Mass Education, International Travel, and Ideal Ages at Marriage

Abstract

Opportunities to document associations between macro-level changes in social organization and the spread of new individual attitudes are relatively rare. Moreover, of the factors generally understood to be influential, little is known about the potential mechanisms that make them so powerful. Here we use longitudinal measures from the Chitwan Valley Family Study (CVFS) to describe the processes of ideational change across 12 years among a representative sample from a rural agrarian setting in South Asia. Findings from lagged dependent variable models show that (1) two key dimensions of social organization—education and international travel—are strongly associated with change in attitudes, net of prior attitudes; (2) reorganization of education and travel are associated with attitudes toward ideal age at marriage; and (3) this association varies by gender. Using the study’s prospective design, we document not only these important associations but also potential mechanisms of education and travel—exposure to the English language and friends’ international travel experience—as potentially powerful social influences on individuals’ attitudes, independent of their own experiences.

Introduction

Demographers have identified changes in social organization and social psychology as key factors influencing demographic outcomes. Both classical and contemporary theories emphasize the role that macro-level contextual changes play in shaping widespread ideational change (Caldwell 1976; Durkheim 1893/1984; Entwisle et al. 2007; Lesthaeghe 2007). Yet opportunities to document associations between macro-level changes in social organization and the spread of new individual-level attitudes are relatively rare in both historical and contemporary studies (Coale and Watkins 1986; Thornton and Axinn 2004). Here we use recent longitudinal measures from rural Nepal to document the processes of change in ideas across 12 years among a systematic representative sample of more than 8,000 individuals. This setting allows us to document the associations with macro-level changes in social organization to provide a unique new window into the potential connection between changes in social organization and changing ideas about demographic processes.

We provide empirical description related to three fundamental questions regarding this relationship. First, which specific dimensions of social organization are associated with population-wide change in attitudes? Following prior demographic research, we investigate mass education in schools and international travel, both of which have been shown to be associated with variations in attitudes (Baker et al. 2011; Bongaarts and Watkins 1996; Caldwell 1980; Lutz and Kebede 2018). We advance the understanding of both education and travel by documenting specific dimensions of each that show strong associations with subsequent attitude change. Second, can we document the specific ideas that change and the direction of the change? Again, building on prior findings from across the field, we reason that ideals about marriage are the most likely to change and that the direction of change is predictable (Caldwell 1980; Lesthaeghe 2007; Thornton and Young-DeMarco 2001). Third, does documentation reveal that the association between social organization and ideas differs across the most pronounced social categories? Population research has repeatedly documented gender as a crucial conditioning category for social processes, making it an ideal category for this test.

Three important new tools make answers to these questions possible. The first is long-term panel measurement of individuals’ attitudes in a representative sample in a setting of watershed change in social organization. We use panel data from rural Nepal, where the opportunities for education and international travel were created within the lifetimes of those being studied. The second is detailed documentation of change over time in social organization and individual-level variation in exposure to new forms of it. The Chitwan Valley Family Study (CVFS) is novel in that it provides continuous histories of community-level access to schools and transportation and complete life histories of family member and individual experience with education and travel; CVFS is unique in this rich array of measures of individual and contextual experiences with education and travel across an individual’s lifetime. The third is well-defined shifts in specific dimensions of social organization occurring largely outside the control of those being studied. International aid to Nepal shaped the location, timing, and content of local schools (shift 1), and international flights out of Kathmandu created a dramatic change in the opportunity for travelling abroad (shift 2). The combination of these tools affords an important opportunity to document in new depth the association between social organization and demographic attitude change. Note that although this opportunity to document these associations is great, it cannot demonstrate cause. Even with the hundreds of measures available in the CVFS, these data do not remove the possibility of endogeneity: the observed associations may be produced by factors unaccounted for in our models. Rather, the CVFS provides a crucial descriptive window into associations among long-term changes that would be difficult to randomize.

Theoretical Framework

We identify potential mechanisms through which key dimensions of social organization—education and travel—are likely to produce attitude change. We then consider the types of attitudes most likely to change—marriage timing preferences—and the direction in which they will change in response to those mechanisms. Last, we identify potential mechanisms that may differ by gender, producing gendered differences in changing attitudes. Our objective is to guide the detailed description of the processes through which new dimensions of social organization may change population-scale ideas about demographic processes.

Education and Population-Wide Attitude Change

The spread of mass education is understood to have strong consequences for social life (Baker et al. 2011; Becker 1991; Caldwell 1980; Lesthaeghe 2007; Lutz and Skirbekk 2017; Montez and Hayward 2014; Thornton and Lin 1994). However, previous studies have struggled to document specific dimensions of education likely to produce this association. Focusing on an individual’s immediate educational context, we identify three specific dimensions through which education is likely to influence attitudes: aspirations, social networks, and school materials.

Education and Aspirations

Classic studies of status attainment found that perceptions of others’ experiences and expectations influence teens’ future educational and occupational outcomes (Sewell et al. 1969; Sewell and Hauser 1975). A key driving force here is the psychology of attainment: exposure to education strongly shapes individuals’ attitudes and expectations by altering their understanding of future labor market success and by instilling in them the idea that educational attainment expands occupational options. Additional years of available education strengthen these ideas: the more years spent in school, the more educated the individual and thus the greater their occupational options. As a result, as the highest grade offered in close proximity to the individual shifts upward, changes in attitudes about options are likely to increase. In Nepal, the proliferation of schools has been recent and rapid (see Fig. 1), so that the proximity of schools offering education to high levels has changed substantially within the lifetimes of current residents (Axinn and Barber 2001).

Education and Networks

Schools expose individuals to a specific, nonfamily social environment during adolescence and young adulthood, increasing social interactions outside the family. This change has been shown to change attitudes toward those emphasizing personal achievement and independence (Alwin et al. 1991; Goldscheider and Waite 1993). Detailed studies of these interactions in the United States indicate that interactions with school faculty change attitudes in ways that have long-lasting consequences well into adulthood (Alwin and Krosnick 1991; Newcomb 1943). In Nepal, most teachers attain just a secondary education, although this is changing because of government expansion of bachelor’s degree–level teacher-training programs and increased teaching standards in the 1990s (UNESCO 2011). Similar to the United States, these college-educated teachers expose students to new potential education and job trajectories as well as spread the ideas to which they were exposed while in college. Increased exposure to teachers with a bachelor’s degree (BA) will likely produce more change in attitudes.

Education and School Materials

Last, education is well known to expose individuals to new ideas through educational materials and instruction (Caldwell et al. 1988; Thornton 2005). This is true across contexts, including low-income regions. For the latter, these materials—which often originate abroad due in part to colonial histories—tend to disseminate ideas and values more prevalent in rich settings, such as those promoting individual growth and economic development (Brock-Utne 2000; Caldwell et al. 1988; Thornton 2005). Nepal is a clear example: its education system is highly influenced by its southern neighbor, India, a former British colony, where all dimensions of education—teaching and learning, textbooks, and scientific research—were established in English (Altback 1989).¹ Today, foreign ideas are still diffused in Nepalese schools through engagement with the English language, with English-only schools often using materials imported from outside Nepal. However, the level of exposure to English varies greatly by school type: government schools expose students to English by grade 5, but privately financed schools—first established in south-central Nepal in 1985—use English as the main medium of instruction beginning in grade 1. Thus, greater exposure to the English language is likely to lead to greater change in individuals’ attitudes.

International Travel and Population-Wide Attitude Change

Like education, international travel exposes individuals to new ideas. In Nepal, as in much of the world, social life historically has occurred close to home and was organized within the family, emphasizing the exchange of family-centric attitudes among kin (Ogburn and Tibbitts 1933; Thornton and Fricke 1987). The opportunity to travel has changed most drastically in recent years because of eased travel policies and a steep increase in international flights from Nepal’s capital, Kathmandu (see Fig. 2). Changing labor demands abroad, particularly in the Gulf region, is a key factor in these changes (Kansakar 2003).

A person’s own travel experiences directly introduces her/him to new networks, thereby accelerating the exposure to new or innovative ideas (Bongaarts and Watkins 1996). These changes are likely especially strong as networks expand across international borders. For rural Nepalese, the sudden increase in international flights and associated labor recruitment abroad has led to increased travel to distant countries, where ideas are expected to differ vastly from those common at home, creating new aspirations for work, consumption, and the family.

Of course, it may not just be individuals’ own exposure to new ideas in foreign countries that matters: indirect exposure to new ideas, through the travel experience of important others, likely affects an individual’s attitudes as well. Family and friends are both relevant. Family members are key social contacts given that the majority of social interaction occurs within the home in early life (Freedman 1979; Thornton and Fricke 1987), whereas friends tend to serve as central sources of ideas as individuals age into young adulthood and spend more time outside the home (Axinn and Yabiku 2001; Montgomery and Casterline 1996). In rural settings like Nepal, the social influence of international travel is likely so powerful that regardless of an individual’s own experience, having a close network member visit a foreign country leads to the transmission of new ideas to individuals at home as well.

What Specific Ideas Change and What Is the Direction of the Change?

Neither historical nor contemporary U.S. and European research has access to documentation of both dramatic change in social organization and long-term individual change in attitudes. Nevertheless, three streams of prior research offer clues about the ideas most likely to change, as well as the likely direction, in response to massive social reorganization: (1) the nature of attitudes themselves; (2) role incompatibility; and (3) rising consumption aspirations. We start with the well-documented link between socialization and a fundamental social unit: the family. And because marriage remains central to family life, we focus on individuals’ attitudes toward ideal ages for men and for women to marry.

First, specific attitudes are expressions of favor or disfavor toward an object that form our evaluations and underlying beliefs about the object. They form throughout the life course as individuals react to and internalize others’ attitudes as a way to understand themselves. Scholars have posited that individuals’ sense of self is a continual process that develops through an understanding of how others experience a situation (Lesthaeghe 1983; Lesthaeghe and Surkyn 1988; Mead 1934). Specific attitudes (e.g., ideal marriage timing) are malleable, influenced by exposure to new ideas through specific experiences and expanding networks.

Second, specific attitudes, such as those toward marriage, are linked to larger ideas about social roles. Young adults are often faced with many paths. Given the difficulty in occupying multiple roles simultaneously, young adults often delay assuming one social role while focusing on another (Thornton et al. 2007)—for example, delaying marriage because of school enrollment (Yabiku 2005). Thus, perceived role incompatibility links educational attainment and preferences for educational attainment to ideas about the ideal timing for marriage.

Third, international travel creates exposure to many new ways of life. However, when the starting point is rural Nepal, foreign travel has an especially powerful influence on consumption aspirations. Nepalese often leave the country to work, and that work is almost always in much richer settings, often the Middle East. Travel to these settings exposes Nepalese to much higher levels of consumption than they have seen before. Demographers have long argued that such dramatic shifts in consumption aspirations are a key motive for changes toward later marriage and fewer children as individuals seek more time to accumulate wealth before family formation (Easterlin and Crimmins 1985; Freedman 1979). Thus, like perceived role incompatibility, rapidly rising consumption aspirations are also likely to link international travel experiences to ideas about the ideal age for marriage.

Does the Relationship Between Education or Travel and Attitudes Differ by Gender?

As noted at the beginning of this article, gender differences influence daily life in nearly every social circumstance (Bem 1993; Risman 2004), and these differences are maintained in two ways. First, perceived differences between the sexes emphasize separate social roles (Acker 1990; Scott 1986), exposing men and women to varying opportunities deemed necessary to fulfill these roles (Charles and Badley 2009). Often as education expands, men increase their human capital, and thus employment opportunities, by investing in schooling (Becker 1993; Coleman 1990)—an investment some have historically seen as less important for women’s roles (Thornton et al. 1983). This is particularly true in rural agricultural societies, such as Nepal, where patriarchal norms historically limit women’s opportunities outside the home. For example, women still marry much younger than men (Allendorf et al. 2017; Yabiku 2005), and school enrollment and educational attainment remain higher for boys than girls (Stash and Hannum 2001). Because the role changes from increased enrollment are likely to be greater for women, the association between education and attitude change will be stronger for women.

Second, social roles reinforce gender differences by altering individuals’ attitudes and expectations of others (West and Zimmerman 1987). In explaining marriage, scholars have argued that gender differentiation of roles within households maximizes gains (Becker 1991). Preferences may mirror these roles. For example, women may seek earning potential and economic success in a partner, while men may seek physical attractiveness (Oppenheimer 2003). So, although school enrollment has increased for both sexes in Nepal, its associations may be gendered: education may raise men’s and women’s expectations that men, more so than women, have a career rather than just a low-skilled labor job (Coleman 1990; Thornton et al. 2007). Thus, increasing educational attainments will increase the gender gap in ideal marriage timing (i.e., the difference in ideal ages for men and women to marry).

Data and Analytic Approach

Sample

The data for this study come from the Chitwan Valley Family Study (CVFS). Initiated in the mid-1990s, the CVFS provides an important opportunity to study associations between changes in educational and travel opportunities on the one hand and family-related attitudes and behaviors on the other. Baseline individual and household interviews began in 1996 with face-to-face interviews conducted with all household members aged 15 to 59 and their spouses, regardless of age or place of residence, in all 151 sampled neighborhoods. The response rate to the baseline survey was 97%. Spouses were interviewed simultaneously to allow for privacy. Individual life history calendars captured annual retrospective measures of schooling, travel, work, and marital status. After the 1996 baseline interview, regular household interviews tracked every 1996 respondent measuring prospective monthly updates of family transitions and travel events. The 1996 respondents were then reinterviewed in 2008. Using the constant contact from 1996 to 2008, the CVFS attempted to interview all original respondents, even those who had moved outside Nepal by 2008 (9.9% of the 1996 respondents were living outside Nepal for the 2008 interview). Very few individuals (<5%) dropped out of the study or died during the study period. Our study includes only those 1,550 respondents aged 15–25 in 1996 who participated in both interviews. We target those aged 25 years or younger in 1996 to focus on the portion of the life course during which novel experiences and expanding social networks are likely to change individuals’ ideas with lasting consequences (Alwin et al. 1991).²

The CVFS also collected school-level data for all schools ever open in the study area. School history calendars captured annual information on multiple characteristics of each school for every year in which the school was open, including grades offered, number and educational attainment of faculty, medium of instruction, and the school’s geocoordinates. Neighborhood history calendars similarly collected geocoordinates for all 151 selected neighborhoods. Again, the CVFS provides an opportunity to study the association between education and travel and attitude change in a new way: never before have we had access to rich measures of both personal and contextual experiences with education and travel across an individual’s lifetime.

Measures

Attitudes

Individuals’ attitudes were measured in individual interviews using exactly the same questions with exactly the same words in exactly the same order in both 1996 and 2008 to the same individuals. We use two of these repeated measures to assess attitudes regarding ideal marriage timing: each asked respondents what they believed is the ideal age for a woman (first question) and the ideal age for a man (second question) to get married these days. Responses were in years of age, and any integer was treated as valid.

Between 1996 and 2008, average attitudes toward ideal marriage timing for both men and women increased, showing preferences for later marriage in general. The differences in both attitudes’ means between these two time points are statistically significant. Respondents in 2008, however, still preferred women to marry at younger ages than men: 21.7 years for women compared with 26.2 years for men. Between 1996 and 2008, the ideal age at marriage for men increased by a mean of 2.1 years, and the ideal age at marriage for women increased on average by a smaller margin of 1.5 years. These mean changes in ideal ages are significantly different (p < .001), with marriage timing preferences increasing more for men during this period than for women. The gap between ideal ages for men and women increased from 4.0 years in 1996 to 4.6 years in 2008. Further analyses (not presented) indicate that these shifts in both attitudes were consistent throughout the distribution, rather than, say, one tail getting more extreme over time. Table 1 presents descriptive statistics for variables used in the analyses.

Education

We use educational attainment measures that assess the highest grade of school or year of college the respondent has completed. They were collected in a context of a complete life history in 1996 as well as an updated life history in 2008 that captured the years of education attained between 1996 and 2008. Importantly, the 2008 interview was conducted in a context of constant re-contact between 1996 and 2008 through the household registry, which improved the accuracy of responses. Analyses include a continuous measure of total educational attainment by 1996 and a second continuous measure of any additional years of education attained between 1996 and 2008. We also include a dichotomous measure indicating whether a respondent is enrolled in school in 1996 (= 1) or not (= 0). Average educational attainment was 6.4 years in 1996, with a small increase between then and 2008 (0.9 years). In 1996, 39% were enrolled in school.

Key Dimensions of the Education Context

We use school history calendars to calculate specific dimensions of education in respondents’ immediate proximity. We include nearby school characteristics in 1995 because it indicates an individual’s regular exposure to education occurring in the time immediately prior to attitudes assessed in 1996. Analyses include two continuous measures: (1) highest grade offered at the closest school in 1995, and (2) percentage of teachers with a BA, also at the closest school in 1995. This is calculated by dividing the number of teachers at the nearest school with a BA by the total number of teachers employed at that school in that year. On average, nearby schools in 1995 offered 6.8 grades, and 18% of teachers had a BA. We also include a continuous measure indicating the percentage of schools within a 1 km radius in which English was used as the only medium of instruction. Distances were calculated “as the crow flies” using the geocoordinates of each school and the center of each neighborhood. Because much of Nepalese social life occurs close to home, the short distance captures those institutions to which a respondent is most likely to be exposed on a regular basis. Of schools within a 1 km radius of respondents, an average of 21% used only English in 1995.³

International Travel

Respondents’ international travel experience between 1996 and 2008 was measured using the 2008 life history calendar. Respondents reported timing and destinations of travel outside Nepal for all spells of one week or more between 1996 and 2008. Again, in the CVFS context of ongoing contact with each household many times a year from 1996 to 2008, even when an individual travelled away, these reports of international travel are exceptionally complete and accurate. We do not include India in our measure of international travel for two reasons. First, because the countries share an open border, cultural similarities, and a history of seasonal migration, travel to India is less likely to expose Nepalese to vastly new experiences or ideas. Second, Nepal’s only other shared border is with China—specifically Tibet, a tightly monitored stretch of mountainous land. Aside from India, the most common foreign destination for Nepalese is the Middle East, a region in which foreigners are overwhelmingly labor migrants signing multiyear labor contracts. So, in the vast majority of instances, international travel captures experiences necessitating airplane travel to a faraway, higher-income country and for an extended period—all factors that increase an individual’s exposure to new ideas and experiences. The measure is dichotomous, with 1 indicating international travel (outside India and Nepal) during this period, and 0 indicating none.⁴ Between 1996 and 2008, 6% of respondents travelled outside Nepal and India (see Table 1).

Key Dimensions of the Travel Context

Family members’ and friends’ travel experience were measured using two separate but similar measures from the individual interview in 2008. These items assessed whether any members of the respondent’s family or any friends had ever travelled outside Nepal and India, with 1 indicating any experience, and 0 indicating none. Again, we conceptualize family and friends’ international experience as travel to destinations other than India given the countries’ geographic proximity and cultural similarities. By 2008, 75% of respondents had family, and 59% had friends who had travelled outside Nepal and India.⁵

Other Key Factors

Research has identified additional factors that might be associated with changes in attitudes about marriage timing. One key predictor is marital status. A vast literature has explored the interrelationship between attitude change and behaviors related to the attitudes of interest (Ajzen 1988). Evidence suggests that individuals report more positive attitudes toward behaviors they have done. For example, persons who have experienced a divorce report more positive attitudes toward divorce after the event than before it (Thornton 1985); similar associations have been found with marriage (Axinn and Thornton 1992). The CVFS assesses marital status in both 1996 and 2008, so we construct two dichotomous measures. The first is coded 1 if the respondent is married by 1996, and the second is coded 1 if the respondent is married by 2008. The reference group for comparisons to these two measures is those individuals who were not married by 2008 (so, never married). Because divorce remains a rare event, and death of a spouse between 1996 and 2008 was also rare, only 13 cases reported being married in 1996 but not in 2008. Those 13 cases were excluded from analyses. Two-fifths (41%) of respondents were married by 1996, and an additional 50% were married by 2008. Nine percent had not married by 2008 (see Table 1).

We also include two measures for socioeconomic status (SES), a factor strongly associated with travel and education. First, we construct an index of household wealth using measurement of ownership. The CVFS assessed household ownership of a variety of goods in a repeated survey in 1996 and 2006. Models include a continuous measure indicating the number of household goods owned in 1996: radio, television, bike, motorcycle, cart, tractor, gobar gas, toilet, and electricity. Prior research has demonstrated this measure captures key variability in the SES of households in Nepal (Link et al. 2012). Second, we include a dichotomous variable indicating an important occupational/earnings divide in the monetized economy of Nepal: working for pay. Models include a measure indicating that the respondent worked for either daily wages or a monthly salary in 1996. Respondents owned an average of 2.7 household goods, ranging from 0 to 8, and 42% worked for pay (daily wages or monthly salary).

Gender, ethnicity, and birth cohort are each fixed at the time of birth and cannot change. Our analyses include a dichotomous measure for gender, with 1 indicating female and 0 as male. Roughly one-half (56%) of the sample was female. We also include a set of dichotomous measures corresponding to five broad ethnicity/caste categories reflecting meaningful distinctions in Nepalese society: Brahmin/Chhetri, Dalit, Newar, Terai Janajati, and Hill Janajati, with Brahmin/Chhetri serving as the reference category. The majority of respondents were Brahmin/Chhetri (51%), followed by Terai Janajati (21%), Hill Janajati (13%), Dalit (9%), and Newar (6%). Last, we measure birth cohort with respondent age in 1996. Average respondent age in 1996 was 19.4 years.

Aside from marital experience and factors established at birth, it is possible that other factors may produce changes in attitudes across these 12 years, such as a broad array of individual, family, and community experiences. Fortunately, the CVFS is a rich source of measures of these experiences. We use two strategies to explore the possibility that other factors create spurious associations between either education or international travel and changes in ideal ages at marriage. Of course, these strategies do not substitute for randomization of education or travel—endogeneity of observed associations is still possible—but they do provide additional evidence that our empirical findings are consistent with general theoretical predictions. The first is to estimate individual-level fixed-effects models, eliminating time-invariant individual-level differences and focusing purely on change over time. This strategy has the benefit of accounting for unobserved and observed differences, but the limitation is that it focuses the analysis on only associations between attitude change from 1996 to 2008 and other factors that also change across that period—specifically, educational attainment and travel. It is a conservative approach to the elimination of the possibility that prior differences produce observed associations that are spurious, so we present results from this approach in the online appendix.⁶

The second strategy is to estimate models of these associations that control for large numbers of other factors measured in the CVFS. Because the CVFS measures so many possible factors, it is possible to construct many models following this strategy. For example, previous literature has indicated that media exposure (Barber and Axinn 2004; Kincaid 2000) and membership in organized social welfare groups (McPherson et al. 2001; Rindfuss et al. 2004) shape information flow, with consequences for attitudes. The CVFS includes measures of exposure to radio, television, movies, and newspapers as well as measures of membership in youth groups, women’s groups, rotating credit groups, political groups, sports groups, and other social groups. We estimated models adding controls for each of these measures as well as many other factors measured in the CVFS, including parental and spouse SES and various conceptualizations of urban context; the results were not at all parsimonious. However, in each case, even when slight changes in estimates occurred, the estimated associations of education and international travel and attitude change remained in the same direction as those presented in our main tables and statistically significant. In all models, the estimated associations were bounded by those presented in our main tables and those presented in the online appendix.⁷

Analytic Approach

In modeling the associations between education/international travel and attitude change, we use a life course approach. The life course perspective acknowledges the importance of the timing, duration, and sequencing of exposures over the lifetime (Elder et al. 2003). Exposure to specific dimensions of social organization and actors at specific life phases—early childhood or adolescence, for example—is expected to relate to attitudes and behaviors not only at that point but in the future as well (Elder 1977). These years constitute a period of life in which individuals are highly susceptible to attitude change (Krosnick and Alwin 1989); experiences occurring earlier in life are thus expected to be particularly powerful in shaping attitudes. Thus, for our sample, educational attainment and exposure to specific school characteristics—highest grades offered, teachers with a BA, and density of English-medium schools in close proximity—occurring prior to 1996, when respondents were (on average) 19 years old, will be especially important for attitudes at that point and later in life. Personal experiences and exposure to specific networks occurring later, between 1996 and 2008, are still expected to shape attitudes, although to a lesser extent. The first set of models includes measures from 1996 and earlier, and models on which the second set builds include measures occurring between 1996 and 2008.

Results

Specific Dimensions of Social Organization and Direction of Attitude Change

Mechanisms of Attitude Change, 1996 and Earlier

Results for the association of education (in 1996 and earlier) with respondents’ change in attitudes between 1996 and 2008 regarding ideal ages for women to marry are presented in the first column of Table 2 (Model 1). This model shows that respondents who went to school during this period came to prefer a later ideal age at marriage than those who did not attend school, regardless of whether they were enrolled in school in 1996 (p < .05). In fact, for each additional year of education attained prior to 1996, an individual supports an ideal marriage timing that is 0.09 years older for women (p < .001). The strong predictive power of educational attainment prior to 1996 on the change in attitudes between 1996 and 2008 suggests that this early school experience continued to be associated with attitude change into the future.⁸

Given previous work, this strong association between education and attitude change is not surprising. We next investigate potential dimensions of the immediate educational context to better understand what it is about education that makes it so powerful. In additional models, we find that characteristics of nearby schools are indeed important: additional grades offered, a higher percentage of teachers with a BA, and a higher percentage of schools that use only English as the medium of instruction (p < .05) are each associated with an individual supporting an ideal marriage timing in 2008 that is older for women, with the same attitude toward marriage in 1996 held constant. Two patterns are noteworthy. First, in each model, the estimated association between individuals’ own attainment and attitude change does not change, regardless of which dimension of the immediate educational context we include in the model. Second, when all three dimensions are included in the same model, all still are positively associated with respondents’ ideal ages for women to marry, although only the percentage of schools using only English remains significant (p < .05). In Table 2, we present results from the model that best illustrates these findings: net of individuals’ own attainment, their exposure to the English language and the foreign ideas diffused through the related materials are particularly powerful in shaping attitudes toward marriage timing—not just at that point in life but also in later years.

The second column of Table 2 (Model 2) presents results from a similar model presented in Model 1 but for change in attitudes between 1996 and 2008 regarding ideal ages for men to marry. Results are nearly identical. As expected, educational attainment has a strong positive association with attitudes toward men marrying (p < .01), regardless of enrollment status in 1996 (p < .05). The specific dimensions of the educational context in close proximity do as well. Findings again show that higher grades offered, a higher percentage of teachers nearby with a BA (p < .05), and a higher percentage of nearby schools using only English (p < .05) are all positively and significantly associated with respondents’ marriage timing preferences for men. We also find that when all dimensions of the educational context are included in one model, only the percentage of nearby schools using only English remains significant (Model 2). This full model suggests that these associations are not independent, once more highlighting that exposure to the English language is particularly meaningful.⁹

Models shown in Table 2 and discussed earlier also show that other key factors are important correlates of attitude change. Strong differences in attitudes in 2008 exist across ethnicity/caste distinctions as well as by age. As expected, marital experience prior to 1996 is negatively associated with attitudes toward marriage timing (those who have already married prefer younger ages of marriage, on average), although this association is not significant. Household and individual SES are associated with ideational change, but other estimated associations are independent of these dimensions of SES.

Table 2 shows that both individual educational attainment and a local context of schools teaching in English are associated with attitudes about marriage timing. Because private schools did not become available in Chitwan until 1985 and after, exposure to the English language and related texts—most of which originate abroad—were extremely limited before this time. We use 1985 as a marker of English language availability (and thus opportunity for exposure) and test its interaction with educational attainment. As shown in Fig. A1 of the online appendix, the association between educational attainment and attitudes toward ideal ages at marriage was significantly greater after 1985 (when English language schools became available) than it was before 1985. In both periods, higher attainment was associated with preference for later marriage, but after English language schools were available, this positive association became significantly stronger.¹⁰

Mechanisms of Attitude Change, 1996–2008

Experiences occurring earlier in the life course are clearly important. We next turn to experiences that occurred later—between 1996 and 2008—and test their association with the change in attitudes occurring during this period, controlling for those experiences that occurred prior to 1996. Table 3 presents results for the associations between education and international travel with changes in attitudes regarding ideal ages for women to marry (1996–2008) in the first column (Model 1), and results of these same associations with changes in attitudes regarding ideal ages for men to marry in the second column (Model 2). For clarity, the top rows show results for education and travel occurring between 1996 and 2008, whereas the remaining rows show results for key predictors discussed in Table 2 that occurred in 1996 and earlier as well as other key factors.

Regarding attitudes toward marriage timing for women (Model 1), we again find that education matters: for each additional year attained during this period, respondents prefer marriage timing that is 0.10 years older for women (p < .01). Similar to education attained earlier in the life course, this is not surprising given prior theoretical and empirical predictions. We next turn to a second dimension of social organization that we expect will be associated with attitudes: international travel. We find that respondents’ experience travelling abroad between 1996 and 2008 is also associated with later ideal ages for marriage. Model 1 shows that individuals who travelled outside Nepal and India prefer later ages at marriage for women that are 0.48 years older than individuals who did not travel to these destinations during this time (p < .05). Experience traveling internationally, particularly to destinations outside of India—a culturally similar country that shares an open border with Nepal—likely exposes Nepalese to new social networks and ideas, creating new aspirations for work, consumption, and family.^11,12

We next examine specific characteristics of an individual’s immediate travel context and estimate the associations between family and friends’ international travel experience with individuals’ attitude change, net of their own travel experience. Results from additional models show that individuals who have family members who travelled outside Nepal and India between 1996 and 2008 prefer significantly later marriage timing for women (p < .05). We find similar results for individuals who have friends with travel experience: these respondents prefer women to marry 0.44 years later (p < .001). Including both family and friends in the same model, we find that having friends with international travel experience is particularly consequential in shaping attitudes, net of individuals’ own experiences abroad. This result supports our expectation that it is not just an individual’s own direct exposure to foreign travel that matters. Rather, exposure to these new ideas and experiences abroad, which differ greatly from those common in rural Nepal, is so powerful that friends’ experiences in foreign destinations change individuals’ attitudes back home as well.¹³

Model 2 in Table 3 presents results for the same potential mechanisms of attitude change as those in Model 1—those occurring between 1996 and 2008—but with a slightly different outcome: change in attitudes between 1996 and 2008 regarding ideal ages for men to marry. We again focus on key mechanisms of attitude change occurring between 1996 and 2008, although models also include 1996 education measures and other key factors. Similar to results in Model 1, international travel is positively associated with marriage timing preferences, although here the association is not significantly different from 0. Having either family (p < .05) or friends (p < .001) with travel experience outside Nepal and India are associated with later ideal ages for men to marry. Again, the full model shows that it is friends with international travel experience that is particularly consequential for attitude change. Specifically, respondents with friends who have travelled abroad between 1996 and 2008 support men marrying 0.65 years later than respondents who do not have friends with such experience (p < .001). Net of individuals’ own travel experience, the travel experience of their close social friends is a strong correlate with attitudes toward marriage timing as well (see Table 3).

Top rows for both models in Table 3 show results for measures occurring between 1996 and 2008; the remaining rows show results for measures occurring prior to 1996 as well as other key factors. Marriage occurring by 2008 supports the expected association of marriage with related attitudes, similar to marriage occurring by 1996: models show that respondents married by 2008 support ideal marriage ages that are younger for women and for men than those respondents who were not married at either time point. We also find that educational attainment in 1996, percentage of English-only schools, and other key factors operate in the same direction as those presented in Table 2. School enrollment status in 1996 is no longer significant.

School enrollment status in 1996 is, as expected, highly correlated with additional educational attainment in 1996–2008 (r = .60). Additional analyses show that in models similar to those presented in Table 3 but without additional attainment, school enrollment status in 1996 retains a strong, positive, and statistically significant association with attitudes in 2008. The adjusted R² values in these additional models, however, are smaller than those presented in Table 3, suggesting that the full model with both measures is a better fit.

Differences Across Social Categories: Gender

Results evidence two key ways in which the relationship between education or travel and attitudes about marriage timing varies by gender. First, findings indicate that new experiences of education and travel are associated with men’s and women’s attitudes to varying degrees, particularly early education. The models shown in Tables 2 and 3 show that female respondents prefer that both women and men marry at later ages than male respondents do (p < .01 and p < .001, respectively). As noted earlier, education and travel relate to later marriage timing preferences. To see whether associations with these mechanisms of attitude change depend on respondents’ gender, we test interaction effects between gender and education and travel. The interaction effect between educational attainment in 1996 and being female is significant for both outcomes. These results, illustrated in Fig. 3 for the ideal age for women to marry, suggest that the association between education in 1996 and attitude change is significantly different for women than for men, with stronger associations of earlier educational attainment for women. In other words, women with more education in 1996 prefer older ages for men and for women to marry than women with little education in 1996. This difference in marriage preferences between more and less educated women is larger than the difference in preferences between men with the same varying levels of education.

Second, in addition to gender differences in marriage age preferences, the change in the gap between these ideal ages at marriage for men and women is also gendered. The gap we refer to here is the ideal age for men to marry minus the ideal age for women to marry, or the difference between responses to our two gender-specific attitudes. This difference is what we describe as a gender gap in the ideal age at marriage; it is measured in both 1996 and in 2008, so the gender gap itself can change over time. And it does change, becoming somewhat larger over time (see Table 1). Men and women also show different levels of change in this gender gap. Specifically, in 2008, the gap in ideal ages is larger for women than for men. Women’s gap in preferred marriage ages is 0.23 years larger than men’s gap (p < .05), indicating that women prefer that both women and men marry at older ages than men do but that the women’s stated ideal age difference between men and women at marriage is larger than men’s stated ideal difference.

We next estimate multivariate models of the 1996–2008 change in the gender gap to learn whether education and international travel are associated with this change. Estimates from these additional models show that education attained between 1996 and 2008 is positively associated with the gap in ideal age between men and women: with each additional year of education attained between 1996 and 2008, the gap in marriage age preferences between men and women also increased by 0.09 years (p < .05). The rise in educational attainment may be associated with other changes in gender-specific ideas of work or financial security before marriage that push both men and women to expect men to marry at later ages than women. Finally, although this finding is not significant, international travel is negatively associated with the gap in ideal marriage ages.

Discussion

Recent decades have seen ideal ages at marriage rise worldwide, with the biggest increases in rural agricultural parts of Asia and Africa (Jayakody et al. 2008). It is not surprising to find parallel changes in ideas about marriage in rural Nepal. Much more important is documentation of the processes that may produce these changes. Here, we develop a theoretical framework drawing from decades of demographic research to understand these processes. Rapid school expansion near rural villagers increases enrollments in school (Beutel and Axinn 2002; Burde and Linden 2012), and we find that educational attainment is associated with significant changes in these attitudes (Table 2). In fact, we find that not only is educational attainment between 1996 and 2008 strongly associated with a change in ideal ages at marriage from 1996 to 2008, but that even educational attainment before 1996 has an independent association with subsequent changes in these ideas after 1996. It is not surprising to find that education is strongly associated with ideas: literature on ideational change has repeatedly turned to education as an explanation for such change (Alwin et al. 1991; Lesthaeghe 2007).

But what is it about education that leads to these changes in ideas? Here we document specific dimensions of nearby schools, including the highest grade offered, the percentage of teachers with a BA at the closest school, and the density of English-only schools. Although each of these dimensions of the educational context is a theoretically plausible explanation, we find that only the density of English-only schools has a significant, independent association with changes in ideal ages of marriage. Even with individual enrollments, we find that education occurring in years in which private English-only schools are available leads to stronger changes in ideas than education attained in years prior. In settings like Nepal, schools using English as the only medium of instruction tend to be strongly influenced by foreign instructional materials, which are to diffuse ideas that are more prevalent in high-income settings (Caldwell et al. 1988; Thornton 2005)—in this case, older ideal ages at marriage.

International travel is a different mechanism of potential influence on ideas. Because it takes the individual to new settings and creates new social interactions, it has the potential to teach individuals novel information and provide ideas about family and demographic behaviors (Thornton 2005). Predicting both the specific ideas that may change and the direction of the change may depend greatly on the social, economic, and cultural differences between the population of origin and the population of destination. For the rural Nepal population, though, ideas about marriage timing are highly salient given that marriage is nearly universal, with historically young ages at first marriage. Nepalese women born in 1952–1956 married at a median age of 14.6 years; for men, this median age was closer to 18 years (Allendorf 2013; Caltabiano and Castiglioni 2008; Fricke 1986). Thus, when individuals from rural Nepal travel to nearly any other setting, they are exposed to higher ages of marriage, higher ideal ages at marriage, and most likely dozens of other ideas and experiences that are consistent with higher ideal ages at marriage. Also important, the specific high-volume international travel streams between 1996 and 2008 were to high-income, high-consumption settings in the Middle East and Gulf region. As Freedman pointed out after studying family change in Asia for decades, rising consumption aspirations can be an important force toward later ideal age at marriage and preferences for smaller family sizes (Freedman 1979). In our analysis, we find that international travel has an independent association with changes in ideal ages at marriage that is not explained by the powerful forces of educational attainment.

Independent of individuals’ personal travel experiences, the travel experiences of those in their immediate social environment may also be associated with their ideas. We find that the international travel experience of friends is associated with individuals’ ideal ages at marriage independent of their own education and travel. This is an important finding: in rural settings like Nepal, the social influence of international travel may be so powerful that regardless of an individual’s own experience, having a close network member visit a foreign country creates new ideas and aspirations for individuals at home as well.

We also find that the association between new experiences and ideal ages of marriage does indeed vary by gender, at least to some extent. Women’s ideas about the ideal age at marriage are more strongly associated with educational attainment than men’s ideas about ideal age at marriage, at least in rural Nepal. Moreover, this same change toward greater educational attainment is also associated with a growing gender gap in ideal ages at marriage, favoring increasingly later ages at marriage for men more than for women. This interesting result demonstrates that education may magnify other gendered expectations about work, pay, and financial independence, such as higher expectations for men to have work success before marriage (Oppenheimer 2003; Thornton et al. 2007). In this case, greater educational attainment appears to lead both men and women to believe men should marry later than women, consistent with other recent work on gender and marriage in Nepal (Allendorf et al. 2017).

Rates of school enrollment and levels of educational attainment continue to rise worldwide, even in settings of relatively high educational attainment (Barro and Lee 2013). Ideal ages at marriage will likely continue to rise as long as educational attainment continues to rise. Educational enrollment has similar delaying effects on the timing of marriage in the United States, Europe, and Nepal; in each case, role conflict is considered the key mechanism (Thornton et al. 1995; Yabiku 2005). Likewise, international travel continues to grow more common. The evidence we document showing that international travel—both personally and by key social contacts—has significant population-level consequences for ideas about marriage timing is entirely consistent with social demographic reasoning in recent decades (Bongaarts and Watkins 1996; Thornton 2005). Our analysis of ideal ages at marriage provides an illustration of how these changes can produce long-term behavioral consequences: ideal ages at marriage influence marital timing decisions, with many long-term consequences for individuals’ life courses and family outcomes (Thornton et al. 2007).

Acknowledgments

This research was supported by an NICHD center grant to the Population Studies Center at the University of Michigan (P2CHD041028). We would like to thank the Institute for Social and Environmental Research in Chitwan, Nepal for collecting the data used here; Jennifer Barber and Dirgha Ghimire for their helpful feedback on earlier versions of this manuscript; and Cathy Sun for assisting with data management. The authors alone remain responsible for any errors or omissions.

Notes

References

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.