The Effect of Parents’ Attitudes on Sons’ Marriage Timing

Abstract

Theories of family stability and change, demographic processes, and social psychological influences on behavior all posit that parental attitudes and beliefs are a key influence on their children’s behavior. Though we have evidence of these effects in Western populations, there is little information regarding this social mechanism in non-Western contexts. Furthermore, comparisons of mothers’ and fathers’ independent roles in these crucial intergenerational mechanisms are rare. This paper uses measures from a ten-year family panel study featuring independent interviews with both mothers and fathers in rural Nepal to investigate these issues. We test the association of specific attitudes, rather than broad ideational domains, about childbearing and old-age care with sons’ subsequent marriage behavior. Our results indicate that both mothers’ and fathers’ attitudes have important and independent influences on sons’ marriage behavior. Simultaneous study of both parents’ attitudes reveals that gender-specific parenting contexts can shape the relationship between parental attitudes and children’s behaviors. This crucial mechanism of intergenerational continuity and change is strong in this non-Western setting, with substantial implications for studies of intergenerational influences on behavior in all settings.

Introduction

Substantial empirical evidence demonstrates that parental attitudes and expectations shape their children’s subsequent behavior. This evidence is limited to rich, industrialized countries with individualistic orientations toward social life and a high degree of personal independence and freedom (Liefbroer and de Jong Gierveld 1993; Sewell et al. 1975; Thompson, Alexander, and Entwisle 1988; Whitbeck, Simons, and Kao 1994). There is little systematic, general population investigation of this relationship among the majority of the world’s population, who live in poorer and more agrarian countries of Asia, Africa, and Latin America. These populations tend to have more collectivist orientations and lower independence than in the US (Jayakody, Thornton, and Axinn 2008; Sastry and Ross 1998; Thornton and Lin 1994). In these societies, some of which are characterized by strong beliefs about the importance of both living parents and deceased ancestors, families play a central role in people’s lives (Goode 1970; Lee, Parish, and Willis 1994; Ogburn and Tibbitts 1933; Thornton and Lin 1994; Westermarck [1891] 1894). Therefore parents’ attitudes and expectations should be especially powerful predictors of their children’s behavior.

Rural Asian populations provide strong examples of societies with more collectivist orientations than Western populations. Asian societies historically emphasize selfless subordination to family and extended kinship, especially to senior kin (Goode 1970; Sastry and Ross 1998). Also, decisions about family formation have historically been considered too important to be left to the young themselves, making parents important decision-makers in their children’s family formation behaviors (Ghimire et al. 2006; Macfarlane 1976; Watkins 1996). Furthermore, in the absence of economic prospects outside the home, parental inheritance remains the primary source of livelihood and wealth for young people, especially sons (Cain 1977; Gertler and Lillard 1994). Hence, parents have a great deal of opportunity to influence their children in these settings. This paper investigates the consequences of parents’ attitudes and expectations for their sons’ marriage timing in such a setting: rural Nepal. We expect parental influences on children to be powerful in these settings relative to societies with more individualistic orientations. Of course both collectivist societies and individualistic societies are characterized by high heterogeneity within populations (Lesthaeghe and Surkyn 1988), making direct comparison across societies complex and beyond the scope of this investigation.

We construct a new framework for the study of intergenerational consequences of parental attitudes and expectations. This framework is specially designed to extend established intergenerational and attitude-behavior frameworks to the study of radically different social and economic contexts. Theoretically, this framework advances the study of intergenerational influences in three ways. First, the framework addresses a crucial issue in the study of ideational influences on behavior – the differences between broad ideational domains and specific attitudes and expectations. Our empirical investigation then focuses on specific attitudes and expectations. Second, we use this framework to identify hypotheses that go beyond linking behavior to attitudes about that specific behavior. Rather, we identify specific attitudes and expectations about family behaviors other than marriage that we expect to shape children’s marriage behavior. Third, we illuminate context-specific mechanisms likely to create important differences in the consequences of mothers’ versus fathers’ attitudes and expectations.

The juxtaposition of mothers’ and fathers’ attitudes is an especially important advance in this area, as the influence of each parent may depend on the role they hold within the family. Because these mechanisms are context-specific, their identification in a radically different family context such as rural Nepal provides an essential contrast to potential parental gender differences in settings such as the US. To further clarify the independent influence of parents’ attitudes we also investigate young peoples’ own attitudes, which are well known to be shaped by – but independent of – their parents’ attitudes, as another key influence on marriage timing.

Empirically, this paper takes advantage of unique intergenerational panel data spanning more than a decade that documents family change and variation in rural Nepal. These data feature a representative sample of linked parent-son triads, measures of attitudes and expectations from key family domains, direct personal interviews with mothers, fathers, and sons, and ten years of monthly panel data on sons’ subsequent marital behavior. Independent measurement of the attitudes of mothers, fathers, and their sons in a panel study with subsequent measures of behavioral consequences is exceptionally rare; this unique empirical resource provides the means to dramatically advance our knowledge of intergenerational attitude-behavior linkages.

Theoretical Framework

In order to extend established intergenerational and attitude-behavior frameworks to radically different social and economic contexts, we first differentiate between broad ideational domains and specific ideas. Both are relevant to behavior, but we focus on specific ideas, acknowledging that these ideas are often linked to broad ideational domains. The framework also identifies key specific ideas that vary within a population and are likely to be closely associated with specific behaviors. We focus on marriage timing behavior among males in rural Nepal – specificity in both behavior and study population that allows for the identification of specific ideas likely to influence that behavior. Finally, the framework also identifies the role that both parents’ ideas may play in behavioral decision-making. Following the central theme of important long-term intergenerational effects on children, we focus on the role of parental attitudes, comparing and contrasting the potential consequences of mothers’ versus fathers’ attitudes.

Broad Ideational Domains versus Specific Attitudes and Expectations

Many social scientists argue that a key difference between South Asian settings such as Nepal and Western Europe or the United States lies in broad ideational domains. These might include individualism, collectivism, fatalism, freedom and equality, materialism, familism, or developmental idealism (Goldscheider and Goldscheider 1987; Goode 1970; Hechter et al. 1999; Lesthaeghe and Surkyn 1988; Thornton 2005). Investigation of these broad ideational domains reveals that within a single population, some behavioral trends and some individual-level variations are associated with variations in these broad ideational domains (Goldscheider and Goldscheider 1987; Lesthaeghe and Surkyn 1988). There is potential, then, that ideational contrasts at a societal level are reflected in individual-level ideational variations within populations, and those individual-level variations are connected to subsequent behavioral choices.

Social psychological scholarship on the mechanisms linking ideas to actions often focuses on specific ideas – specific attitudes or expectations – rather than broad ideational domains. The leading research in this area argues that attitudes toward a particular behavior, along with subjective norms, predict intentions, and intentions predict behavior (Ajzen 1988; Fishbein and Ajzen 1975). Thus, positive attitudes toward marriage, coupled with social influence favoring marriage, increase the likelihood and speed of marriage (Liefbroer and de Jong Gierveld 1993; Thornton, Axinn, and Xie. 2007). Following this perspective, the strongest ideational predictor of most behaviors is likely to be the attitude toward that specific behavior. For example, attitudes toward childbearing predict childbearing behavior (Coombs 1974; Freedman, Freedman, and Thornton 1980; Thomson 1997), attitudes toward contraception predict contraceptive use (Biddlecom, Casterline, and Perez 1997; Odimegwu 1999), attitudes toward marriage predict marital behavior (Bayer 1969), attitudes toward premarital cohabitation predict cohabiting behavior (Clarkberg, Stolzenberg, and Waite 1995), attitudes toward divorce predict divorce behavior (Thornton 1985), and attitudes toward educational attainment predict educational attainment (Astone and McLanahan 1991; Sewell et al. 1975).

The second contribution of our framework, in its extension of established attitude-behavior frameworks, is the identification of both closely related and opposing attitudes as equally important predictors of behavior. This theoretical extension of the reasoned action and planned behavior theory is crucial to our framework because it maintains a focus on specific attitudes but also broadens the range of specific attitudes that may be relevant. For example, Barber (2001) shows that attitudes toward educational attainment shape behaviors that influence pregnancy, and Barber and Axinn (1998) show that attitudes about childbearing shape the timing of marriage. We expect that people enter adulthood with numerous specific attitudes and expectations that may shape marriage behavior in addition to their attitudes about marriage itself (Ajzen 1988; Fishbein and Ajzen 1975; Thornton et al. 2007).

Of course specific attitudes may well be a reflection of broader ideational domains. For example, those who are highly individualistic may hold more negative attitudes toward marriage while those who are highly familistic may hold more positive attitudes toward marriage. But specific attitudes may also be independent of broader ideational domains. We are fortunate to have measures from a study that was designed to measure both specific attitudes and broad ideational domains. This unique measurement resource gives us the means to focus our empirical investigation on specific attitudes rather than broad ideational domains.

A third extension of our framework, in addition to the identification of which specific attitudes are likely to be important, is the context-specific reasoning in identifying whose attitudes are likely to be influential. We limit our reasoning to the marriage timing of sons rather than the marriage timing of both sons and daughters because the marital experiences of sons and daughters differ greatly in Nepal, as in South Asia more generally. For example, though the majority of daughters (76%) experience arranged marriage in this setting, half of sons enjoy considerable input into the choice of both whom they marry and when they marry (Ghimire et al. 2006). As a result, the context of marriage is quite different for men and women, especially in which and whose specific attitudes are likely to be influential. In fact, because daughters’ marriages are arranged by parents in such high proportions, observed correlations with parental attitudes need not reflect intergenerational mechanisms. The remainder of our theoretical framework is designed to generate testable hypotheses regarding the consequences of specific attitudes for sons’ marriage behavior.

Which Attitudes?

The setting for this research is Chitwan Valley in south central Nepal, near the Indian border. The majority of people in this region identify as Hindu. They practice arranged marriage, where older generations of relatives seek spouses and negotiate marriages for younger generations. However, sons have been gaining autonomy in spouse choice (Ghimire et al. 2006). Marriage timing is also changing rapidly, with the mean age at marriage for males rising from 16.8 years for those married between 1950 and 1959 to 21.9 years for those married between 1990 and 1996 (Yabiku 2005). Marriage and childbearing are closely linked and both are nearly universal.¹ Children are needed for both short term economic productivity and longer term old-age security, and marriage is an essential step toward childbearing (Cain 1977; Caldwell 1982; Yabiku 2004). Sons are especially valued for this economic assistance and old-age care, and also for the performance of death rituals (Bennett 1983; Fricke 1986).²

In rural Asian settings like this one, specific attitudes about ideal family size and the responsibility of sons to care for elderly parents are particularly likely to influence decisions about marriage timing. The timing of the transition to marriage drives the production of children, and both the pace and total number of children have consequences for parents, grandparents, and the wider kin network (Bennett 1983; Fricke 1986). In US settings, parental attitudes about family size and their children’s childbearing are powerful in shaping their desires for their children to marry quickly (Barber and Axinn 1998). Attitudes about family size and children’s responsibility to care for parents when they reach old age are likely to be just as powerful in this kind of setting.

A desire for many children or the pressure to have many children is likely to lead to earlier marriage in most settings. In settings of high variation in marital behavior, such as the US, the desire for many children speeds marriage because couples prefer to be married for childrearing (Liefbroer and de Jong Gierveld 1993; Manning 1995). In Nepal, a setting of relatively low variation in marital behavior and virtually no premarital births, childbearing attitudes emerge as even more important. Marriage is not only desirable for childrearing, but it is also a necessary status for childbearing.

Although essentially everyone marries in Nepal, those who desire many children may be among the fastest to marry while those who prefer few children will wait for later marriage (Fricke 1986; Gipson and Hindin 2007; Hirschman and Rindfuss 1982). Because sex outside of marriage is rare in Nepal, and pregnancy outside of marriage is even rarer (Retherford and Thapa 1998), those who wish to have many children are motivated to marry quickly in order to become sexually active (Gipson and Hindin 2007). Literature in the US suggests that people will prefer to marry if they have a desire for childbearing, not only because marriage is the most socially acceptable union in which to do so, but also because marriage facilitates childrearing (Barber and Axinn 1998; Manning 1995). Thus, both of these mechanisms work in the same direction in both generations, potentially making attitudes about family size an exceptionally strong influence on marriage timing in Nepal.

Analytic implications: Specific attitudes regarding family size preferences are likely an important influence on marriage timing in this setting.

Research on fertility transition in general, and fertility behavior in South Asia in particular, frequently identifies old-age security needs as a motivation to have children (Cain 1991; Caldwell 1982; Lillard and Willis 1994; Willis 1980). In fact, the demand for kin care networks is so high in rural South Asia that we expect the need for old-age security to relate to marriage timing (Bennett 1983; Fricke 1986; Niraula 1995). In a setting like Nepal, where the political economy is unstable, the government social security scheme is minimal, and private pensions and health insurance are rare, parental need for old-age care from the younger generation is exceptionally high (Caldwell 1982; Lillard and Willis 1994; Niraula 1995; Willis 1980).³ Children’s marriages are the first step toward construction of a kin care network for old age support. Therefore, variations in attitudes toward responsibility for care of elderly parents are also likely to be a strong influence on variations in marriage timing. In Nepal, sons have historically taken on the care of their own parents, while daughters often co-reside with and provide care for their husbands’ parents (Bennett 1983; Goldstein, Schuler, and Ross 1983; Niraula 1995). Thus, this specific attitudinal domain is likely to be associated with sons’ marriage timing because sons’ marriage usually brings an additional caregiver into the home.

Analytic implications: Specific attitudes regarding sons’ responsibility for parental care in old age are likely an important influence on marriage timing in this setting.

Whose Attitudes?

Mothers, fathers, and sons will not always hold the same attitudes, and sons’ behaviors may not respond to each person’s attitude equally. In fact, parents’ attitudes may have more influence than sons’ own attitudes in domains in which parental attitudes are particularly important to children (Axinn and Thornton 1993). Of course, children’s attitudes may also be independent of parents’ attitudes, especially in a setting such as Nepal, where the younger generation has exposure to a wider range of non-family experiences than the older generation (Ghimire et al. 2006). In the following section, we discuss the theoretical reasons for studying mothers and fathers separately, as well as the reasons that sons may make decisions independent of their parents’ desires.

Parents’ Attitudes

Previous evidence clearly demonstrates that parental attitudes, values, and beliefs about a range of issues can influence children’s behavior (Barber 2001; Bengtson 1975; Glass, Bengston, and Dunham 1986; Longmore et al. 2009).This could happen through social control, where children behave in the way that their parents want them to behave in order to please them rather than behaving that way because it is how the children prefer to behave (Gecas and Seff 1990; Goldstein et al. 1983; Liefbroer and de Jong Gierveld 1993; Smith 1988). Children may behave in ways that they believe will make their parents happy even when parents are not invoking social control techniques (Barber 2000; Liefbroer and de Jong Gierveld 1993). On the other hand, this can happen through socialization, where children conform to their parents’ attitudes by internalizing those attitudes (Manning, Longmore, and Giordano 2007; Starrels and Holm 2000).

In this setting there is strong value placed on obedience and respect for elders and parents, and parents often have complete control over the means of production and household decisions (Goldestin et al. 1983; Niraula 1995). Thus, it is especially likely that parents influence their children through the mechanism of social control. There are many ways that parents may exert social control over their sons’ marriage (Goldstein et al. 1983; Lee et al. 1994). For example, parents who value marriage may use their resources to steer their children toward marriage (Thornton et al. 2007). This may entail refusal to invest in sons’ education, as educational enrollment is a deterrent to marrying quickly (East 1998; Thornton, Axinn, and Teachman 1995; Yabiku 2004). In Nepal, parents also exert social control on marriage by finding potential spouses for their children. In an effort to encourage marriage, they may begin the process of arranging their sons’ marriage early. These mechanisms are examples of ways in which parents’ attitudes may influence their sons’ marriage behavior directly and independent of the sons’ own attitudes.

Socialization may also be a powerful parental influence on children (Mead 1934; Peterson and Rollins 1987; Starrels and Holm 2000). Parents who prefer early marriage are likely to display this value in consistent ways, thereby teaching their children similar attitudes. They might express negative impressions of young adults who remain unmarried, or enthusiasm when celebrating the marriages of friends and family. Parents might also influence children’s attitudes in their choice of surroundings (Bengston 1975). For example, parents who value work and education over marriage may choose to live in a place where they can better expose their children to the workplace and schools. Having greater exposure to these settings, in turn, may lead children to value things like work and education over marriage (Mead 1934). In these ways, parents may socialize their children to hold the same attitudes and values. Though the relevance of parental preferences for children’s behaviors may decline across age, the considerable influence that parents have over their children’s behavior is likely to shape children’s opportunities and constraints in adulthood (Caldwell 1982; Elder 1975; Glass et al. 1986). Of course socialization and social control are potentially closely interrelated, so both may simultaneously influence children.

Analytic implications: Sons’ behavior is likely to be directly related to their parents’ attitudes.

Mothers’ and Fathers’ Attitudes

There is ample evidence that parents’ attitudes influence their children’s behavior, but less is known about the different influences exerted by each parent. Although a great deal of intergenerational research has documented important effects of mothers on their children, much less has examined the effects of fathers (Forum on Child and Family Statistics 1998; Thornton 2001). Even less research has directly compared the effects of mothers to the effects of fathers. The absence of research on fathers’ role was recognized as a major weakness of demographic research in the 1990s, and since that time research on fathers has become an especially high scientific priority (Forum on Child and Family Statistics 1998; Nock 1998; Thornberry, Smith, and Howard 1997). Although these priorities center on family research in the US and other Western settings, we have even greater reason to investigate fathers’ roles in Asian settings like Nepal, where intergenerational property and social transmission is paternal in nature (Bennett 1983; Caldwell 1982; Lee et al. 1994). Under these circumstances the effects of fathers may be independent of the effects of mothers. Documentation of such independence constitutes a crucial theoretical advance in intergenerational studies, likely to motivate greater research on the mother and father roles.

Mothers and fathers may share similar attitudes. If their attitudes are extremely similar then it would be difficult to empirically distinguish independent influences of each. In this case, it would make sense to study parents’ attitudes, combined, as a predictor of children’s behaviors. This type of similarity might be expected given that spouses are typically matched on a wide range of characteristics before they marry (Mare 1991). For example, married couples in Nepal typically come from the same or a similar caste, and caste standing is related to both culture and socioeconomic position. This similar positioning within the social system may lead to similar attitudes throughout parents’ young lives, up until marriage (Bengtson 1975). After marriage they share similar experiences as a result of their shared lives: they reside within the same surroundings, they share neighbors and friends, and, furthermore, their attitudes likely influence one another. These similar and shared experiences, along with the transmission of attitudes through spousal communication, lead us to posit that each parent might have the same influence on children.

On the other hand, if mothers’ and fathers’ attitudes do not overlap completely, the attitudes of each parent may affect their child independently. Though they may be similar, no two people share the exact same experiences through their life course, and differences in background or events experienced may lead them to hold different attitudes. For example, a mother may come from a large family with many siblings, leading her to place value on having many children. Her husband, however, may have been raised in a smaller family, leading him to place less value on having many children. Moreover, the more gendered that social experiences are, the more independent we might expect husbands’ and wives’ attitudes to be (Mason and Smith 2000). There is strong reason to believe that social life in rural Nepal is highly gendered (Bennett 1983; Fricke 1986), creating substantial independence in the life experiences of husbands and wives. Thus, we investigate the independent effects of parents’ attitudes in order to determine these potentially independent consequences. In the following paragraphs we outline the reasons to expect both parents to influence sons. Because there are compelling reasons for each parent to be influential, we do not argue that either will be more influential than the other.

If parents’ attitudes are indeed different, we would expect fathers’ attitudes to have an important influence on their sons. First, children may be more likely to model their behaviors after the attitudes of their same-sex parent (D’Angelo, Weinberger, and Feldman 1995; Starrels and Holm 2000; Treiman and Terrell 1975). For this reason, sons may more intentionally identify with and model after their fathers than their mothers (Kaufman and Uhlenberg 1998). Second, they might exhibit behaviors more similar to fathers’ attitudes due to biological inheritance (Udry 1995), which may be partly determined by the same-sex parent (D’Angelo et al. 1995). Third, sons may develop attitudes more similar to their fathers than their mothers, as they are likely to have more exposure to similar kinds of gendered and non-family activities that their fathers are exposed to, such as employment, military service, and travel. Fourth, because men historically hold much of the decision-making authority in households (Bennett 1983), sons may see their father as the authority figure and behave in accordance with their fathers’ attitudes.

We also have reason to believe that mothers’ attitudes will have an independent effect on sons’ behaviors. Mothers, in their role as caregivers and nurturers, generally spend more time with their young children than do fathers (Bianchi, Robinson, and Milkie 2006). Because of this, children are likely to interact with their mothers as their primary source of socialization. Through this regular and constant interaction, sons will become aware of mothers’ attitudes and this may influence their behavior. This pattern has been found in the US, and there is reason to expect it in rural South Asia, where mothers are even more likely to be the primary caregivers for their children (Bennett 1983; Lee et al. 1994; Paneru 1981). In fact, mothers may be closer with their sons in this setting, where men typically spend more time outside the home to provide financial support for the family. Mothers may focus on building especially close bonds with their sons so that they can ensure support for their future. For these reasons, mothers may have a strong influence on their sons.

Analytic implications: Both mothers’ and fathers’ attitudes are likely to influence sons’ behavior, although the influence from each parent may not necessarily be the same.

The relative influence of mothers and fathers may also vary across substantive domains. For example, an effect of mothers spending more time in caring for children than fathers is that mothers may have significant influence on their sons in the attitudinal domain related to childrearing. Compared to fathers, mothers are likely to both have stronger feelings about the ideal family size and to more adamantly impress these attitudes onto their sons. Likewise, sons may see mothers as having greater expertise than fathers in childrearing, so that mothers’ attitudes about family size may be treated with more credibility. Of course, for the reasons described above, fathers’ attitudes about ideal family size may affect their sons’ behaviors independent of the mothers’ attitudes, even in a domain associated with mothers.

Analytic implications: Both mothers’ and fathers’ attitudes in favor of childbearing are likely to speed sons’ marriage timing.

Just as mothers have expertise in the domain of childrearing, Nepalese fathers have expertise in the domain of sons’ responsibilities to care for elderly parents. This is because elderly parents in Nepal usually depend on sons for old age care. Thus, fathers are much more likely than mothers to have had personal experience with this responsibility for the care of their elderly parents, and, therefore, are more likely to be treated as experts in this substantive domain. As explained above, this experience is also quite likely to increase fathers’ appreciation for the role a wife can play in performing the actual caregiving tasks for his elderly parents. As a result, fathers may have an influence on their sons in this domain, speeding the sons’ acquisition of a wife.

There is some theoretical reason to expect that mothers’ attitude toward sons providing elderly care will have the opposite effect of fathers’ attitude. Both psychological theory (Freud 1937) and research on relationship quality (Rossi and Rossi 1990) point toward especially close and affectionate relationships between mothers and their sons. The rural Nepalese context is likely to heighten this. Virtually all mothers in Nepal breastfeed their sons for as long as they are lactating or until their next pregnancy, physically reinforcing these intense affectionate relationships (Paneru 1981; Retherford et al. 1989). In fact, Nepalese language includes a saying common among men, “Ama Ko Dhudh ko Bhara,” which means “the son can never repay the debt of mother’s milk”. The conflict between this intense mother-son relationship and the son’s relationship with his new bride is symbolized as the mother feeds yogurt to her son during his marriage ceremony (Paneru 1981), signifying her last breastfeeding before giving him to his new bride. Though rural Nepalese mothers generally want their sons to marry and produce grandchildren, these mothers have reason to wish to delay their sons’ marriages.

Even more crucial for our context-specific reasoning, these maternal desires to slow their sons’ marriages are likely to be stronger when the expectation of old-age care from their sons is high. As social changes emphasizing the conjugal bond spread through settings like rural Nepal, maternal expectation for poorer quality relationships with their married sons than with their unmarried sons are likely to be exacerbated (Allendorf 2009; Caldwell 1982; Niraula 1995; Paneru 1981). This expectation may not be limited to rural South Asian contexts—there is even some evidence that mothers in the US report less positive relationships and less care from their sons once sons are married (Merrill 2011; Sarkisian and Gerstel 2008). But in the rural Nepalese context, it is not surprising that mothers sometimes view their sons’ wives with competitive feelings, including concerns that the new wife may threaten a son’s willingness and ability to provide high quality old-age care (Bennett 1983; Stone 1978; Watkins 1996). While a mother may transmit the idea that it is a woman’s task to provide caregiving—motivating sons to marry—she also transmits the preference for her son to delay marriage so that his caregiving remains focused on mother rather than wife.

Analytic implications: Mothers’ attitudes in favor of sons holding the responsibility of caring for their elderly parents may slow sons’ marriage timing, but fathers’ attitudes may lead sons to marry faster.

Sons’ Attitudes

As described above, sons’ attitudes are likely to be similar to their parents’ attitudes because parental socialization plays a powerful role in shaping young people’s attitudes and beliefs. There is also theoretical reason to expect that parents’ and sons’ attitudes have independent effects on sons’ behaviors. Three important mechanisms may produce independence of parent and child attitudes. First, though parents and children share similar surroundings and experience similar events, they are experiencing these things at different points in their life course. Having the same experience at a substantially different point in the life course is likely to yield differences in attitudes (Elder 1975; Hagestad and Neugarten 1985). Second, macro-level changes can create considerable differences in social context between parental and child generations. Rapid growth in schooling, non-family work, government services, and mass media outlets in Nepal has created substantial differences between the contexts of parental life experience and their children’s life experience (Axinn and Barber 2001; Axinn and Yabiku 2001). These differences between parents’ childhood experiences and children’s experiences are also likely to produce independence between parent and child attitudes. Third, the new experiences spreading in Nepal introduce ideas from outside of the parental home (Ghimire et al. 2006). Exposure to the media, schooling, and peer groups allow transmission of ideas to children that are different from parents’ ideas (Ghimire et al. 2006). For these reasons, we expect parents’ and sons’ attitudes to be different and to potentially have independent consequences for sons’ behaviors.

Analytic implications: Sons’ own attitudes are likely to influence their marriage timing, independent of their parents’ attitudes.

Data

The Chitwan Valley Family Study CVFS offers measures from identical 72-minute individual interviews with both mothers and fathers, in addition to sons, allowing us to predict sons’ marriage timing over the subsequent ten years. Special care was taken during the data collection to interview mothers and fathers simultaneously in two different locations to enhance the independence of their responses. These detailed and comparable measures from parents provide an unprecedented opportunity to investigate the intergenerational influences from both mothers and fathers.⁴

The CVFS includes a baseline interview, conducted in 1996, that consists of a structured questionnaire interview and a semi-structured Life History Calendar interview. These interviews collected information on both the attitudinal and experiential measures that we employ in our analyses. The CVFS interviewed all members of households aged 15–59 in sampled neighborhoods and their spouses (even if outside this age range or living elsewhere). The overall response rate was 97%. Beginning in 1997, monthly follow-up interviews were conducted that collected information about household members on a range of demographic events, including marriage. Drawing on the CVFS we analyze all the unmarried men, aged 15–17 in 1996, whose mother and father were both alive and interviewed in 1996 (n=205 son, mother, father triads)⁵. Sixteen of these sons have brothers in the sample. Our analysis follows their monthly hazard of marriage for 126 months. We study men aged 15–17 in 1996 because less than 4% of Nepali men under age 18 were married in 1996. This percentage climbs to 12% by age 18 and continues to rise rapidly thereafter. Thus, the rates of entering marriage are highest for this group of under-18-year-old males, maximizing our opportunity to examine the consequences of parental attitudes for marriage behavior (Yabiku 2006).

Measures

Marriage Timing

We focus on first marriage, as remarriage in Nepal is a rare event (Yabiku 2006). We use 126 months of data on marriage to operationalize the monthly hazard of marrying in discrete time. The discrete time approach yields results similar to a continuous approach because the incidence of marriage in any one month is quite low, but the discrete time approach allows us to avoid making any parametric assumptions regarding the distribution of the underlying baseline hazard (Yamaguchi 1991). The measure of marriage is coded as 0 for every month in which the person is not married and 1 for the month in which the person marries, at which time they cease to contribute to person-months of exposure to risk of marriage.

Attitude Measures

Prior to fielding the 1996 study, great effort was taken to construct measures of attitudes specific to the rural Nepalese context. Five pilot studies were used in sequence with ethnographies and cognitive interviews to arrive at Nepalese language measures of core family attitudes. Some of these measures are often used in US studies of family attitudes and some are designed to be completely tailored to the Nepalese context. The attitudinal measures in this round of the study were deliberately designed to represent specific attitudes and not broad ideational domains. A second round of the CVFS attitude survey was conducted in 2008, with the aim of measuring attitudes that fit into broad ideational domains. In 2008, the CVFS also repeated the 1996 baseline measures designed to measure specific attitudes about specific behaviors. Results from factor analyses of these 2008 measures reveal that the 1996 specific measures repeated in 2008 do not load strongly with broader attitudinal domains. What we present here are English language translations of the two specific attitude items from the 1996 Nepalese instrument. The original Nepalese question wordings and response alternatives can be found at {perl.psc.isr.umich.edu}.

Ideal Family Size

We use the most widely investigated measure of ideal family size: the Coombs scale (Axinn, Clarkberg, and Thornton 1994; Coombs 1974, 1978, 1979). This measure indicates the number of children that respondents want to have and is calculated based on responses to a series of questions. Respondents were first asked how many children they would want if they could have the exact number they wanted. Respondents who already had children were asked how many children they would like to have if they could start life over. They were then asked a series of follow-up questions, indicating whether they would prefer fewer or more children if they were unable to have the exact number previously indicated. The Coombs scale has been found to produce the most accurate indicator of fertility preferences compared to other survey measures (Coombs 1974, 1978, 1979). On the single-item measure of ideal family size, a majority of people in the CVFS (60%) report a preference for exactly two children. The Coombs scale measure produces variance in responses, pushing respondents to report whether they would prefer a minimum of two children or a maximum of two children. This variance allows for high accuracy in estimating the effects that family size preference has on marriage timing (Barber and Axinn 1998).

Responses are coded into a scale with values ranging from 1–25, with a higher value indicating a preference for more children. For example, a respondent who says he prefers two children, and two is the maximum number of children he would prefer, receives a code of 4, as compared to a code of 7 for a respondent who says that two is the minimum number of children he would prefer. See Figure 1 for an illustration of the coding scheme. Among the CVFS sample, those who have higher values on the Coombs scale have a greater desire for more children (single-item measure correlates at r=0.92), actually have more children (achieved fertility correlates at r=0.33 among those ever married), and prefer a younger age at marriage (ideal age at marriage correlates at r=−0.16). As shown in Table 1, the mean value is 7.80 for mothers, with a standard deviation of 2.90, and 7.54 for fathers, with a standard deviation of 3.17. The mean value for sons is lower, at 5.96, with a standard deviation of 1.65. Mothers’ and fathers’ attitudes are not strongly or significantly correlated, at r=0.07. The correlations between mothers’ and sons’ attitudes and fathers’ and sons’ attitudes are also weak and insignificant, at r=0.09 and r=0.07, respectively.

Figure 1.

Table 1.

Descriptive Statistics

Measure	Mean	Standard Deviation	Minimum Value	Maximum Value
First Marriage (proportion marrying)	0.65	0.48	0.00	1.00
Ideal Family Size (Coombs Scale)
Mother	7.80	2.90	1.00	19.00
Father	7.54	3.17	2.00	25.00
Son	5.96	1.65	2.00	16.00
Married Son Should Care for Parents in Old Age
Mother	2.31	0.79	1.00	3.00
Father	2.25	0.79	1.00	3.00
Son	2.54	0.62	1.00	3.00
Mothers’ Level of Spouse Choice (had more choice)	1.61	1.39	1.00	5.00
Mothers’ Age at Marriage	15.00	2.69	7.00	24.00
Mothers’ Years of Education	0.70	1.97	0.00	12.00
Mother Ever Worked	0.43	0.50	0.00	1.00
Fathers’ Level of Spouse Choice (had more choice)	2.24	1.69	1.00	5.00
Fathers’ Age at Marriage	20.37	4.12	12.00	35.00
Fathers’ Years of Education	3.67	4.52	0.00	16.00
Father Ever Worked	0.41	0.49	0.00	1.00
Parents’ Level of Spouse Choice (had more choice)	1.93	1.21	1.00	5.00
Parents’ Age at Marriage	17.69	2.54	10.00	25.00
Parents’ Years of Education	2.18	2.79	0.00	13.00
Parents Ever Worked	0.84	0.80	0.00	2.00
Bhramin/Chhetri	0.53	0.50	0.00	1.00
Dalit	0.14	0.34	0.00	1.00
Newar	0.04	0.21	0.00	1.00
Terai Indigenous	0.16	0.36	0.00	1.00
Hill Indigenous	0.12	0.32	0.00	1.00
Sum of Services within 5 minute walk	1.83	1.54	0.00	5.00
Age in 1996	16.03	0.80	15.00	17.00
Has coresident brother	0.36	0.48	0.00	1.00
Birth Order	3.40	2.14	1.00	11.00
Total sons 205
Total sons marrying 133

Responsibility for Old-Age Care of Parents

Our measurement of attitudes about sons caring for their parents in old age comes from a survey item that was specifically developed and tested for this Nepalese population. Respondents were asked: “Some people think that a married son should take care of his parents in their older age. Do you strongly agree, somewhat agree, or don’t agree at all?” The item was coded from 1–3, so that a higher value indicates a more positive attitude. Mothers and fathers have similar mean values on this measure, at 2.31 and 2.25, respectively, with a standard deviation of 0.79 for each. Sons’ mean value is slightly greater, at 2.54, with a standard deviation of 0.62. Mothers’ and fathers’ attitudes are not strongly correlated, however, at r=0.13. Parents’ attitudes are not strongly nor significantly correlated with sons’ attitudes, at r=−0.07 for mothers and sons, and r=0.03 for fathers and sons.⁶

Controls

We control for a range of parental characteristics that may influence both parents’ attitudes and sons’ marriage timing.⁷ First, we control for parents’ marital experiences, starting with whether parents had an arranged marriage. A scale ranging from 1–5 was created to indicate the level of choice that mothers and fathers had in their marriage, from having no choice of their spouse (1) to having complete choice (5) (Ghimire et al. 2006). We use the mean of this scale for full models that include both mothers’ and fathers’ attitudes because, although they are in the same marriage, husbands and wives often have different levels of choice in their spouse. We also control for parents’ age at marriage. Again, in the full models, we use the mean of mothers’ and fathers’ age at marriage.

Next, we control for parents’ experiences with non-family activities. We use an indicator of each parent’s years of education to control for the first generation’s educational attainment. We use the mean of mothers’ and fathers’ years of education in the full models. We also control for whether parents ever worked. In the full models, we use a variable that is the sum of two dummy variables, again reported by parents themselves, indicating whether the mother and/or father ever worked for pay outside of the home. Many mechanisms link parents’ education and work to both their own attitudes and their children’s marital timing (Thornton et al. 2007; Yabiku 2005).

We also control for key indicators of family circumstances that may shape both parental attitudes and sons’ marital timing. First, we control for the ethnicity of the family using dummy variables for the different ethnic groups in Chitwan: Bhramin/Chhetri, Dalit, Newars, Terai Indigenous, and Hill Indigenous⁸. In our analyses, Bhramin/Chhetri is the reference category to which the other ethnic groups are compared. Second, we control for the community context in which families resided in 1995. Following the practice of previously published research using these data, community context is measured with a neighborhood-level indicator that is the sum of the number of services – bus stop, health center, employer, school, and/or market – that are within a five-minute walk of respondents’ neighborhood (Axinn and Yabiku 2001; Yabiku 2004, 2006). Third, we also control for sons’ age in 1996. Fourth, we control for the presence of an unmarried brother in the household. This measure also serves to account for assignment of the same parents to multiple sons in our sample. Fifth, we control for the respondents’ birth order. Both birth order and having a coresident brother can influence parental motivations for a particular son to marry in this setting.^{9, 10}

Finally, in order to control for the duration of exposure to marriage risk we use a control for time, which is the duration of time, in months, since the first monthly interview. Previous research shows this functional form fits the baseline hazard of marriage in Nepal well (Yabiku 2004, 2005, 2006).

Analytic Approach

We employ discrete-time event history analysis to model the risk of marriage, with person-months of exposure as the unit of analysis (Peterson 1993). We use multilevel logistic regression to account for the clustering of the CVFS sampling design at the neighborhood level. We estimate the monthly hazard of marrying, controlling for key family and parental characteristics as discussed above (Barber et al. 2000). Our logistic regression analyses produce multiplicative coefficients. We discuss the results as odds ratios, which is the anti-log of the coefficient. These odds ratios can be interpreted as the amount by which the odds are multiplied for each unit change in the respective independent variable. If the odds ratio is greater than 1, the effect is positive and if the odds are less than 1, the effect is negative. We test our unidirectional hypotheses for attitudinal influences using one-tailed tests of significance and we test the control measures using two-tailed tests of significance.

Results

We begin by investigating the consequences of attitudes toward childbearing, then move on to an investigation of attitudes about responsibility for old-age care. We present both models with each family member’s attitudes analyzed separately and models in which all three family members’ attitudes are analyzed together. In separate estimates (not shown here), we find results consistent with the planned behavior/reasoned action framework: sons’ attitudes toward the ideal age for men to marry predict sons’ marriage timing. However, those analyses do not reveal any direct effects of parents’ attitudes about the ideal age for men to marry. This result highlights the importance of investigating different specific attitudes to advance our understanding of intergenerational associations. Below we focus on the specific substantive domains in which context-specific reasoning predicts intergenerational associations.

Attitudes about Ideal Family Size

Table 2 displays the odds ratios from models estimating the effects of attitudes about ideal family size on the hazard of first marriage. Because the theoretical expectations outlined for parents’ and sons’ attitudes are unidirectional, we test the influences of these attitudes with one-tailed tests. As expected, we find that mothers’ attitude in this domain does have a substantial positive association with sons’ subsequent marriage timing that is independent of the effect of fathers’ or sons’ attitudes. In fact, mothers’ attitude has a significant positive association both when included as the sole attitude to predict sons’ marriage timing (Table 2, Model 1) and when included in a model with fathers’ and sons’ attitudes (Table 2, Model 4). When included in the full model with other family members’ attitudes, mothers’ attitude in favor of large families speeds sons’ marriage by 14% for each unit increase on the Coombs scale. Because the Coombs scale has a large range, this association is substantial: a 6-point increase in mothers’ attitude (e.g., the difference between wanting two children at most and wanting three children at least) leads sons to marrying 2.19 times faster. Thus, the association of mothers’ attitude about desired family size with sons’ marriage timing is substantively large, statistically significant, independent of fathers’ attitude, and not explained by including sons’ own attitude in the model.

Table 2.

Odds Ratios from Logistic Regression Estimates of Attitudes Predicting the Hazard of Sons’ First Marriages (Sons aged 15–17 in 1996)

	Model 1	Model 2	Model 3	Model 4
Attitude
Ideal Family Size (Coombs Scale)
Mother†	1.10** (2.53)			1.14*** (3.12)
Father†		1.07* (2.13)		1.07* (2.03)
Son†			1.07 (1.16)	1.05 (0.88)
Mothers’ Characteristics
Level of Spouse Choice (had more choice)	1.13 (1.68)
Age at Marriage	1.10* (2.22)
Years of Education	0.82** (−2.67)
Ever Worked	1.96** (3.26)
Fathers’ Characteristics
Level of Spouse Choice (had more choice)		1.24*** (3.53)
Age at Marriage		0.99 (−0.34)
Years of Education		0.85*** (−5.40)
Ever Worked		0.94 (−0.27)
Parents’ Characteristics
Level of Spouse Choice (had more choice)			1.37*** (3.33)	1.41*** (3.60)
Age at Marriage			1.05 (1.20)	1.05 (1.04)
Years of Education			0.75*** (−5.54)	0.77*** (−4.99)
Ever Worked			1.21 (1.31)	1.14 (0.88)
Sons’ Characteristics
Dalita	0.95 (−0.14)	0.80 (−0.62)	0.64 (−1.22)	0.68 (−1.01)
Newara	0.78 (−0.42)	0.44 (−1.19)	0.60 (−0.82)	0.50 (−1.00)
Terai Indigenousa	1.51 (1.12)	1.41 (0.87)	1.16 (0.38)	0.89 (−0.27)
Hill Indigenousa	1.60 (1.40)	1.09 (0.25)	1.01 (0.04)	1.04 (0.11)
Sum of Services within 5 minute walk	0.86 (−1.60)	0.91 (−0.93)	0.92 (−0.80)	0.92 (−0.81)
Age in 1996	1.34* (2.28)	1.25 (1.79)	1.32* (2.21)	1.33* (2.17)
Has coresident brother	0.60* (−2.18)	0.77 (−1.11)	0.69 (−1.54)	0.74 (−1.26)
Birth Order	0.94 (−1.04)	0.82** (−3.21)	0.82** (−3.19)	0.79*** (−3.74)
Time	1.03*** (10.48)	1.03*** (11.14)	1.03*** (11.31)	1.03*** (11.74)
N (person-months)	18465	18465	18465	18465
N (persons marrying)	133	133	133	133

Note: Estimates are presented as odds ratios. T-ratios are given in parentheses.

^aBhramin/Chhetri is the reference category.

^† Denotes that one-tailed tests were used, all other variables used two-tailed tests of significance

^*p<.05

^**p<.01

^***p<.001

Fathers also have a positive influence in this domain, though the magnitude of the association with fathers’ attitude is not statistically different from the association with mothers’ attitude.¹¹ In the full model (Table 2, Model 4) each unit increase on fathers’ Coombs scale increases the hazard of sons marrying by approximately 7%. This effect is again substantial: a 6-point increase in fathers’ attitude on the Coombs scale leads sons to marry 1.50 times faster. As is the case with mothers, fathers’ attitude has a significant association that is independent of mothers’ attitude, and unexplained by sons’ own attitude.

Sons’ attitude, on the other hand, does not have independent associations with marriage timing in this domain. This is true in both the models using only sons’ attitude to predict their marriage timing (Table 2, Model 3) and in the full models. These results are consistent with the conclusion that sons whose parents want larger families will marry faster, independent of sons’ own attitude.

As discussed above, sons with mothers who prefer larger families are motivated to marry more quickly so that they can begin childbearing. Mothers’ attitudes related to childbearing have been found to have a similar socializing influence on their children’s marriage timing in the US (Barber and Axinn 1998). In this radically different Nepalese setting we find the same result. However, measures to compare the consequences of mothers’ and fathers’ childbearing attitudes are not available from the US. In the Nepalese setting we find that fathers’ childbearing attitudes also have significant association with their sons’ marriage timing independent of mothers’ and sons’ attitudes. This result is consistent with our hypothesis that attitudes of both parents have the potential for independent influences on their children’s marital behavior.

Many of the controls have independent associations with sons’ marriage timing, as well. The sons of parents who experienced more participation in spouse choice and of mothers who have had work experience marry more quickly. On the other hand, sons of mothers who married at older ages and of parents who have more education marry more slowly. Sons’ own characteristics are also associated with their marriage timing. Their sibling composition has an important influence: both having more siblings and having a coresident brother slows their rate of marriage. This result may indicate that sons experience less pressure to marry if they have siblings to absorb some of that pressure. Finally, the parameters we use to specify the baseline hazard are also significant.

Attitudes about Old-Age Care

Based on the highly gendered relationships between each parent and their son in this setting, we expect the independent associations of specific parental attitudes may have opposing consequences for marital timing in the domain of old-age support. The results of our analyses are consistent with this expectation. Mothers’ attitude about old-age care slows the rate of sons’ marriages (though not statistically significant in Model 1, Table 3), but fathers’ attitude about old-age care speeds the rate of sons’ marriages (Table 3). When these two opposing forces are estimated in the same model (Model 4, Table 3), both associations are statistically significant and substantial, though the measurement metric for this attitude only has three units. A one-unit increase in mothers’ belief that sons should care for parents reduces sons’ marriage rate by 21%, but a one-unit increase in fathers’ belief increases sons’ marriage rate by 50%. Also interesting, in this domain sons’ own attitude appears to have independent consequences for their rate of marriage in a similar magnitude as fathers’ attitude. A one-unit increase in sons’ belief that sons should care for elderly parents increases their marriage rate by 51% (Table 3, Model 4).

Table 3.

Odds Ratios from Logistic Regression Estimates of Attitudes Predicting the Hazard of Sons’ First Marriages (Sons aged 15–17 in 1996)

	Model 1	Model 2	Model 3	Model 4
Attitude
Married Son Should Care for Parents in old Age
Mother†	0.85 (−1.29)			0.79* (−1.72)
Father†		1.38** (2.40)		1.50** (3.00)
Son†			1.49** (2.59)	1.51** (2.61)
Mothers’ Characteristics
Level of Spouse Choice (had more choice)	1.14 (1.77)
Age at Marriage	1.12** (2.72)
Years of Education	0.80** (−2.96)
Ever Worked	1.92** (3.13)
Fathers’ Characteristics
Level of Spouse Choice (had more choice)		1.21** (3.00)
Age at Marriage		0.99 (−0.47)
Years of Education		0.85*** (−5.75)
Ever Worked		1.00 (−0.02)
Parents’ Characteristics
Level of Spouse Choice (had more choice)			1.38*** (3.46)	1.44*** (3.79)
Age at Marriage			1.05 (1.22)	1.05 (1.11)
Years of Education			0.76*** (−5.25)	0.75*** (−5.45)
Ever Worked			1.24 (1.50)	1.14 (0.89)
Sons’ Characteristics
Dalit a	0.92 (−0.25)	0.74 (−0.82)	0.61 (−1.36)	0.59 (−1.42)
Newar a	0.65 (−0.72)	0.63 (−0.73)	0.65 (−0.69)	0.60 (−0.80)
Terai Indigenous a	1.66 (1.41)	1.57 (1.11)	1.12 (0.29)	1.12 (0.26)
Hill Indigenous a	1.66 (1.53)	1.04 (0.10)	1.01 (0.04)	1.00 (0.00)
Sum of Services within 5 minute walk	0.85 (−1.72)	0.91 (−0.94)	0.92 (−0.86)	0.92 (−0.84)
Age in 1996	1.34* (2.31)	1.22 (1.55)	1.29* (2.04)	1.25* (1.74)
Has coresident brother	0.61* (−2.14)	0.70 (−1.48)	0.69 (−1.55)	0.68 (−1.57)
Birth Order	0.97 (−0.52)	0.85* (−2.43)	0.82** (−3.25)	0.85* (−2.57)
Time	1.03*** (10.31)	1.03*** (10.94)	1.03*** (11.39)	1.03*** (11.88)
N (person-months)	18465	18465	18465	18465
N (persons marrying)	133	133	133	133

Note: Estimates are presented as odds ratios. T-ratios are given in parentheses.

^aBhramin/Chhetri is the reference category.

^† Denotes that one-tailed tests were used, all other variables used two-tailed tests of significance

^*p<.05

^**p<.01

^***p<.001

To investigate whether this attitude has independent associations from attitude about ideal family size, we tested a model in which both attitudes for all three family members were included (not shown). The direction and significance of each family member’s attitudes did not change compared to the final models presented in Tables 2 and 3. Thus, the two attitudes are independently associated with sons’ marriage timing.¹²

The statistical independence of the consequences of these three different family members’ attitudes toward the same specific topic – sons’ care for their parents in old age – highlights the complexity of these ideational influences on behavior even within a single substantive domain. Sons’ marriage timing responds similarly to both fathers’ and sons’ attitudes about old-age care. The independence of these two estimates is important. It means that, while substantial socialization may produce this similarity, paternal attitudes can have an influence independent of sons’ own attitudes, even when sons’ attitudes are quite relevant to the outcome. Both fathers and sons may anticipate the advantages of adding a wife to the family if sons are responsible for providing old-age care to their parents, but each person exerts an independent influence on sons’ marriage timing. Parental social control mechanisms are likely in play even when the fathers’ and sons’ motivations are in the same direction. As a result, the sum of these two independent associations is that sons who both expect to care for elderly parents and have fathers who expect them to provide such care marry much more rapidly than other sons.

By contrast, mothers’ positive attitude about sons providing old-age support slows their sons’ marriages. This opposing consequence of mothers’ attitude is both independent of and statistically different from the consequences of both fathers’ and sons’ own attitudes on this topic. It may be that the potential competition between mothers and their daughters-in-law leads mothers who expect their sons to provide their old-age care to be apprehensive about bringing a new wife into the household. Historically, daughters-in-law have been an important source of labor for young men’s mothers (Bennett 1983). However, more recent years have been characterized by more affection-based relationships between men and their new wives (Ghimire et al. 2006), which may change the way that mothers-in-law feel about daughters-in-law. With this type of marriage on the rise, mothers may perceive that married sons will put more effort toward caring for their wives than their mothers (Bennett 1983; Stone 1978; Watkins 1996). If this is the case, mothers who expect old-age care from their sons may actually have a motivation to slow their sons’ marriages. More investigation of this intriguing result is required to determine the mechanisms responsible for this observed association.

In the context of our investigation, this result demonstrates two crucial points. First, mothers’ attitudes can have a socializing influence on their sons’ behaviors independent of both fathers’ attitudes and sons’ own attitudes. In spite of assortative marriages of mothers and fathers, parental socialization of their children, and a shared social environment, there are enough differences in experience and circumstances to produce independence in the consequences of these three different family members’ specific attitudes. Second, mothers’ attitudes may actually shape sons’ behavior in the opposite direction of fathers’ attitudes on the same topic. In this particular context, with respect to an attitude about a specific behavior, even when mothers and fathers agree on the topic, the consequences for their sons’ behavior can be in opposing directions.

Conclusion

We use innovative measures from multiple family members to estimate the influence of parents’ attitudes on sons’ marriage timing. We find that parental influences are strong in rural Nepal, as in other parts of the world with more individualistic and less collectivist orientations (Liefbroer and de Jong Gierveld 1993; Sewell et al. 1975; Thompson, Alexander, and Entwisle 1988; Whitbeck, Simons, and Kao 1994). A major limitation of previous research on the influence of parental attitudes on their children’s behavior is the absence of measures of attitudes directly from fathers to compare with mothers (Forum on Child and Family Statistics 1998; Thornton 2001). Our results reveal that both mothers and fathers can simultaneously influence their sons, and these influences can be in opposing directions.

Theoretically, we identify setting-specific mechanisms likely to produce not only independence in the consequences of mothers’ and fathers’ attitudes, but for some specific attitudes, consequences in opposing directions. Parental attitudes toward childbearing constitute an important influence in children’s marriage timing (Barber and Axinn 1998; Thornton et al. 2007). In this setting, we have reason to expect mothers’ and fathers’ attitudes to each have an independent influence. But we also have strong reasons to expect parental expectations for care from children in old age to be an important influence on children’s marital timing. With no other institutional forms of old-age care, children and their families are the exclusive source of protection, support, and care for the elderly in Nepal (Bennett 1983; Brauner-Otto 2009; Fricke 1986). However, the role of a new daughter-in-law in helping a son provide that care is differentiated by the gender of the parent – mothers may feel competition for sons’ care, but for fathers this competition is unlikely (Merrill 2011; Sarkisian and Gerstel 2008). For this reason, we argue that parental attitudes toward children’s care of the elderly will be consequential in Nepal, but in opposing directions.

Empirically, we find that associations between fathers’ attitudes and sons’ behavior are independent of associations between mothers’ attitudes and sons’ behavior. In the domain of ideal family size these independent associations are in the same direction, but in the domain of old-age care they are in opposing directions. Fathers’ positive attitudes toward sons’ care of elderly parents speeds sons’ marriage timing, but mothers’ positive attitudes toward sons’ care of elderly parents slows sons’ marriage timing. This finding is consistent with our reasoning that fathers with a strong belief in sons’ responsibility for old-age care focus on the advantage of having a daughter-in-law to perform caregiving tasks, whereas mothers with the same belief are more hesitant to admit a new wife into the home who may reduce the quality of the mother-son relationship (Merrill 2011; Sarkisian and Gerstel 2008). These associations are important to worldwide consideration of parental influences on children’s behavior because they demonstrate that mothers and fathers not only have the potential for independent influences on their children, but these influences can be in opposing directions.

Why is the study of parental influences on marriage timing in Nepal relevant to scientists studying marriage and family in any other setting, especially settings as different as the United States or Western Europe? Actually, the remote Himalayan setting provides a comparative perspective on intergenerational family dynamics, making it exceptionally relevant to all research on these topics. First, our findings reveal the importance of employing specific attitudes to study parental influences on behavioral outcomes across a wide range of social, economic, and cultural circumstances. Specific parental attitudes influence children’s behavior independent of the children’s own attitudes, consistent with both substantial social control mechanisms and broad socialization that reaches beyond a single specific attitude (Barber 2000, 2001; Gecas and Seff 1990; Smith 1988). This appears just as true in Nepal as in the US, and it is likely true in every social, economic, and cultural circumstance in between – and probably those not “in-between” as well.

Second, we document that attitudes about family behaviors other than marriage can have an important influence on marriage behavior in a radically different setting. Evidence from the US demonstrates that one of these attitudes – the attitude toward childbearing – accelerates children’s marriage just as it does in this analysis of Nepal (Barber and Axinn 1998). In the US, parental attitudes on topics like education and work also influence marriage timing (Thornton et al. 2007). In Nepal we find attitudes about old age care do, too. This means that a substantively wide range of specific parental attitudes can be important in predicting any particular child behavior across a wide range of settings. Moreover, they may be different attitudes in different settings or populations. This is all consistent with theory (Ajzen 1988; Barber 2001), but empirically the door is wide open – all manner of specific attitudes may be highly relevant, depending on the specific behavior and setting.

Third, by adding direct measures of both mothers’ and fathers’ attitudes we are able to compare the role of each parent’s attitudes. Our results reveal that each parent has a substantial independent influence on their sons’ marriage timing, but that both the magnitude and direction of the influence varies depending on the specific substantive domain. Because Western literature also reveals a strong parental influence, this result motivates the expansion of family studies in these individualistic contexts to consider the individual pressure that each parent may exert on their children. These results also motivate advances toward the consideration of specific family relationships in intergenerational influences, adding substantial complexity. The attitudes of many other closely-connected people – including not only the individual and his or her parents, but also siblings, other relatives, and close friends—deserve careful investigation. Again, the implications for research across all settings are mammoth: multiple different others’ attitudes may be simultaneously relevant to each specific behavioral decision, but perhaps differently in different settings.

Though the context-, attitude-, and relationship-specific nature of our investigation inherently poses a limitation to the generalizability of our findings, the findings point toward many new, high priority topics of investigation. Parallel investigations of specific attitudes juxtaposing comparable measures from mothers and fathers across a wider range of settings are among the highest priorities. Even in comparable settings, future investigations of similar models of daughters’ behaviors have the potential to contribute greatly to our understanding of same-sex parenting. Our findings point toward substantial complexity in these intergenerational and ideational influences on behavior. Nevertheless, a clearer understanding of the independent consequences of mothers’ and fathers’ ideas has the potential to significantly advance models of parental influences on children.