Marriage Migration and Inequality in India, 1983 2008 Smriti Rao Kade Finnoff Most scholars of migration study the flows of people between countries. Official UN estimates suggest that in 2013 about 231 million people were living in countries other than their country of birth. While large, this number is comparable to estimates of internal migration in countries such as China or India. Migrants in China totaled between 150 million and 440 million people in 2011 (Chan 2013), while in India recent estimates suggest that up to 400 million people could be internal migrants (UNESCO 2013). Thus, internal migrants in China and India alone may be triple the total number of global international migrants. They are a significant aspect of the socioeconomic transformations these countries are undergoing. In both countries, large numbers of these internal migrants are women. There is a growing body of research on female migrants in China, who for the most part are seeking employment in cities (Huang 2001; Liang and Chen 2004; Chan 2013). In India, a large and growing proportion of migrants are temporary or circular migrants (Breman 1996; Garikpati 2008; Deshingkar and Akter 2009). Among India s permanent migrants, women make up 83 percent (208 million) of the total according to 2008 National Sample Survey (NSS) data, up from 75 percent in 1983. Yet there is little research on this group, primarily because most women state their reason for moving as marriage (among the few studies see Premi 1980; Rosenzweig and Stark 1989; Bhattacharya 2000; and Fulford 2013). The assumption appears to be that marriage migration is explained by unnamed socio-cultural factors (Kundu 2009) and is therefore less relevant to understanding economic change in India today. Until recently, similar assumptions resulted in the neglect of research on female international migrants, who were assumed to be tied to fathers and spouses. Feminist researchers have since found that women s international POPULATION AND DEVELOPMENT REVIEW 41(3): 485 505 (SEPTEMBER 2015) 485 Published on behalf of the Population Council by Wiley-Blackwell.
486 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a migration has important economic implications for Western labor markets, for developing-country growth rates via remittances, and for understanding how and why cultural notions of femaleness and maleness do or do not change in source and origin countries (Parreñas 2001; Donato et al. 2006; Benería, Deere, and Kabeer 2012). In the present study of marriage migration in India, we find little evidence that increases in marriage migration are caused by disguised economic migration by women. We hypothesize that the changing patterns of marriage by socioeconomic status are a more likely cause. In particular, we are interested in the intersection between economic inequality and marriage migration. Customs related to marriage exogamy in much of India mean that large numbers of women have always migrated to their marital homes after marriage, so we are interested not only in the prevalence of female marriage migration, but also in its increase over time. Given that rates of male migration have remained largely unchanged, 1 it is unlikely that the increase in female migration is a result of more women following mobile spouses. We approach marriage as a site of social production and reproduction, including the production and reproduction of economic inequalities (see Lee 2012 for a recent example of such an approach). We thus see changing marriage migration trends as shaping and shaped by wider changes in the economic context. We examine the socioeconomic correlates of marriage migration from 1983 to 2008 and decompose it by both marriage distance and sectoral streams. Our most significant finding, robust to a number of different specifications, is that marriage migration is increasingly most likely to occur among households with lower levels of per capita consumption. A sectoral decomposition of marriage migration suggests, furthermore, that urban inequality is an important driver of female marriage migration. Most studies on inequality in India report widening inequality both between urban and rural areas and within urban areas after the 1980s, and the marriage migration data help us to see the effect of this inequality on marriage. Internal migration in India The main sources of data on migration in India, the NSS and the census, are unable to capture temporary or circular migration of less than 6 months. As a result, the literature, including the present study, concentrates on permanent, rather than circular, migration. According to the NSS, permanent internal migrants have risen slightly as a share of the Indian population, increasing from 23 percent in 1983 to 29 percent by 2007 08. Some 87 percent of this increase was accounted for by the growing share of female permanent migrants and, in particular, female marriage migrants. Marriage migrants increased from 76 percent to 84 percent
S m r i t i R a o / K a d e F i n n o f f 487 of all female migrants over this period, even as other forms of female permanent migration either declined or remained unchanged in both absolute and relative terms. The share of women migrating to follow family members has fallen from 15 percent in 1983 to 11 percent of all female migrants in 2007 08, which corresponds to a slight increase in absolute numbers from 18 million to 23 million over this period. The proportion and numbers of women migrating for studies have increased slightly from 0.5 percent to just under 1 percent of female, working-age married permanent migrants (an increase of 1.6 million), but this is an almost insignificant increase compared to the increases in marriage migration. The proportion of women who report migrating for economic reasons is also small and declining, dropping from 2.6 percent in 1983 to 1.1 percent in 2008. 2 Based on this data, India has lower overall migration rates than many other Asian countries, prompting some attempts to explain this by referring to the role of caste networks in India in reducing mobility, the government s hostility to urban in-migrants, and/or the relative underdevelopment of laborintensive manufacturing in India (Munshi and Rosenzweig 2007; Kundu 2009). The explanation referring to the underdevelopment of manufacturing appears to be supported by considerable evidence on the exclusionary nature of Indian growth. While the 1980s was a period of high growth with stagnant inequality, the post-1993 period has been marked by rising growth and also, most studies agree, by rising economic inequality post-liberalization (Sen and Himanshu 2004; Chaudhari and Ravallion 2006; Vakulabharanam and Motiram 2012; Thorat and Dubey 2012). While results for rural inequality differ somewhat across studies, inequality between rural and urban areas has clearly risen, as has inequality within urban India. Vakulabharanam (2010) uses a class-based analysis to show that the most significant beneficiaries of post-1993 economic growth have been those in highly skilled (and highly educated) service occupations in urban areas, with the majority of urban workers losing out. In rural areas he finds that the rentier classes and very large farmers have gained while rural peasants and agricultural laborers have clearly lost ground in relative and absolute terms. The enclave nature of urban growth means that the rural dispossessed have little prospect of finding alternative livelihoods in the city. This conclusion is confirmed by studies of Indian economic migrants. Permanent economic migrants in India are largely working-age men who move from rural to urban areas or within urban areas. Rural rural migration accounted for only 32 percent of all male migration in 2007 08, while it accounted for 70 percent of female migration in the same period (NSSO 2010). Members of higher castes and those who are better educated are also more likely to engage in economic migration within and to urban areas in India (Dubey, Palmer-Jones, and Sen 2006; Mitra and Murayama 2008; Vakulabharanam and Thakuratha 2012). Kundu and Sarangi (2007)
488 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a also find that these mostly male urban in-migrants tend to be relatively well-off, although they are neither the richest nor poorest, and argue that urban growth in India is exclusionary along economic and non-economic markers of status. 3 The question we address is whether and why the class, caste, and other correlates of female migration differ from those for male migration and whether these patterns have been shifting over the years. Disguised economic migration One initially plausible explanation for rising rates of female marriage migration is that it is the result of disguised economic migration (Krishnaraj 2005). This would mean that married women move with their husbands with the intent of working at the new destination, much as the husband does, but then report their move as being for marriage. There are two ways to investigate this possibility. First, if workforce participation was driving female marriage migration, we would expect to see higher economic activity rates for married female migrants. Given the wide range of economic activities women engage in, we focus on married working-age women who report that their principal economic activity was attending to domestic duties (in some cases alongside free collection of goods ) that is, women who do not report the various forms of self and paid employment that are generally considered to be participation in the workforce. Migrant women do report slightly lower rates of attending to domestic duties (and thus higher rates of other kinds of workforce participation), but the gap between migrants and non-migrants on this metric has shrunk over the NSS rounds. In 1983, 62 percent of female migrants reported that they mainly attended to domestic duties, as compared to 66 percent of non-migrants. By 2007 08 these figures were 70 percent and 73 percent. The clearest pattern in these data is the increasing proportion of married women in both groups who report that their principal activity is attending to domestic duties. There is thus no compelling evidence that increases in marriage migration rates can be correlated with increasing economic activity rates for marriage in-migrants, since these rates have been falling. Second, if marriage migration is in fact disguised economic migration, then the actual journey made by the man and woman in the household would be identical they are both, for example, moving to the same destination, and we would expect the underlying geographic and socioeconomic correlates of the move to be similar if not identical. However, according to the NSS data, only 5 percent of female working-age marriage migrants in 2008 had spouses who were also migrants, down from 7 percent in 1983. We also find a negative and strengthening statistical correlation between the husband being a migrant and the in-migration of his wife. These results suggest that
S m r i t i R a o / K a d e F i n n o f f 489 there are few couples within the NSS dataset of permanent migrants who make the joint economic journey described above. Based on the evidence so far, it is hard to argue that the increases in female marriage migration rates within the NSS data are a result of increases in disguised economic migration by women. The explanation, it seems, must lie elsewhere. The nature of marriage migration in India The literature on Indian marriage, and therefore on marriage migration, takes for granted some stylized facts that we also adopt here. First, marriage in India is still overwhelmingly arranged by parents and other family members. Second, most marriages in India are within-jati (sub-caste) and thus constrained by the geographical distribution of jati members. Third, the vast majority of Indian marriages are virilocal that is, the wife moves to the husband s household (usually also the location of the husband s parents). Thus marriage and migration for many Indian women go hand in hand. From an economic perspective, improvements in transportation and communication infrastructure have lowered costs of migration over time, accounting for at least some of the increases in the migration rates of interest to us. (In the regression analysis that follows, we use time dummies to control for those improvements.) The search process in an arranged marriage also implies that women living in larger population centers, such as urban areas, would be more likely to find a spouse locally and not have to move. A possible demographic explanation for rising rates of marriage migration derives from the rising ratio of males to females at birth in India. Since Indian men usually marry younger women, this would create a shortage of girls of marriageable age, increasing the need for in-migration of brides. However, any impact of rising sex ratios has been lessened by India s current age structure, in which younger cohorts are larger than older ones (Rao 1993). As a result, women in these younger cohorts are still more numerous than the men they seek to marry in the older cohort. Sex ratios have also become more similar over time across districts in India, making it less likely that they would be the primary drivers of marriage migration (Basu 1999; Fulford 2013). Nevertheless, we control for the impact of sex ratios in the analysis below. From an economic perspective, marriage constitutes a transfer of labor from the natal to the marital home (Agarwal 1994). Given India s cultural and economic diversity, both domestic and non-domestic work may require specific local skills that increase the demand for local brides (Ramamurthy 2011). Thus, in regions where female labor force participation is high and women s skills are required for specific agricultural tasks, local women with such skills are apt to be preferred over outsiders, and marriage in-migration is thus less likely. The agrarian crisis in much of rural India has been linked to an
490 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a ongoing feminization of agriculture, whereby women become custodians of the failing family farm, while men seek better-paying non-agricultural work (Jackson and Rao 2009). This would again increase the demand for women with the willingness and ability to perform agricultural work in rural households. In both cases, female in-migration would be less likely in districts with higher female relative to male labor force participation rates. Another hypothesis relies upon changing norms of village exogamy in rural India. Anthropologists and sociologists writing about rural society in the 1950s and 1960s argued that village exogamy was more common in North India, as was hypergamy, where families traditionally married their daughters into better-off families within the same jati. In contrast, village and kin endogamous marriages were more common in South India (Karve 1965; Trautmann 1995; Dyson and Moore 1983). More recently, however, some field studies have documented increases in kin and village exogamous marriages in the South (Agarwal 1994; Kapadia 1995). Rahman and Rao (2004) found that village endogamy was uncommon in the areas they studied and the average distance between natal and marital family (marriage distance) did not vary significantly between rural areas of North and South India. Thus changing norms of exogamy, particularly in the South, could be one explanation for increases in rural marriage migration (Basu 1999). As we will see, the NSS data show an increase in marriage migration in the South, but they also show a much larger increase in the North where village exogamy was already common. There is far less research on how norms of exogamy play out in urban India. Does village exogamy translate into city exogamy? Given the wide inter- and intra-state variations in language and culture, this is unlikely to be the case. We find below that it is urban-to-urban and rural-to-urban marriage migration that are driving the increases in rates of marriage migration. Changing norms of village exogamy in the traditional sense thus do not help us understand these shifts. Status, inequality, and marriage migration Marriage in India is an extremely important means of creating, undermining, or strengthening class and/or caste alliances and of signaling and consolidating social and economic status (e.g., Bloch, Rao, and Desai 2004). Traditional economic models of assortative mating predict that richer households may have to search across a wider territory to find a spouse within an equally rich household (Becker 1981). On the other hand, our study, as well as two prior analyses of marriage migration in India, find the opposite effect: richer families are less likely to engage in marriage migration (Rosenzweig and Stark 1989; Bhattacharya 2000). The more detailed study by Rosenzweig and Stark argues that virilocal marriage practices have evolved as an insurance
S m r i t i R a o / K a d e F i n n o f f 491 mechanism in the absence of other well-developed insurance markets. Since in-laws can provide loans to smooth out income shocks such as those due to the weather, the daughters of poorer households that cannot afford access to formal insurance markets are more likely to be married to men in more distant, ecologically distinct areas. This argument requires that loans and transfers between families flow both ways from natal to marital households as well as the other way around. The latter seems unlikely based on what we know from sociologists about Indian marriage networks. On the basis of recent survey data, Fulford (2013) finds no evidence of post-marriage transfers between households in either direction, throwing doubt on this possible explanation. The vast literature on dowry and marriage in India suggests that income and wealth are now explicitly important variables in marriage negotiations (Rao 1993; Agarwal 1994; Anderson 2003). While sub-caste or jati remains a crucial marker of both economic and non-economic status in India, there is growing evidence of class-based differentiation within castes. As discussed earlier, most of the recent research on this subject is focused on rising economic inequality in India, for which there is considerable evidence, particularly for urban rural differentials and urban inequality in the period from the 1993 NSS to the 2004 05 NSS (Sen and Himanshu 2004; Jayadev, Vakulabharanam, and Motiram 2011; Zacharias and Vakulabharanam 2011). We hypothesize that greater stratification by class (defined in the Weberian sense) also implies fewer cross-class alliances. Richer families are able to negotiate local alliances, while those that are worse off are forced to widen their search range, increasing rates of marriage migration by women. Empirical analysis Our empirical analysis relies on the four most recent rounds of the NSS Employment-Unemployment Survey that have asked detailed questions about migration. These are the 38th round conducted in 1983, the 43rd round in 1987 88, the 55th round in 1999 2000, and the 64th round in 2007 08. 4 This means we have data for the 1980s, no data for most of the 1990s, and data for the twenty-first-century post-liberalization period. In each case the NSS asks whether the current place of enumeration of each household member differs from the last usual place of residence and, if so, the reason for leaving the last usual place of residence. The answer to the first question helps us define the category of migrants those who answer yes; while the category of marriage migrants comprises those who give marriage as the reason for migration. The descriptive statistics below are reported using NSS population weights. We then included in the NSS dataset statelevel averages for the sex ratio of the population under six years of age from successive rounds of the census (1981 for round 38, 1991 for round 43, 2001
492 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a for round 55, and 2011 for round 64). The adult sex ratio is endogenous to migration, so we use the sex ratio only for the under-6 population. We also used the NSS household-level data to calculate (separately for rural and urban areas) state-level average per capita income as well as the rural and urban Gini coefficients for each state. Bell and Muhidin (2009) note that rates of migration can be influenced by demographic shifts if there are increases in the relative size of the group most likely to migrate. To minimize this problem, we restrict our analysis to the married population of working age (15 64 years), eliminating those who are generally too young or old to migrate. Note that when women report being marriage migrants, they are already in their marital homes. Female marriage migration statistics thus represent in-migration to the marital household and region. Given that the NSS does not ask about the woman s natal family, our information on the marital household into which the woman has migrated is limited to income, occupation, and other socioeconomic characteristics. We do, however, know the educational attainment of the woman as well as her employment status before migration. Using the NSS data on the sex and relationship of each member of the household to the household head, we were also able to uniquely identify the male spouse of 98 percent of working-age married women and thus also identify his age, level of education, and migration status. To control for the fact that local brides may be in high demand where female employment is high (assuming that region-specific skills are required for agricultural work in particular), we used NSS data on employment to calculate the share of women and men employed in each state (separately for rural and urban areas) and included the ratio of the share of women employed to the share of men employed for that region as one of our control variables. We also control for the effect of different marriage norms between North and South by including a dummy for northcentral India, or the Hindi heartland states of Rajasthan, Punjab, Haryana, Chandigarh, Delhi, Himachal Pradesh, Uttar Pradesh, Bihar, Madhya Pradesh, and, in the 64th round, Uttaranchal, Chhattisgarh, and Jharkhand. The southern region (Andhra Pradesh, Karnataka, Tamil Nadu, Kerala, and Pondicherry) had higher household consumption and higher shares of urban populations but also higher Gini coefficients than the northcentral region in each of these four rounds. Real expenditure was derived by deflating monthly per capita household consumption expenditure by the appropriate value of the consumer price index (CPI) (agricultural laborers) for rural households and CPI (industrial workers) for urban households. These values are reported in 2011 12 Rs. Median real monthly per capita expenditure (in 2011 12 Rs) for the population has risen slowly over these four rounds, but the mean real monthly per capita expenditure fell in 1987 88 and rose again in 1999 2000 and 2007 08 (Table 1).
S m r i t i R a o / K a d e F i n n o f f 493 TABLE 1 Monthly per capita expenditure in four NSS rounds Monthly Real monthly per capita per capita expenditure expenditure Round (year) (Current Rs) (2011 12 Rs) 38 (1983) Mean 144.4 1,278.3 Median 100.7 835.4 43 (1987 88) Mean 384.2 1,150.0 Median 138.8 836.8 55 (1999 2000) Mean 525.6 1,183.2 Median 423.3 890.0 64 (2007 08) Mean 843.4 1,484.2 Median 644.6 1,074.2 NOTE: Number of weighted observations by round: 38 = 85,627; 43 = 138,906; 55 = 130,202; 64 = 134,229. SOURCE: Authors calculations based on NSS data. Descriptive statistics Female marriage migrants accounted for 71 percent of the married workingage female population in 2007 08, up from 55 percent in 1983 (see Table 2). This increase was greater in the northcentral states (+28 percentage points between rounds 38 and 64 compared with +10 in the rest of India), and in rural areas (+21 percentage points). The result for rural areas is expected since the greater search options within a large urban population increase the likelihood of finding a spouse within the city. However, the share of the rural population has fallen over the years, and in absolute terms the number of marriage migrants to and within urban areas has risen by 158 percent over this period while the number migrating to and within rural areas has risen by 110 percent. Urban marriage migration is thus driving the increase in marriage migration over time. Interestingly, while Hindus had higher levels of migration, the rate of increase did not differ between Hindus and non-hindus (+19 points each). While there is evidence that jati-like hierarchies exist within non-hindu communities in India, this widespread increase in marriage migration suggests that it may be driven by factors other than jati membership alone. In contrast to previous findings for male permanent economic migrants across caste groups, rates of migration were higher among Dalits while the increase in rates was greatest for Tribals. Looking across expenditure quintiles in each round, rates of marriage migration among women were quite similar across quintiles in 1983 and 1987 88.
494 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a TABLE 2 Share of marriage migrants among married working-age women in four NSS rounds (percent) 1983 1987 88 1999 2000 2007 08 India 55 61 65 71 By rural/urban Rural 57 65 71 78 Urban 39 44 47 51 By region Northcentral 53 65 74 81 Rest of India 53 57 56 63 By religious group Hindu 54 62 66 73 Non-Hindu 45 54 59 64 By caste group Tribal 52 57 59 72 Dalit 57 66 70 75 Other 52 60 65 70 By expenditure quintile Lowest 54 62 66 75 Second 54 61 67 73 Third 53 61 67 72 Fourth 53 61 67 70 Highest 51 59 62 65 By woman s educational attainment Illiterate 55 64 69 77 Primary or less 50 56 64 70 Secondary or less 44 50 58 66 Above secondary 36 40 43 52 SOURCE: Authors calculations based on NSS data. However, while the marriage migration rates of the top quintile increased by 14 percentage points over the four rounds, up to 65 percent in 2007 08, the rates for the bottom three quintiles increased by almost 20 percentage points, the increase being largest for the bottom quintile. The fourth, or second richest quintile, saw a slightly smaller increase of 17 percentage points. Thus, by 2007 08 there was a negative relationship between expenditure quintile and the rate of female marriage migration, with women in the bottom three quintiles much more likely to report being marriage in-migrants. A negative relationship also existed between female educational status and the rate of marriage migration by women within each round, although this negative relationship was strong even in 1983 and changed little over this period. To summarize, unlike most male migration, female marriage migration seems to be more likely among women who are less privileged in terms of household consumption and educational attainment.
S m r i t i R a o / K a d e F i n n o f f 495 Regression analysis of marriage migration We next use regression analysis to test whether the negative class effect on migration that emerged in the descriptive statistics persists after controlling for age, education, caste group, rural population centers, average sex ratios in the state, the ratio of female to male employment shares, and time and state fixed effects. We conduct a logistic regression with the dependent variable being a binary variable that takes the value 1 if a woman has in-migrated for marriage and 0 if she has not. We restrict the regression to married working-age women and merge the data from the four NSS rounds, creating a time dummy to represent each of the first three rounds. 5 We do not use sample weights since we are combining data across all four rounds. Because traditional norms of marriage exogamy do not translate as easily into urban contexts, we split the sample into rural and urban and report regression results for each separately. We include as independent variables the woman s age in years, the caste group of the marital household (Dalit or Adivasi, the comparison being with all other castes 6 ), the rural and urban Gini coefficient for the state in which the marital household is located, and the average sex ratio for the state. We also include dummies for three educational categories: at least some primary education, at least some secondary education, and at least some post-secondary education, the comparison category being no schooling. We include the log real per capita expenditure of the marital household and time dummies for the first three rounds, the comparison category being the 2007 08 round. We introduce interaction terms between the log of real consumption and the time dummies to capture any changes in the impact of economic status on female marriage migration over time. It is well known that logit coefficients and the associated odds ratios do not accurately capture the effect of interactions between independent variables, and average marginal effects (AMEs) are recommended in such situations (Norton, Wang, and Ai 2004). We use STATA s margins command to compute the average marginal effect of a unit change in log real expenditure on the probability of marriage migration at different values of the time dummies that represent the four rounds, and we rely on the AMEs rather than coefficients for our analyses of variables that are included in these interaction terms. Robust standard errors are reported, with the villages (or FSUs in the NSS sampling frame) being the cluster unit. Supplemental Table S1* shows the extremely large effect of being part of the northcentral region of India as hypothesized by Fulford (2013). The reported coefficients correspond to odds ratios of 497 for rural and 55 for urban India. Thus being part of the northcentral region increased the likelihood of marriage migration by 497 times for rural Indian women and by 55 times for urban Indian women. * Supplemental tables are available at the supporting information tab at wileyonlinelibrary.com/journal/pdr.
496 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a Turning to the impact of economic status, the average marginal effects for both rural and urban samples confirm that there has been a steady increase in the negative relationship between the economic status of the marital household and the likelihood of female marriage in-migration (Table S1). For the urban sample, the coefficient was negative but strengthening across the period. For the rural sample, the impact of household consumption levels went from positive to negative, indicating a decline in the probability of female marriage in-migration. Among rural Indians (who are also poorer on average), this negative correlation between consumption and marriage migration is thus more recent than it is for the richest Indians. In terms of size, the AME of log expenditure for rural India went from a small positive coefficient of 0.01 in 1983 to a negative coefficient of 0.061 by 2007 08, holding all other variables constant at their mean values. Thus by 2007 08, a unit increase in log real monthly per capita expenditure (MPCE) in rural India decreased the likelihood of marriage migration by 6 percentage points, a 7.5-point change from the marginal effect in 1983. Thus the same one-log MPCE unit gap between rich and poor households generated an additional 7.5 percentage points of marriage migration on the part of the poorer household in 2007 08, or about 28 percent of the 26-point increase in marriage migration rates within the poorest quintile of rural households. For the urban sample, the coefficient increased by almost 0.07 units, or 7 percentage points. Thus a one-unit gap in the expenditure of poor and rich households contributed an additional 7 percentage points of marriage migration in 2007 08, around a third of the 21-point average increase in marriage migration rates among the poorest quintile of rural households over this period. Women with the highest level of education were less likely to migrate in both rural and urban areas, although the results for other education levels were ambiguous. Adivasi women are also less likely to migrate, perhaps reflecting the legacy of more endogamous marriage traditions. In terms of household occupation categories, agricultural households (both laborer and self-employed) in rural areas, and households whose primary occupation was reported as self-employed in non-agriculture in both rural and urban areas, were more likely to report in-migration of brides than those households engaged in non-agricultural regular or casual wage work. We examine this effect more closely below when we decompose migration by sectoral stream. The ratio of female to male employment share was negatively and statistically significantly related to the probability of marriage migration for rural but not for urban areas. This outcome fits the hypothesis that local women in rural areas are most likely to be highly valued for their work skills. The rural results differed from the urban in one significant way: the state-level rural Gini coefficient was negatively related to the probability of
S m r i t i R a o / K a d e F i n n o f f 497 migration, while for urban areas the relationship was positive. Again, this is a result we return to in our sectoral decomposition. Decomposing marriage migration by distance The NSS allows us to determine whether female migration was within the same district, to a different district in the same state, or to a different state, giving us an imprecise estimate of marriage distance. As seen in Table 3, migration to a different district in the same state accounts for the largest percentage increase over this period, although a majority of marriages are still within the same district. Cross-state migration has seen little change, remaining at about 5 6 percent of the total across all four rounds. Given the inter-state variations in language, caste, and customs in India, it is not surprising that migration to a different state has not greatly increased. The results from an ordered logistic regression on the marriage distance (Table S2) are very similar to those obtained earlier. The overall marginal effect of income again has a negative effect on the probability of migration. In the urban case this is true across the entire period, while in the rural case this switches from a positive to a negative effect in the 1987 88 round. The results on the Gini coefficients are also the same, with greater rural inequality discouraging rural rural marriage migration, but greater urban inequality encouraging it. Given that the marriage distance analysis confirms the robustness of our initial results, we turn to a sectoral decomposition of marriage migration. TABLE 3 Numbers of marriage migrants among married working-age women in four NSS rounds (millions) Percent change (1983 to 1983 1988 1999 2007 08 2008) By marriage distance Same district 64 78 101 121 89 Same state, different district 15 18 33 47 213 Different state 5 5 8 11 120 Total 85 101 142 179 111 By sectoral stream Rural rural 67 82 111 138 106 Rural urban 9 10 15 20 122 Urban urban 5 5 10 13 160 Urban rural 4 5 6 8 100
498 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a Sectoral decomposition of marriage migration Over time, rural rural marriage migration has been growing slightly more slowly than overall marriage migration, but still accounts for the largest share of marriage migrants. However, rural urban and urban urban marriage migration have grown faster over the 25 years studied here and now account for 18 percent of working-age marriage migrants, as compared to 16 percent in 1983 (Table 3). Urban rural migration, unsurprisingly, continues to be small. What makes the sectoral decomposition interesting, though, is that it reveals the distinct socioeconomic characteristics of each stream. The very small numbers of urban rural migrants are much more privileged and better-educated than the comparison group of other rural women (Tables 4 and 5). In the rural social hierarchy, it would appear that those who do not report migrating for marriage are the next best off, with rural rural migrants marrying into the poorest rural quintiles and having the lowest educational attainment. The pattern is largely repeated in urban India, with rural urban migrants seeming to occupy the bottom rungs of the socioeconomic hierarchy. Interestingly, Tables 4 and 5 suggest that the pattern for education and TABLE 4 Average monthly per capita expenditure of marital households by sectoral stream in four NSS rounds (2011 12 Rs) 1983 1988 1999 2000 2007 08 Rural rural 980 925 909 1,089 Urban rural 1,031 1,056 1,093 1,333 Non-migrant, rural 1,153 926 925 1,206 Rural urban 1,934 1,603 1,714 2,129 Urban urban 2,612 2,107 2,202 2,738 Non-migrant, urban 2,073 2,132 2,403 2,992 TABLE 5 Average educational level of female marriage migrants by sectoral stream in four NSS rounds 1983 1988 1999 2000 2007 08 Rural rural 0.22 0.25 0.44 0.63 Urban rural 0.53 0.56 0.94 1.05 Non-migrant, rural 0.25 0.32 0.52 0.74 Rural urban 0.61 0.67 0.93 1.14 Urban urban 1.16 1.25 1.53 1.72 Non-migrant, urban 0.87 0.99 1.32 1.48 NOTE: Educational status takes the value 0 for no formal education; 1 for some primary education; 2 for some secondary education; and 3 for some post-secondary education. Thus a mean value of 0.22 indicates that in 1983 the average rural rural marriage migrant had no formal education; by 2007 08 the average rural rural marriage migrant had some primary education.
S m r i t i R a o / K a d e F i n n o f f 499 income seems to diverge slightly, with non-migrants belonging to better-off households but possessing lower average levels of education. Regression analysis by sectoral stream To control for some of these overlapping effects, we conducted a similar regression analysis on the probability of being part of each sectoral stream of migrants, for rural and urban areas separately. In this regression analysis, rural rural and urban rural marriage migrants are each in turn being compared against all other rural married, working-age women. Similarly, rural urban and urban urban marriage migrants are each in turn being compared against all other urban married, working-age women. As before, standard errors were clustered by village. We also included an interaction term between the Gini coefficient and household per capita consumption to capture any differences between the impact of inequality at the regional level on the likelihood of marriage migration among women from poor and rich households (see Table S3). 7 For rural rural, urban rural, and rural urban migrants, the regression results once again show an increasingly negative relationship between the economic status of the marital household and the likelihood of female marriage in-migraton. The change in size of the average marginal effects of consumption in the case of rural rural, urban rural, and rural urban migration also remained approximately the same as discussed in Table S1. However, urban urban migrants do not appear to be differentiated from non-migrants by economic status in the same way. Tables 4 and 5 suggest that urban urban marriage migrants and non-migrants have higher average consumption levels and more education than rural urban marriage migrants. It would appear that both urban urban migrants and non-migrants are part of the entrepreneurial and managerial classes that have garnered most of the benefits of economic liberalization over this period. Indeed, urban inequality does not seem to be related to the likelihood of marriage migration among urban urban migrants. This is unsurprising given the similarity between urban urban migrants and their non-migrant peers in being among the economic beneficiaries of the last few decades, and suggests that these two groups are largely unaffected by the precarious circumstances facing less-advantaged urban workers. Urban workers have been among the losers in the unbalanced economic growth of this period (Vakulabharanam 2010), and yet the persistence of the rural urban income gap allows them to attract brides from rural areas. In what might be a social mechanism that compensates for their economic disempowerment within urban India, some of these men can find rural brides who are perceived to be marrying up in terms of income and education and thus translate their urban location into social and economic capital. These
500 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a women find themselves part of the so-called urban precariat, and it is an open question whether their lives are improved by this move, but their migration enables them to expand their family network into the city. Since most marriage migration is still within-state, this stream of rural urban out-migration also restructures the rural marriage market. The bestoff rural households attempt to attract brides from urban areas, while others marry locally. The worst-off rural households, the regression results tell us, are those whose brides in-migrate. The marginal effect on the Gini coefficient confirms that higher urban inequality raises the likelihood of rural urban marriage migration into cities and towns. A 0.01-unit change in the Gini coefficient results in a 2.4-percentage-point increase in rural urban marriage migration. Thus the lower the relative status of the poorest urban households, the more likely they are to attract brides from rural areas. Rising urban inequality over the past decades could thus generate the impetus for more marriage migration. This result is reversed for rural inequality. A 0.01-unit increase in rural inequality reduces the likelihood of rural rural marriage migration by 8 percentage points on average. One possible explanation is that unlike in the case of the worst-off urban households, the rural poor cannot draw upon a more dispossessed group for the supply of brides, given that marriage is still largely within-jati and within-district. Marriage therefore cannot be used as a strategy to counter low economic status. High levels of rural inequality may simply mean continued adherence to traditional norms, whether of exogamy or endogamy, while low levels of rural inequality may suggest greater flux in the social order and a greater likelihood of using marriage and marriage migration as a way to signal higher educational or income status. We conclude that urban inequality helps drive the trend toward greater marriage migration in ways that increase the likelihood of within-class alliances in both rural and urban India. For some families the poorest families in urban India in particular this migration may compensate for their relative economic dispossession. For the poorest rural families, this potentially positive impact is unlikely to exist. And in all cases, it is unclear whether the migrating woman s own well-being improves with the economic and noneconomic fortunes of her natal and marital families. Marriage migration and gender inequality Our dataset does not allow us to link marriage migration to women s wellbeing at the individual level, but research suggests that the effects are likely to be uncertain and contradictory. The older literature on kinship in India assumed that proximity to and continuous contact with the natal family benefited married women (Dyson and Moore 1983), while some recent studies have challenged the existence of any negative correlation between marriage
S m r i t i R a o / K a d e F i n n o f f 501 distance and women s well-being. Some field studies report a preference among women themselves for greater distance from the natal family, because it is said to minimize quarrels (Kapadia 1995). If parental obligations to care for married daughters and their children are stronger when they are closer, having a daughter marry close by might be costly (Agarwal 1994). Living close to kin may also mean greater pressure on young brides to conform to dominant patriarchal norms, while migration to a distant village or city may enable the renegotiation of such norms (Chaudhari 2007). A set of studies by Das Gupta (Das Gupta et al. 2003; Das Gupta and Li 1999) reframes this argument to focus on the extent to which women remain connected to and are considered part of their natal families even after marriage. The studies present a comparative analysis of the determinants of son preference in Asia. The authors argue that in China, North India, and South Korea very different regions except for their strong son preference women are cut off from their natal families after marriage, reducing parents incentives to invest in and value their daughters. In countries like Thailand, on the other hand, less restrictive norms allow married daughters to come and go from natal homes, regardless of distance from their marital homes, and allow daughters and parents to draw on one another s material and emotional support. Their argument is thus that marriage migration per se does not affect women s well-being as much as it affects the norms that govern interactions between women and their natal families after marriage. Of course, greater distance may reduce the frequency of such interactions in a setting where travel is expensive and household resources are constrained. If what we are seeing is higher rates of marriage migration among the poorest social groups, then it is these women who experience the greatest isolation from their natal families, while more highly educated women from higher income groups remain close to their better-off natal families. This would suggest convergence rather than divergence between intra-household gender inequality and economic inequality and a worsening of the status of young brides in poorer households. While our analysis does not allow us to reach conclusions about the relative strengths of these different effects, we recognize the need for greater understanding of how Indian families in contemporary urban contexts negotiate norms of exogamy and endogamy and what the impacts are upon women s well-being as a result. Conclusion Using NSS data on migration, we argue that marriage migration is associated with economic inequality in India, with potentially contradictory outcomes for intra-household gender inequality and the status of women within their marital homes. Research on economic inequality in India has tended
502 M a r r i a g e M i g r at i o n a n d I n e q u a l i t y i n I n d i a to look to the market and state to understand how inequality is generated and transmitted and how patterns of inequality have changed over time. This research has not, however, explored the ways in which changing family structure or changing forms of marriage may reinforce or undermine economic inequality. We find that different sectoral streams of marriage migration are clearly distinguished from each other by levels of household per capita consumption. Urban households in which women do not report migrating for marriage have the highest per capita consumption levels, while the poorest urban households are most likely to have brides who are in-migrants and, importantly, inmigrants from rural areas. Thus, poor urban households may lead precarious lives at the margins of the urban economy, but they are able to hold out the promise of access to the urban economy to families in rural India. That this access is attained through marriage and the displacement of women, rather than directly through the labor market, is indicative of the enclave nature of urban economic growth in India. For poorer urban households, marriage may yield concrete material benefits in the form of dowries, and, as the correlation with urban inequality suggests, bolster their non-economic status in ways that compensate for their relatively disadvantaged economic position. This stream of rural urban marriage migration, we posit, then has ripple effects on the rural economy so that higher urban inequality acts as a trigger for greater rural urban and ultimately rural rural marriage migration. Our analysis leads us to three conclusions. First, marriage migration in India is at least partly economic, in that it is correlated with wider changes in the Indian economy. Second, as many observers have pointed out, the urban economy exerts a tremendous pull on the rest of India. Our study confirms that this urban bias extends to marriage and the family, with marriage being one way to gain access to the urban economy through the migration of women. Third, the degree to which levels of household per capita consumption demarcate these various marriage migration streams also suggests a marriage market that is increasingly segmented by class. If cross-class marriage alliances are less common, marriage serves to reinforce rather than undermine larger patterns of class and caste inequality. Marriage may thus be considered a critical factor in the consolidation of economic inequality in India.
S m r i t i R a o / K a d e F i n n o f f 503 Notes This article was much improved by comments from Vamsi Vakulabharanam, Smita Ramnarian, Amit Basole, and by participants in the World Cultures Group, Women s Studies Research Center (WSRC), and WSRC Brandeis lecture series. We also gratefully acknowledge research support from the Mind and Society Institute, Azim Premji University, Bengaluru. 1 The rate of male migration (of all kinds) has been steady over the period we study, 1983 2008, falling slightly from 12 percent to 11 percent of all men and from 18 percent to 15 percent of married working-age men. This means that the bulk of the research on permanent migration in India has been conducted on a relatively small and shrinking proportion of the population. 2 It is possible that the NSS surveys are mis-classifying some economic migration by women as marriage migration. This is because the surveys ask migrants to provide only one reason for their move, and women may find it more socially acceptable to give marriage as the answer (Krishnaraj 2005). We investigate this hypothesis as best we can, even as we note again that the share of another form of more socially acceptable migration that of following family members has also decreased. 3 This line of research acknowledges that reclassifications of rural areas as urban across successive rounds of the census may be an underlying problem with the data. This is true of the sectoral analysis in our study as well. 4 The first three surveys are thick rounds of the quinquennial Employment- Unemployment Survey and the 2007 08 survey is a thin round of the same. 5 We also conducted the same regression analysis for the age group 20 30 only, given that their marriages are more likely to have been affected by economic conditions in the relatively recent past. The results were unchanged and are available on request. We report only the results for all working-age women here because changes in shares of married women over time in younger and older age groups make it difficult to determine which age sub-group (20 30 versus 15 35, for example) would be most appropriate to analyze. 6 The terms Adivasi and Dalit correspond to the official state categories of scheduled tribe and scheduled caste, respectively. The latter categories were first developed by the colonial state to identify members of groups most affected by caste discrimination and continue to be used by the modern Indian state as the basis for caste-based affirmative action policies. 7 As a robustness check, we also ran this regression without the interaction term for the Gini coefficient and log consumption and the results were similar. References Agarwal, Bina. 1994. A Field of One s Own: Gender and Land Rights in South Asia. Cambridge University Press. Anderson, Siwan. 2003. Why dowry payments declined with modernization in Europe but are rising in India, Journal of Political Economy 111(2): 269 310. Basu, Alaka Malwade. 1999. Fertility decline and increasing gender imbalance in India, including a possible South Indian turnaround, Development and Change 30(2): 237 263. Becker, Gary. 1981. A Treatise on the Family. Cambridge, MA: Harvard University Press. Bell, Martin and Salut Muhidin. 2009. Cross-national comparisons of internal migration, Human Development Research Paper 2009/30. Benería, Lourdes, Carmen Diana Deere, and Naila Kabeer. 2012. Gender and international migration: Globalization, development, and governance, Feminist Economics 18(2): 1 33. Bhattacharya, Prabir C. 2000. An analysis of rural-to-rural migration in India, Journal of International Development 12(5):12: 655 667.