	.01
Female	1.20	.12
College education +	1.22	.16
Country indicatorsc
Norway	1.32	.20
England	2.64***	.43
Spain	.98	.15
Germany	1.32	.20
Pseudo R2		.043
χ2 (df)		109.02 (17)

^*p < .05.

^**p < .01.

^***p < .001

^aResults are presented as oddsratios.

^bReference group consists of those somewhat or moderately religious.

^cThe reference group is Israel.

Other variables also predicted likelihood of class membership. In terms of national contrasts, respondents in England were more than 2.5 times likely to be independent than their counterparts in Israel. When effect coding rather than dummy coding is used for national context (not shown) results show that England is exceptional in this regard compared to all countries taken together.

Relationship factors also had an influence on class membership. Quality of relationship and frequency of contact with parents were significant in the equation. Those with stronger emotional ties with parents and those with more frequent contact with them were less likely to be independent. Competing demands in the forms of having a spouse or having a young child at home did not affect the likelihood of class membership.

DISCUSSION

Our first goal in this investigation was to understand the relationship between what respondents think the role of children should be with respect to aging parents and what they actually do to support their parents. To investigate this issue we created a typology based on filial norms and filial behavior. The results indicate that most respondents across the countries studied are congruent with regard to norms and behavior in the domains of geographic proximity and instrumental support. Little evidence of cognitive dissonance in the sample was found, as there were few respondents whose commitment was strong but unfulfilled, or weak despite their involvement. To what degree, this correspondence reflects filial norms as an attribution or a motivation of behavior will await longitudinal research on this topic. Nevertheless, these results help resolve discrepancies in the literature regarding the relationship between norms and behaviors in intergenerational relationships. The strength of our findings could be due in part to innovations in the design and method used in the study such as matching norms and behaviors in specific domains of involvement, using classification analysis as tool to categorize individuals based on several dimensions simultaneously, and examining the issue from the child’s rather than the parent’s perspective.

Our second goal was to examine the representation of classes in the pooled sample and by national context. The largest group—about half of the pooled sample—consisted of adult children who expressed strong normative commitment and were engaged with their older parents in terms of living close and providing support. About a third did not feel strongly about the role of children nor did they live close or provide support to their parents. The third class was the most variegated, with strong commitment to and enactment of support, but little intent and realization of living in close proximity to their parents. Although these respondents provided support and did so from afar, they were congruent in their norms and behavior within each domain. In other previous research we found that long-distance care providers had extraordinarily strong normative commitment to withstand the costs of negotiating care at a distance (Silverstein, Conroy & Gans, 2008). The consequences of such support arrangements–in terms of caregiver stress and the quality of distant care–require further research.

There is wide variation across countries in the distribution of the three indentified types. As expected, national differences followed a north-south divide with more committed-supporters in the more familistic nations of Spain and Israel, and greater independence in England and (geographically) in Norway. Germany, the nation in the middle of the north-south axis is equally distributed across the three types. The large proportion of long-distance supporters in Norway is consistent with research showing that intergenerational proximity is not a strong goal in Norwegian families and that multigenerational households are extremely rare (Daatland, Herlofson, 2003; Daatland & Lowenstein, 2005). In part, the generosity of formal services and universal access to services allows Norwegian children to be involved in ways that are less onerous and demanding such that they may not need to live close to their parents in order to exchange support (Daatland, 1997).

Both in the cross-classification and logistic regression analysis England was unique in its over-representation in the independent class. While this finding is somewhat difficult to interpret in light of the findings in Norway, it is possible that England represents a mixed system in which nationalized health care comes with increasingly strict eligibility requirements. Thus, adult children may be more likely to prefer formal arrangements but often face the prospect of parent care given restrictions in the public system (explaining why over half were in the committed-supporters class). While it possible that cultural factors are also at play we are reluctant to speculate about this at present without additional evidence.

The third goal of this paper was to examine the role religiosity plays in developing normative obligations and enacting parent-care roles. The current results indicate that religiosity plays a part in how adult children are involved in the support and care of their older parents. Further, the findings suggest a linear trend rather than a threshold effect, whereby the least religious are most likely and the most religious least likely to be independent of their parents. In other words, the religious extremes represent distinct groups with respect to the outcome variable, with religiosity following a dose-response pattern with respect to commitment and enactment of support to older parents.

These results demonstrate that religiosity is an important context within which filial norms are developed and are enacted. However, the means by which religiosity exerts its influence is not identified by this analysis. As discussed earlier, most religions proclaim it a virtue to take care of older people–particularly older parents–reflecting their concern for supporting the more vulnerable of society and for preserving the sanctity of family life. Religious activities instill values of collectivism that encourage caring behaviors to others and provide a context that promotes both informal and formal venues for helping others, all of which may promote a strong sense of family cohesiveness and commitment leading to higher levels of commitment and support to one’s family members (Chatters and Taylor, 2005; Ellison, 1997).

Given that this study is cross-sectional, one cannot state with certainty that religiosity predicts the development of filial norms and their enactment into actual support, nor can one propose that religiosity promotes congruence between norms and behavior. One cannot rule out the possibility that other factors such as cultural conservatism are in fact explaining both religiosity and caring commitment and behavior. That said, the results indicate that religiosity is indeed significantly associated with filial norms and parental care across five different national contexts. As such, the study contributes to the literature and opens the door for future investigations. Future research should utilize longitudinal data to address the temporal relationship between the development of religiosity, filial norms and actual supportive behavior towards parents. Future research in the area of religiosity and exchange of intergenerational support in other societies and cultures will further be needed to establish whether the relationship between religiosity and filial norms and behavior is universal.

We recognize that a multidimensional approach to measuring religious belief, practice, and identification, along the lines suggested by Levin, Taylor, & Chatters (1995) will best shed light on the pathways by which religion promotes familistic attitudes and behaviors in later life families. From the current study one can only speculate how the world-view and institutional structures provided by religions collude to reinforce altruistic motivations and enable actions of adult children toward their aging parents in need.

Table 2.

Latent Class Models of Intergenerational Relationships Using Four Dichotomous Indicators of Filial Behavior and Norms.

	L2 (df)	p-value	BIC	Classification error rate
One class –
complete independence	699.75 (71)	4.2e-104	−149.34	0
Two class	330.99 (62)	2.2e-38	−149.65	.05
Three-class	218.60 (53)	5.4e-22	−192.22	.17
Four-class	146.88 (44)	5.3e-13	−194.22	.22
Five-class	79.76 (35)	2.4e-5	−191.57	.30
Six-class	43.52 (26)	0.017	−158.03	.31