Emotional influences on fertility behaviors are an understudied topic that may offer a clear explanation of why many couples choose to have children even when childbearing is not economically rational. With setting-specific measures of the husband-wife emotional bond appropriate for large-scale population research matched with data from a long-term panel study, we have the empirical tools to provide a test of the influence of emotional factors on contraceptive use to limit fertility. This article presents those tests. We use long-term, multilevel community and family panel data to demonstrate that the variance in levels of husband-wife emotional bond is significantly associated with their subsequent use of contraception to avert births. We discuss the wide-ranging implications of this intriguing new result.
Although research on fertility decline has preoccupied demographers for nearly a century (Thompson 1929), and scores of studies have provided empirical evidence on the topic, many fundamental hypotheses about fertility decline remain untested. Emotional factors have received particularly little empirical attention in fertility research and other areas of demography (Massey 2002); however, the emotional bond between spouses is a critical component of family life that could powerfully influence couples’ fertility-limiting behaviors. Virtually all societies view the marital relationship as a functional dimension of social networks, but the dominant Western European view emphasizes the emotional basis of marriage (Goode 1963; Thornton et al. 2007). This view has now spread to parts of the world that are not historically characterized by an emotional view of entry into marriage (Ghimire et al. 2006; Rindfuss and Morgan 1983; Thornton and Lin 1994). However, even in societies that historically did not view emotions as the basis for entering marriage, there is widespread acknowledgement that within marriage, emotional dimensions of the relationship are core defining features of marriage (Ahearn 2001; Allendorf and Ghimire 2013; Fuller and Narasimhan 2008; Goode 1959; Pasupathi 2002). Variation in emotional dimensions of marriage may motivate adoption of behaviors to avert births, such as contraceptive use. In this study, we use measures from a long-term panel study to explore the empirical evidence for a link between indicators of emotional bond within a couple’s relationships and their subsequent use of contraception.
Suitable measurement of emotional factors has proved a formidable obstacle to empirical testing of their possible association with fertility-related behaviors. Measurement of the emotional bond is not well defined, nor is it simple to execute. In fact, the lack of standardized measures of variation in the emotional bond in the general population is undoubtedly a key reason that population-based social sciences have given little empirical attention to the emotional basis for behavior (Massey 2002). The study that we report here is the product of several years of ethnographic, focus group, survey design, and pilot study research to devise a general population measure of the emotional bond between husbands and wives for a specific South Asian setting: Nepal.
By integrating this measure into the Chitwan Valley Family Study (CVFS) in rural Nepal, we are able to describe the observed empirical relationship between variations in the husband-wife emotional bond and their subsequent fertility-limiting behavior. This panel study is ideal because it was designed to measure community, family, and individual influences on fertility behavior; and it features a multilevel, comprehensive baseline measurement combined with a detailed record of fertility behaviors spanning more than 10 years after the baseline. These data allow us to establish the hypothesized time order between measures of emotional bond and subsequent behaviors to limit childbearing and to control for a broad range of key determinants of fertility limitation that could produce a spurious relationship between couples’ emotional bond and fertility behaviors. Also important, this unique panel study features measures of other subjective phenomena that may covary with variations in the emotional bond, including spousal communication and spousal conflict. By accounting for these highly relevant yet distinct dimensions of couples’ relationships, we are able to examine the extent to which variation in the emotional bond has an independent association with subsequent fertility behavior. Together, these empirical resources provide an unprecedented opportunity to examine associations between emotional variation and fertility behavior.
Although our objective is to test a single hypothesis, the complexity of empirical research on the behavioral consequences of emotional variation requires specific theoretical guidance to formulate an appropriate empirical test. The framework that we propose is simple, but it establishes three crucial steps: (1) conceptualization of the specific emotional dimension that we investigate; (2) consideration of the mechanisms likely to link that emotional dimension to the specific subsequent behavior that we investigate; and (3) consideration of the known determinants of the subsequent behavior that may also shape the emotional dimension. This third step is critical because without randomization of emotional factors—a research tool likely to remain unavailable for the foreseeable future—all observational studies must address the potential that observed associations with emotional factors are the spurious product of other strong associations.
Conceptualizing the Husband-Wife Emotional Bond
Research on the husband-wife relationship has identified many dimensions, but the dimension of positive emotional bond—love—is commonly deemed one of the important elements of marriage. Research on marriages in historical Western Europe has often treated the emotional bond as a prerequisite and the singular core defining feature of marriage, on which social institutions such as church and state build other dimensions of marriage (Goode 1959; Hamon and Ingoldsby 2003; Hart 2007; Khandelwal 2009; Patico 2010; Thornton et al. 2007). The literature not only sometimes identifies love as unique to Western marriage (Coontz 2005; Goode 1959; Khandelwal 2009; Thornton and Lin 1994) but also sometimes characterizes non-Western arranged marriage as loveless, empty, patriarchal, and without choice (Khandelwal 2009; Pasupathi 2002; Patico 2010).
Recent critics of the ethnocentric characterization of arranged marriage have recognized “love” in general—and the “husband-wife bond” in particular—as universal psychological phenomena. Scholars have seen the loveless characterization of arranged marriage as discourse of exaggerated difference that incorrectly renders romance unique to Western contexts (Khandelwal 2009; Pasupathi 2002). These scholars’ conclusions are consistent with the Hindu philosophy of arranged marriage that dominates the central parts of South Asia. Hinduism defines marriage as the union of two individuals for life so that they can pursue dharma (duty), artha (possessions), kama (physical desires, sexual pleasure), and moksa (ultimate spiritual release) (Ramachandran 2010; Vātsyāyana 2009). In fact, as is evident from the amazing variety of mythical love stories that abound in the Sanskrit literature, Hinduism often glorifies the idea of love between spouses. In Hinduism, love consists of four components: kama (pleasurable, sexual love), prema (elevated love), karuna (compassion and mercy), and bhakti (devotion) (Ramachandran 2010; Vātsyāyana 2009). Kamasutra, one of the oldest texts on human sexual behavior, was written by Vātsyāyana around the fourth century AD and provides substantial evidence of the glorification of love and sex in Hinduism (Vātsyāyana 2009).
Although Hinduism glorifies love between spouses, it conflicts with the Western notion of love in one important way. Although the Western perspective conceives of romantic love as springing out of individual’s desire and attraction and often considers this a prerequisite for marriage, Hinduism professes that the bond between husband and wife (love) begins after the marriage and grows between the two spouses with years of collaborative life struggle (Fuller and Narasimhan 2008; Goode 1959; Medora 2003; Pasupathi 2002). In the Hindu belief system, marriages are made in heaven and celebrated on Earth. In this philosophy, arranged marriage is explicitly designed to enhance the strength of the emotional bond between husband and wife—to grow love.
In this investigation, we follow the key foundational schema of the universality of love (not limited to those situations understood as Western) with cultural variability (Patico 2010). As in many other societies, Nepali society has a long history of cultural practices, music, poetry, and literature that glorifies the emotional bond between marriageable youth or between husbands and wives (Bennett 1976; Fricke 1986; Macfarlane 1976; Matthews 1989; Pasupathi 2002). Ethnographic research on Nepalese in Nepal and India has verified the local cultural construction, meaning, and centrality of love for husband-wife bonding (Ahearn 2001; Allendorf 2009; Fricke 1986; Macfarlane 1976; Pasupathi 2002). In her research of marital relationships in Nepal, Allendorf (2009:136) asserted that “people conceive of love between a husband and wife as a powerful emotion, felt inside heart and mind (man).”1
Although Hinduism is not universal in Nepal, it informs many aspects of the local views of marriage; however, important ethnic variation continues at the intersection of culture, social change, and love (Ahearn 2001) or expressions of love (Bennett 1976; Fricke 1986; Macfarlane 1976). Moreover, Ahearn’s (2001) investigation of the meaning and centrality of love offers a vivid illustration of how recent dramatic changes characterizing Nepali society—especially proliferation of schools—have shaped this emotional phenomenon. As a result, variation in the strength of the emotional bond between husbands and wives is part of daily life for married couples in Nepal.
Emotional Bond and Fertility-Limiting Behaviors
Husband-wife relationships are multidimensional, including facets such as love and affection, communication, and conflict (Allendorf and Ghimire 2013). Full consideration of the husband-wife relationship reveals multiple emotional dimensions that may influence fertility behavior. More frequent communication between husbands and wives is likely to directly affect contraception and childbearing behavior. Close communication between sexual partners increases the likelihood of effective contraceptive use and the achievement of childbearing intentions (Lasee and Becker 1997; Link 2011; Satayavada and Adamchak 2000; Sharan and Valente 2002). Couples who have generally higher levels of communication are more likely to discuss issues of contraceptive use and desired number of children (Salway 1994), and marital relationships that have a stronger emotional bond are likely to also be characterized as more communicative (Ahearn 2001; Allendorf 2009; Gottman 1979). Therefore, it is important to consider this related yet distinct dimension of the marital relationship.
Disagreement is another dimension of the husband-wife relationship that is likely to be associated with both the strength of the emotional bond and fertility-limiting behaviors. Of course, high marital conflict that leads to divorce reduces fertility by terminating sexual relationships, thereby removing couples from exposure to risk of contraceptive use or pregnancy (Bongaarts 1982). However, a number of studies have shown that even among couples who are currently married, pregnancies are less likely to take place when the marriage is characterized by a high level of conflict or instability (Lillard and Waite 1993; Peters 1986; Thornton 1978), which thus may lower their need for contraception. Husband-wife conflict may also affect fertility-limiting behaviors because the couple is less interested in engaging in mutual activities, including sex and childrearing. As a result of these mechanisms, high-conflict marriages may produce lower overall fertility and higher rates of fertility limitation. On the other hand, high levels of conflict are likely to leave couples less able to use contraception effectively, much like poor communication reduces effective contraceptive use. Either way, because positive emotional bonds and disagreement likely covary, albeit inversely, it is crucial to consider husband-wife disagreement in evaluating the consequences of the husband-wife emotional bond.
The strength of positive emotional bonds is also likely to have an influence on fertility behavior independent of these other dimensions of the relationship. There is reason to believe that a strong positive emotional bond between husband and wife may influence fertility-related behaviors by shifting views of the role of childbearing and childrearing within marriage. That is, stronger husband-wife emotional bonds may change perceptions of childbearing and childrearing from being primarily an obligation to the larger kin network to instead being primarily an expression of love and affection toward one other (Caldwell 1982; Degler 1980). Some theories of population-level change in fertility-limiting behaviors have identified this type of shift as a watershed change promoting contraceptive use to limit fertility (Caldwell 1982). Of course, this shift to perceiving childbearing as an expression of love could motivate couples to bear more children. However, the broader fertility literature implies that as couples invest more of their emotional bond in childrearing, couples will also invest more time and resources into their children and begin trading high investments in few children for low investments in many children. This emotional connection to childrearing promotes what economists have described as the quality-quantity tradeoff in childbearing and childrearing (Becker 1991; Easterlin and Crimmins 1985; Willis 1973). This emotional dimension of fertility behavior may explain why couples continue to have children even as circumstances change the costs and benefits of childbearing. That is, even when having no children at all is economically rational, the emotional motivation for limiting childbearing predicts that couples will still consider having a small number of children as an expression of their love and affection.
The relationship between marital dynamics and fertility is complicated by the reciprocal influences of fertility on relationship dynamics, which is especially true in the Nepalese setting. Absence of childbearing—sometimes because a spouse is unable to conceive—is known to produce higher levels of conflict and lower levels of satisfaction within marriages in Nepal (Bennett 1983; Stone 1978). Thus, any empirical investigation into the influence of marital relationship dynamics on fertility must consider temporal ordering to address the possible reciprocal effects of fertility on marital relationship dynamics. As a result, detailed time-ordered measures of marital relationship, contraceptive use to avoid pregnancy, and childbearing are crucial.
Additional Factors Confounding the Association Between the Emotional Bond and Fertility-Limiting Behaviors
The widespread adoption of fertility-limiting behaviors is such an important element of demographic change that extensive population research has identified many factors that influence fertility, such as individual-level education, employment, exposure to media, religion, and individuals’ orientations about family or family formation processes (Caldwell 1982; Cleland and Wilson 1987; Lesthaeghe 1983; Lesthaeghe and Wilson 1986; Notestein 1953; Rindfuss et al. 1988; Thornton 2001, 2005). These factors also include family and household-level factors, such as parents’ education, work, and exposure to media (Axinn and Yabiku 2001; Barber 2000; Caldwell 1982; Caldwell et al. 1983, 1988; Prasad et al. 2015; Shakya and Gubhaju 2016; Skirbekk 2008). Finally, these factors also include various dimensions of social context, such as the spread of nonfamily services (Axinn and Yabiku 2001; Thornton and Lin 1994), mass education (Axinn and Barber 2001; Cochrane 1979), family planning policy (Entwisle and Mason 1985), and family planning programs (Brauner-Otto et al. 2007; Entwisle et al. 1997; Knodel 1987). This body of literature has produced numerous insights regarding the factors affecting dimensions of fertility behavior, especially couples’ use of contraceptives to limit their childbearing.
Studies have also demonstrated the important influence of change and variation in individuals’ community context and nonfamily experiences on marital processes, including the quality of the husband-wife relationship (Allendorf and Ghimire 2013; Ghimire et al. 2006; Hoelter et al. 2004; Thornton and Lin 1994; Yabiku 2004, 2005, 2006). Thus, failure to consider the exogenous consequences of community, family, and individual experience for both the husband-wife relationship and fertility-limiting behavior could drive a spurious relationship between husband-wife relationship and subsequent contraceptive use.
Our effort to document the association between emotional variations and subsequent fertility limitation in rural Nepal benefits from two crucial factors: (1) we are able to use measurement from the Chitwan Valley Family Study (CVFS), which was explicitly designed to estimate consequences of community, family, and individual experiences on marriage, fertility, and contraceptive use; and (2) dozens of completed studies have used CVFS data, which document the community, family, and individual factors affecting marriage and fertility in this specific study population (Axinn and Barber 2001; Axinn and Yabiku 2001; Barber and Axinn 2004; Barber et al. 2002; Brauner-Otto 2011, 2012; Brauner-Otto et al. 2007; Yabiku 2004, 2005, 2006). Drawing on those studies, we incorporate the known determinants of couples’ subsequent contraceptive use to limit childbearing. The measures themselves span the domains of health services, education, media exposure, social welfare groups, travel, parents, children, spouses, ethnic groups, religion, values, and beliefs (Axinn and Barber 2001; Axinn and Yabiku 2001; Barber and Axinn 2004; Barber et al. 2002; Brauner-Otto 2011, 2012; Brauner-Otto et al. 2007; Yabiku 2004, 2005, 2006). To organize this broad range of factors into our empirical models, we rely heavily on the life course approach, beginning with those factors determined earliest in life and then incorporating factors from later in life (Axinn and Yabiku 2001; Elder 1983, 1985, 1994).
Finally, although demographic research on fertility processes has historically focused on women, recent research on men has demonstrated important gender differences in these processes (Axinn 1992; Barber 2001; Ghimire 2015; Ghimire and Axinn 2006; Malhotra 1991; Mason and Smith 2000; Morgan and Niraula 1995; Sorenson 1989). Thus, we also took advantage of the unique CVFS individual-level measures, collected independently from both the husband and wife, to include the effects of the husband’s and wife’s experiences on the couple’s fertility-limiting behavior.
Data and Methods
This study used data from the Chitwan Valley Family Study (CVFS), which features substantial mixed-method, multilevel measurement of couples, marriage, and childbearing (Axinn and Pearce 2006) and is based on a systematic probability sample of 171 neighborhoods in South Central Nepal. In these neighborhoods, all individuals aged 15–59 and their spouses were interviewed in 1996 (N = 5,271, response rate = 97 %). A structured survey interview provided measures of individuals’ values, attitudes, and marital relationship; and a life history calendar (LHC) provided rich retrospective measurement of the timing of individual life events, including marital and contraceptive use events. The LHC method provides more accurate retrospective measurement of life events than alternatives (Axinn et al. 1999; Belli 1998; Freedman et al. 1988).
In February 1997, the CVFS launched a monthly prospective study of contraceptive use in 151 neighborhoods of the 171 originally sampled neighborhoods. Nearly all (95 %) respondents who were interviewed in 1996 participated in a monthly contraceptive use survey through 2008 (144 months).2 We focused on the 814 currently married women aged 15–44 in the 1996 baseline interview who were also followed through 2008.3 We limited the sample to married women because the measure of their emotional bond with their partner is contingent on their marriage. Moreover, very few unmarried women use contraceptives in this setting. The CVFS includes measures of these women’s family, households, and neighborhoods.
We focused on behavior aimed at limiting fertility by studying the use of any contraception to avert births. Contraceptive prevalence in Nepal, which was very low up until the late 1970s, became widespread by the late 1990s (Satayavada and Adamchak 2000). The current use of contraceptive methods increased from 28.5 % in 1996 to 49.7 % in 2011 (MOHP et al. 2012). The patterns of contraceptive use are even more dramatic in Chitwan. Among women born between 1942 and 1951, less than 5 % had used permanent methods of contraceptives before they reached age 25. But among the women born between 1962 and 1971, more than 35 % used those methods by age 25, with more than 62 % of those women in our current sample (Axinn and Barber 2001). Further, there was a concurrent dramatic decline in total fertility rate (TFR) from 6.3 to 2.6 between 1971 and 2011 (MOHP et al. 2012).
We operationalized contraceptive use as the first reported use of any contraceptive methods during the observation period (months 1–144 in a monthly prospective study), while controlling for prior use. Both husbands and wives were asked, “Did you, your (husband/wife), or your partner ever use any kind of contraceptives or any method for delaying or avoiding pregnancy?” These methods include oral contraceptive pills, Depo-Provera (injectable, hormonal contraceptive), condoms, foams, intrauterine device (IUD/“loop”), Norplant (subdermally implanted, long-acting contraception), abstinence, respondent sterilization, spouse sterilization, and any other method such as withdrawal. Responses were coded as 1 for “yes,” and 0 if otherwise. We coded a time-varying, dichotomous measure equal to 0 for the months the couple did not use any contraceptive methods, and 1 for the first month the couple used any contraception within the study period. This measure was used to estimate the hazard of contraceptive use to limit fertility.
More than two-thirds of married couples began using contraceptives to avoid pregnancy during the observation period. Table 1 presents descriptive statistics for all measures used in these analyses. The distributions refer to the respondent’s last person-month contributed to the analysis; for couples who used any contraceptive methods, this is the first month they used the method during the study period, and for couples who did not use any contraceptive methods, this is the final month (month 144) of data collection. Likewise, slightly more than one-third of the women reported using any contraceptive methods prior to start of the observation period.
Emotional Bond With Husband
Although emotions are expressed and shared, the internal nature of the husband-wife emotional bond makes it especially difficult to measure. We took this challenge seriously and invested substantial effort in constructing culturally appropriate, context-specific measures of values, attitudes, and emotions (Ahearn 2001; Allendorf 2009).
To construct these measures, we applied a mixed method approach in an iterative process through a series of steps (Axinn and Pearce 2006). First, we used unstructured interviews and direct observations focused on marital relationships to investigate multidimensional aspects of the husband-wife emotional bond in the local setting. Several investigators, including the authors, spent extensive time with people in Chitwan in their yards, farms, and nearby markets observing rituals and related activities, attending public meetings, and engaging in casual discussions. These informal meetings gave us a baseline understanding of how these complex feelings are expressed and communicated—an understanding that agrees with subsequent ethnographic research in other parts of Nepal (Ahearn 2001).
Second, we used the insights from this unstructured investigation to create multiple structured survey measures of emotional bonds between husbands and wives. Using these initial measures, we conducted more than 100 cognitive interviews and dozens of focus group interviews for more than a year with systematically selected sections of the study population to create potential measures. We then used results from these interviews to eliminate some of the survey measures and revise others.
Third, these measures were then pretested sequentially among four samples of 100 married individuals who were representative of the full study sample. The pretests were conducted by an experienced local research staff that included men and women and was religiously, ethnically, and racially diverse. Once again, we used the results from these pretests to eliminate measures and to revise the strongest of the measures.
Fourth, we then administered the final versions of survey measures of husband-wife emotional bond to a pilot sample in order to assess the measurement properties of these survey items. We then used the results of this pilot study as the basis for choosing a set of standardized survey measures included in the 1996 interviews with married couples. The final version of the questionnaire administered to all CVFS respondents in 1996 covered six topics: (1) love/affection; (2) disagreements; (3) criticism; (4) intimate partner violence (IPV); (5) communication regarding the number of children to have; and (6) communication about contraception. Recent research on Nepalese marriages has indicated that satisfaction is a single dimension including love/affection (topic 1), disagreements and criticisms are elements of a single negative dimension of marriage (topics 2 and 3), and communication is a single dimension (topics 5 and 6) (Allendorf and Ghimire 2013). Although these dimensions of marriage overlap, they are also empirically distinct (Allendorf and Ghimire 2013).
Because of the significant potential for husbands and wives to influence each other’s responses, we also made special effort to maintain independence in their reporting. We sent multiple interviewers to each household for separate, simultaneous husband and wife interviews to ensure independent responses.
Although more than one-half dozen Nepali words/phrases that can be used to describe the husband-wife emotional bond, the word maya (a type of love) was the most commonly used term. We used this same Nepali word in our interviews to measure the intensity of the emotional bond between spouses. In the baseline interview, all married women were asked, “How much do you love (maya) your (husband)? Very much, some, a little, or not at all?” For statistical analyses, we coded the responses “very much” as 3, “some” as 2, and “little” or “not at all” as 1. Because of the time-invariant nature of the indicator, the value assigned at the time of the baseline interview is carried forward through the entire hazard analysis. This item is similar in scope to the global items often employed to measure the positive aspect of marital quality (Gottman 1998; Gottman and Notarius 2002; Norton 1983) and is correlated, albeit weakly, with other dimensions of relationship quality (discussed later herein) as anticipated.
Although we focus on women’s responses in the current study, the CVFS also asked the same question to all married men at the time of the baseline survey. In ancillary analyses, we used men’s responses to confirm that (1) husbands’ responses are consistently correlated with their wives’ responses; (2) wives’—not husbands’—perceptions of the emotional bond are most predictive of fertility-limiting behavior; and (3) the findings based on women’s reports are robust in models that account for (a) husbands’ reports and/or (b) a binary indicator of discrepancy between the husband’s and wife’s report. These additional analyses are available upon request.
Other Dimensions of Marital Relationship Quality
Although the focus of our study is on the emotional bond between husbands and wives, we also controlled for other dimensions of marital quality that are associated with contraceptive use and are likely highly correlated with the strength of couples’ emotional bond. The first measure is an indicator of conflict within the marriage: the frequency of disagreements between the spouses. Respondents were asked, “How often you do you have disagreements with your (husband/wife)? frequently (3), sometimes (2), or seldom/never (1)?”
The second measure is an indicator of spousal communication, a dimension of marital relationships that is known to be highly related to contraceptive use. Women were asked, “How often do you discuss contraceptive methods with your husband? Often, sometimes, or never?” With this information, we included an indicator of frequency of communication about contraceptives, ranging in value from 1 (never) to 3 (often). Note that in the design phase of the study, we developed and tested these measures using the same rigorous design and implementation process that we did for the measure of emotional bond.
Although the measure of emotional bond is correlated with these two other dimensions of the relationship, they appear to be distinct dimensions of relationship quality. The measure of emotional bond is negatively correlated with spousal disagreement (r = –.103) and is positively correlated with spouse’s frequency of communication about contraception (.108). The independence of these dimensions of Nepalese marriages is consistent, which corroborates recent research on an independent sample of married Nepalese couples (Allendorf and Ghimire 2013). Of course, we anticipate that these additional dimensions of marital relationship quality are associated with contraceptive use; however, the independence across these three dimensions of marital quality is consistent with our expectation that associations with subsequent fertility-limiting behaviors are also independent.
A number of prior family experiences may shape both the strength of the emotional bond between women and their husband as well as the couple’s subsequent use of contraception. We measured couples’ previous use of contraception (ever used contraception before study period) using data from the detailed LHC that records any contraceptive use preceding the start of the study. We also measured the couple’s living arrangement, including whether they live with extended family members and/or whether they live with one another (in 1996). We also measured the length of the couple’s marriage using a continuous measure of the number of months the couple has been married. Given the practice of arranged marriage in Nepal, we also included a measure estimating the level at which women participated in choosing their spouse. Because the degree of involvement in spouse choice varies tremendously among those who do participate in arranged marriages in Nepal, this measure ranges in value of 1 to 5 denoting whether only the respondent’s parents (1), mostly parents (2), parents and respondent equally (3), respondent mostly (4), or only the respondent (5) chooses her spouse (Ghimire et al. 2006). Finally, because number of children—especially sons—strongly influences a couple’s desire for subsequent children, we also accounted for the number of children ever born and whether the respondent has at least one son (Axinn and Yabiku 2001).
Attitudes Toward Contraception
We also accounted for women’s receptivity to birth control using an indicator of whether she reports that birth control is “okay.” The measure ranges in value of 1 to 4 denoting whether the respondent (1) strongly disagrees, (2) disagrees, (3) agrees, or (4) strongly agrees.
Many prior nonfamily experiences may also shape both the strength of women’s emotional bond with their husband and contraceptive use. We accounted for two nonfamily experiences that are closely associated with women’s contraception behavior: schooling and media exposure (Axinn and Yabiku 2001; Ghimire and Axinn 2013). In terms of schooling, we account for the number of years that women have attended school as well as the total number of years that her spouse attended school. Because media exposure is a key source of information about family planning, and may also influence women’s and men’s marital quality and satisfaction, we included a summative measure of the number of media sources that each respondent and spouse has ever used: radio, television, or movies. Finally, we accounted for whether women have ever been a member of a community group—an experience that could influence her perception of the emotional bond with her husband as well as her willingness and openness to subsequently use contraception. Finally, we accounted for whether the woman has ever worked and/or traveled outside Nepal.
Because women’s parents’ experiences and behaviors are likely to shape both women’s own marital relationships and their subsequent contraceptive use, we included several measures of these in our multivariate models, including whether the respondent’s mother and/or father ever went to school or ever used contraception, as well as an indicator of the respondent’s mother’s total number of children.
Age (Birth Cohort)
Because reproductive stage is likely to be associated with women’s contraceptive use, we also included a measure for the respondent’s birth cohort, which is coded in three categories: Cohort 1, born between 1972 and 1981; Cohort 2, born between 1962 and 1971; and Cohort 3, born between 1952 and 1961.
We also accounted for the ethnic diversity in the research site. Although Nepalese society is ethnically complex, we adopted the standard categorization of ethnic groups into five categories: Brahmin/Chhetri (high caste Hindus), Dalit (low caste Hindus), Hill Janajati (Hill Tibeto-Burmese), Newar, and Terai Janajati (indigenous to Chitwan) (Bista 1972; Blaikie et al. 1980; Fricke 1986; Guneratne 1994; Gurung 1980; Macfarlane 1976).
Both theory and empirical evidence have demonstrated the potential of characteristics of the community to shape fertility behavior. Building directly on past research on this specific study site (Axinn and Yabiku 2001; Brauner-Otto et al. 2007), we accounted for community context in both childhood and adulthood (Axinn and Yabiku 2001). In terms of exposure to nonfamily experiences in childhood, respondents were asked to report the types of services and infrastructure available within an hour of walking distance from the community where they lived during childhood (before age 12). With this information, we created an index measure for the number of nonfamily services in a respondents’ community during childhood (a count of 1, 0 for each school, health service, employment center, market, or bus service, as measured in Axinn and Yabiku 2001).
In addition to measures of childhood community context, the CVFS has detailed information on respondents’ community context in adulthood. We focus on two measures of contemporary community context that have been shown to be highly related to contraceptive use among the study population: (1) access to schools, and (2) access to health services (Axinn and Barber 2001; Brauner-Otto et al. 2007). These measures come from the CVFS neighborhood history calendar (NHC) data, which recorded the walking time (in minutes) to the nearest of each of these services (Axinn et al. 1997). We created two binary indicators for whether the respondent does or does not live within a five-minute walk of a school and a health service.
We used event-history methods to model the risk of adopting contraceptive methods. Because the data are precise to the month, we used a discrete-time approach (Allison 1984; Petersen 1991). Person-months of exposure were the unit of analysis.4
where p is the yearly probability of using any contraceptive method, p/1 – p is the odds of using any contraceptive method, a is a constant term, βk represents the effects parameters of the explanatory variables, and Xk represents the explanatory variables in the model. This approach to estimating the discrete-time hazard model is described in detail elsewhere (Allison 1982, 1984; Petersen 1986, 1991). The results presented in this article properly specify the multilevel nature of the data (individual as Level 1 and neighborhood characteristics as Level 2). Estimating multilevel, discrete-time hazard models requires three major assumptions regarding the model: conditional independence, noninformative covariates, and coarsening at random (for a full description of these assumptions and the application of these techniques to these data from Nepal, see Barber et al. 2000).
We estimated three models. First, we estimated the zero-order association between the strength of the emotional bond and contraceptive use (Model 1). Second, we controlled for the robust set of controls that could confound the relationship: prior family experiences, nonfamily experiences, community context (childhood and contemporary), parents’ experiences, age (birth cohort), and ethnicity (Model 2). In the third model, we also included the two additional measures of relationship quality—spousal disagreement and communication—to confirm whether the relationship between the strength of the emotional bond and contraceptive use is robust in comparison with these other dimensions of the marital relationship (Model 3). Because multiple dimensions of spousal relationships co-occur, it is not possible to adjudicate the role that each plays in mediating the effects of other dimensions. By estimating the relationship between the strength of the emotional bond with contraceptive use with and without measures of these other dimensions of spousal relationships in our models (Model 2 and Model 3, respectively), we conceptualized the two coefficients as providing a range of the possible upper and lower bounds of the relationship between the strength of the emotional bond and contraceptive use.
Table 2 presents results from the multilevel, discrete-time hazard models. The table shows the coefficients as well as the exponentiated values expressed as odd ratios. An odds ratio of 1.00 represents no effect, an odds ratio greater than 1.00 represents a positive effect, and an odds ratio less than 1.00 represents a negative effect on the odds of contraceptive use.
Model 1 in Table 2 shows the zero-order association between the strength of the emotional bond and contraceptive use, accounting for women’s previous use of contraception. We find that the emotional bond has a strong, positive, statistically significant relationship with the rate of using contraceptives to prevent pregnancy. The odds ratio of 1.22 means that for each one-point increase in the strength of the couples’ emotional bond, the odds of using contraception increases by approximately 22 %.
In Model 2, we included a robust set of individual and community controls, finding that the observed association of this positive emotional bond with contraceptive use operates net of these factors (Axinn and Yabiku 2001; Brauner-Otto et al. 2007). Even after accounting for a wide range of controls, we find that the estimated association between the emotional bond and contraceptive use remains highly significant, actually increasing in strength. The estimated odds ratio of 1.36 means that for each one-unit increase on the scale measuring the strength of the husbands’ and wives’ emotional bond (maya), the odds of subsequently using contraception to avert births increases by 36 %. This translates into a large difference when considering the full range of the measure: the difference in the odds of contraceptive use is 71 % higher if the emotional bond is strong (3) versus weak (1).
Note that the effects of other factors that are known to affect contraceptive use are estimated as expected. For example, the strong, positive relationship between the number of children born by 1996 and contraceptive use suggests that each additional child increases the rate of contraceptive use by approximately 72 %—an effect similar to that reported in previous studies of this setting (Axinn and Yabiku 2001; Barber and Axinn 2004; Brauner-Otto et al. 2007). Likewise, having given birth to a son has a strong positive effect on the rate of contraceptive use, consistent with the effect reported in previous studies (Axinn and Yabiku 2001; Barber and Axinn 2004). Also, as expected, family and nonfamily experiences early in life are not strongly associated with contraceptive use given our inclusion of fertility experiences before the baseline interview.
Model 3 displays our estimates of the effects of a positive emotional bond between a husband and wife on contraceptive use, independent of other dimensions of spousal relationship. Adding measures of other dimensions of the spousal relationship to the model leads to almost no change in the effects of the husband-wife emotional bond, suggesting that it operates independently from these other aspects of marital quality. Moreover, corroborating vast literature on fertility behaviors and contraceptive use, the results show a strong positive association between spousal communication and contraceptive use (Link 2011; Salway 1994; Sharan and Valente 2002; Stone and Ingham 2002). However, we find no evidence that marital conflict, as measured by the frequency of disagreements, is associated with contraceptive use. Together, the results suggest that these related dimensions of the marital relationship operate to influence contraceptive use independently from the emotional bond.
Nearly a century of demographic research on the process of fertility decline has tested hundreds of hypotheses and provided empirical evidence for dozens of factors promoting the transition from high fertility and no use of birth control to widespread birth control and low fertility (Axinn and Yabiku 2001; Bulatao and Lee 1983; Cleland and Hobcraft 1985; Coale and Watkins 1986; Davis 1955; Easterlin and Crimmins 1985; Freedman 1979; Freedman et al. 1988; Knodel 1987; Notestein 1953; Thompson 1929). In spite of fertility transition’s lofty status as one of the most sustained areas of empirical investigation in the social sciences, many fundamental hypotheses about fertility decline remain untested. The potential for variations in the emotional dimensions of couples’ relationships to shape their fertility-limiting behavior is an important example.
The measurement demands required to empirically test this intriguing hypothesis are daunting. Random assignment to emotional states is not within our means. Observational studies of variations in emotional states linked to records of fertility behaviors are not only rare but without longitudinal designs and extraordinary measurement: the likelihood of observing a spurious correlation between an observed emotional state and childbearing behavior is high. Of course, our study does not fully eliminate this possibility. However, using the Chitwan Valley Family Study (CVFS)—a 15-year panel study of communities, families, couples, and individuals that was specifically designed to study fertility transition—we have a rare opportunity to gain some empirical insight into the possibility that emotions influence fertility. As demonstrated in our analysis in this article, our study yields strong evidence that the strength of the husband-wife emotional bond influences fertility-limiting behaviors: specifically, the subsequent use of contraceptives to avoid pregnancy.
Of course, this investigation is predicated on adequate measurement of variations across couples in the levels of the husband-wife emotional bond. Such measurement is a significant challenge. Thus, a first notable limitation of our measure is that it is taken at a single point in time, although we recognize—and anticipate—that the strength of a couple’s emotional bond is likely to vary over time. However, the fact that we document a strong, robust association between the strength of couples’ emotional bond and subsequent contraceptive behavior when relying on a single measure suggests that we may document an even stronger relationship if we had more precise, time-varying measures that better captured couples’ emotional bond over time.
A second study limitation is that we do not identify the precise causal pathways by which a couple’s emotional bond leads to their fertility-limiting behaviors. We both theoretically and empirically acknowledge that other specific dimensions of couples’ relationships may act as possible mechanisms linking the strength of their bond to their fertility-related behaviors (i.e., spousal communication, spousal disagreement). However, our analysis confirms that—at least in the instance of Nepal—these other relationship dimensions are weakly correlated with the strength of the emotional bond and do not explain its independent association with couples’ fertility-limiting behavior. Thus, we anticipate that other factors contribute to the association that we observe here. Theories of fertility change focused on parents’ investments in the “quality” of children rather than the “quantity” of children are especially relevant (Willis 1973). The quantity-quality tradeoff view of fertility transition is consistent with the prediction that marital change toward stronger, positive emotional husband-wife bonds may alter parents’ childrearing behaviors to increase parents’ investments of time and money in fewer children (Caldwell 1982). The associations that we document here provide strong motivation for future research to investigate this possibility in greater detail and identify the specific childrearing changes that may account for the association that we observe.
A third limitation is that our study focuses on only one direction of the potentially bidirectional association between couples’ emotional bond and their fertility-related behaviors. Although there is little reason to hypothesize that using contraception in particular will lead couples’ emotional bond to be stronger, based on past research on both American (Glenn and McLanahan 1982; Waite and Lillard 1991; White et al. 1986) and Nepalese (Bennett 1983; Stone 1978) couples, other aspects of fertility—including the experience of bearing children together—may have a powerful influence on the strength of a couple’s emotional bond. Thus, given the association that we document here, dynamic models of the potentially complex reciprocal association remain an important priority for future research.
Despite these study limitations, the study results provide a significant new insight into the potential for variation in married couples’ emotional bonds to shape their subsequent fertility-limiting behaviors. The measure itself is the product of years of ethnographic, cognitive, and survey research on the husband-wife emotional bond in the context of rural Nepal. Empirically, it correlates with reports of other dimensions of the relationship as expected. This measure provides new evidence of the consequences of this emotional dimension of couples’ relationships for couples’ subsequent decisions to stop having children.
Of course, this area of research can only benefit from greater scientific efforts to produce improved measures of husband-wife emotional bonds. We argue that this study should serve as motivation for substantial new efforts to construct such measures. This emotional dimension of the marital relationship has an empirically independent association with couples’ subsequent use of contraception to limit childbearing. Given this, it may also have other demographically significant consequences for the trajectory of marital relationships.
Finally, other dimensions of the husband-wife relationship covary with the emotional bond but may independently influence contraceptive use to avoid pregnancy. We investigate husband-wife conflict and husband-wife communication as two of the most likely to shape subsequent contraceptive use. Measures of these dimensions of the husband-wife relationship produce no change in our estimate of the consequences of variations in the husband-wife emotional bond. The emotional bond association that we observe appears to be independent of these other observed dimensions of husband-wife relationships.
Of course, many other factors in our models remain strongly associated with fertility limitation even when this emotional dimension is included. It is unlikely that the emotional dimension is strong enough to explain fertility decline by itself. The idea that many forces are working simultaneously to produce fertility decline is much more plausible (Axinn and Yabiku 2001; Freedman 1979). However, this emotional dimension may also operate independently of these many other forces.
Stepping back from the narrower question of the emotional influences on fertility per se, the results also point toward the need for an even broader view of demographic research that includes empirical attention to emotions. In recent years, several leading demographers have argued that closer engagement with the psychology of interpersonal relationships and emotions will yield a more comprehensive understanding of demographic behaviors (Basu 2006; Hobcraft 2006; Massey 2002). There are both theoretical and empirical reasons to agree (Massey 2002). We add to those here, providing empirical evidence that is consistent with that hypothesis. Certainly, as we move forward to study change and variation in couples’ decisions to have children, this emotional dimension deserves greater scientific attention. But we also join these demographers in advocating both theoretical and empirical attention to emotional factors as we investigate the full range of demographic topics that include decisions such as educational enrollment, job choice, geographic moves, or health behaviors.
This research was jointly supported by the National Institute of Child Health and Human Development (R01HD32912 and R03HD055976) and the Fogarty International Center (5D43TW000657). The authors thank Cathy Sun for assistance with data management and analyses, the research staff at the Institute for Social and Environmental Research in Nepal for collecting the data reported here, and the CVFS respondents who continuously welcome us into their homes and share their invaluable experiences, opinions, and thoughts. All errors and omissions remain the responsibility of the authors. Conflict of interest: Dr. Ghimire is also the Director of the Institute for Social and Environmental Research in Nepal (ISER-N) that collected the data for the research reported here. Dr. Ghimire’s conflict of interest management plan is approved and monitored by the Regents of the University of Michigan.
Man is a Nepali word that represents both the heart and mind; it does not directly translate into English.
Individuals who moved out of the study area were tracked and interviewed throughout this period.
In supplementary models, we limited the analytic sample to married women who had never used contraception at the time of the survey. The results were consistent with those produced in the analyses shown here, confirming that prior contraceptive use is not what is driving both the strength of the couple’s emotional bond and their subsequent uptake of contraception.
Although it may appear that the discrete-time method of creating multiple person-months for each individual inflates the sample size resulting in artificially deflated standard errors, this is not the case (Allison 1982, 1984; Petersen 1986, 1991). In fact, the estimated standard errors are consistent estimators of the true standard errors (Allison 1982:82).