Scholars traditionally argued that industrialization, urbanization, and educational expansion lead to a decline in extended families and complementary rise in nuclear families. Some have suggested that such transitions are good for young married women because living in nuclear families benefits their health. However, extended families may also present advantages for young women’s health that outweigh any disadvantages. Using the Indian National Family Health Survey, this article examines whether young married women living in nuclear families have better health than those in patrilocal extended families. It also examines whether young married women’s living arrangements are changing over time and, if so, how such changes will affect their health. Results show that young married women living in nuclear families do not have better health than those in patrilocal extended families. Of eight health outcomes examined, only five differ significantly by family structure. Further, of the five outcomes that differ, four are patrilocal extended-family advantages and only one is a nuclear-family advantage. From 1992 to 2006, the percentage of young married women residing in nuclear families increased, although the majority remained in patrilocal extended families. This trend toward nuclear families will not benefit young women’s health.
For decades, scholars have theorized about two central aspects of family structure around the world: the distribution of nuclear versus extended-family structures over time and the ways in which family structure influences well-being. The literature has generally argued that industrialization, urbanization, and educational expansion lead to a decline in extended families and complementary rise in nuclear families (Adams 2010; Bongaarts 2001). Some have further suggested that such transitions benefit young married women because their health is better in nuclear families (Santow 1995; Vlassoff and Manderson 1998). In this article, I discuss the theoretical basis of the second argument and suggest possible important benefits of living in patrilocal extended families that counteract any negative effects. Using data from India, I then empirically evaluate whether young married women’s family structure is indeed changing over time and whether their health is the better for it. I do not examine whether any changes in family structure are due to industrialization, urbanization, educational expansion, or other factors.
Although it has long been believed that societies with historical patterns of extended-family living will shift to nuclear forms as industrialization, urbanization, and educational expansion take place, this belief has been challenged. In a review of household composition in non-Western regions, Bongaarts (2001) concluded that if there is a rising trend toward nuclear families, it is proceeding slowly. More recently, using longitudinal data from 15 non-Western countries, Ruggles and Heggeness (2008) concluded that family structure is not associated with economic development. Further, they found that extended families are actually becoming more—not less—common over time in many non-Western countries.
In India, the disintegration of the extended family has reached the level of “popular cliché” (Shah 1996), but there, too, scholars have challenged this belief. Some have found evidence of an increase in nuclear families (Niranjan et al. 2005; Ram and Wong 1994), but others have found growth of extended families (Freed and Freed 1982; Wadley and Derr 1993) or have suggested that any changes are only fluctuations in the life cycle of the extended family (Caldwell et al. 1984). Others have contended that although the extended family was featured in historical Hindu texts, it has never been the dominant practice (D’Cruz and Bharat 2001; Shah 1996). Thus, this raises the question of how relevant extended-family living actually is for young married women in India. Do most of them live in extended families, and are their residence patterns changing over time?
The belief that extended-family living has a negative influence on young married women’s health arises from the position of daughters-in-law in a patrilocal extended family. In the patrilocal extended-family system, young women join their husband’s family after marriage. In their marital home, they occupy the bottom of the gender and generational hierarchies of the family (Das Gupta 1999; Zuo 2009). Kandiyoti (1988) contended that the “hardship and deprivation” of a daughter-in-law is the price of the patriarchal bargain that women pay in exchange for the comparative ease and power of the mother-in-law position that they hold later in life. The constraints of the daughter-in-law position are specific to patrilocal extended families and are unlikely to apply to matrilocal or ambilocal extended families. In fact, there may be an opposite effect in which young women in matrilocal or ambilocal families are better off than those in nuclear families. Thus, I use the term “patrilocal extended family” to reinforce that the links discussed here refer only to patrilocal extended families and not all types of extended families.
The negative impact of being a daughter-in-law in a patrilocal extended family has been both explicitly and implicitly expressed in the literature. Santow (1995) explicitly identified young women’s position as daughters-in-law in patrilocal extended families as one cause of women’s poor health in low-income countries. Vlassoff and Manderson (1998) also included daughter-in-laws’ position in the patrilocal extended family in their framework of the gendered causes of vulnerability to infectious diseases. Das Gupta (1999) made a similar suggestion in reference to the patrilocal extended family in India specifically. She argued that the North Indian patrilocal extended-family system contributes to poor health for young women. Das Gupta’s point of comparison was the European stem family, and she did not suggest that the same benefits would accrue to women residing in nuclear families in India. However, many of the constraints she outlined are reduced or absent in nuclear families.
There are also more implicit statements about patrilocal extended-family living harming young women’s health. Jeffery et al. (1989:221) stated that “poor maternal and child health reflect, in microcosm, the problems facing the bahu [daughter-in-law].” They never explicitly stated that young women would be better off in nuclear families, but their account suggested a counterfactual in which young women are not primarily daughters-in-law but instead reside in nuclear families, and their health is the better for it. Similarly, descriptions of young women’s positions as daughters-in-law in patrilocal extended families figured prominently in background explanations for young women’s poor health in India (Barua and Kurz 2001; Chorgade et al. 2006; Das Gupta 1995).
Although the daughter-in-law position in patrilocal extended families figures prominently in the literature as one reason for young women’s poor health, there appears to be no direct examination of whether there actually is such an effect. Thus, using data from India, this article examines whether young married women living in nuclear families have better health than those in patrilocal extended families. It also examines the relevance of patrilocal extended-family structure for young married women by exploring how common it is for young married women to live in such families and whether their family structure is changing over time. It then examines the implications of any changes in family structure for young women’s health.
Family Structure and Young Married Women’s Health
A Nuclear Advantage? The Position of the Daughter-in-Law
Earlier in this article, I noted that a daughter-in-law is at the bottom of both the gender and generational hierarchies of a patrilocal extended family. Further, the fundamental tie is the patrilineal blood tie among men (Das Gupta 1999). Daughters-in-law play an essential role in continuing the family line and are an important source of labor (Jacobson and Wadley 1977; Jeffery et al. 1989). However, any individual daughter-in-law is not a member of the patriline and can be replaced by another woman.
This position shapes how family members interact with daughters-in-law in ways that may harm their health. First, family members may be reluctant to invest resources that are required to secure health care (Das Gupta 1999). Maternal health services, including antenatal care and delivery assistance, can be costly (Jeffery and Jeffery 2010). Second, given the importance of their labor, daughters-in-law engage in long hours of work (Das Gupta 1995; Jacobson and Wadley 1977; Seymour 1999). This work often includes onerous tasks, such as planting rice, carrying water, and collecting fodder for livestock. This amount of labor itself may harm young women’s health. Further, the importance of this labor provides an incentive for families to refuse daughters-in-law time to rest or travel to access health services during pregnancy and illness (Barua and Kurz 2001; Barua et al. 2004). Third, daughters-in-law may receive less and lower-quality food, which harms their nutrition. It is customary for family members to eat at different times and for food to be unevenly distributed by senior family members. Daughters-in-law often eat after others in the family (Chorgade et al. 2006; Das Gupta 1995; Madan 1965). These routines are often maintained when young women are pregnant or nursing, and their nutritional needs are greater than normal. Patrilocal extended-family living also provides a disincentive for husbands to buy special foods for their wives because it may have to be shared among family members (Jeffery et al. 1989).
The daughter-in-law position also prevents young women from asserting their agency to protect their own health. Compared with women in nuclear families, daughters-in-law have little control over family decisions (Allendorf 2007a; Balk 1997; Bloom et al. 2001). Daughters-in-law are usually not able to challenge the decisions of mothers-in-law and other in-laws who may not be concerned with protecting their health (Barua and Kurz 2001; Griffiths et al. 2002). Further, some young women feel uncomfortable asserting their needs out of a desire to be an obedient daughter-in-law (Brunson 2010). Thus, daughters-in-law usually lack the high levels of power over family decisions and their own mobility that is associated with greater use of maternal health care (Allendorf 2007b; Bloom et al. 2001; Mumtaz and Salway 2007).
This discussion suggests the following hypotheses (Hypotheses 1b, 1c, and 2b to follow in later sections):
Young married women living in nuclear families have better health than young married women living in patrilocal extended families.
The health advantage of living in a nuclear family is mediated in part by young women’s decision-making power.
An Extended Advantage? Economic Status and Social Support
Although patrilocal extended-family living may present disadvantages, it also presents benefits that may compensate for these disadvantages. Many of the pathways linking the daughter-in-law position to poor health are contingent upon a family’s economic resources. Living in a patrilocal extended family may minimize such pathways because it provides greater economic status. Extended families own more assets and have better living conditions, including higher-quality housing materials, sanitation facilities, electricity, and piped water (Gage et al. 1996; Mberu 2007; Niranjan et al. 2005). Living conditions and assets can be shared among family members; thus, these analyses usually do not adjust for family size. Even in research that adjusted for family size, though, extended families were found to be better off than nuclear families (Lloyd 1999). Further, one Indian study found that extended households have greater daily income per person than nuclear households (Murthy et al. 1985).1
This economic advantage in patrilocal extended-family households is largely due to more earners and producers pooling their labor and resources. Pooling may also increase the total income earned and reduce consumptions costs. First, the inheritance custom in India is usually to divide assets evenly among sons. Thus, if family members operate a business or farm land collectively, they can do so more efficiently. Second, the larger number of family members provides economies of scale (Becker 1981). Further, extended families with income from multiple sources may be better able to smooth their income, making resources available at any time they are needed.
The economic benefits of patrilocal extended-family living should have several advantages for daughters-in-law. Like any household member, they should benefit from common resources, such as sanitation and piped water. Other resources, such as food, must be divided among household members, but even if they are the least-valued member, they may still get as much as women in nuclear families if there is more to go around. Greater economic resources should also reduce the need for their labor. So, daughters-in-law may be able to work less and be better able to take time off to rest during illness or childbirth and use health care.
Daughters-in-law may also benefit more from emotional and social support of family members when they live in the same household. Coresident sisters-in-law and mothers-in-law provide a ready labor backup (Jacobson and Wadley 1977; Jeffery and Jeffery 1997; Wadley and Derr 1993) and reduce the burden of child care (Albrecht et al. 1994; Takeda et al. 2004). In places where women do not travel alone, these family members can also help women access health services by accompanying them to facilities (Barua et al. 2004; Mumtaz and Salway 2007). Some women also form supportive relationships with sisters-in-law and mothers-in-law with whom they live (Brunson 2010; Sharma 1980). These relationships can provide companionship that in turn reduces stress and benefits their health (Ross and Mirowsky 2002; Umberson et al. 2010; Walton and Takeuchi 2010). Such companionship may be especially important when women practice purdah and have limited interactions outside their household (Jacobson and Wadley 1977; Wadley and Derr 1993).
In keeping with the potentially supportive role of in-laws, patrilocal extended-family living may also protect women from domestic violence, which is a risk factor for poor health (Ellsberg et al. 2008; Kishor and Johnson 2006). Koenig et al. (2003) found a protective effect of patrilocal extended-family living on physical violence in Bangladesh. They speculated that senior members in extended families mediate and suppress violence. However, family structure is typically not included as a determinant in studies of domestic violence; and among studies that did include family structure, some found extended families to be protective while others identified them as a risk factor (Clark et al. 2010; Tang and Lai 2008). Studies that identified patrilocal extended-family living as a risk factor suggested that in-laws perpetrate and instigate violence or are a source of conflict.
This discussion suggests the following alternative hypotheses:
Young married women living in patrilocal extended families have better health than young married women living in nuclear families.
The health advantage of living in a patrilocal extended family is mediated by economic status and emotional and social support.
No Causal Advantages? Selection Into Family Structure
Any observed differences in the health of young women by family structure may be due to selection into family type. Previous research has shown that several of the determinants of family structure are also associated with women’s health (Dasgupta et al. 1999; D’Cruz and Bharat 2001; Kolenda 1987; Niranjan et al. 1998, 2005; Ram and Wong 1994). These factors include region, religion, caste, education, urban residence, landholding, and occupation. Extended families are more common in the North, among higher-caste Hindus, higher education levels, and higher economic classes. Overall, this pattern suggests that women in extended families are more likely than those in nuclear families to have characteristics that are associated with better health.
The presence of economic status in this list bears further explanation. Earlier in this article, I described how living in a patrilocal extended family itself may lead to greater economic status. However, the relationship between family structure and economic status most likely operates in both directions (Gage et al. 1997). Economic status may also affect family structure. Fathers may persuade their sons to stay in extended families through their control over land and other assets (Jeffery and Jeffery 1997). Keeping land and businesses together may also be a powerful incentive for all family members to live together (Caldwell et al. 1984).
This discussion suggests one final hypothesis:
Family structure has no causal effect on young married women’s health. Any observed differences in young women’s health by family structure are due to selection.
The data come from the Indian National Family Health Survey (NFHS), collected in 1992–1993 (IIIPS 1995), 1998–1999 (IIPS and ORC Macro 2000), and 2005–2006 (IIPS and Macro International 2007). The NFHS is a cross-sectional, nationally representative survey of households and women of reproductive age. I limited the sample of women to those who were of an appropriate age and marital status to be daughters-in-law. Specifically, the analytical sample comprises currently married women aged 15–29 who were usual residents of the household and were living with their husbands, but not their natal families. The sample is limited to women living with husbands to ensure that all marriages were intact and the dynamic of having a husband present was consistent across family types. I dropped the small number of women living with natal families because living with natal family members is substantively different from living with in-laws. The analytical sample consists of 32,814 women interviewed in 1992–1993; 32,207 women interviewed in 1998–1999; and 29,907 women interviewed in 2005–2006. For the sake of simplicity, I refer to this analytical sample as “young women.”
The results of the health analysis presented here are limited to the sample of 29,907 women interviewed in 2005–2006. Only results from the third survey wave are presented because the measures of health are not consistently available across all waves. Two health measures are available for all three waves, four are available for two waves, and two are available for only one wave. However, the results presented here are indicative of the effects of family structure on health in 1992–1993 and 1998–1999. Earlier analyses undertaken with all three waves provided identical substantive results. Further, interactions of family structure and survey wave were not significant for the six health measures that were available in multiple waves (results not shown).2
Nuclear households are households in which the only family members with whom the young woman lives are her husband and children. Patrilocal extended households include all households in which the young woman also resides with one or more adult in-laws.
Health is operationalized with eight commonly used measures of women’s health in non-Western countries. Two measures are indicators of maternal health care. “Antenatal care” refers to whether the young woman had three or more antenatal check-ups, beginning in the first trimester of pregnancy.3 “Delivery assistance” refers to whether the birth was attended by a health professional. Both variables are limited to women who had a birth in the three years previous to the survey.
Four of the health measures are indicators of women’s diet and nutritional status. “Anemia” refers to whether the woman had a low level of hemoglobin in her blood, defined as less than 10 grams per deciliter. “Underweight” refers to whether the young woman’s body mass index (BMI) is less than 18.5. “Dairy consumption” indicates whether the woman consumes milk or curd at least weekly. Similarly, “meat consumption” indicates whether the woman consumes meat, fish, or eggs at least weekly. Milk and curd, as well as meat, fish, and eggs, are all foods that are nutritious yet also valued as special foods in India. Many high-caste Hindus are vegetarians who do not eat meat but will eat fish or eggs. Thus, these foods have been grouped.
The last two health measures are measures of domestic violence, which is a risk factor for injury, poor self-reported health, and other health problems (Ellsberg et al. 2008; Kishor and Johnson 2006). “Physical violence” refers to whether the woman reported one or more acts of physical violence by a family member in the year before the survey. “Sexual violence” refers to whether the woman reported being physically forced to have sex or forced to engage in any other sexual act by her husband in the year before the survey.4
Mediating variables include measures of economic status and decision-making power. No measures of emotional and social support were available. Economic status is measured with respondents’ occupation; husband’s occupation; household wealth; and, for dairy consumption only, whether the household owns a cow. Household wealth is a measure of the assets owned by the household, as well as the housing materials. Wealth items include, for example, the amount of land and livestock owned by the household; the roofing material of the home; and whether the household owns a television, motor bike, and other consumer durables. The measure was created with a weighted polychoric correlation matrix using principle components analysis (Kolenikov and Angeles 2009). The resulting wealth factor is the first factor retained, has an eigenvalue of 11.0, and accounts for 48 % of the total variance in the wealth items.
Decision-making power is measured with the first factor retained from a principle components analysis of nine items on the woman’s ability to make decisions. These include whether the woman usually participates in decisions on her own health care, large and daily household purchases, visits to family or relatives, and what to do with her husband’s earnings. They also include whether she is usually allowed to go to the market, health facility, and places outside her village or community and whether she has any money of her own that she can decide how to use. These items are commonly used to measure women’s decision-making power, autonomy, and empowerment (Malhotra and Schuler 2005). This decision-making factor has an eigenvalue of 3.15 and accounts for 63 % of the total variance in these items.5
Control variables that should be associated with both family structure and health include region, urban residence, religion/caste, age, education, husband’s education, parity, and the accessibility of a health facility.6
The analysis proceeded in four main steps. First, I explored the relevance of patrilocal extended-family living for young women and the extent to which their family structure has changed over time. This step compared the distribution of young women by family structure across survey waves. I also disaggregated this distribution by region, urban residence, religion/caste, and age.
Steps 2 and 3 examined the connections between family structure and health, using a Heckman two-stage model approach to adjust for selection into family type (Heckman 1979). All models were also adjusted for the survey design, including weighting and clustering on primary sampling units. The selection model in the Heckman two-stage model is a probit model with patrilocal extended-family structure (versus nuclear) as the dependent variable. The independent variables used to predict patrilocal extended-family residence are husband’s education, husband’s age, household wealth, religion/caste, region, landholding, urban residence, and the interaction of landholding and urban residence. This selection model was used to calculate the predicted probability of living in a patrilocal extended family. This predicted probability was transformed into an inverse Mills ratio, which was used as a control variable for selection in substantive health models.
The substantive health models were run jointly with the selection model using a bivariate probit model. They were run jointly to further adjust for selection into family type. If there are unobserved factors that affect both selection into family structure and health, then the error terms in the selection and substantive models will be correlated. In turn, this correlation, known as rho, can lead to biased estimates of the substantive model if standard regression techniques are used. Thus, I jointly modeled the selection and substantive models using bivariate probit models that estimate the correlations in the errors. A Wald-type test was used to evaluate whether the correlations in the errors are equal to zero.
Identification of a Heckman two-stage model does not require an exclusion criterion, but its precision is largely dependent on including one (Winship and Mare 1992). In this case, the selection model should include variable(s) that predict selection into family type but are not associated with health. Husband’s age, landholding, and the interaction of landholding and urban residence are used to fulfill the exclusion criterion. (Household wealth is also included in the selection model and not the main substantive models, but is associated with the health outcomes.) I examined whether these variables are associated with the health outcomes by including them in the substantive health models. For six of the eight health measures, one or two of these variables were significantly associated with the outcome but never all of them (results not shown). Thus, each set of models includes at least one variable that satisfies the exclusion criterion. Overall, this approach should be closer to estimating a causal effect than standard models, but the selection model is probably not strong enough to fully control for selection. Ideally, the selection model would include additional variables that better family structure and are not associated with health. Such variables would include whether the husband’s father is alive, whether he has brothers, and the brothers’ ages and marital status. Unfortunately, such variables are unavailable.
In Step 2, I used the Heckman two-stage models to address the main question of whether there is a nuclear or patrilocal extended-family advantage. Thus, this second step tests Hypotheses 1a, 1b, and 1c. These models include exogenous controls that the theoretical model predicts are associated with both family structure and health. The controls are region; urban residence; religion/caste; age; education; husband’s education; and for maternal health care, parity and facility accessibility. As described earlier, these models also included the inverse Mills ratio as a control for selection into family type.
Next, in Step 3, I addressed the questions of whether any nuclear-family advantages are mediated by young women’s decision-making power and whether any patrilocal extended-family advantages are mediated by economic status. Thus, Step 3 tests Hypotheses 2a and 2b. The models used in Step 3 make up the main models used in Step 2 plus the measures of economic status and decision making. If economic status or decision making are mediators, the effect of family structure should be substantially reduced in these models. Because there are no measures of emotional and social support, I am unable to test it as a mediator, as proposed in Hypothesis 2b.
In Step 4, I explored the implications of any changes in family structure for young women’s health. Predicted probabilities simulate the percentage of young women that would experience the health outcomes as family structure changes over time. I calculated predicted probabilities of the health outcomes under different distributions of family structure, using the main set of models from Step 2 (appearing in upcoming Tables 3 and 4). These probabilities were calculated only for the outcomes that differ significantly by family structure, and the calculations assumed that all other characteristics in the model are average at 2005–2006 levels.
During 1992–1993, 64.3 % of young women resided in patrilocal extended-family households, while 35.7 % lived in nuclear-family households (Table 1). The percentage living in patrilocal extended families decreased slightly from 64.3 % to 61.5 % in 1998–1999 and then to 56.5 % in 2005–2006. Conversely, the percentage living in nuclear-family households increased from 35.7 % in 1992–1993 to 43.5 % in 2005–2006. All these changes are statistically significant. Overall, in this roughly 14-year period, the percentage of young women living in nuclear households rose by almost 8 %. Throughout the period, however, the majority of young women remained in patrilocal extended families.
This trend toward nuclear families holds across regions, urban and rural residence, religion/caste, and age groups (Table 1). With two exceptions, the percentage of women living in patrilocal extended families declined significantly between 1992–1993 and 2005–2006 in each category. The two exceptions are the Northeast region and members of other religions. For members of other religions, the percentage of young women living in patrilocal extended families did not differ significantly over time. The Northeast presents more of an exception. The percentage of young women living in extended families in the Northeast rose significantly from 1992–1993 to 1998–1999 and then declined significantly from 1998–1999 to 2005–2006.
The distribution by family structure across these basic characteristics reinforces well-known demographic patterns in India. Although patrilocal extended-family living declined in nearly every region, it was less common throughout the period in the South and the Northeast. The South and the Northeast often differ from the other regions, which can be grouped as the North (Dyson and Moore 1983; Mishra et al. 2004). These regions differ in part because they include ethnic groups that did not customarily have a patrilocal extended-family system (Kolenda 1987). Further, reflecting the life cycle pattern of the patrilocal extended family, younger women are more likely to live in patrilocal extended-family households. Women living in rural areas and those who are general-caste Hindus are also more likely to live in patrilocal extended-family households. However, the difference between rural and urban areas is small.
An Extended or a Nuclear Advantage?
At the bivariate level, young women living in patrilocal extended families have better health than those in nuclear families (Table 2). Young women in extended families are significantly more likely to use antenatal care and delivery assistance and to consume dairy weekly. They are also less likely to experience physical violence and be underweight and anemic. There are only two health outcomes for which there is not a patrilocal extended-family advantage at the bivariate level: meat consumption and sexual violence. Meat consumption is a nuclear advantage. Young women in nuclear families are significantly more likely to consume meat, fish, and eggs weekly. Reporting of sexual violence, on the other hand, does not differ significantly by family type. Overall, simple observed differences between young women in nuclear and patrilocal extended families provide support for Hypothesis 1b that there is an extended advantage. However, as suggested by Hypothesis 1c, this pattern may be due to selection into family type.
Substantive probit models that adjust for selection into family type appear in Tables 3 and 4. One-half of the health measures are positive in nature, with a positive coefficient indicating that patrilocal extended-family living is beneficial. The other half are negative, with a positive coefficient conversely indicating that patrilocal extended-family living is harmful. To aid interpretation, the models for positive outcomes are presented in Table 3, and those for negative outcomes appear in Table 4. The results also include estimates of the correlation in the errors (rho) between the selection and substantive models. This correlation is significantly different from 0 for five of the eight outcomes. The results of the selection model are presented in Table 6 in the appendix.
After adjusting for selection into family type, the differences by family structure seen at the bivariate level are diminished yet persist. There is no longer a significant difference in underweight status and anemia. There is also still not a significant difference by family structure in sexual violence. A significant effect of family structure does remain, though, for five of the eight health outcomes: antenatal care, delivery assistance, dairy consumption, meat consumption, and physical violence (Tables 3 and 4).
Meat consumption is the only nuclear advantage. The coefficient for patrilocal extended-family living on meat consumption is a significant −0.82 (Table 3). This coefficient translates into a marginal effect of 0.30. Thus, the predicted probability of young women consuming meat, fish, or eggs weekly is greater by 0.30 in nuclear families.
The other four health outcomes that differ significantly by family type are extended advantages. Young women in extended families are significantly more likely to use antenatal care and delivery assistance and consume dairy (Table 3). They are also significantly less likely to report physical violence (Table 4). The predicted probabilities of women using antenatal care and delivery assistance are greater by 0.30 and 0.32, respectively, for those residing in extended families. Similarly, the predicted probability of women consuming dairy is greater by 0.32 for women in extended families. The largest marginal effect is for physical violence. The predicted probability of reporting physical violence is reduced by 0.40 for young women living in extended families. Thus, overall, the results still provide support for Hypothesis 1b that there is a patrilocal extended-family advantage.
I also ran additional models that examined differences in young women’s health among different types of patrilocal extended families (results not shown). Descriptions of the hardship of the daughter-in-law position often focus on living with a mother-in-law. Thus, I ran similar models with extended families divided into those with mothers-in-law and those without. According to adjusted Wald tests, the coefficients for extended families with mothers-in-law did not differ significantly from those for extended families without mothers-in-law. Thus, there is no evidence of a special or an additional effect of living with a mother-in-law. Earlier analyses also tested whether the effect of patrilocal extended-family living differs among other subtypes. Following Caldwell and colleagues (1984), I subdivided patrilocal extended families into joint, stem, joint-stem, and other. However, according to adjusted Wald tests, the effects of these extended-family subtypes do not differ significantly.
Next, the models presented in Table 5 test Hypotheses 2a and 2b by examining whether the differences in health by family structure are mediated by economic status and decision-making power. The results suggest that the greater economic status of extended families is responsible for the patrilocal extended-family advantages in antenatal care, delivery assistance, and dairy consumption. The coefficient for patrilocal extended-family residence is reduced from a significant 0.83 in the main model of antenatal care (Table 3) to an insignificant −0.01 in the model with measures of economic status (Table 4). Similarly, the coefficient for patrilocal extended-family residence is reduced from a significant 0.81 in the main model of dairy consumption (Table 3) to an insignificant 0.07 in the model with economic status (Table 5). The results for delivery assistance suggest that the greater economic status of extended families is overcoming an underlying nuclear-family advantage. In the main model of delivery assistance, the coefficient for patrilocal extended-family residence is a significant 0.84 (Table 3). When the measures of economic status are added, the coefficient remains significant at −0.61 (Table 5). The change from a positive to negative coefficient suggests that if it were not for the greater economic status of extended families, women in nuclear families would be more likely to use delivery assistance. Thus, these results provide support for Hypothesis 2b that economic status is indeed a mediator of patrilocal extended-family advantages.
The other two significant differences by family structure in physical violence and meat consumption are due in part, but not entirely, to the mediating effects of economic status and decision-making power. When the model of physical violence includes economic status, the coefficient for patrilocal extended-family living is reduced from −1.14 (Table 4) to −0.87 (Table 5). Thus, economic status explains only some of the patrilocal extended-family advantage in physical violence. Similarly, when the decision-making measure is added to the model of meat consumption, the coefficient for patrilocal extended-family living increases from −0.82 (Table 3) to −0.72 (Table 5). The coefficient is closer to zero, suggesting that decision making is operating in the expected direction. Women in nuclear families have greater decision-making power, which in turn increases the likelihood of meat consumption. However, this difference is small, indicating that the effect of family structure on meat consumption is operating largely through other pathways. Thus, this result provides only weak support for Hypothesis 2a that decision making is a mediator of nuclear-family advantage.
What Are the Implications of Going Nuclear for Women’s Health?
Figure 1 presents how health outcomes in the population of young women would change if the percentage of young women residing in nuclear families rose from the 1992–1993 level of 35.7 % to the 2005–2006 level of 43.5 %, and then to a potential future level of 60 %. Predicted probabilities from the main models in Tables 3 and 4 indicate that the nearly 8 % rise in nuclear families from 1992–1993 to 2005–2006 translates into changes of 2 % to 3 % in the health outcomes that differ significantly by family structure. If the percentage of young women in nuclear families rose further from the 2005–2006 level of 43.5 % to the hypothetical level of 60 %, the models predict additional changes of about 8 % in the health outcomes. Specifically, if the percentage nuclear rose from 43.5 % to 60 %, the percentage of young women using antenatal care would fall by 7.4 %, delivery assistance would fall by 8.1 %, and dairy consumption would fall by 7.8 %. At the same time, such a rise in nuclear families is associated with an increase of 10.0 % in physical violence and 7.8 % in meat consumption. Overall, the results suggest that the trend toward living in nuclear families will result in small changes in young women’s health, but most of the changes are not beneficial.
Discussion and Conclusion
The percentage of young married women residing in nuclear households did increase from 1992 to 2006. This result suggests that young women in India are indeed “going nuclear.” However, despite this trend, the majority of young women remained in patrilocal extended families throughout the period. This analysis confirms that patrilocal extended-family living is indeed relevant for young married women in India. The continued relevance of such families is consistent with analyses of other non-Western countries, which found substantial proportions of people residing in extended families in the 1990s and early 2000s (Bongaarts 2001; Ruggles and Heggeness 2008). However, the trend toward nuclear families found here is not consistent with a broader pattern of change. Nuclear families became more prevalent over time in only four of the 15 non-Western countries in the Ruggles and Heggeness (2008) sample.
Contrary to the literature, young women living in nuclear families do not have better health than those in patrilocal extended families. For three of the eight outcomes examined, there is not a significant difference in young women’s health by family structure. Of the five outcomes for which there is a significant difference, four of the advantages are for patrilocal extended families, and only one is for nuclear families. Young women living in patrilocal extended families are more likely to use antenatal care and delivery assistance, more likely to consume milk and curd at least weekly, and less likely to report physical violence. On the other hand, young women in nuclear families are more likely to consume meat, fish, and eggs on a weekly basis.
These results support the role of economic status as a mediator of the patrilocal extended-family advantage, but also point to other pathways as well. The higher economic status of patrilocal extended-family households accounts for advantages in antenatal care, delivery assistance, and dairy consumption. However, economic status does not fully account for the advantage in physical violence. It appears that in India, it is more common for in-laws to suppress violence than it is for them to instigate it. The null effect for sexual violence further supports the role of family members in suppressing violence when they are able. Sexual violence may often take place in privacy at night when family members are not monitoring and less able to prevent it.
Other explanations may also account for these patrilocal extended-family advantages. Part of the explanation may lie in limitations of the models and family structure measure. As noted earlier, the models may not fully adjust for selection into family type. Further, the boundary between family types, including extended and nuclear households, is often ambiguous (Brown and Manning 2009; Lloyd 1999). Family members can share resources and influence decision making across households, especially when they live nearby (Das Gupta 1999; Jeffery et al. 1989; Seymour 1999). Thus, even young women residing in nuclear families may be exposed to the influence of the patrilocal extended family.
Another explanation may lie in changes in family relations over time. Put simply, the patrilocal extended family may have been kinder and gentler for daughters-in-law in the 1990s and 2000s than it was in the past. Many studies that describe the plight of daughters-in-law in patrilocal extended families are based on ethnographies from the 1980s, 1970s, 1960s, and 1950s or earlier (Jacobson and Wadley 1977; Jeffery et al. 1989; Madan 1965). Ethnographers who revisited their field sites describe changes in family relations over time (Seymour 1999; Wadley 1994). Minturn (1993), for example, noted a few cases in 1950 of daughters-in-law lapsing into comas because of the stress and hardship of patrilocal extended-family living, but such behavior was unheard of in 1975. Further, in the 1980s, Caldwell et al. (1984) noted that the increasing strength of the marital bond was leading to a more even distribution of food in the patrilocal extended family. Others have suggested that young women’s new high levels of education and earning power have shifted the balance of power toward daughters-in-law (Saavala 2001; Vera-Sanso 1999).
These results are more consistent with previous studies that examine the effect of family structure on adult health in Japan and among Asian Americans (Takeda et al. 2004; Turagabeci et al. 2007; Walton and Takeuchi 2010). These studies did not focus on daughters-in-law, but they did examine the effect of patrilocal extended-family living on adult women and men of all ages. In Japan, Takeda et al. (2004) found mixed results: women who lived with in-laws were less likely to exercise and get a health check-up, but also less likely to smoke. The other two studies found no significant differences in a variety of health measures by family structure (Turagabeci et al. 2007; Walton and Takeuchi 2010).
These findings also present an intriguing counterpoint to research on family structure and health in the United States. Several American studies find that family structure, including marital status, affects well-being of children (Astone and McLanahan 1991; Bronte-Tinkew et al. 2009; McLanahan 1985; Thomson et al. 1994) and young women (Albrecht et al. 1994; Kimbro 2008; Waite and Gallagher 2000). These studies present the intact nuclear family as the gold standard of family structure. The present study highlights that in other contexts, an intact nuclear family is not necessarily the standard. Patrilocal extended families may be at least as beneficial as nuclear families in contexts with a preference for such living arrangements or even more beneficial. Similar to the studies on adults described earlier, studies that examined the impact of extended versus nuclear families on children’s health in non-Western countries either found no difference or found an extended-family advantage (Bronte-Tinkew and DeJong 2004; Gage et al. 1997; Griffiths et al. 2002). In keeping with studies from the United States and elsewhere, however, I also found that the economic advantages some family structures confer are part of the link between family structure and well-being (Bronte-Tinkew and DeJong 2004; McLanahan 1985; Thomson et al. 1994).
As one of the first direct examinations of family structure and young married women’s health in a patrilocal extended-family context, this study presents an important contribution. However, this analysis has three main limitations that should be addressed with future research. First, this study did not provide an ideal adjustment for selection into family structure. Gathering more information on family members’ characteristics and interactions inside and outside the household will allow for a more rigorous adjustment of selection, as well as more nuanced measurement of family structure. Such information would include the strength and nature of family ties across households, whether parents-in-law are alive and living nearby, and the existence and characteristics of husbands’ brothers.
Second, the health outcomes examined here are limited to physical health. It is possible that the stresses of patrilocal extended-family living are not captured in such measures. Future research should examine the effect of family structure on other aspects of young women’s health. Measures of subjective well-being, self-rated health, and mental health (including stress and depression) may be particularly relevant.
Finally, given the constraints of cross-sectional data, this study presents a static view where young women are observed in one family type at one point in time. In reality, this is a dynamic process in which women usually start in a patrilocal extended family and later transition into a nuclear family. Previous research finds that the timing of trajectories across family types, as well as the number and nature of transitions, affect well-being (Fomby and Cherlin 2007; Meadows et al. 2008; Williams et al. 2011). Further, early-life conditions affect health later in life (Palloni 2006; Wen and Gu 2011). Thus, future research should collect longitudinal data and explore the impact of trajectories across family types on women’s health.
The author would like to thank Arland Thornton, Tim Liao, and two anonymous reviewers for their helpful comments. A previous version of this article was presented at the 2011 Annual Meeting of the American Sociological Association in Las Vegas, NV.
The pattern of extended families having higher economic status than nuclear families may be unique to contexts with a cultural preference for living in extended families. Van Hook and Glick (2007) noted that in the United States and some Latin American countries, where there is no cultural preference for patrilocal extended-family living, extended families are formed because of extreme economic need and, thus, are worse off than nuclear families.
There are two exceptions to this statement. The interactions of wave and family structure are significant for physical violence and meat consumption. However, the substantive interpretation is still identical across waves. Further, the questions on physical violence differed substantially between waves. Thus, it is likely that the difference between waves in the effect of family structure on physical violence is due to the change in measurement.
International recommendations provided by the World Health Organization (WHO) recommend four antenatal check-ups. However, national Indian policy—namely, the Reproductive and Child Health Programme—recommends three check-ups.
The physical and sexual violence measures are available for the 23,964 women selected for the family relations module, which make up 80 % of the analytical sample. This raises the question of whether the data are representative. Only a limited sample of women was selected for the family relations module in order to ensure strict confidentiality (IIPS and Macro International 2007). Only one woman per household (rather than all eligible women) was selected, and she was interviewed only if complete privacy was attained. If women who experienced violence were less able to secure private interviews in extended households, the sample could be biased. However, there appears to be little to no bias in the sample. Less than 1 % (0.6 %) of women selected for the family relations module were unable to complete the interview because of a lack of privacy (IIPS and Macro International 2007). Further, the selected women who were not interviewed because of privacy were “virtually identical” to those who completed the interviews in terms of age, residence, education, religion, caste/tribe, and wealth. Reporting of domestic violence, on the other hand, differs by age, education, and wealth. Many of the other health outcomes also further limit the analytical sample. However, these limitations restrict the sample in ways that should not bias the data. The samples are restricted to women for whom an outcome is recent and relevant (e.g., antenatal care and delivery assistance), accurate (e.g., underweight), or available (e.g., anemia).
This eigenvalue and percentage of variance explained for the principle components analyses of household wealth and decision-making power refer to the 2005–2006 wave only. The survey waves included different items on household wealth and decision-making power. Therefore, the principle components analyses were run separately by wave.
Accessibility of a health facility refers specifically to physical location. The question that this measure is derived from is, “When you are sick and want to get medical advice or treatment, is the distance to the health facility a big problem, a small problem, or no problem?” Within both rural and urban areas, women in patrilocal extended families are significantly more likely than those in nuclear families to say that the distance to the health facility is not a problem. This pattern may be another reflection of patrilocal extended families’ greater economic status. The homes of patrilocal extended families may be located in more central locations near health facilities, schools, roads, and other facilities.