|Home | About | Journals | Submit | Contact Us | Français|
Is young people’s marriage behavior determined by their socioeconomic characteristics or their endorsement of developmental idealism? This article addresses this question using a unique, longitudinal data set from Nepal and provides the first individual-level test of developmental idealism theory. We find that unmarried individuals with greater endorsement of developmental idealism in 2008 were more likely by 2012 to choose their own spouse, including a spouse of a different caste, rather than have an arranged marriage. Those with salaried work experience were also less likely to have arranged marriages, but urban proximity and education were not significant. We conclude that both developmental idealism and socioeconomic characteristics influence marriage and their influences are largely independent.
For decades, sociologists have theorized about sources of variation in family behavior including marriage, fertility, and family structure. Early theories on the family emphasized the importance of socioeconomic factors as sources of variation (Thornton 2005; Inglehart and Baker 2000; Adams 2010). For example, those living in urban areas, working in industrial occupations, and with high levels of education were expected to choose their own spouses, live in nuclear families, and give birth to only a few children (Burgess and Locke 1960; Goode 1963; Nimkoff 1965). By contrast, those living in rural areas, working in agriculture, and with low levels of education were expected to have arranged marriages, live in extended families, and have many children.
More recently, scholars have emphasized the importance of ideational factors, including beliefs, values, attitudes, cultural scripts, norms, and schemas, as influences on family behavior (Bachrach 2013; Fricke 1997; Johnson-Hanks et al. 2011; Watkins 2000). Johnson-Hanks and colleagues (2011) describe how schemas influence individual behavior by providing people ways to understand and interact with the world. Schemas tell people, for example, what marriage is, how to marry, and what kinds of marriage behaviors are desirable. Bachrach (2013) further suggests that schemas are linked together in dynamic networks, which at their highest level make up culture itself.
Empirical work has also begun to support the influence of ideational factors on family behavior (Buttenheim and Nobles 2009; Cunningham 2005; Hayford and Morgan 2008). For example, American men with more positive attitudes towards marrying someone with children are more likely to marry single mothers (Goldscheider, Kaufman and Sassler 2009) and Nepali men who believe they should care for their parents in old age marry more quickly (Jennings, Axinn and Ghimire 2012). Similarly, at the macro level, descriptions of the second demographic transition suggest that the rise of secularism and individualism are key drivers of sub-replacement fertility, divorce, and cohabitation (Lesthaeghe 2010; Van de Kaa 1987).
Thornton (2001, 2005) points to a particular collection of beliefs and values, which he terms developmental idealism (DI), as a powerful influence on family behavior. These values and beliefs relate family behaviors to development and identify certain behaviors as more desirable than others. For example, according to developmental idealism, choosing one’s own spouse, living in nuclear families, and having a small number of children are not only inherently good, but will help make people wealthier, healthier, and happier. By contrast, having an arranged marriage, living in extended families, and having a large number of children are not inherently good and do not bring wealth, health, and happiness. Developmental idealism theory suggests that individuals who endorse developmental idealism are expected to engage in family behaviors that are compatible with the beliefs and values of developmental idealism.
While Thornton (2001, 2005) suggests that developmental idealism is an important influence on family behavior, this theory has only been evaluated at the macro level. Previous studies that explore the influence of developmental idealism on family behavior, analyze demographic trends in combination with social histories of regions and countries. These macro studies point to developmental idealism as a source of fertility decline worldwide (Thornton et al. 2012b), nuclear family structure in the Middle East (Yount and Rashad 2008), divorce in Indonesia (Cammack and Heaton 2011), and an array of changes in marriage and childbearing in Iran, Turkey, and Eastern Europe (Kavas and Thornton 2013; Loeffler and Friedl 2014; Thornton and Philipov 2009).
No previous individual-level studies have examined whether endorsement of developmental idealism does indeed influence family behavior. This lack of empirical evidence arises in part from the challenge of rigorously linking ideational factors to individual behavior. Unlike socioeconomic factors, such as education and occupation, high quality ideational measures cannot be collected retrospectively. Individuals’ beliefs and values must be measured before the relevant behavior has occurred and then individuals must be followed prospectively so that their behavior can be observed later in time. Family behaviors, such as marriage and childbearing, are often spread out over long periods. Thus, data collection must continue for several years. This type of longitudinal research design was recently implemented in Chitwan Valley, Nepal. Measures of developmental idealism were administered to a population sample in a baseline survey in 2008. Data on these respondents’ family behaviors were then collected prospectively through 2012. This survey is the only source of comprehensive, longitudinal data on family behaviors and values and beliefs related to developmental idealism that has ever been collected and analyzed. Thus, these data are uniquely designed to provide ground-breaking insights into family processes.
This article uses these innovative, longitudinal data from Nepal to examine whether developmental idealism influences individual marriage behavior and, if so, how it compares to key socioeconomic characteristics – comprising nonfamily work, education, and urban proximity. In other words, we evaluate whether developmental idealism is more or less influential than these socioeconomic characteristics and whether developmental idealism exerts a unique influence that operates outside of socioeconomic characteristics. Our evaluation draws on a cohort of young people who were never married in 2008 when data collection began, but were married by 2012 when data collection ended. We evaluate whether those with greater endorsement of developmental idealism when they were unmarried in 2008 were more likely to choose their own spouse and choose a spouse of a different caste, rather than marry a same-caste spouse chosen by their parents. We focus on whether young people choose their own spouse and whether the spouse is a different caste because these behaviors are empirically intertwined and theoretically salient. These behaviors are the focus of many of the values and beliefs found in developmental idealism, yet previous research points to the importance of socioeconomic characteristics in leading to young people choosing their own spouses, including different caste spouses.
This is the first study to use these unique, longitudinal data to provide a rigorous test of whether there is a link between developmental idealism and any family behavior at the individual level. We also expand conceptualization of developmental idealism theory by drawing on other theoretical traditions. We draw on the network model of culture (Bachrach 2013) and the theory of conjunctural action, or conjuncturalism, (Johnson-Hanks et al. 2011) to conceptualize the connections among the values and beliefs that comprise developmental idealism. We also use the theory of reasoned action to explain how endorsement of developmental idealism is translated into behavior (Fishbein and Ajzen 2010). Thus, drawing on an exceptionally high quality, longitudinal data source, this article provides important contributions to understanding variation in family behavior and, by extension, theories of family change. Our focus on values and beliefs also contributes to a broader sociological discussion about the role of ideational influences in social processes (Hitlin and Piliavin 2004; Manza and Brooks 2012; Ryo 2013; Vaisey 2009). We contribute to the growing evidence that values and beliefs do act as independent and important influences on social outcomes.
This is also the first study to examine influences on intercaste marriage using longitudinal data from a population sample. Caste marital endogamy is an important part of the caste system, which is interwoven with stratification, culture, and religion in South Asia and its diaspora. It is endogamy that supports and defines castes by ensuring the reproduction of demarcated caste groups (Gore 1965; Karve 1965; Mody 2008). Substantial numbers of intercaste marriages would lead to people of ambiguous caste status in the short term and the blurring of caste identities and distinctions in the long term. Thus, the identification of influences on intercaste marriage has broad implications for inequality and social change in South Asia – a region that is home to nearly one quarter of the world’s population.
Marriage customs vary within Nepal and the broader South Asian region by caste, religion, locality, and over time (Karve 1965; Kolenda 1987; Majupuria and Majupuria 1978). Despite variation in this large and diverse region, there are also many family customs that apply to large proportions of the population. One such custom is arranged marriage. Customarily, parents, along with other senior family members, chose spouses for their children on the basis of caste, as well as religion, social and economic standing, and other characteristics of prospective spouses and their families (Bennett 1983; Karve 1965; Madan 1965; Uberoi 2006). Arranged marriages reflect the importance of the status and well-being of the family as a whole over the satisfaction of individual preferences and affections. Arranged marriages are also important to South Asian society beyond the family. Marital caste endogamy supports the caste system, which is a defining institution of South Asia. It is caste endogamy that ensures the reproduction of separate, rather than melded, caste groups.
Historically, intercaste marriages did exist in South Asia, but were usually rare and stigmatized. For example, during fieldwork in Kashmir in the 1950s, Madan (1965) recorded a marriage between a high caste Brahmin man and a woman believed to be of a lower caste from outside the region. Villagers refused to accept food from the woman and did not visit the house after the marriage. Stronger ostracism, public punishments, and even suicide and honor killings have been documented in response to intercaste elopements (Chowdhry 2007; Kolenda 1987; Mody 2008). In Nepal, some amount of intercaste marriage has been tolerated and there were special caste categorizations for the children of such marriages (Bennett 1983; Guneratne 2002; Majupuria and Majupuria 1978). In general, higher caste men were sometimes allowed to marry lower caste women, especially as a second wife or “concubine”. Thus, even intercaste marriages that were practiced historically were not contracted solely according to young people’s affections.
In recent years, customary marriage practices have been eroding in the region. In India, the vast majority of marriages are still arranged by parents (Desai and Andrist 2010), but parents show an increased willingness to take children’s preferences, attractions, and potential romantic compatibility into account (Caldwell, Reddy and Caldwell 1983; Fuller and Narasimhan 2008; Uberoi 2006). In Nepal, there has been even greater change. In Chitwan Valley, virtually none of the 1936–45 marriage cohort participated in the selection of their own spouse, while just over half of the 1986–95 marriage cohort participated in choosing their spouse (Axinn, Ghimire and Barber 2008; Ghimire et al. 2006). Thus, choosing one’s own spouse is now common in Chitwan Valley. This decline of arranged marriage suggests there may also be substantial variation in caste endogamy. Intercaste marriages usually occur when young people choose on their own to marry someone of a different caste. In other words, they are a result of elopements and what are known in the region as “love marriages,” rather than arranged marriages. For the most part, parents follow the custom of caste endogamy when arranging marriages for their children.
Previous studies on marriage in South Asia emphasize the role of socioeconomic characteristics in allowing young people to choose their own, sometimes different caste, spouses. In particular, ethnographic and case studies identify high levels of education, working outside the family, and living in urban areas as key influences (Corwin 1977; Kannan 1963; Ross 1961). These socioeconomic characteristics are believed to provide young people greater economic and social independence from their families. In turn, greater independence increases young people’s ability to convince parents to approve of their chosen partner or marry without parental approval. These characteristics are also expected to provide more opportunity for young people to mix with other castes and, thus, greater exposure to potential intercaste spouses.
In contrast, developmental idealism theory suggests that some people choose their own, sometimes different caste, spouse because they endorse the values and beliefs of developmental idealism. The origins of developmental idealism lie in modernization theory, which is a model of social change that was established among scholars in northwest Europe by the 1700s to understand and explain how the world works (Thornton 2001; Thornton 2005). According to modernization theory, as originally formulated in the 18th and 19th centuries, societies move along a developmental trajectory from traditional and less developed to modern and more developed. These early scholars believed that their own region of northwest Europe occupied the pinnacle of the development trajectory and was “modern.” Conversely, “traditional” societies were those located outside of northwest Europe, which scholars viewed as occupying earlier stages in the development trajectory.
In this modernization model, the socioeconomic characteristics of northwest Europe were defined as modern. Thus, a modern society was an industrialized, urbanized society with high levels of education and economic productivity. On the family side, this modernization model defined a modern family as embodying the family practices of northwest Europe, which include, for example, nuclear family households, individualism, and marriages chosen largely by young people on the basis of affection. Conversely, the modernization model defined traditional family behaviors as those practiced in societies outside northwest Europe, including, for example, extended family households, strong family solidarity, and marriages arranged largely by parents without regard to the existence of affection between the couple before marriage. Furthermore, this model suggested that modern societies and modern families were causally interconnected: modern socioeconomic structures helped produce modern families, and modern families helped produce modern societies.
Developmental idealism is the collection of beliefs and values that emerged from modernization theory and provides societies and individuals a model for both understanding the world and living in the world (Thornton 2001; Thornton 2005). As such, developmental idealism provides both goals to be attained and mechanisms for attaining them. The central themes of developmental idealism were summarized by Thornton in four propositions. The first proposition identifies modern society, as defined earlier, as being good and attainable, possessing wealth, well-being, and happiness. The second proposition is that the modern family, as defined earlier, is good and attainable. The third proposition of developmental idealism is that modern family is both a cause and effect of modern society. The fourth proposition is that individuals have a right to be free and equal and relationships should be based on consent. This fourth proposition points to underlying values of individualism, freedom, and equality. To these four propositions of developmental idealism, we add a fifth—that societies are dynamic and moving from having traits defined as traditional towards traits defined as modern.
To conceptualize how the values and beliefs that comprise developmental idealism are related to each other, we draw on Bachrach’s (2013) network model of culture and conjuncturalism (Johnson-Hanks et al. 2011). Thornton (2005) originally described developmental idealism as a “package” of values and beliefs. The term package, however, can imply a cohesive, singular unit, in which people either accept or reject all of the values and beliefs that constitute developmental idealism as a cohesive whole. Thus, developmental idealism is better described as a network of values and beliefs. The schemas that comprise developmental idealism are not a singular package, yet neither are these schemas unrelated and independent. As Bachrach (2013) describes, schemas are grouped together in networks, which comprise “connected bits of meaning,” with the interconnections ranging from tight to loose. We contend that the values and beliefs of developmental idealism are connected together in such a network of loose and tight connections.
Empirical studies support this conceptualization of developmental idealism as a network of loosely and tightly related schemas. Factor analyses of survey data from Nepal point to a multidimensional structure of developmental idealism (Thornton, Young-DeMarco and Ghimire 2012). Individuals who endorsed one schema of developmental idealism were more likely to endorse certain other schemas. However, the correlations among different schemas were not strong enough to support the measurement of developmental idealism with one, single factor. Similarly, using survey data from China, Lai and Thornton (Forthcoming) found that individuals who valued development and believed that development is associated with modern family behaviors were, in turn, more likely to view modern family behaviors positively. However, these dimensions were moderately, rather than tightly, interconnected.
As noted earlier, the original formulation of developmental idealism emerged from the modernization theory found among northwest European elites. From there, developmental idealism spread both geographically around the world and from the elite to ordinary people in everyday life where it can affect people’s decisionmaking and behavior. This diffusion occurred through colonization, mass education, mass media, international aid programs, social networks, international tourism, and migration (Thornton 2005). Recent surveys document that developmental hierarchies and developmental idealism schemas are now widely accepted among populations in 13 diverse countries (Abbasi-Shavazi, Askari-Nodoushan and Thornton 2012; Melegh et al. 2013; Thornton et al. 2012a; Thornton et al. 2012b; Thornton, Ghimire and Mitchell 2012). For example, the majority of respondents to surveys in Argentina, China, Egypt, Iran, Nepal, and the United States believe that development reduces fertility and that fertility decline increases development (Thornton et al. 2012b). Ethnographic studies point to the relevance of developmental idealism in an even wider range of settings (Cole and Thomas 2009; Frye 2012; Hirsch 2003; Hirsch and Wardlow 2006; Rebhun 1999). Endorsement of developmental idealism is also correlated with individual characteristics in ways that are consistent with diffusion mechanisms (Abbasi-Shavazi, Askari-Nodoushan and Thornton 2012; Melegh et al. 2013; Thornton, Ghimire and Mitchell 2012). Those who live closer to urban areas, have higher levels of education, work outside the family, and have greater mass media exposure are more likely to endorse developmental idealism.
According to the network model of culture and conjuncturalism, people reproduce and change schemas over time (Bachrach 2013; Johnson-Hanks et al. 2011). Specifically, as people come into contact with schemas, some are adopted, some are discarded, and others are reframed through group and individual processes. Further, due to their centrality in the human experience, family behaviors usually have many overlapping and related schemas, some of which are reinforcing, while others are discordant. The spread of developmental idealism around the world fits with this conceptualization of group and individual processes shaping local schemas. As ordinary people came into contact with developmental idealism they incorporated those values and beliefs into local networks of schemas that already included other indigenous, family-related schemas. In these processes, many of the schemas of developmental idealism were accepted, some were modified to fit with indigenous schemas, and others were rejected entirely (Thornton 2001; Thornton 2005). In China, for example, the value of low fertility and its power to hasten development was strongly embraced, while individualist beliefs about aging parents being responsible for themselves were rejected in favor of a customary emphasis on children supporting their parents in old age.
Some of the developmental idealism schemas that were adopted were kept alongside discordant schemas that contradict them (Yount and Rashad 2008). For example, Palestinians in Israel embraced the value of a small family, but this value exists side-by-side with the schema that a large family is highly valuable for reproducing the Palestinian nation (Kanaaneh 2002). Survey data from Iran provides another example of such discordant schemas. In keeping with developmental idealism, the majority of Iranian respondents believed that development decreases early marriage, arranged marriage, and fertility (Abbasi-Shavazi, Askari-Nodoushan and Thornton 2012). Contrary to developmental idealism though, a majority also believed that development increases family unity and respect for elders.
As these examples imply, the adoption of developmental idealism is not random. The cultural, historical, demographic, and economic contexts of particular localities shape how developmental idealism is adopted and incorporated into local schema networks. For example, Watkins (2000) describes how Luos in 1960s Kenya adopted developmental idealism schemas that many children are a burden and a small family is progressive, yet they resisted the value of contraceptive use. This resistance was rooted in their indigenous values that fertility control was morally wrong. However, over time, as contraception became widely available through the expansion of family planning services, economic troubles made children’s schooling fees especially burdensome, and urban elites embraced family planning, Luos eventually embraced the value of contraceptive use. In a South Asian example, Allendorf (2013) argues that developmental idealist schemas about “modern” marriage have been readily adopted in the Darjeeling Hills of India, unlike many other parts of India, because the Darjeeling Hills are dominated by ethnic groups that historically practiced forms of marriage that were similar to British, “modern” practices.
The importance of local context is what makes marriage a particularly salient behavior for testing the influence of developmental idealism in South Asia. Under developmental idealism, in “modern” marriages people choose their own spouses, usually on the basis of affection. Further, under developmental idealism, individuals are seen as equal and without differential standing on the basis of caste. Choosing one’s own spouse, and going even further to choose one of a different caste, is thus consistent with developmental idealism and “modern” marriage, as well as profoundly inconsistent with South Asian custom. Such behavior is in keeping with the values of individualism, freedom, and equality that underlie developmental idealism. Conversely, arranged marriages contradict key aspects of developmental idealism by supporting inequality on the basis of caste, the importance of the family and caste collective over the individual, and the importance of caste and other considerations over love as the basis of marriage. In contrast, developmental idealism schemas about intercaste marriage and spouse choice would be unremarkable and irrelevant in contexts, such as Latin America, that did not customarily practice arranged marriages.
Before proceeding further it is important to make other aspects of our approach explicit. We recognize that modernization theory is no longer widely accepted among scholars, and we are neither endorsing nor critiquing it. Further, our intention is not to reflect on whether the values and beliefs found within developmental idealism are good or bad, accurate or inaccurate. Instead, our goal is to test whether individuals’ endorsement of these values and beliefs that emerged from modernization theory influence their marriage behavior.
It is also important to conceptualize the connections between developmental idealism and socioeconomic factors, which are also expected to influence marriage behavior. Developmental idealism and socioeconomic factors are causally intertwined (Thornton 2005: 214–229). On one hand, socioeconomic characteristics influence people’s exposure to and adoption of developmental individualism. Education provides the clearest example of this pathway since schooling is purposefully designed to inculcate students with new skills, ideas, and information. Thus, education directly encourages individuals to encounter and adopt developmental idealism. Working outside the family or residing in urban areas provides additional ways for people to encounter developmental idealism, but only indirectly. Thus, these pathways suggest that some, but not all, of the influence of socioeconomic factors on marriage behavior is mediated by developmental idealism.
On the other hand, causality also goes in the opposite direction, with developmental idealism influencing socioeconomic characteristics. People may attend school, work in nonfamily occupations, and move to urban areas because they and their parents view them as good and modern. Going further back in time, the schools that people attend, the factories and offices they work in, and the cities they reside in exist because other people created them. One important reason that people created those institutions was their belief that schools, factories, and cities were good and modern and would hasten development. Thus, these pathways indicate that the influence of developmental idealism on family behavior is also mediated by socioeconomic characteristics, especially in the longer term. This conceptualization of ideational and socioeconomic factors as interdependent is further consistent with conjuncturalism, which characterizes schemas and materials as causally interwined (Johnson-Hanks et al. 2011).
Developmental idealism comprises the content of the schemas that are the focus of our analysis. As described above, we draw on the network model of culture and conjuncturalism to conceptualize how the schemas that make up developmental idealism are related to each other. However, the central contribution of this article is the examination of the influence of developmental idealism on individual behavior, a goal that was first set out in Thornton’s (2005: 242) original research agenda for evaluating developmental idealism theory. Thus, our theoretical framework must not only describe the content of our schemas of interest, it must also conceptualize how such schemas influence individual behavior.
We use the theory of reasoned action to conceptualize how endorsement of developmental idealism increases the likelihood that an individual chooses their own, potentially different caste, spouse (Ajzen and Fishbein 1980; Fishbein and Ajzen 2010). The theory provides a general framework connecting schemas to behavior. We use it to connect a particular network of schemas, those of developmental idealism, to marriage behavior. In the theory of reasoned action, behavior is a result of a process in which individuals draw consistently on the schemas available to them. The process can range from a lengthy and consciously deliberate procedure to a spontaneous, seemingly instantaneous action.
We focus on two kinds of schemas: beliefs and values. Beliefs are cognitive statements about the way the world works; they include attributing certain characteristics, qualities, or consequences to objects, concepts, and behaviors. For example, thinking that most people in the community marry people of the same caste is a belief. Values constitute positive or negative evaluations of a specific object, concept, or behavior. For example, disliking intercaste marriages or thinking small families are good are both values. The broader ideals of freedom, individualism, and equality are also values. (We use the term schema to simultaneously refer to both values and beliefs.)
We describe how these values and beliefs influence behavior in three steps. First, an individual encounters and adopts schemas from a variety of sources, including personal observation, the media, education, and contact with family and friends. Which elements an individual encounters and adopts are influenced by a range of external factors, such as age, gender, education, place of residence, and religion. Second, individuals draw on these schemas to create an intention, or readiness, to perform a behavior or not. People are more likely to intend to perform a behavior when they hold values that favor it and beliefs that are consistent with it. Finally, their intentions influence whether they perform the behavior, but this connection is moderated by the amount of control the individual has over the behavior.
We suggest that developmental idealism presents an influential network of beliefs and values about caste and marriage, as well as family and development more broadly, that are consistent with choosing one’s own spouse, including a different caste spouse. Contrary to customary South Asian culture, developmental idealism suggests that sustained interactions with people of other castes, choosing one’s own spouse, choosing on the basis of affection, and choosing a spouse of another caste are acceptable and maybe even desirable. Developmental idealism is particularly powerful because it suggests that such “modern” behaviors are not only inherently good, but bring health, wealth, and happiness. In turn, individuals who endorse developmental idealism will be more likely to engage in a range of behaviors that make such marriages more likely. Specifically, they will be more likely to interact with people from a wide range of castes, form attractions and romantic relationships regardless of caste, make special efforts to convince their parents to approve of their chosen partners, and marry even without the approval of their parents. We do not expect those who endorse developmental idealism to specifically intend, or plan, to marry someone of a different caste, although some probably do. Instead, we emphasize the readiness aspect of intentions. People who endorse developmental idealism are more ready and willing to have an intercaste marriage if the right circumstances arise; they are also more likely to put themselves in such circumstances.
The theory of reasoned action also predicts that the more closely related in content a value or belief is to a behavior, the more influential the value or belief will be for that behavior. Schemas about spouse choice and intercaste marriage are more closely related to marriage than schemas about other family behaviors, such as the desirability of a small family or divorce. Thus, we expect that developmental idealism schemas concerning intercaste marriage and spouse choice will be more influential than other schemas of developmental idealism.
The data were collected from 2008 to 2012 as part of the Chitwan Valley Family Study. Data collection for the project began in 2008 with a baseline survey of all individuals aged 15–59 residing in households located in a representative sample of 151 neighborhoods in the Chitwan Valley of Nepal. Current residents aged 12–14 in 2008 were also part of the data collection, but were not administered the baseline interview until they reached 15 years of age. Thus, the final sample for the baseline survey comprises 5,840 residents aged 12–59 in 2008 and had a response rate of 97.1%.
The baseline survey included a large battery of developmental idealism items. A number of additional ideational items were also administered a few months after the initial baseline survey in an extended baseline survey. The ideational items were formulated through extensive use of ethnographic fieldwork, in-depth interviews, focus groups, multiple pretests, and a pilot study. Analyses of the data indicate that they are very high quality. There is little missing data and the reliability of measures is high (Thornton, Ghimire and Mitchell 2012; Thornton et al. 2010; Thornton et al. 2012c; Thornton, Young-DeMarco and Ghimire 2012). Further, acquiescence in answering questions is low; respondents were able to accurately distinguish between questions worded in opposite directions, such as agree versus disagree (Thornton, Ghimire and Mitchell 2012).
The 2008 baseline survey was followed by ongoing collection of demographic events, including marriages, through 2012 in a household registry. Any recording of a respondent’s marriage in the household registry was followed by the administration of an additional short survey with the newly-wed called the marital supplement. This marital supplement comprised a set of questions about the nature of the respondent’s marriage and the characteristics of their spouse.
Our analytical sample comprises 465 individuals who were never married and aged 12–34 in 2008 and then subsequently married between 2008 and 2012. Thus, the respondents in our analytical sample were all married in slightly more than a four year period and constitute a marriage cohort. This analytical sample was administered questions on developmental idealism at the start of the study period, before they married, and then administered questions about their marriage in the marital supplement after they married. This longitudinal design provides a rigorous basis for identifying the influence of developmental idealism on marriage behavior.
Our dependent variable draws on two questions from the marital supplement survey, one on spouse choice and the other on the spouse’s caste. In reference to spouse choice, respondents responded to the following question: “People marry in different ways. Sometimes our parents or relatives decide who we should marry, and sometimes we decide ourselves. In your case, who selected your husband/wife? Your parents/relatives, yourself, or both?” In reference to caste, respondents were asked, “Was your marriage to someone who is in your own caste (aphno jaat) or not?” It should be noted that this question measures the respondent’s own report of whether their spouse is the same caste or not. We did not compare the caste (jaat) of the respondent to that of their spouse, nor was the spouse’s caste collected.
As described above, intercaste marriage is intertwined with spouse choice in South Asia. Generally, people only marry someone of a different caste if they choose their own spouse. While there are exceptions, when parents or relatives choose a spouse in an arranged marriage they choose someone of the same caste. In our sample, 42% of young people who chose by themselves married a spouse of a different caste, compared to only 5% of those whose parents or relatives were involved in the choice. Only 12 young people (2.6% of the total sample) were in the rare position of their parents or relatives participating in the choice of a spouse who was a different caste.
Given this overlap between choosing a spouse on one’s own and intercaste marriage, we combined responses on these questions to create one dependent variable with three categories: 1) Parents/relatives participated in choosing a same caste spouse (ie. arranged marriage); 2) Respondent alone chose a same caste spouse; and 3) Respondent alone chose a different caste spouse (ie. intercaste marriage). The 12 respondents whose parents participated in choosing different caste spouses are included in the third category of intercaste marriage, even though they did not choose their spouses on their own. However, the results are not affected by this categorization. We repeated the analyses with these 12 respondents placed in the first category of arranged marriage and another analysis dropped them from the analytical sample entirely. In both cases, the results are substantively identical to those presented here.
The majority of respondents had arranged marriages, but sizeable proportions of youth did not follow South Asian custom. Specifically, 53% of respondents had an arranged marriage, 26% chose a same caste spouse on their own, and 21% chose a different caste spouse. Those with arranged marriages practiced a family behavior that is consistent with South Asian custom and contrary to developmental idealism. The third category of those with intercaste marriages present the opposite; their behavior is in keeping with developmental idealism and contrary to South Asian custom. The middle category of young people who chose a same caste spouse on their own are a mix – the caste of their spouse is consistent with South Asian custom, but their choice behavior is consistent with developmental idealism.
The baseline survey included five modules on developmental idealism. The first module measured beliefs about development causing family change. These questions asked whether various family behaviors would increase, decrease, or stay the same if Nepal became richer in the future. The family behaviors asked about included, for example, intercaste marriage, parents choosing children’s spouses, premarital sex, never marrying, and divorce. The second set of items measured the converse belief, whether family change causes development. These questions asked respondents what would happen if certain family behaviors became more or less common in Nepal: would such changes make Nepal richer, poorer, or have no effect. The third module measured expectations of future changes in family behavior. These questions asked whether the family behaviors would increase, decrease, or stay the same in Nepal during the next 20 years. The fourth set of items measured respondents’ values concerning such future changes. After respondents were asked if they expected a particular change, they were asked whether that particular change would be good, bad, or not matter. Finally, the fifth set of items directly measured respondents’ values about family behaviors. They were asked which of an opposing set of behaviors is better for most people in Nepal today. For example, one question asked whether youth choosing their own marriage partners or parents’ choosing on their children’s behalf is better.
We looked for underlying dimensions of developmental idealism with exploratory factor analyses using tetrachoric correlation matrices. We began by separating out the ten items on caste and spouse choice and estimated factor analyses on those sets of items separately because they are particularly relevant to our dependent variable of marriage behavior, and, as predicted by the theory of reasoned action, should be particularly influential. Given the extremely large number of remaining items, we were not able to put them all into a single factor analysis simultaneously. Thus, we estimated several factor analyses, using different groupings of items to determine which items loaded together. We began by putting together items that were similar based on face value, as well as putting together those that the first author suspected would load together based on her experience of ethnographic fieldwork with a similar population (Author). Three additional developmental idealism factors emerged inductively from the remaining items. Items with low loadings, those below 0.4, were not retained.
We measured developmental idealism with five scales that are standardized versions of the factors retained from the factor analyses. We refer to these five developmental idealism (DI) scales as: 1) caste DI; 2) spouse choice DI; 3) modern family is good, coming, and caused by development, or “modern family” for short; 4) new modern family is coming and caused by development, or “new modern family is coming” for short; and 5) “new modern family is good”. The first two scales on caste and spouse choice include all developmental idealism items on the two behaviors that are particularly relevant to self-choice and intercaste marriage. The third “modern family” scale is about behaviors practiced historically in northwest Europe and, thus, those characterized as “modern” under developmental idealism. These modern behaviors include, for example, low fertility, widow remarriage, and high status of women. Respondents’ values about these modern behaviors are correlated well with their beliefs about how behavior will become more like the modern family in the future in Nepal. Specifically, those who think these behaviors are going to increase in the future also view the behaviors as good. Thus, endorsement of the modern family loaded onto one factor. Because we had separate scales for caste and spouse choice, we did not consider the caste and spouse choice items for this “modern family” scale.
The last two developmental idealism scales are about family behaviors that were not practiced historically in northwest Europe, but have become widespread in Western countries in recent decades. We refer to these behaviors, which include divorce, premarital sex, and cohabitation, as “new modern.” Unlike modern behaviors, respondents’ beliefs about whether new modern behaviors will increase in the future are not correlated with their values about them. Thus, beliefs and values about new modern behaviors loaded onto two separate scales, one measuring beliefs and the other values.
Details on the exploratory factor analyses used to create these five scales, including the loadings of individual items, eigenvalues, and Cronbach’s alphas, are shown in Table 1. The scales account for 27% to 47% of the variance in the individual items, while the alphas range from 0.45 to 0.74. The alphas for three of the scales – caste, spouse choice, and “new modern family is good” – are below the typical cutoff of 0.7 for scale creation. However, these three scales are based on five items, while the other two scales are based on ten and fifteen items respectively. Cronbach’s alpha is sensitive to the number of items; a relatively small number of items can result in a low alpha, while 20 or more items can itself lead to a high alpha (Cortina 1993). In this case, the average inter-item covariance is not systematically lower for the three scales with low alphas, which suggests that it is indeed the small number of items that is driving the low alphas. Additionally, reinforcing the multi-dimensional nature of developmental idealism, the scales have low correlations, ranging from nearly zero to 0.38.
The five developmental idealism scales are standardized versions of the factors retained from the factor analyses and have a mean of zero and a standard deviation of one. We standardized the factors to facilitate comparison across scales and interpretation of results. Thus, when we examine the effect of a one unit difference in a developmental idealism scale on intercaste marriage, we are examining the effect of a one standard deviation unit.
The results of the factor analysis also reinforce the conceptualization of connections among the values and beliefs of developmental idealism described above. On one hand, there is clear evidence that developmental idealism is not a singular package. The items do not load all together onto one single factor. Instead, the items load onto multiple factors and some items do not load at all. On the other hand, there is also clear evidence that many values and beliefs of developmental idealism are indeed related to each other. Many of the items load together well onto a relatively small number of factors.
The factor analyses and the percent of people who endorse particular developmental idealism schemas further point to some of the particularities of how developmental idealism has been adopted into schema networks among this marriage cohort in Nepal. For example, this marriage cohort has fully embraced many values and beliefs about the modern family, but are ambivalent about the “new modern” family (Table 1). While the majority of them believe that divorce, cohabitation, and premarital sex will increase over time and are caused by development, only a small minority positively value these behaviors.
We included measures of the three socioeconomic factors that figure prominently in the literature: educational attainment, nonfamily work experience, and distance from the urban center. All three of these socioeconomic measures refer to the time of the baseline survey. Distance from the urban center is a standardized, continuous variable. We standardized the distance to facilitate comparison with the developmental idealism scales, which are also standardized. Educational attainment is a categorical variable that distinguishes among key thresholds, including elementary schooling (0–6 years), secondary schooling (7–9), a school leaving certificate (10–11), and college or more (12+ years).
Nonfamily work experience is a categorical variable with three categories: 1) never worked outside the family; 2) worked outside the family, but never for a salary, and 3) worked outside the family for a salary. Salaried work, such as working as a teacher, in a government office, or for a non-governmental organization, comprises high status work that is highly and regularly paid. By contrast, unsalaried nonfamily work, which is paid in daily or weekly wages, is lower status and often irregular. Unsalaried work includes, for example, agricultural labor, carpentry, and working as a porter. Thus, holding a salaried job should confer the greatest independence and put people in the most contact with local elites and foreigners. In turn, we expect that salaried work will have the largest influence on marriage behavior.
The temporal ordering of our independent and dependent variables minimizes the potential for reverse causation. However, the association could still be inflated by correlations with other factors that are associated with both marriage behavior and endorsing developmental idealism. Thus, we also included three additional control variables: gender, age at marriage, and caste/ethnicity. Gender and caste/ethnicity are exogenous to both developmental idealism and the socioeconomic variables. The descriptive statistics for all variables appear in Table 2.
Caste/ethnicity is a four category variable denoting four main caste groupings: 1) Chhetri-Bahun, 2) Dalit, 3) Hill Janajati, and 4) Terai Janajati. Chhetri-Bahuns are high castes, also known as Brahman and Kshatriya in the broader region. Dalits are low castes, such as Kami, Sarki, and Damai castes. Hill Janajati refers to groups that historically spoke Tibeto-Burman languages and are indigenous to the Himalayan foothills, such as Tamang, Magar, Newar, Gurung, and Rai. Finally, Terai Janajati includes members of ethnic groups that are indigenous to the Terai, a flat plain that runs the length of Nepal’s southern border with India. Terai Janajati largely consists of only one group, the Tharu, but also includes individuals from other castes. Thus, each of these four categories includes more than one individual caste (jaat).
Since our dependent variable is trichotomous we use multinomial logit models and present a series of seven models. The first model includes only socioeconomic and control variables, which allows us to examine the influence of socioeconomic characteristics on marriage behavior without being affected by the inclusion of developmental idealism in the model. Thus, this first model provides an estimate of the largest possible effect of the socioeconomic factors, one that is not dampened by the potential mediating role of developmental idealism. The following five models comprise one model for each of the five developmental idealism scales. Each of these models includes one developmental idealism scale along with the socioeconomic variables and controls. We put one developmental idealism scale in a model at a time to estimate the influence of each scale when it is unaffected by the inclusion of other scales in the model. Finally, in the last model, we put all five developmental idealism scales in the model simultaneously, along with socioeconomic and control variables. All models also adjust for the survey design by including robust standard errors that control for the clustering of respondents within neighborhoods.
We present our effect coefficients as relative risk ratios that indicate multiplicative increase or decrease in the risk of a respondent 1) choosing a same caste spouse on their own and 2) having an intercaste marriage versus 3) the reference category of an arranged marriage. For example, a relative risk ratio of 1.0 would indicate that developmental idealism has no effect, a relative risk ratio of 1.50 would indicate that being one standard deviation higher in a developmental idealism scale raises the odds of an intercaste marriage by 50% relative to an arranged marriage, and an odds ratio of .5 would indicate that one standard deviation higher in a developmental idealism scale lowers the odds of an intercaste marriage by 50% relative to an arranged marriage.
To provide a more holistic and intuitive comparison of the influence of developmental idealism versus socioeconomic characteristics, we also present predicted probabilities of intercaste marriage for various values of the predictor variables. These predicted probabilities are based on the final model, which includes all five developmental idealism scales, socioeconomic characteristics, and controls, as well as an earlier model that includes only socioeconomic characteristics and control. We use these predicted probabilities to measure the influence of developmental idealism as a whole by comparing the likelihood of an intercaste marriage for an individual with low endorsement of developmental idealism to that of someone with high endorsement of developmental idealism. Similarly, we measure the influence of socioeconomic characteristics as a group by comparing the predicted probability of an intercaste marriage for someone with little education, no nonfamily work experience, and greater distance from an urban center to the predicted probability for someone with extensive education, salaried work experience, and close proximity to the urban center. Finally, we use the resulting differences in the predicted probabilities to compare the influence of developmental idealism as a whole to that of socioeconomic characteristics as a whole.
We also provide supplementary analyses. First, as discussed above, developmental idealism is not a singular package, yet neither does it consist of unrelated schemas. Thus, we have measured developmental idealism with five scales that draw on multiple developmental idealism schemas. However, within the network of schemas that comprise developmental idealism, it is likely that some schemas about caste and spouse choice are more influential than others. For example, valuing intercaste marriage as good may be more influential than believing that if Nepal becomes richer, self-choice marriages will become more common in the future. Thus, we also include models that show the influence of individual caste and spouse choice developmental idealism schemas. Specifically, these ten models include one caste or spouse choice item, all three socioeconomic variables, and all three controls.
We also provide additional analysis that shows the associations between developmental idealism and socioeconomic characteristics, as well as controls. This supplementary analysis comprises five ordinary least squares regression models in which the five developmental idealism scales comprise the dependent variables – with one model for each developmental idealism scale. The independent variables in these models are the socioeconomic and control variables listed above. We also performed robustness checks, which are described below, that further explore associations among the variables in the analysis.
The relative risk ratios from the multinomial models are shown in Table 3. These models show that young people with greater endorsement of developmental idealism are more likely to choose their own spouse and have an intercaste marriage, rather than an arranged marriage. In particular, as expected, two scales stand out – those for caste and spouse choice developmental idealism. The relative risk ratios for intercaste, versus arranged, marriage are 1.43 and 1.45 respectively for the caste and spouse choice developmental idealism scales (Models 2 & 3, Table 3). Thus, scoring one standard deviation higher on caste or spouse choice developmental idealism increases the relative risk of having an intercaste marriage, rather than an arranged marriage, by 43% and 45% respectively. Note that in Model 7, with all developmental idealism scales included, these effects are reduced to 1.30 and 1.33 respectively.
Further, the relative risk ratio for a respondent choosing a same caste spouse on their own is 1.28 for the caste developmental idealism scale, indicating that endorsement of developmental idealism increases the odds of a young person choosing their own same caste spouse by almost a third (Model 2, Table 3). Similarly, the relative risk ratio for spouse choice developmental idealism for a respondent choosing a same caste spouse is 1.21 (Model 3), although this effect is not statistically significant and is reduced to 1.17 in the full model (Model 7). These models with caste and spouse choice developmental idealism scales (Models 2 & 3) also fit the data significantly better than the model with only the socioeconomic and control variables (Model 1), further indicating that developmental idealism uniquely accounts for part of the variation in marriage behavior.
Endorsement of a third dimension of developmental idealism, namely “new modern family is coming,” increases the likelihood of an intercaste marriage, but not the risk of choosing a same caste spouse. The relative risk ratio for the “new modern family is coming” scale is a significant 1.33, indicating that scoring one standard deviation higher in endorsement of this scale raises the risk of an intercaste marriage, rather than arranged marriage, by a third (Model 5, Table 3). This effect is reduced to 1.24 in the full model (Model 7, Table 3). The relative risk ratio for this scale on respondent choosing a same caste spouse versus an arranged marriage is a small and insignificant 1.03 (Model 5, Table 3).
The other two developmental idealism scales – modern family and “new modern family is good” – have no effect on marriage behavior. The relative risk ratios for these two scales are all close to one and statistically insignificant (Models 4 & 6, Table 3). Specifically, the relative risk ratios for modern family development idealism are 0.91 for choosing a same caste spouse and 1.12 for an intercaste spouse. The relative risk ratios for “new modern family is good” are 1.10 for choosing a same caste spouse and 0.99 for choosing a different caste spouse.
To delve deeper into which developmental idealism schemas are most influential, we also examined multinomial logit models in which the explanatory variables of interest were the ten individual developmental idealism items on caste and spouse choice that are listed in Table 1. The relative risk ratios indicate that there is substantial variation in the influence of individual developmental idealism schemas (Tables 4 & 5). Some individual items had strikingly large influences that more than double the relative risk of choosing a same caste spouse and having an intercaste marriage. Substantively, these highly influential items tap into values – whether intercaste marriage and spouse choice are good or bad – and beliefs about development causing family change. Other individual items had a moderate level of influence or no influence at all. The items with no influence were the beliefs that intercaste marriage and spouse choice cause development.
The results of the socioeconomic variables provide limited support for the influence of socioeconomic factors. On the supportive side, salaried work experience substantially increases the likelihood of an intercaste marriage. Specifically, the relative risk ratio of 2.27 indicates that young people who worked for a salary have more than double the relative risk of an intercaste marriage compared to those who never worked outside the family (Model 1, Table 3). The relative risk ratio of 1.47 for unsalaried work experience suggests that unsalaried work also increases the likelihood of intercaste marriage, but is not statistically significant. Further, neither salaried, nor unsalaried, work experience significantly affects choosing a same caste spouse. The relative risk ratios of 1.25 for salaried work is suggestive of a modest effect, but is not statistically significant, while the relative risk ratio of 0.95 for unsalaried work is not significant and very close to one.
The results are less supportive for urban proximity. A relative risk ratio of 0.84 indicates that those who live farther away from the urban center are less likely to have an intercaste marriage (Model 1, Table 3). However, while this effect is sizeable and in the predicted direction, it is not statistically significant. Further, those who live farther from the urban center are not less likely to choose a same caste spouse on their own. The relative risk ratio of 1.14 is insignificant and in the wrong direction.
The remaining socioeconomic characteristic – education – appears to have the opposite effect from that predicted, but also lacks statistical significance (Model 1, Table 3). Those with high levels of education are consistently less, not more, likely to choose a same or different case spouse on their own. For example, a youth who has completed seven or more years of schooling has just over two-thirds the odds of having an intercaste marriage compared to someone with 0–6 years (Model 1, Table 3). While these point estimates indicate sizeable effects, none of them are statistically significant.
A comparison of the influence of the socioeconomic factors versus that of developmental idealism indicates that both groups are influential, but with variation within each group. Both developmental idealism and the socioeconomic characteristics include highly influential items, which more than double the relative risk of an intercaste marriage. Both groups also include items that have no influence on intercaste marriage. Further, as seen in the model fits, both developmental idealism and the socioeconomic characteristics provide unique contributions to explaining variation in marriage behavior.
We use predicted probabilities of intercaste marriage to provide a more intuitive and holistic comparison of the influence of developmental idealism as a whole to that of socioeconomic characteristics as a whole (Table 6). These predicted probabilities are calculated using the full multinomial logit model, which includes all five developmental idealism scales, socioeconomic characteristics, and controls (Model 7, Table 3). To measure the influence of developmental idealism as a whole, we compare the predicted probability of an intercaste marriage for an individual with low endorsement of developmental idealism to that of someone with high endorsement. We define low endorsement as one standard deviation below the mean on all five scales, while high endorsement is one standard deviation above the mean on all five scales. All other variables in the model are left at the mean values present in the data. The predicted probability of an intercaste marriage is 0.148 for someone with low developmental idealism endorsement and 0.294 for someone with high endorsement (Table 6). This indicates that moving from a low to high level of developmental idealism raises the probability of an intercaste marriage by 0.146 (0.294 – 0.148).
To provide a comparable estimate of the influence of socioeconomic characteristics as a whole, we compare the predicted probability of someone with 0–6 years of education, no family work experience, and one standard deviation above the mean for distance from the urban center to the predicted probability of someone with 12+ years of education, salaried work experience, and one standard deviation below the mean for distance from the urban center. The predicted probability of an intercaste marriage is 0.204 for someone with little education, no nonfamily work experience, and living far away from the urban area, while it is 0.281 for someone with the opposite characteristics (Table 6). This implies that socioeconomic characteristics as a whole raise the probability of an intercaste marriage by 0.077 (0.281 – 0.204). This difference of 0.077 for the socioeconomic characteristics does not differ significantly from the difference of 0.146 for developmental idealism. However, this comparison suggests that the influence of developmental idealism is just as large, if not larger, than that of socioeconomic characteristics.
As described above, some of the developmental idealism scales and socioeconomic characteristics do not have a statistically significant effect on marriage behavior. Thus, we also calculated another set of predicted probabilities that restricted the variables we manipulated to those with significant effects. For developmental idealism, significant scales include caste DI, spouse choice DI, and “new modern family is coming” DI. For socioeconomic characteristics, the only significant variable is work experience; neither distance from the urban center, nor education were significant. We measure the influence of developmental idealism through a comparison of low versus high endorsement on the three noted developmental idealism scales and that of socioeconomic characteristics through a comparison of those with no nonfamily work experience to those with salaried work experience.
When limiting the manipulations to significant variables, the predicted probability of an intercaste marriage increases by 0.168 when developmental idealism endorsement rises from low to high, while it increases by 0.106 when nonfamily work experience goes from never to salaried work (Table 6). The influence of the socioeconomic characteristics in this case is larger than the previous 0.077 because education – which unexpectedly works to reduce, rather than increase, the probability of intercaste marriage – is not included. However, again, the influence of developmental idealism as a whole does not differ significantly from that of the socioeconomic characteristics. Thus, this comparison provides additional evidence that the influence of developmental idealism is at least as large as that of socioeconomic characteristics.
It is useful to note that our comparisons of the effects of developmental idealism and the socioeconomic variables in Table 6 are based on the full model with all variables in the equation (Model 7, Table 3). The Model 7 estimates are the direct effects of each variable included net of the other variables. It might be argued that these estimates may not completely capture the effects of the socioeconomic characteristics because some of the effects of the socioeconomic characteristics may operate through developmental idealism, and these indirect effects are not reflected in Model 7 of Table 3 or in the estimates of Table 6. This line of reasoning suggests that the estimated effects of the socioeconomic characteristics in Model 1 of Table 3 indicate the total effects of the socioeconomic characteristics (net of each other), that the estimated effects of the socioeconomic characteristics in Model 7 of Table 3 indicate the direct effects of the socioeconomic characteristics net of each other and the developmental idealism scales, and that the differences between the two estimates indicate the indirect effects of the socioeconomic characteristics through the developmental idealism scales (Alwin and Hauser 1975).
A comparison of the socioeconomic coefficients in Models 1 and 7 of Table 3 indicates that, for the most part, the total and direct effects of these variables are quite similar, suggesting that, if this model is true, there is relatively little of the total effects of socioeconomic characteristics operating through the developmental idealism scales. For example, the relative risk ratio for salaried work experience is 2.27 in the model without developmental idealism (Model 1) and 2.25 in the model with all five development idealism scales (Model 7).
The most marked differences in coefficients between Models 1 and 7 in Table 3 are for education. With developmental idealism controlled, the estimated effect of education becomes larger; that is, the relative risk ratios for education become even farther away from one (the no effect comparison) when developmental idealism is added to the model. For example, the relative risk ratio for 12+ years of education on intercaste marriage declines from 0.70 (Model 1) to 0.43 (Model 7) when developmental idealism is added to the model. This difference in the relative risk ratio indicates that there is an indirect effect of education operating through developmental idealism. However, since education has an unexpectedly negative effect on intercaste marriage and developmental idealism has the expected positive effect, this means that the indirect effect of education operating through developmental idealism is positive. This also means that the direct effect of education is even more negative than the total effect.
To obtain further insights into the potential indirect effects of the socioeconomic characteristics through developmental idealism, we estimated equations with the socioeconomic characteristics predicting the developmental idealism scales (Table 7). In these equations, education is the only socioeconomic predictor to have a consistently significant and positive association with developmental idealism (Table 7). Clearly, on the assumption that education affects developmental idealism and not vice versa, education increases endorsement of developmental idealism and, with the positive influence of developmental idealism on the marriage variable, there is a positive indirect effect of education through developmental idealism.
Having unsalaried work experience and distance from the urban center affects one of the developmental idealism variables—spouse choice developmental idealism. However, the combination of these effects and the effects of these socioeconomic characteristics on marriage are small enough that the indirect effects of these socioeconomic characteristics through developmental idealism are very small.
Earlier, we compared the effects of socioeconomic characteristics and developmental idealism by estimating predicted probabilities of intercaste marriage based on different values of the predictor variables. Those comparisons were based on the direct effects reported in Model 7 of Table 3 and are reported in Table 6. We also estimated predicted probabilities of intercaste marriage based on different values of the socioeconomic characteristics using the total effects estimates reported in Model 1 of Table 3. These estimates of the additional predicted probabilities are also reported in Table 6. When we use the total effects model without developmental idealism scales as covariates (Model 1, Table 3) our estimate of the effect of socioeconomic characteristics does increase slightly, but our substantive conclusion remains the same.
As shown in Table 6, when we use the model with just socioeconomic characteristics and the controls (Model 1, Table 3), the predicted probability of an intercaste marriage is 0.165 for someone with little to no education, no nonfamily work experience, and who lives far from the urban center. For a person with the opposite characteristics – 12+ years of education, salaried work experience, and living close to the urban center – the predicted probability of an intercaste marriage rises to 0.313. Thus, the total effect of all socioeconomic characteristics is 0.148 (0.313 – 0.165). When we restrict our manipulations to the only significant socioeconomic characteristic, nonfamily work experience, the difference in the predicted probabilities is reduced to 0.111 (0.285 – 0.174). These estimates of a rise in the probability of intercaste marriage of 0.148 for all socioeconomic characteristics and 0.111 for the significant work experience variable are slightly smaller than the estimates of 0.146 and 0.168 for developmental idealism discussed previously, but the differences are still not statistically significant. Thus, even when we include the indirect effects of socioeconomic characteristics on marriage that operate through developmental idealism, the influence of developmental idealism on marriage is at least as large as that of socioeconomic characteristics.
Although they are not the focus of theories on family variation, the three control variables also present striking influences on marriage behavior. First, there are large differences in marriage behavior by gender. Women are substantially less likely than men to report choosing a same caste spouse on their own (RRR=0.29), as well as intercaste marriage (RRR=0.21) (Table 3, Model 1).
There are also substantial differences in marriage behavior by age at the time of the marriage, with age having a strong, negative influence (Table 3, Model 1). Older people are significantly less likely than younger people to choose a spouse of a same or different caste on their own. Specifically, each additional year of age reduces the relative risk of choosing a same caste spouse by 11% (1/0.90) and an intercaste marriage by 33% (1/0.75). Thus, for example, the predicted probability of an 18 year old having an intercaste marriage is 0.37, compared to 0.11 for a 25 year old.
Finally, caste/ethnicity also has a strong effect on marriage behavior (Table 3, Model 1). Compared to high caste Chhetri-Bahuns, Hill Janajati have double the relative risk of choosing a same caste spouse on their own (RRR=2.45) and having an intercaste marriage (RRR=2.04). Terai Janajati are even more likely to choose a same caste spouse on their own with a relative risk ratio of 4.24. A relative risk ratio of 1.41 suggests that Terai Janajati are also more likely than Chhetri-Bahuns to have an intercaste marriage, but this effect is not statistically significant. Dalits’ marriage behavior does not differ significantly from Chhetri-Bahuns, but the point estimates of the relative risk ratios suggest sizeable effects. Namely, Dalits appear more likely than Chhetri-Bahuns to choose a same caste spouse on their own (RRR=1.46), but less likely to have an intercaste marriage (RRR=0.56).
The results we have just reported show that caste/ethnicity has exceptionally strong effects on marriage behavior. It is also well-known that caste/ethnicity shapes several other dimensions of life in Nepal and elsewhere in South Asia. This led us to investigate the magnitudes of the correlations of caste/ethnicity with the other controls and socioeconomic status variables in our analyses and to investigate the effects of the other variables in models that excluded the caste/ethnicity variable.
We calculated polychoric correlations between each of the variables used in the analysis and four dichotomous caste variables. These polychoric correlations show well known advantages among higher castes and disadvantages among lower castes (results not shown in tables). Education stands out as having the largest correlations with the caste/ethnicity variables: .52 with Chetri-Bahun; -.14 with Hill Janajati; −.33 with Dalit; and −.39 with Terai Janajati. There are also substantial correlations of distance from the urban center with caste/ethnicity: −.25 with Chetri-Bahun and .28 with Terai Janajati (showing settlement patterns by caste, as well as caste advantages). Caste has relatively little correlation with nonfamily work, the only socioeconomic variable to have a significant and large effect on marriage behavior. Importantly, the correlations of caste/ethnicity with the other variables are all well below standard cutoffs of concern about multicollinearity of around .7.
We also estimated additional models predicting marriage that excluded caste from the equations (results not shown in tables). Overall, the models with and without caste have very similar results. Most of the fluctuation of the relative risk ratios are small and the substantive conclusions remain the same. The only exceptions are for education and distance from the urban center, as one would expect from their correlations with caste discussed above. For both education and distance from the urban center, the effects are larger without caste/ethnicity in the equation than with caste/ethnicity in the equation. In addition, removing caste/ethnicity from the equations results in the relative risk ratios for the caste DI scale to become smaller and lose significance. Overall though, this comparison indicates that whether or not the model controls for caste, the effects of the predictor variables do not differ substantially. In addition, because of the exogenous nature of caste/ethnicity in Nepalese society, it is appropriate to include caste/ethnicity in the equations.
The results support our hypothesis that those who endorse developmental idealism are more likely than those who do not, to choose a spouse on their own, including one of a different caste. Further, this influence operates independently of socioeconomic characteristics and is comparable in size, if not larger than, the influence of socioeconomic characteristics. Given the longitudinal nature of the research design and temporal ordering of the developmental idealism and marriage measures, these results provide strong evidence for an important influence of developmental idealism on family behavior.
Our results also show that some aspects of developmental idealism are more influential than others. Three of the five developmental idealism scales – caste, spouse choice, and beliefs about new modern family behaviors – significantly and substantially increase the likelihood of engaging in “modern” marriage behavior. Further, they appear to have stronger effects on intercaste marriage, which combines the two “modern” behaviors of young people choosing their own spouse and choosing one of a different caste. The other two developmental idealism scales – beliefs and values about the modern family and values about new modern behaviors – have no effect on marriage behavior. Additionally, values about caste and spouse choice, as well as beliefs about development causing change in intercaste marriage and spouse choice are particularly influential. This suggests that developmental idealism schemas which are directly relevant to marriage are the most influential. This result is consistent with the theory of reasoned action, which predicts that the beliefs and values which are the most relevant to the behavior have the strongest effect (Fishbein and Ajzen 2010).
Of the three developmental idealism scales that are less relevant to marriage, only one – “new modern family is coming” – has a significant influence. Individuals who believe that new modern behaviors, like premarital sex, cohabitation, and divorce, will increase in the future are more likely to have an intercaste marriage. Believing that new modern family behaviors will increase and are caused by development may be influential because people see their intercaste marriage as compatible with a future that is characterized by many new and previously forbidden forms of family behavior.
Schemas about the modern family and values about new modern behaviors may have no effect because they are too peripheral to marriage. It is also possible that values about new modern family behaviors have no influence overall because these values work in contradictory directions. As hypothesized here, such values should make people more open to “modern” marriage behaviors. At the same time though, some of the small number of people who approve of premarital sex and cohabitation may maintain a non-marital sexual relationship with a person of another caste, rather than marry them, at least in the short term. Premarital sexual relationships are rare and secretive in Nepal, but do exist. This alternative link would make such people less likely to have an intercaste marriage.
The mixed results for the socioeconomic characteristics also provide insights into earlier theories on family variation. As predicted by early theories and ethnographic studies, salaried work experience does increase the likelihood of an intercaste marriage. Further, the influence of nonfamily work experience on marriage is not mediated by developmental idealism. In other words, salaried work is not increasing the likelihood of an intercaste marriage by exposing young people to developmental idealism. Instead, it appears that the influence of nonfamily work lies in providing greater exposure to potentially different caste spouses and greater independence which allows young people to sway parental approval or marry without it. However, the influential roles of education and urban proximity, which also figure prominently in early theories, was not supported.
While our focus is on the influence of developmental idealism and its comparison with socioeconomic characteristics, the results for the control variables also provide important insights. Other studies from Nepal and elsewhere indicate that older ages and higher levels of education are associated with greater endorsement of developmental idealism (Abbasi-Shavazi, Askari-Nodoushan and Thornton 2012; Thornton, Ghimire and Mitchell 2012). Our analyses also support these positive associations between developmental idealism and age, as well as developmental idealism and education. In keeping with the theory of reasoned action, it appears that schooling and age inculcate developmental idealism in people by giving them greater contact with developmental idealism. At the same time, we find that people with older ages and higher levels of education are unexpectedly less likely to have intercaste marriages, although the association with education is not significant.
We suggest that this pattern can be attributed to the spouse selection context and what we call the “hasty effect” of not delaying gratification. In this context, openly dating and maintaining premarital relationships, especially those that might involve premarital sex, is not widely accepted. Thus, premarital relationships between young people are often secretive and short lived (Allendorf 2013). Further, there is an ideal in the region for arranged love marriages that simultaneously embody the schemas of both “modern” marriage and customary South Asian culture (Allendorf 2013; Chowdhry 2007; Mody 2008; Raval, Raval and Raj 2010). In these arranged love marriages, young people choose a spouse based on affection, but then only marry after gaining their parents’ approval and going through the steps of an arranged marriage. Thus, there are tensions between wanting to realize the sometimes contradictory schemas of both developmental idealism and customary Nepali culture, while also trying to maintain often secretive premarital relationships without having sex.
We suggest that marriages among young people and less educated people are more “hasty” – younger, less educated people are less able to manage these tensions and delay gratification (cf. Allendorf 2013). Due to their age, young people lack the maturity to manage these tensions for long periods of time. So they are more likely to end up marrying if they are in a relationship, even when their partner is of another caste. Conversely, older people are better able to wait to gain parents’ approval or until they meet someone who is caste appropriate or otherwise matches both their own and their parents’ preferences. We believe education plays a similar role, although through the process of selection, rather than maturation. We suggest that more educated people are selected to be less “hasty.” Individuals who successfully pursue high levels of education are more concerned with doing well in society and pleasing parents, community members, and themselves simultaneously. Schooling itself may also inculcate skills of self-control and long term planning. Thus, more educated people are less likely to hastily enter into marriage with someone of another caste.
The influence of caste/ethnicity on intercaste marriage reflects historical differences in marriage practices among castes, as well as the relative numbers of different castes. Hill and Terai Janajati are the two groups in which young people are more likely to choose their own spouse and to choose someone of another caste. However, the risk of intercaste marriage is greater among Hill Janajati. Historically, Hill and Terai Janajati maintained marriage practices that involved more choice of young people (Ahearn 2001; Guneratne 2002; Jones and Jones 1976; Macfarlane 1976; Majupuria and Majupuria 1978; Niraula 1994). So, the greater amount of choice in these two groups is consistent with historic practices. The differences between these groups are probably due to the relative availability of different castes. As noted earlier, Terai Janajati largely consist of only one caste (jaat), the Tharu. By contrast, Hill Janajati comprise several castes, including Magar, Tamang, Newar, Gurung, Rai, and others. Thus, a Hill Janajati individual is more likely to encounter potential spouses of different castes than a Terai Janajati individual.
By contrast, the large gender difference does not lend itself to a straightforward explanation. On one hand, women reporting less choice than men is consistent with past research (Caldwell, Reddy and Caldwell 1983; Ghimire et al. 2006; Malhotra 1991). Marital choice is an individual-level characteristic and it’s common for a man to choose his wife, while the wife’s parents negotiate the marriage with the prospective husband. On the other hand, whether a marriage is intercaste or not, should be a couple-level characteristic. Thus, we expected that men and women would be roughly equally likely to have an intercaste marriage. Instead, we found a large gender differential, with women being less likely than men to report an intercaste marriage.
This gender differential may be due to the intercaste measure coming from a self-report. It is possible that women are taking a larger grained view of caste. Castes that are similar in culture and rank are sometimes perceived as being the same caste. For example, a Chhetri may be viewed as the same caste as a Brahmin and a Gurung may consider herself or himself to be the same caste as a Tamang. Thus, women may be more likely than men to report a spouse of a similar caste to be the same caste.
Women may also be underreporting intercaste marriages. Intercaste marriages can still carry a stigma in Nepal, especially when between castes of very different ranks. Further, while marriage is important to the social standing of both men and women, it may have even greater salience for women. Thus, women may be reluctant to report that their marriage is intercaste. Further, unlike men, women are often incorporated into their spouse’s caste after marriage if the intercaste marriages is accepted, just as they are incorporated into their husband’s family (Allendorf 2013; Guneratne 2002). This practice may also prompt women to report intercaste marriages as being the same caste.
More broadly, this analysis goes far beyond earlier studies by providing the first evidence linking developmental idealism to family behavior using systematic individual-level panel data. Previous studies comprise discussions about the connections between ideational and behavioral trends at the macro level (Kavas and Thornton 2013; Thornton and Philipov 2009; Yount and Rashad 2008), ideational interpretations of cross-sectional correlations among individual attributes (Cammack and Heaton 2011; Yount 2005), and observations about the widespread dissemination of developmental idealism beliefs and values (Abbasi-Shavazi, Askari-Nodoushan and Thornton 2012; Melegh et al. 2013; Thornton et al. 2012a; Thornton et al. 2012b). Thus, this article provides an important contribution by providing the first direct, systematic, and detailed empirical support for the theorized link between developmental idealism and “modern” family behavior at the individual level.
By extension, our analysis also supports the role of developmental idealism in theories of family change. The influence of developmental idealism on marriage behavior implies that the spread of developmental idealism in a population will lead to family change. Specifically, as endorsement of developmental idealism increases over time, young people choosing their own spouses, including different caste spouses, should increase in response. These findings are consistent with Thornton’s (2001, 2005) and others’ suggestions (Cammack and Heaton 2011; Kavas and Thornton 2013; Yount 2005; Yount and Rashad 2008) that the spread of developmental idealism creates changes in family behaviors. Further, our finding that the influence of developmental idealism is operating largely independently of socioeconomic influences is consistent with the suggestion that the spread of new ideas itself is changing family behavior (Bongaarts and Watkins 1996; Lesthaeghe 2010; Meyer et al. 1997; Pierotti 2013; Thornton et al. 2012b; Thornton and Philipov 2009). Thus, these results add developmental idealism to the list of values and beliefs that influence family behavior (cf. Buttenheim and Nobles 2009; Goldscheider, Kaufman and Sassler 2009; Hayford and Morgan 2008). More broadly, we also add marriage to the list of social outcomes that are influenced by ideational factors (Hitlin and Piliavin 2004; Manza and Brooks 2012; Vaisey 2009).
The analysis also provides support for the utility of the theory of reasoned action as an explanation of how ideational factors influence family behavior. Our results suggest that individuals’ behavior is at least somewhat consistent with their values and beliefs. Further, values and beliefs that are more relevant to the behavior are the most influential. Thus, like Ajzen and Klobas (2013), we conclude that the theory of reasoned action is a useful and appropriate explanation of how values and beliefs influence family behaviors more generally.
Finally, these results also have important implications for stratification. Our analysis suggests that developmental idealism is a powerful influence on family behavior, but since one aspect of the behavior is intercaste marriage, it further implies that developmental idealism may have an important impact on inequality. If developmental idealism becomes widely accepted, intercaste marriages may become widespread and, in turn, caste distinctions would be challenged. Group identities of castes can be maintained when intercaste marriages are rare, as they have been in the past. However, if marrying without regard to caste becomes common, or even universal, it would result in profound changes in the caste system and stratification more broadly in South Asia.
Future research should replicate and extend our findings to other family behaviors. Marriage is just one family behavior among many. Thus, additional studies should examine the influence of developmental idealism on family structure, fertility, and other behaviors, as well as the timing of when people marry and if they marry at all. Such research will not only further estimate the influence of developmental idealism, but reflect on which aspects of developmental idealism are most influential. Future research should also directly examine change over time and evaluate whether increases in the endorsement of developmental idealism over time in a population do indeed result in changes in family behaviors, as well as related social institutions.
We thank Brian Powell, Clem Brooks, Dina Okamoto, Dirgha Ghimire, Colter Mitchell, and three anonymous reviewers for their helpful comments on earlier versions of this article. We also thank J. Scott Long for his help with the analysis using predicted probabilities. This research was supported by a research grant from the National Institute of Child Health and Human Development (R01 HD054847, Ideational Influences on Marriage and Childbearing) and by an NICHD center grant to the Population Studies Center at the University of Michigan (R24 HD041028).
1Fishbein and Ajzen (2010) use the term attitude to refer to this evaluative component for specific objects and behaviors. We use the more general term of value to refer to favorableness, or unfavorableness, towards specific objects and behaviors, as well as broad ideals.
2The factor analyses used tetrachoric correlation matrices because the developmental idealism items were all recoded to be dichotomous. A one indicates that the respondent’s response was consistent with developmental idealism and a zero indicates it was not. A dichotomous recoding of endorsing developmental idealism was appropriate because the vast majority of respondents gave responses that were consistent with developmental idealism or in direct contradiction to developmental idealism. Only a very small number of respondents gave in-between responses, such as wouldn’t matter or neither good nor bad.
3It should be noted that our analytical sample is 66% female. The original baseline sample of unmarried people aged 12–34 in 2008 was 48.7% female. Thus, the skewed gender distribution in the analytical sample reflects gender differences in marriage during 2008–12. During the study period, 33.7% of the women in the baseline sample married, compared to 16.8% of men.
4There is one exception to this statement. The models predicting endorsement of developmental idealism in Table 7 include age at the baseline interview, instead of age at marriage. Age at marriage is an appropriate predictor of marriage behavior, but not of endorsement of developmental idealism at the baseline interview because the baseline interview occurred before they married. Age at marriage and age at the baseline interview have a correlation of 0.94. So, this change in the age measure does not impact results.
5There are two minor exceptions to this statement. Distance from urban center has a significantly positive effect on spouse choice developmental idealism with a coefficient of 0.09 (Table 7). Unsalaried work experience also has a small significantly positive effect on spouse choice developmental idealism with a coefficient of 0.23 (Table 7).
6When we use the total effects model (Model 1, Table 3) to calculate the predicted probabilities the estimate of the total effect of the socioeconomic characteristics decreases from 0.148 when we manipulate all three variables to 0.111 when we manipulate only the significant work experience variable (Table 6). By contrast, as discussed earlier, when we use the full model (Model 7, Table 3) to calculate the predicted probabilities the estimate of the total effect of the socioeconomic characteristics increases from 0.077 when we manipulate all three variables to 0.106 when we manipulate only the significant work experience variable (Table 6). This seeming inconsistency is due to the effect of education. Education is the only socioeconomic variable with an important correlation with developmental idealism. This correlation results in a sizeable difference in the effect of education on the predicted probabilities from Model 7 versus Model 1. In Model 7, when developmental idealism is controlled, education has a large negative effect on intercaste marriage; it individually reduces the probability of an intercaste marriage by 0.110 (Table 6). However, in Model 1, without developmental idealism, the large negative effect of education is reduced because it takes on part of the positive effect of developmental idealism; education now individually decreases the probability of an intercaste marriage by only 0.038. Thus, the predicted probability for the socioeconomic variables as a group is 0.148 when using Model 1, but only 0.077 when using Model 7.
Keera Allendorf, Assistant Professor of Sociology and International Studies, Indiana University, 1020 E. Kirkwood Ave., Bloomington, IN 47405, Phone: (812), 855-1540, Fax: (812) 855-0781.
Arland Thornton, Professor of Sociology, University of Michigan.