WPS5096 Policy Research Working Paper 5096 Missing Women and India's Religious Demography Vani Borooah Quy-Toan Do Sriya Iyer Shareen Joshi The World Bank Development Research Group Poverty and Inequality Team October 2009 Policy Research Working Paper 5096 Abstract The authors use recent data from the 2006 National lower levels of infant mortality (particularly female Family Health Survey of India to explore the relationship infant mortality). This effect is robust to the inclusion of between religion and demographic behavior. They controls for non-religious factors such as socio-economic find that fertility and mortality vary not only between status and area of residence. This result has important religious groups, but also across caste groups. These policy implications because it suggests that India's groups also differ with respect to socio-economic status. problem of "missing women" may be concentrated in The central finding of this paper is that despite their particular groups. The authors conclude that religion and socio-economic disadvantages, Muslims have higher caste play a key role in determining the demographic fertility than their Hindu counterparts and also exhibit characteristics of India. This paper--a product of the Poverty and Inequality Team, Development Research Group--is part of a larger effort in the department to understand the relations between religion and demography. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at pflewitt@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Missing Women and India's Religious Demography Vani Borooah Quy-Toan Do Sriya Iyer§ Shareen Joshi¶ 1 Introduction Two features of India's demography have recently received a great deal of attention. The first is gender bias - the small number of females compared to males (Visaria, 1971; Dyson and Moore, 1983; Sen, 1992, 2001; Kishor, 1993; Das Gupta, 2005). According to the 2001 Census, India has 933 females for every 1000 males, which implies that as many as 35­37 e million women in India may be "missing" (Dr`ze and Sen, 1996; Klasen, 1994; Agnihotri, 2000; Sen, 2003; Oster, 2005).1 The second feature is that demographic variables in India vary sharply by religious group. Fertility and the population growth rate for example, are higher among Muslims than Hindus (Basu 1997; Jeffery and Jeffery 1997; Iyer 2002; Dharmalingam and Morgan, 2004).2 In the literature on Indian demography, both these issues have received enormous at- tention, but they have typically been studied independently of each other. The adverse sex For helpful discussions we are grateful to Robert Barro, Eli Berman, Edward Glaeser, Timothy Guinnane, Monica Das Gupta, Partha Dasgupta, Larry Iannaccone, Siddiq Osmani, Abusaleh Shariff, Chander Velu, David Voas, and to participants at the American Economic Association 2007 meetings in Chicago, the `Religion, Economics and Culture' 2004 Annual Meetings in Kansas City, Missouri, and the 2005 ESRC seminar on `Demography and Religion' in Lancaster, UK. We are also grateful to Sarina Siddhanti for excellent research assistance, We are especially grateful to the Editor and two anonymous referees for their invaluable comments. Faculty of Economics and Politics, University of Ulster, Northern Ireland. vk.borooah@ulst.ac.uk Development Economics Research Group, The World Bank. qdo@worldbank.org § Corresponding author: Faculty of Economics and St. Catharine's College, University of Cambridge, Sidgwick Avenue, Cambridge CB3 9DD, United Kingdom. Tel: 44 1223 335257; Fax: 44 1223 335475; Email: Sriya.Iyer@econ.cam.ac.uk ¶ The School of Foreign Service, Georgetown University. shareen.joshi@gmail.com 1 The estimate of the number of missing women is based on comparisons with Europe and North America which have 1050 females per 1000 males. 2 An often-cited figure from the National Family Health Survey conducted in 1998-99, shows that the Total Fertility Rate (TFR) for Hindus and Muslims was 2.8 and 3.6 respectively. 1 ratio is mainly discussed in the context of the preference that many South and East Asian families have for sons over daughters (`son preference') and related issues that concern the marriage market, fertility, and dowry (Edlund, 1999; Bhat and Zavier, 2003; Sen, 2001; Rao, 1993; Bloch and Rao, 2002; Botticini and Siow, 2003; Das Gupta, 2005; Jacoby and Mansuri, 2006). The demographic differences between Hindus and Muslims, on the other hand, have usually been discussed in terms of the number of children in Muslim families and the higher population growth rates of this community. However, the debate often over- looked the fact that infant mortality among Muslims (at 59 per 1000) is much lower than that among Hindus (at 77 per 1000) as documented in Borooah and Iyer (2005a). Similarly child mortality, which is 83 per 1000 for Muslims, is substantially lower than child mortality among Hindus, at 107 per 1000 (IIPS and ORC Macro International, 2000). The survival advantage of Muslim children despite the higher fertility of their mothers is particularly puzzling in light of the fact that Muslims in India are typically poorer than Hindus as highlighted by the Sachar Committee Report (Government of India, 2006). Even a cursory glance at these facts and figures suggests that religious differentials in infant mortality need to be examined more closely. In this paper, we link these two separately-studied demographic realities into a com- parative analysis of infant mortality across religious and caste groups, using data from the newly released National Family Health Survey of 2005-06. To do so, we first divide our population into three subgroups: non-Dalit Hindus (also known as upper-caste Hindus), Dalit Hindus (also known as lower-caste Hindus), and Muslims. Our preliminary compara- tive analysis finds that with respect to observable socio-economic characteristics, Muslims are similar to Hindu Dalits in that they have lower levels of education, are poorer than non-Dalit Hindus. They also have more children, are less likely to experience the death of a child (particularly a girl), have higher female-male sex ratios among children alive as well as among children ever-born, are less likely to use contraception, and have preferences for a greater number of girls as well as boys. The next step of our analysis is to estimate survival probabilities for children and control for a number of important socio-economic, geographic and demographic variables. Our findings confirm that Muslim infant and child mortality is considerably lower than that for the Hindus, even after accounting for the wide range of socio-economic characteristics available to us. The effect is particularly strong for girls. Female infant mortality rates are lower in Muslim families than in Hindu families even with the inclusion of numerous controls for household and community characteristics. The final 2 step of our analysis involves a series of robustness checks. We find that our results are upheld. Although we cannot rule out all possible alternative rationales, or the possibility that socio-economic status may be insufficiently captured by our controls, these results lend support to the conjecture that religion might play a direct role in explaining the differences in gender differences in mortality between the Hindus and Muslims of India. This paper proceeds as follows: Section 2 discusses the role of religion in Indian demog- raphy and establishes our conjecture; Section 3 conducts a comparative analysis of non-Dalit Hindus, Dalit Hindus and Muslims. Section 4 concludes. 2 The Role of Religion in Indian Demography The idea that religion can shape demographic behavior is not new, particularly in the con- text of South Asia. Most of the existing literature on religion and demography however, focuses on fertility alone. Two main theories have dominated the discussion of religious differences in Indian fertility rates. The `characteristics hypothesis' proposes that the rel- ative poverty and lower education levels of Muslims drive fertility and mortality patterns that are different than those observed among Hindus (Iyer, 2002). In contrast to this, the `particularized theology hypothesis' - or the 'pure religion effect' argues that the intellec- tual content of religion affects fertility directly. Proponents of this hypothesis point out that religious injunctions in favor of multiple wives, large numbers of children, and a ban on the use of contraception and abortion may encourage higher fertility among Muslims. Such arguments however, have also been made for Hindus. The Mysore Population Study conducted in 1961, for example, concluded that Hindu religious traditions in Indian society favored having many offspring (United Nations, 1961). We believe that the focus on fertility, to the exclusion of other demographic variables such as mortality and the sex-composition of off-spring, is incomplete. For the remainder of this section, we document the differences in beliefs and norms between Hindus and Muslims that may impact their perceived benefits of sons versus daughters. In Islam, the institutional requirements of the religion are specified in the Sharia or Islamic law which is derived from two main sources ­ the Koran and the Sunnah, and also the writings of the medieval theologian Al Ghazzali, often cited by Muslim clerics, who 3 summarized Sunni and Shia positions on demography-related issues such as marriage and birth control (Al Ghazzali, 1909).3 In the case of Hinduism we consider religious texts such as Vedas, and Upanishads (Radhakrishnan, 1923); epic poems such as the Ramayana and Mahabharata (Deshpande, 1978); social commentaries such as Kautilya's Arthasastra (Shamasastry, 1951); and verse-poems in praise of Hindu goddesses such as the Lalita- sahasranama and the Sri-sukta (Suryanaraya Murthy, 2000) in the context of Indian de- mography (see Iyer, 2002 for a more detailed discussion of this literature). A close and comparative reading of the above texts suggests that the costs and benefits of sons compared to daughters may be different within Islam compared to Hinduism. While both religions encourage marriage, the nature of the contract differs. An Islamic marriage or the nikah, is defined not really as a sacrament, but more as a civil contract (Azim, 1997).4 Parents and guardians exercise control over the selection of marriage partners, and a dower or `bride price' is paid to the bride or her guardian (Youssef, 1978, p.78). The Koran recognizes the possibility of divorce and encourages remarriage of divorced or widowed women (Qureshi, 1980: 564; Youssef, 1978: 88, Coulson and Hinchcliffe, 1978: 37-38). In the context of India, this is important for Muslim families as the investments in daughters are frequently recoverable after marriage. As in Islam, Hinduism also encourages all Hindus to enter married life.5 The marriage of a daughter for Hindus is described as kanyadaan ­ this can be translated as the `donation' of a daughter. Such a donation is believed to benefit Hindu families both socially and religiously (Niraula and Morgan, 1996). Once married, women cease to be members of their natal home, and are generally not permitted to remarry in the event of divorce or widowhood. Hindu marriages are often accompanied by the giving and taking of significant dowries, i.e. payments from the family of the bride to the groom (Edlund, 1999; Bhat and Zavier, 2003; Rao, 1993; Bloch and Rao, 2002; Botticini and Siow, 2003). Women do not have property rights over these dowries and they are not returned to her or her family in the event of divorce or widowhood. 3 The Sunnah are the interpretations of the words of Mohammad and their application to various situa- tions. 4 For a Muslim marriage to be legally valid, it needs to meet four conditions: proposal by one party; acceptance by the other; the presence of a sufficient number of witnesses (two in Sunni law); and a formal expression of both the proposal and the acceptance at the same meeting (Azim 1997). 5 For example Shakuntala, a princess from Hindu mythology tells Dushyanta her beloved that `when a hus- band and wife are carrying on smoothly, then only pleasure, prosperity and piety are possible' (Deshpande, 1978: 91). 4 Such distinctions between the `contractual' versus the `donational' notion of marriage in Islam compared to Hinduism have implications for the relative costs and benefits of having sons and daughters. From a purely economic perspective, the net benefit of having daughters may be more `costly' for Hindu than for Muslim parents in India. Investments in daughters are also not recoverable. The preference for sons over daughters is also emphasized by other areas of Hindu ritual and philosophy. For example, securing a good `rebirth' is believed to be directly related to whether the eldest son of a deceased individual lights the funeral pyre. This sentiment is echoed in many writings: `At the end of the (Sraddha) death ceremony the performer asks, "Let me, O fathers, have a hero for a son!"' (Radhakrishnan, 1927, pp. 59-60). Additionally, sons are believed to be a vital source of security in old-age. Daughters are generally considered to `belong' to the family that they marry into, and so cannot be viewed as a source of support to her natal family. It must be emphasized that son-preference has also been documented in Islamic society. Islamic law is patriarchal and patrilineal. Sons are given twice as large an inheritance as daughters are and a man's testimony in court is worth twice that of a woman (Coulson and Hinchcliffe, 1978). Women in Islamic societies have typically been restricted to a lifestyle that guaranteed preservation of family honor and prestige and had limited opportunities to participate in labor markets (Coulson and Hinchcliffe, 1978: 38; Obermeyer, 1992). We simply argue that son-preference may less unequivocal as it is with Hindus. More recently, some sociological evidence has also emphasized this. For example, in his study of 378 Muslim women and men in Mangalore, Azim found that over two-third of respondents in his sample did not prefer sons, over daughters (Azim, 1997, p. 187); moreover, a large proportion of those who did were from poor and illiterate households (Azim, 1997: 189). Other nationally representative data from the National Family Health Surveys of India also show that about one-third of Muslims do not prefer sons over daughters (Kishor, 1993). 6 The relative importance of sons and daughters may stem not only from theology, but also from the differential socio-economic circumstances of Hindus and Muslims (Iyer, 2002; 6 An early example of the recognition of the contrast in gender norms comes from Maulana Ashraf Ali Thanavi who wrote a compendium of useful knowledge for women. He condemned the practice of blessing a woman by wishing her husband, brother, or children long life, or wishing for her many sons and grandsons (Minault 1998, p. 62). He argued that this was not suficiently Islamic because to view women blessed in terms of their relation to men and sons devalues their relationship to God, and hence goes against the tenet that all are equal in his sight. 5 Dharmalingam and Morgan, 2004). For example, in India today, it seems important to have educated sons, but in order to get daughters married, it is equally important to have educated daughters. If the average levels of education among Muslim men are for example lower than among Hindu men, then there may be lower educational investments also required of Muslim women (Borooah and Iyer, 2005b). A related issue is land ownership ­ there may be a greater desire on the part of Hindus to have sons in order to keep land within the patrilineal family line. It is documented that land ownership among Muslims is less than among Hindus (Shariff, 1999). In summary, for reasons that stem both from theological considerations, and from the socio-economic characteristics of religious groups, there may be important differences between Hindus and Muslims in their fertility and mortality of sons compared to daughters. We next turn to a formal investigation of the socio-economic determinants of fertility and mortality differences between Hindus and Muslims. 3 Fertility, Mortality and Religion in India India today has a total population of just over 1.1 billion people. Census data shows that Hindus form over 80 percent of India's population, while Muslims form approximately 15 percent. Muslims are the most significant `minority' population and consist of approxi- mately 150 million people. Our description of contemporary India uses data from the newly released 2005-06 National Family Health Survey (NFHS-3).7 The NFHS interviewed a total of 123,385 women in 29 states of India. The survey is based on a sample of housesholds that is representative at the state as well as national level. The religious composition of the households is consistent with the findings of the Indian Census (2001): 82 percent of households are Hindu, 13 percent are Muslim, 3 percent are Christian, 2 percent are Sikh, and the remainder are other religions. Among Hindus, the caste composition of our popu- lation also mirrors the census. 44% of Hindu respondents reported that they belonged to a "Scheduled Caste" or a "Scheduled Tribe" (henceforth referred to as Dalits), and 44% of 7 The first NFHS survey was conducted in 1992-93, and the second in 1998-99. All three surveys were conducted in conjunction with the Ministry of Health and Family Welfare (MOHFW),Government of India. Funding for the survey was provided by the United States Agency for International Development (US- AID), the United Kingdom Department for International Development (DFID), the Bill and Melinda Gates Foundation, UNICEF, UNFPA, and the government of India. Technical assistance was provided by Macro International Maryland, USA. 6 the Hindu population was higher caste or non-Dalit.8 We restrict our sample by including only those households that are Hindu and Muslim, and in which at least one female respondent had borne children. This leaves us with a sample of a total of 81,021 women. Panel A of Table 1 presents additional summary statistics for the female sample. The basis of our analysis is a comparison between the demographic and socio-economic differences between three subgroups of India's population: the Dalits, Muslims, and Hin- dus. Since these categories are conceptually overlapping, we define three mutually exlusive categories: non-Dalit Hindus; Muslims; and Dalit Hindus.9 We begin with a comparison of socio-economic variables for the three groups in our sample. The first three columns of Table 2 explore differences between non-Dalit Hindus and Muslims, and the next three columns explore differences between non-Dalit and Dalit Hindus. Note that in nearly all respects, non-Dalit Hindus appear richer than Muslims and their Dalit counterparts. They also have higher rates of schooling, lower fertility, higher ownership of farmland and a lower chance of falling below the poverty line. In other words, Muslims and Dalits both appear to be socio-economically disadvantaged relative to upper-caste Hindus. The differences are statistically significant. Where Hindu Dalits and Muslims diverge from each other significantly is in the rates of female labor-force participation. Muslim women work less than their non-Dalit Hindu counterparts, while Dalit Hindus work more. The differences are also visible in other indicators of female labor force participation rates. Muslim women (Dalit women) are less (more) likely to be self- employed, and less (more) likely to work away from home. These differences are again statistically significant. While Muslims are similar to Dalit Hindus on a range of socio-economic characteristics, 8 The term 'Dalit' translates as 'the downtrodden', and refers specifically to India's Scheduled Castes and Sceduled Tribes - these are those castes and tribes recognised by the Indian Constitution as deserving special recognition in respect of education, job reservation in employment, and political representation. 9 While Dalits are also found among Muslims in India (Sachar Committee Report, 2006), we restrict our attention to the Hindu Dalits. This is for two reasons. First, while 30% of Hindus are Dalits, the proportion of Muslims reporting Dalit-status is approximately 2­5% (Sachar Committee Report, 2006). Second, among Hindus, the criteria for being a Dalit are widely recognized. As mentioned in the footnote above, the Indian government maintains official lists of "scheduled castes" and "scheduled tribes". Such benchmarks do not yet exist among other groups, making the categorization of particular houses in a survey somewhat subjective. In our sample, only 452 Muslim women reported themselves as Dalits. We included this group in our Muslim sample, but excluded them from the Dalit sample. 7 they appear to be quite similar to upper-caste Hindus in that they are less likely to expe- rience the death of a child (particularly a girl), have higher female-male sex ratios among children alive as well as among children ever-born, are less likely to use contraception and have preferences for greater number of girls as well as boys, as evidenced by the greater numbers of girls and boys that they regard as "ideal".10,11 In all these cases, the differences are statistically significant. Age-specific fertility also confirms that Muslim women bear larger numbers of children at earlier ages than women from other religious groups (Figure 1). 3.1 Fertility Differences To explore the often-cited fact about higher Muslim fertility, we construct from the female sample the total number of children the woman has had (those born alive as well as those who have died, but excluding miscarriages and stillborns). We first examine the determinants of fertility on the entire sample of women, and then restrict the sample to women over the ages of 30 and 40. Estimation on the restricted sample permits us to examine the relationships for women who have most likely completed their child-bearing. We control for a variety of other individual, family and regional characteristics. First, we control for a woman's age because of the well-known fact that children born to mothers at very young or very old ages are more likely to die in infancy than children born to mothers in prime childbearing ages. A second set of controls includes variables on whether a woman (the child's mother) and her husband (the child's father) had completed primary school. We also include a dummy variable indicating whether they had never attended school. Third, we control for economic status using a wealth index. This index has been developed and tested in a large number of countries and has been shown to be consistent 10 Sex-ratios are measured at the level of the cluster/village rather than at the level of a woman. Since a sex-ratio is defined as the number of females relative to the number of males, it can only be constructed for a woman who has had at least one male birth. When aggregating at the level of the village or cluster (which was the NFHS primary sampling unit) however, this problem is alleviated, since it is an average of female and male deaths for a broad group of women. 11 The low female-male ratio at birth may be attributed to the fact that the practice of sex selective abortion is higher among upper-caste Hindus. This group has been shown to engage in this practice more than other groups (Arnold, Kishor and Roy, 2002). However, since the NFHS data do not contain information on the prevalence and practice of abortion, this is one aspect that we are not able to investigate comprehensively. However, we do acknowledge, readily, the importance of these practices for the subject of this research. 8 with expenditure and income measures (Rustein, 1999; Filmer and Pritchett, 2001).12 We also include a control for whether the family resides in a rural area. Rural areas in India typically have significantly higher rates of infant mortality than urban areas. We finally add dummy variables for states and the region in which an individual resides. This is intended to capture state- or region-fixed effects. The results of the fertility regression are presented in Table 3. As expected, and based on the summary statistics seen in Table 2, the coefficients for Muslim and Hindu Non-Dalit take opposite signs: Muslims have more children, and Hindu Non-Dalits have fewer children than the omitted group (Dalits) and these differences remain statistically significant even when all the control variables are included. Muslims have about 1 child more than Dalits and Non-Dalits have about 0.3 fewer children than Dalits. We interepret this as evidence that Muslim fertility is higher than Hindu fertility overall. Additionally, most measures of socio- economic status have the expected negative sign in the fertility regressions. The household wealth index as well as parental age and education are all associated with decreased fertility. As widely seen in the empirical literature, maternal attributes have a stronger effect on fertility than paternal attributes. In results not shown here, we examine the robustness of the relationship between religion and fertility by employing additional socio-economic and demographic variables such as the use of contraception (traditional as well as modern), duration of the use of contraception, exposure to mass media and nutritional indicators. Inclusion of these additional explanatory variables had little to no effect on the two variables Muslim and Hindu Non-Dalit that are central to our interest in this paper. 12 The wealth index is constructed by combining information on 33 household assets and housing charac- teristics such as ownership of consumer items, type of dwelling, source of water, and availability of electricity into a single wealth index. These 33 assets are as follows: household electrification, type of windows, drink- ing water source, type of toilet facility, type of flooring, material of exterior walls, type of roofing, cooking fuel, house ownership, number of household members per sleeping room, ownership of a bank account, own- ership of a mattress, a pressure cooker, a chair, a cot/bed, a table, an electric fan, a radio/transmitter, a black and white television, a color television, a sewing machine, a mobile phone, another other telephone, a computer, a refrigerator, a watch or clock, a bicycle, a motorcycle or scooter, an animal-drawn cart, a car, a water-pump, a thresher, and a tractor. Each household asset is assigned a weight (factor score) that is generated through principal components analysis. The resulting asset scores are standardized in relation to a normal household and is then assigned a score for each asset, and scores were summed for each household; individuals are ranked according to the score of the household in which they reside. 9 3.2 Infant Mortality Differences We now turn to religious differences in mortality risks. Our main specification relies on the child sample, which includes all children ever born to women in the female sample described above. The information was gathered from birth-histories and includes children who are alive as well as those who died, and also includes children living within the household as well as those who do not reside in the household any more. As child-death is a censored variable and since the risk of death appears not to be limited to any particular age group, our empirical analysis of the determinants of mortality is based on the Cox Proportional Hazard model: p i (t) = 0 (t) expg(Mi )+Xi , (1) where the dependent variable is the mortality hazard, or rather, the risk of death for a particular child i. 0 (t) is the baseline hazard, Mip is the dummy variable that indicates whether the child's parents are Muslim, g is a general function and Xi is a vector of observ- able characteristics that may affect a child's mortality risk. This model is a semi-parametric model in the sense that it does not impose any functional form on the baseline specification. Our working sample includes 218,769 children who are born to 79,118 Hindu and Muslim mothers in the NFHS-3. Of the 218,769 children who were included in our sample, 25,784 (10.4 percent of the total) had died. The children's ages range from 0­37.13 For children who died, the average age at death was 20 months. The set of control variables is very similar to what was used in the female-sample. We also include some additional controls: a dummy indicating whether the child is female, his/her birth order, a dummy indicating whether the child born just previously was female, indicators for whether any meat, dairy or plant-protein was consumed in the 24 hours prior to the survey, and cluster-level averages of vaccination rates for measles, polio, BCG and DPT.14 Cluster-level averages of female labor force participation rates and female self- employment rates were also included. The subset of observations that has information on 13 For those children who had died, the age variable was coded as age-at-death. 14 It is important to note that the average immunization rates will only serve as a rough proxy for health- care services in the cluster. Immunization rates could have changed considerably in many areas (though not in the aggregate for the country) before or after the death of some of the children in the sample. In future work, we will control for immunization levels by year of birth. 10 all these observations gives us a final sample of 195,080 children. The results from the basic Cox-Hazard model (in the form of exponentiated coefficients) are presented in Table 4. Six models are estimated, starting with the simplest version, with no control variables.15 Since the model is non-linear, the coefficients of the variables Muslim, Hindu Non-Dalit, Female, MuslimXFemale and Hindu Non-Dalit X Female do not provide a measure of the relative risk of belonging to these groups. We calculate these separately and present them in the top columns of Table 4 under the heading "Total Effects".16 Note that the coefficients that are of most interest to us ­ Muslim and Hindu Non-Dalit ­ take a negative sign and are statistically significant at the 1 percent level in all models, indicating that individuals in both these groups face lower mortality risks than Dalit Hindus. It is also noteworthy that the coefficient for the dummy variable Female is positive and significant at the 1 percent level in all six models, confirming that there is a strong female disadvantage in survival probabilities in India. Interestingly the interaction term Female X Muslim is negative and significant at the 1 percent level in all models, indicating that the survival disadvantage is mitigated somewhat among Muslim girls. The coefficient for Hindu Non-Dalit X Female is negative in the first two models (columns 1 and 2), but then loses significance (columns 3 to 6). This important result suggests that while caste differences are largely explained by differences in socio-economic characteristics, the same cannot be said of religious differences. The robustness of the interaction Female X Muslim is indicative that something specific to "being of the Muslim community" and not captured by survey variables is partially explaining differences in mortality rates between boys and girls.17 We interpret this robust finding as supporting evidence for a religion-based theory of the documented group differences in demographic variables. A close look at the estimates for the state-dummy variables provides some additional intersting insights about the variation of mortality risks by state and region. The omitted 15 STATA 10. The basic unit of time is one year. 16 STATA 10 calculates these total effects using the command "lincom". The total effect for the variable Muslim X Female for example, is calculated by multiplying 'the relative risk of the main effect of Muslim==1' × 'the relative risk of the main effect of Female=1' × 'the relative risk of the interaction term (Muslim, Female)=(1,1)). 17 First, we interact the variable Female with additional controls (mother's and father's educational at- tainment, wealth, rural-urban residence, and age at marriage) and then test whether the double interaction terms Muslim X Female continued to have explanatory power. Second, we separate the sample of Hindus and Muslims and perform the same tests. Both sets of results confirm that while these variables may have considerable explanatory power, the differences between groups remain significant. 11 state in the analysis is Jammu and Kashmir, a state where 56% of the population reports that it is Muslim. Relative to this state, mortality is lower (and statistically significant) in the neighboring mountainous state of Himachal Pradesh and Uttaranchal. Mortality is also lower in the south. The states of Goa and Kerala in particular display significantly lower risks of mortality. This is entirely expected based on the well-documented progress by these states in the area of human-development (Dreze and Sen, 2001). The mortality risks are higher than the baseline (and statistically significant) in some states in the North and North- East. These include Delhi, Uttar Pradesh, Bihar, Arunachal Pradesh, Tripura, Assam and Jharkhand. Mortality is also higher (and statistically significant) in Orissa, Chhatisgarh and Madhya Pradesh. All other states display no statistically significant mortality risks. There is no evidence that the size of the Muslim population explains these effects. In a separate analysis that is not shown here, we also explore the regional patterns of mortality. Among our group of explanatory variables, we include interactions of the variable Muslim with dummy variables corresponding to different regions. Results indicate that the interaction, and the overall effect of region, is most significant in the states of the North West (Jammu and Kashmir, Himachal Pradesh, Uttaranchal, Haryana, Punjab, Delhi, Rajasthan and Uttar Pradesh). The regional effect is infact insignificant in the South (Goa, Karnataka, Kerala and Tamil Nadu) as well as the North-East (Sikkim, Arunachal pradesh, Nagaland, Manipur, Mizoram , Tripura, Meghalaya, Assam, West Bengal and Jharkhand). In other regions, the Muslim effect as well as the regional effect remain significant, but less so than in the Northwestern states. 3.3 Robustness Checks In order further to investigate the robustness of our findings, we examine the question from two different but complementary angles. First, we construct two mortality measures defined at the level of the mother: (1) The fraction of ever-born boys who have died, and (2) The fraction of ever-born girls who have died. In our regressions, we use the same set of control variables as the one used in the regressions of the determinants of fertility (see Table 3). The results of the mortality regressions are presented in Table 5. We see that the coefficient Muslim is negative and statistically significant in all eight columns, indicating that Muslim women are less likely than Hindus to experience the loss of a child, even when we control for their poorer socio-economic status and location (columns 3­5 and then 8­10 include state 12 fixed effects). The coefficient for Hindu Non-Dalit was negative and significant in the case without control variables (columns 1 and 2) but the variable lost significance once other controls and fixed effects were included, indicating that they were statistically speaking no different from their Dalit counterparts once we include measures of socio-economic status. The Muslim effect however was robust and significant, suggesting that the effect may not be driven by socio-economic status, at least to the same extent as the Dalits. As an additional robustness check, we perform a similar analysis using village-level data.18 Our sample now consists of the 3644 villages in the NFHS-3, and our left-hand side variable of interest will be average fertility of women in a village (defined as the total number of children divided by the number of women who were interviewed in a village). The dependent variables are as follows: the fraction of ever-born boys who had died before the age of 5 and the fraction of ever-born girls who had died before the age of 5. The set of our control variables is similar to that in the previous section, except that they are constructed as village-level averages. We control for the fraction of a village's population that is Muslim, as well as upper-caste Hindus. We also control for the average female age, age at marriage, primary school completion and the fraction of the population that has never attended school. Similar variables for men ­ average age, fraction of men who completed primary school and fraction of men who never attended school ­ are also included as controls. We also include a set of control variables that focus on wealth. Measures of average landholdings, the average household wealth index and the fraction of households that are rural, are included in this group of control variables. Finally, we also include a set of community level averages for vaccination rates for measles, polio, DPT and BCG, nutritional intakes and female labor force participation rates. The results for male mortality are presented in columns 4­6 of Table 6 and the results for female mortality are presented in columns 5­9. Note that the coefficient for Muslim pop- ulation is not significant in the male mortality regressions, and is negative and significant in the female mortality regression that includes fixed effects (column 9). On the contrary, the coefficient for Hindu Non-Dalit Population is negative and significant in the male mortality regressions without controls (column 4) and positive and significant in the female mortality regression with controls (column 8). We interpret this as evidence that Muslim girls exhibit lower mortality rates than the Dalits, but upper-caste Hindu girls display higher mortality 18 Ideally, we would like to have performed the analysis at the level of a district, but this was not possible with the NFHS-3 dataset. District information was not included in the public release of the data. 13 rates than the Dalits. Muslim boys however, do not display any significant difference in mortality rates compared to the Dalits. Upper-caste Hindus however, diplay lower mor- tality rates in this group, although this appears to be explained by their socio-economic characteristics. This set of findings is consistent with the results from the child sample. Overall, all the results confirm that Muslims experience lower levels of female mortality. The difference between Hindus and Muslims is not attributable to differences in education, landholdings, wealth, rural-urban residence or state of residence. While Dalits may also display lower levels of female mortality, the Muslim effect appears to be stronger than the Dalit effect. New research from immigrant communities in North America corroborates our finding on the distinctive demographic characteristics of certain religious groups. Recent work by Almond, Elund, and Milligan (2009), who study sex-ratios among immigrant communities in Canada, finds relatively normal sex-ratios in the Muslim and Christian community, but highly skewed sex-ratios among other religious groups. We once again acknowledge that we can not rule out the possibility that unobserved variations in socio-economic status may indeed be driving these results. However, unless more national data become available, we must seriously consider the possibility that religion may indeed be a driver of preferences for males and females in India. The survival advantage of Muslim girls ­ as confirmed by our results from the child sample, woman sample and village sample ­ suggest that India's "missing women" problem may be concentrated in certain religious groups. This has major policy-implications for it suggests that the status of women within particular caste- and religious-groups deserves greater analysis and attention from academics as well as policy-makers. 4 Conclusion In this paper, we use recent data from the 2006 National Family Health Survey to explore the relationship between religion and demographic outcomes in India. We find that fertility and demographic behavior vary not only across religious groups, but also across caste groups. A comparison of socio-economic variables suggests that Muslims are similar to Dalit (lower- caste) Hindus in that they are poorer and have more children, but unusually also exhibit 14 lower infant mortality rates. Our econometric analysis confirms that these differences persist even when we control for socio-economic dcharacteristics, community characteristics and location. Results from samples at the level of individual children, adult females and entire villages all suggest that total infant mortality, and in particular, female infant mortality is lower among Muslims than Hindus. This is an important result for it suggests that India's "missing women" may be most concentrated in particular caste and religious groups and may not be a general problem in the Indian population. The results of this paper also suggest that the tendency to focus merely on differences in fertility between religious groups may be simplistic. While we can not rule out the possibility that unobserved aspects of socio-economic status may be driving our results, we highlight the possibility that religion and religious customs may have a direct effect on how daughters are valued in their families. We believe that there is much scope for further research on the interations between religion, fertility, gender and mortality in India. References 1. Al Ghazzali, M. H. (1909), The Alchemy of Happiness. Translated from the Hindus- tani by Claud Field, London: John Murray. 2. Agnihotri, S. (2000), Sex Ratio Patterns in the Indian Population, New Delhi: Sage Publications. 3. Almond, D., Edlund, L., and Milligan, K. (2009), `Son Preference and the Persistence of Culture: Evidence from Asian Immigrants to Canada', NBER Working Paper 15391. 4. Arnold, F., Kishor, S. and Roy, T.K. (2002), 'Sex-Selective Abortions in India,' Pop- ulation and Development Review, 28 (4), pp. 759-785. 5. Azim, S. (1997) Muslim Women: Emerging Identity. New Delhi: Rawat Publications. 6. Basu, A. (1997), `The "Politicization" of Fertility to Achieve Non-Demographic Ob- jectives', Population Studies, Vol. 51, pp. 5-18. 7. Bhat, M and Zavier, F (2003), `Fertility Decline and Gender Bias in Northern India', Demography, 40:4, pp. 637- 657. 15 8. Bloch, F. and Rao, V. (2002), 'Terror as a Bargaining Instrument: A Case Study of Dowry Violence in Rural India', American Economic Review 93(4), pp. 1385-98. 9. Borooah, V. (2004) `Illuminating the Politics of Demography: A Study of Inter Com- munity Fertility Differences in India', European Journal of Political Economy, 20, pp. 551-578. 10. Borooah, V and S. Iyer (2005a), `Religion, Literacy, and the Female-to-Male Ratio in India', Economic and Political Weekly, Special Issue on Religion and Fertility, January. 11. Borooah, V and S. Iyer (2005b), `Vidya, Veda and Varna: The Influence of Religion and Caste on Education in Rural India', The Journal of Development Studies, Vol. 81, Number 4, November, pp. 1369-1404. 12. Borooah, V and S. Iyer (2005c), 'The Decomposition of Inter-group Differences in a Logit Model: Extending the Oaxaca-Blinder Approach with an Application to School Enrolment in Rural India,' Journal of Economic and Social Measurement, November, Vol. 30, No. 4, pp. 279-293. 13. Borooah, V. and S. Iyer (2004), 'Religion and Fertility in India: The Role of Son Preference and Daughter Aversion', Cambridge Working Papers in Economics 436, Faculty of Economics, University of Cambridge. 14. Botticini, M and A. Siow (2003) `Why Dowries?', American Economic Review, 93(4), pp. 1385-98 15. Caldwell J.C., Reddy P.H. and Caldwell P. (1983), `The Causes of Marriage Change in South India', Population Studies 37 (3), pp. 343-361. 16. Chen, L. C. (1982), `Where have the Women Gone? Insights from Bangladesh on the Low Sex Ratio of India's Population.', Economic and Political Weekly, 17. 17. Coulson, N. and Hinchcliffe, D. (1978), `Women and Law Reform in Contemporary Islam', in Women in the Muslim World, edited by L. Beck and N. Keddie, Cambridge: Harvard University Press, pp. 37-49. 18. Das Gupta, M. (2005), `Explaining Asia's `Missing Women': A Look at the Data', Population and Development Review, 31(3). 16 19. Deolalikar, A. B. and Rao V. (1998), `The Demand for Dowries and Bride Charac- teristics in Marriage: Empirical Estimates for Rural South-Central India', in Gender, Population and Development, edited by Krishnaraj, M., Sudarshan, R. and Shariff, A. Delhi: Oxford University Press, pp. 122-40. 20. Deshpande, C. R. (1978), Transmission of the Mahabharata Tradition. Vyasa and Vyasids. Simla: Institute of Advanced Study. 21. Dharmalingam, A. and S. P. Morgan (2004), `Muslim-Hindu Fertility Differences in India', Demography, Vol. 41, No. 3, pp. 529-545. 22. Dreze J. and Murthi, M. (2001) `Fertility, Education and Development: Evidence from India', Population and Development Review 27(1), pp. 33-63. 23. Dreze, J. and Sen, A.K. (1996), Economic Development and Social Opportunity, New Delhi: Oxford University Press. 24. Dyson, T. and M. Moore (1983), `On Kinship Structure, Female Autonomy, and Demographic Behaviour in India', Population and Development Review, Vol. 9, pp. 35-60. 25. Edlund, L.E. (1999), `Son Preference, Sex Ratios and Marriage Preference', Journal of Political Economy, vol. 107, pp. 1275-1304. 26. Filmer, D. and L. Pritchett (1999), The Effect of Household Wealth on Educational Attainment: Evidence From 35 Countries, Population and Development Review, Vol. 25(1), pp.85-120 27. Goldschielder, C. and Uhlenberg, P. (1969), 'Minority Group Status and Fertility', American Journal of Sociology, 74 (January), pp. 261-272. 28. Government of India (2006), Social, Economic and Educational Status of the Muslim Community of India, Prime Minister's High Level Committee, Cabinet Secretariat, Government of India. 29. International Institute of Population Sciences, Mumbai and ORC Macro International (2000), India 1998-99 National Family Health Survey (NFHS-2). IIPS, Mumbai. 30. International Institute of Population Sciences, Mumbai and ORC Macro International (2007), India 2005-06 National Family Health Survey (NFHS-3). IIPS, Mumbai. 17 31. Iyer S. (2002) Demography and Religion in India. Oxford University Press: Delhi. 32. Jacoby, H. G. and Mansuri, G. (2005), 'Watta Satta: Exchange Marriage and Women's Welfare in Rural Pakistan', mimeograph, The World Bank. 33. Jeffery R. and Jeffery P. (1997), Population, Gender and Politics: Demographic Change in Rural North India. Cambridge University Press: Cambridge. 34. Jensen, R. (2003), Equal Treatment, Unequal Outcomes: Generating Sex-Inequality Through Fertility Behavior, mimeograph, Watson Institute. 35. Kishor, S. (1993), `"May God Give Sons to All": Gender and Child Mortality in India', American Sociological Review, vol. 58, No. 2, April, pp. 247-265. 36. Klasen, S. (1994), "Missing Women Reconsidered", World Development, vol. 22, pp. 1061-1071. 37. Landes, D. S., 1998. The Wealth and Poverty of Nations. Little, Brown and Company, London. 38. Minault, G (1998), Secluded Scholars. Delhi: Oxford University Press. 39. Moulasha, K and Rao, G. R. (1999) `Religion-Specific Differentials in Fertility and Family Planning', Economic and Political Weekly, 34 (42), pp. 3047-51. 40. Murthi, M., Guio, A-C, and Dreze, J. (1995) `Mortality, Fertility, and Gender Bias in India: A District-Level Analysis', Population and Development Review, 21(4), pp. 745-82. 41. Niraula, B. B. and S. Philip Morgan (1996), `Son and Daughter Preferences in Be- nighat, Nepal: Implications for Fertility Transition', Social Biology, Vol. 42, pp. 256-73. 42. Obermeyer C. M. (1992) `Islam, Women and Politics: The Demography of Arab Countries', Population and Development Review 18 (1), pp. 33-60. 43. Oster, E. (2005), `Hepatitis B and the Case of the Missing Women', Journal of Political Economy 113(6), pp. 1163-1216. 44. Qureshi, S. (1980) `Islam and Development: The Zia Regime in Pakistan', World Development 8, pp. 563-575. 18 45. Radhakrishnan, S. (1923) Indian Philosophy. Volume 1. Reprinted in 1997 by Oxford University Press, USA. 46. Radhakrishnan, S. (1947) Religion and Society. London: George Allen and Unwin. 47. Rao V. (1993) `The Rising Price of Husbands: A Hedonic Analysis of Dowry Increases in Rural India', Journal of Political Economy 101, pp. 666-677. 48. Rustein, S. (1999), Wealth versus expenditure: Comparison between the DHS wealth index and household expenditures in four departments of Guatemala. Calverton, Mary- land: ORC Macro. 49. Shamasastry, R (1951), Translation into English of Kautilya's Arthasastra. Mysore: Sri Raghuveer Printing Press. 50. Sen, B. (1998) `Why Does Dowry Still Persist in India? An Economic Analysis Using Human Capital', in South Asians and the Dowry Problem, edited by Menski, W., Trentham Books; London: University of London, School of Oriental and African Studies, pp. 75-95. 51. Sen, A. K. (1992), 'Missing Women', British Medical Journal, 304, March, pp. 586- 587. 52. Sen, A.K. (2003), 'Missing Women - Revisited', British Medical Journal, 327, Decem- ber, pp. 1297-1298. 53. Shariff, A. (1999), India Human Development Report, Oxford University Press: New Delhi. 54. Suryanarayana Murthy, C (2000), Sri Lalita Sahasranama with Introduction and Commentary. Mumbai: Bharatiya Vidya Bhavan Publications. 55. United Nations. (1961) Mysore Population Study. Department of Social and Eco- nomic Affairs, New York. 56. Visaria, P. (1971) The Sex Ratio of the Population of India, Census of India, Mono- graph No. 10 (New Delhi: Manager of Publications). 57. Youssef, N. H. (1978) `The Status and Fertility Patterns of Muslim Women' in Women in the Muslim World, edited by L. Beck and N. Keddie, Cambridge: Harvard Univer- sity Press, pp. 69-99. 19 Tables and Figures: Empirical Section Figure 1: Age-specific fertility rates for women of all religious groups. 20 Table 1: Summary statistics for variables used in regressions Variable Mean Std. Dev. N Panel (A): Female Sample Percent of ever-born boys died before age 5 3.484 11.229 79118 Percent of ever-born girls died before age 5 2.997 10.271 79118 Muslim 0.131 0.338 79118 Hindu Non-Dalit 0.564 0.496 79118 Total number of children 3.034 1.789 79118 Woman's age 32.883 8.068 79118 Woman's age at marriage 17.828 3.801 79118 Woman never attended school 0.401 0.49 79118 Woman completed primary school 0.444 0.497 79118 Husband's age 38.58 9.142 79118 Husband never attended school 0.221 0.415 79118 Land (in acres) 4.706 20.811 79118 Wealth index 0.002 0.1 79118 Rural 0.562 0.496 79118 Panel (B): Child Sample Muslim 0.155 0.362 218769 Female 0.479 0.5 218769 Muslim × Female 0.075 0.264 218769 Hindu Non-Dalit 0.524 0.499 218769 Hindu Non-Dalit × Female 0.249 0.433 218769 Birth order 2.531 1.669 218769 Previous sibling female 0.319 0.466 218769 Total number of siblings 3.062 2.117 218769 Mother's age 34.931 7.723 218769 Mother's age at marriage 17.08 3.469 218769 Mother never attended school 0.514 0.5 218769 Mother completed primary school 0.326 0.469 218769 Father's age 40.646 8.856 218769 Father never attended school 0.281 0.449 218769 Land (acres) 3.415 17.733 218769 Wealth index -0.016 0.097 218769 Rural 0.618 0.486 218769 AteMeat 0.026 0.16 218769 AtePlantProtein 0.03 0.171 218769 AteDairy 0.019 0.136 218769 Hemelevel 116.099 17.669 218769 Average measles vaccinations 0.547 0.262 218769 Average BCG vaccinations 0.795 0.234 218769 Average polio vaccinations 0.885 0.152 218769 Average DPT vaccinations 0.761 0.245 218769 Average female LFP 0.376 0.237 218769 Average female self-employed 0.069 0.104 218769 Panel (C): Cluster-level data Fraction of boys died before age 1 0.087 0.072 3646 Fraction of girls died before age 1 0.077 0.073 3646 Muslim population 0.123 0.26 3628 Dalit population 0.313 0.312 3628 Average female age 37.311 2.291 3628 Continued on next page 21 Table 1: Summary statistics for variables used in regressions Variable Mean Std. Dev. N Average female age at marriage 17.288 2.043 3628 Average female education 2.683 0.959 3628 Average male education 3.079 0.736 3628 Rural population 0.570 0.495 3628 Average land(in acres) 2.492 4.622 3628 Average wealth index scores -0.257 83.431 3628 Average of meat 0.015 0.038 3628 Average of plant protein 0.016 0.036 3628 Average of dairy 0.01 0.029 3628 Average of hemoglobin level 117.049 7.627 3628 Average of current female LFP 0.376 0.236 3644 Average of female self-employment 0.071 0.103 3644 Table 1: Summary statistics for variables used in the regressions. 22 Table 2: Differences between Hindus, Muslims and Dalits Non Dalit Hindus Muslims Diff Non-Dalit Hindus Dalits Diff (std. err.) (std. err.) Demographic Profile Number of children born 2.9 3.65 -0.76 2.9 3.38 -0.47 (-0.02) (-0.015) Fraction of ever-born boys who died 0.036 0.039 -0.0026 0.037 0.048 -0.012 (-0.0014) (-0.001) Fraction of ever-born girls who died 0.032 0.033 -0.0012 0.032 0.042 -0.0100 (-0.0012) (-0.00096) Sex-ratio of children alive 980.8 1053.4 72.56 981.5 1033.8 52.29*** (20.12) (15.06) Sex-ratio of children ever-born 972.7 1029.8 57.05 972.7 1007.6 32.49*** (18.72) (14.1) Sterilized 0.44 0.26 0.18 0.44 0.39 0.048 (-0.0058) (-0.0043) Female contraceptive use 0.62 0.45 0.17 0.62 0.51 0.11 (-0.0058) (-0.0043) "Ideal" number of girls 1.02 1.22 -0.20 1.02 1.24 -0.22 (-0.0054) (-0.0045) "Ideal" number of boys 1.31 1.6 -0.28 1.31 1.58 -0.26 (-0.0067) (-0.0052) 23 Ideal sex-ratio 0.85 0.84 0.0095 0.85 0.86 -0.0080 (-0.0041) (-0.0031) Age at first marriage 18.2 17.1 1.12 18.1 17.3 0.81 (-0.04) (-0.03) Socioeconomic Status Female completion of primary school 0.52 0.34 0.18 0.51 0.32 0.19 (-0.0052) (-0.0038) Female completion of secondary school 0.12 0.032 0.084 0.12 0.03 0.086 (-0.0032) (-0.0022) Male completion of primary school 0.69 0.48 0.21 0.69 0.5 0.20 (-0.0049) (-0.0037) Male completion of secondary school 0.21 0.085 0.12 0.21 0.093 0.12 (-0.0041) (-0.0029) Land (acres) 4.9 4.84 0.051 4.9 3.99 0.91 (-0.22) (-0.16) Household below poverty line 0.53 0.57 -0.045 0.53 0.57 -0.046 (-0.016) (-0.012) Women currently working 0.35 0.24 0.11 0.35 0.46 -0.12 (-0.0049) (-0.0038) Women self-employed 0.072 0.04 0.032 0.072 0.08 -0.0085 Continued on next page Table 2: Differences between Hindus, Muslims and Dalits Non Dalit Hindus Muslims Diff Non-Dalit Hindus Dalits Diff (std. err.) (std. err.) (-0.0026) (-0.0021) Women working away from home 0.31 0.16 0.15 0.31 0.46 -0.15 (-0.0047) (-0.0037) Nutrition and Health Inputs Eat Meat 0.026 0.055 -0.029 0.024 0.036 -0.011 (-0.0019) (-0.0014) Eat Plant protein 0.037 0.035 0.0024 0.037 0.037 -0.00044 (-0.0021) (-0.0016) Eat Dairy 0.019 0.04 -0.022 0.018 0.026 -0.0084 (-0.0016) (-0.0012) Women's hemoglobin level 117 116.2 0.89 117 114.9 2.09 (-0.19) (-0.15) Vaccinated against measles 0.62 0.46 0.16 0.62 0.5 0.12 (-0.0063) (-0.005) Vaccinated against BCG 0.83 0.68 0.16 0.83 0.72 0.11 (-0.0051) (-0.0042) Vaccinated against polio 0.92 0.86 0.059 0.92 0.85 0.066 (-0.0038) (-0.0032) Vaccinated against DPT 0.81 0.63 0.17 0.81 0.69 0.11 24 (-0.0054) (-0.0043) Table 2: Notes: (i) Individual data are based on a total sample of 51217 Hindu women and 9269 Muslim women who are married and have had at least one male birth prior to the survey; (ii) The sex-ratio is defined per cluster as the (number of females)/(number of males) × 1000 and is calculated as an average over clusters; (iii) Standard errors in parentheses: p < 0.05, p < 0.01, p < 0.001 Table 3: Fertility Regressions (female sample), Dependent Variable: Number of children born All Women Age30 Age40 (1) (2) (3) (4) (5) Muslim 0.344 0.448 0.539 0.780 0.971 (0.110) (0.110) (0.0585) (0.0884) (0.127) Hindu Non-Dalit -0.530 -0.280 -0.216 -0.285 -0.325 (0.0433) (0.0720) (0.0269) (0.0361) (0.0483) Mother's age 0.124 0.114 0.0824 0.0766 (0.00852) (0.00642) (0.00398) (0.00539) Mother's age at marriage -0.116 -0.114 -0.106 -0.106 (0.00669) (0.00453) (0.00486) (0.00472) Mother never attended school 0.367 0.231 0.235 0.212 (0.0733) (0.0319) (0.0385) (0.0436) Mother completed primary school -0.121 -0.113 -0.226 -0.340 (0.0306) (0.0222) (0.0305) (0.0464) Father's age -0.0153 -0.00217 -0.00624 -0.0137 (0.00432) (0.00229) (0.00290) (0.00353) Father never attended school 0.0934 0.124 0.133 0.133 (0.0408) (0.0262) (0.0359) (0.0566) Land (in acres) -0.00178 -0.00194 -0.000826 -0.00239 (0.000383) (0.000627) (0.000586) (0.00105) Wealth index -3.328 -3.471 -4.107 -4.420 (0.405) (0.303) (0.367) (0.403) Rural population -0.0319 -0.0261 -0.00946 -0.0238 (0.0458) (0.0401) (0.0480) (0.0626) Constant 3.288 1.628 1.288 2.447 2.957 (0.110) (0.142) (0.211) (0.166) (0.217) State Fixed-Effects No No Yes Yes Yes Observations 79118 79118 79118 49342 19518 Adjusted R-squared 0.034 0.420 0.463 0.414 0.380 Table 3: Number of children per woman, inclusive of any who died. Standard errors in parentheses: p < 0.10, p < 0.05, p < 0.01 25 Table 4: Cox Proportional Hazard Model (child sample) (1) (2) (3) (4) (5) (6) Total Effects: Female 0.430 0.339 0.359 0.357 0.364 0.360 (0.105) (0.105) (0.105) (0.105) (0.105) (0.105) Muslim -0.351 -0.558 -0.375 -0.372 -0.372 -0.366 (0.0694) (0.0704) (0.0696) (0.0697) (0.0709) (0.0724) Hindu Non-Dalit -0.472 -0.312 -0.144 -0.144 -0.149 -0.178 (0.0479) (0.0470) (0.0466) (0.0467) (0.0470 (0.0477) Muslim X Female -0.180 -0.482 -0.271 -0.268 -0.264 -0.254 (0.0693) (0.0705) (0.0704) (0.0705) (0.0713) (0.0730) Non-Dalit Hindu X Female -0.301 -0.236 -0.0409 -0.0401 -0.0417 -0.0662 (0.0477) (0.0469) (0.0474) (0.0474) (0.0476) (0.0481) Coefficients: Muslim -0.458 -0.707 -0.513 -0.509 -0.511 -0.499 (0.0720) (0.0730) (0.0728) (0.0729) (0.0739) (0.0754) Female 0.171 0.0757 0.104 0.103 0.107 0.112 (0.0462) (0.0466) (0.0466) (0.0466) (0.0466) (0.0467) Muslim X Female 0.107 0.149 0.138 0.137 0.139 0.133 (0.0946) (0.0952) (0.0947) (0.0947) (0.0947) (0.0947) Hindu Non-Dalit -0.625 -0.426 -0.261 -0.260 -0.266 -0.293 (0.0503) (0.0502) (0.0504) (0.0505) (0.0507) (0.0510) Hindu Non-Dalit X Female 0.153 0.114 0.117 0.116 0.117 0.115 (0.0651) (0.0652) (0.0651) (0.0651) (0.0651) (0.0650) Birth order 0.138 0.143 0.143 0.143 0.143 (0.00905) (0.00916) (0.00916) (0.00917) (0.00921) Previous sibling female -0.00499 0.00357 0.00274 0.00570 0.00510 (0.0315) (0.0314) (0.0314) (0.0314) (0.0314) Total number of siblings 0.171 0.0811 0.0811 0.0735 0.0649 (0.00888) (0.0102) (0.0102) (0.0105) (0.0107) Mother's age 0.0227 0.0228 0.0227 0.0215 (0.00473) (0.00474) (0.00473) (0.00471) Mother's age at marriage -0.0284 -0.0284 -0.0285 -0.0265 (0.00604) (0.00607) (0.00608) (0.00619) Mother never attended school 0.192 0.192 0.171 0.149 (0.0509) (0.0509) (0.0510) (0.0517) Mother completed primary school -0.244 -0.244 -0.243 -0.259 (0.0714) (0.0714) (0.0713) (0.0715) Father's age 0.000984 0.000917 0.00211 0.00400 (0.00365) (0.00366) (0.00367) (0.00377) Father never attended school 0.105 0.105 0.103 0.116 (0.0363) (0.0363) (0.0364) (0.0368) Land (acres) -0.00104 -0.00103 -0.00113 -0.00148 (0.00108) (0.00108) (0.00108) (0.00108) Wealth index -3.621 -3.591 -3.501 -3.065 (0.285) (0.286) (0.290) (0.308) Rural -0.0523 -0.0508 -0.0615 -0.0762 (0.0443) (0.0444) (0.0457) (0.0468) Ate Meat -0.290 -0.279 -0.263 (0.173) (0.173) (0.173) Ate Plant Protein 0.0861 0.0792 0.0578 (0.128) (0.129) (0.128) Continued on next page 26 Table 4: Cox Proportional Hazard Model (child sample) (1) (2) (3) (4) (5) (6) Ate Dairy 0.0981 0.125 0.118 (0.178) (0.178) (0.177) Hemelevel -0.00155 -0.00152 -0.000893 (0.000925) (0.000926) (0.000954) Average measles vaccinations -0.435 -0.282 (0.110) (0.114) Average BCG vaccinations 0.0919 0.158 (0.157) (0.163) Average polio vaccinations 0.160 -0.0404 (0.127) (0.136) Average DPT vaccinations 0.0702 0.0599 (0.151) (0.156) Average female LFP -0.00575 0.194 (0.0756) (0.0871) Average female self-employed 0.0659 0.267 (0.151) (0.165) State Dummies No No No No No Yes N 195080 195080 195080 195080 195080 195080 Chi-squared 286.7 1977.8 2449.8 2471.8 2502.2 2912.8 Table 4: Cox Proportional Hazard Model. Standard errors in parentheses: p < 0.10, p < 0.05, p < 0.01 27 Table 5: Mortality Regressions (female sample), Dependent Variable: Fraction of woman's children who died Boys Girls All mothers Age30 Age40 All mothers Age30 Age40 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Muslim -0.758 -0.986 -1.045 -1.491 -1.898 -0.586 -0.863 -1.086 -1.286 -1.516 (0.365) (0.240) (0.227) (0.281) (0.377) (0.338) (0.205) (0.197) (0.264) (0.393) Hindu Non-Dalit -1.166 0.00612 -0.171 -0.201 -0.463 -0.916 0.245 -0.0565 0.0199 0.0654 (0.210) (0.158) (0.145) (0.184) (0.305) (0.207) (0.139) (0.100) (0.105) (0.166) Total Children 1.130 1.145 1.163 1.019 1.120 1.098 1.094 1.055 (0.0676) (0.0622) (0.0627) (0.0693) (0.0703) (0.0565) (0.0678) (0.0699) Woman's age -0.0204 -0.0246 0.0308 0.0437 -0.0473 -0.0557 -0.00307 0.0257 (0.0135) (0.0127) (0.0169) (0.0336) (0.0143) (0.0131) (0.0170) (0.0330) Woman's age at marriage -0.0237 0.000853 0.000619 0.00818 -0.00182 0.0228 0.0269 0.00974 (0.0128) (0.0116) (0.0130) (0.0204) (0.0150) (0.0117) (0.0135) (0.0231) Woman never attended school 0.457 0.391 0.516 0.381 0.540 0.369 0.506 0.265 (0.126) (0.132) (0.143) (0.268) (0.116) (0.125) (0.149) (0.315) Woman completed primary school -0.196 -0.196 0.0540 -0.190 -0.321 -0.323 -0.143 0.0183 (0.119) (0.113) (0.147) (0.205) (0.131) (0.134) (0.151) (0.212) Husband's age -0.0157 -0.0131 -0.00659 0.00839 0.000259 0.0125 0.0102 -0.00000417 (0.0110) (0.0114) (0.0163) (0.0236) (0.0133) (0.00985) (0.0110) (0.0194) Husband never attended school 0.175 0.177 0.147 -0.165 0.510 0.527 0.384 0.459 (0.125) (0.132) (0.131) (0.249) (0.135) (0.133) (0.163) (0.238) 28 Land (in acres) 0.00549 0.00418 0.00327 -0.00131 0.00650 0.00451 0.00802 -0.00466 (0.00241) (0.00230) (0.00249) (0.00675) (0.00221) (0.00227) (0.00346) (0.00516) Wealth index -7.354 -5.680 -6.353 -10.96 -5.585 -4.633 -5.393 -7.930 (0.810) (0.728) (0.844) (1.270) (0.624) (0.658) (0.775) (1.256) Rural -0.177 -0.0107 -0.0757 -0.207 -0.0855 0.0341 -0.0685 -0.0822 (0.127) (0.136) (0.157) (0.264) (0.115) (0.115) (0.111) (0.173) Constant 4.241 1.835 1.007 -1.313 -1.134 3.591 0.995 0.422 -1.796 -1.681 (0.368) (0.379) (0.384) (0.477) (1.383) (0.368) (0.388) (0.374) (0.471) (1.363) Observations 79118 79118 79118 49342 19518 79118 79118 79118 49342 19518 Adjusted R-squared 0.002 0.047 0.050 0.071 0.076 0.002 0.054 0.057 0.079 0.086 Table 5: Notes:(i) Standard errors in parentheses: p < 0.10, p < 0.05, p < 0.01. Table 6: Cluster-level mortality regressions (village sample) Fraction of Boys Died Fraction of Girls Died (1) (2) (3) (4) (5) (6) Muslim population -0.00719 -0.00519 -0.00362 -0.00635 -0.00502 -0.00556 (0.00552) (0.00318) (0.00324) (0.00614) (0.00319) (0.00322) Hindu Non-Dalit population -0.00913 0.00144 0.000537 -0.00450 0.00503 0.00114 (0.00354) (0.00235) (0.00212) (0.00360) (0.00270) (0.00211) Average female age 0.000199 0.000172 0.0000599 0.0000116 (0.000119) (0.000113) (0.000116) (0.000112) Average female age at marriage -0.000514 -0.000307 -0.000230 0.0000772 (0.000279) (0.000261) (0.000237) (0.000259) Average female education -0.0135 -0.0119 -0.0108 -0.00779 (0.00187) (0.00184) (0.00221) (0.00183) Average male education 0.000496 0.000180 -0.00182 -0.00176 (0.00229) (0.00204) (0.00290) (0.00202) Rural population -0.00170 0.00149 0.00264 0.00575 (0.00248) (0.00233) (0.00295) (0.00231) Average land (in acres) 0.0000233 0.0000229 0.0000659 0.0000560 (0.0000850) (0.0000642) (0.0000641) (0.0000638) Average wealth index scores -0.0000970 -0.0000811 -0.0000782 -0.0000637 (0.0000140) (0.0000134) (0.0000126) (0.0000133) Average of meat -0.0287 0.0142 -0.0812 -0.00396 (0.0176) (0.0206) (0.0149) (0.0205) Average of plant protein -0.0145 -0.0163 -0.0242 -0.0296 (0.0152) (0.0179) (0.0189) (0.0178) Average of dairy -0.0595 -0.0430 -0.0228 0.00292 (0.0227) (0.0228) (0.0237) (0.0226) Average of hemoglobin level -0.000639 -0.000539 -0.000719 -0.000568 (0.000164) (0.000147) (0.000164) (0.000146) Average of measles vaccinations -0.00463 -0.000318 0.000182 0.00416 (0.00530) (0.00533) (0.00487) (0.00529) Average of BCG vaccinations -0.0151 -0.00282 -0.0186 -0.00678 (0.0115) (0.00952) (0.0115) (0.00946) Average of polio vaccinations 0.0322 0.0200 0.0370 0.0116 (0.0126) (0.00827) (0.0128) (0.00822) Average of DPT vaccinations -0.0159 -0.0127 -0.0287 -0.0186 (0.00835) (0.00949) (0.00935) (0.00943) Average female LFP 0.000960 0.00657 -0.00298 0.00778 (0.00654) (0.00476) (0.00607) (0.00473) Average female self-employment 0.00112 -0.00559 0.0162 0.0114 (0.0112) (0.00943) (0.0118) (0.00938) Constant 0.0766 0.185 0.158 0.0656 0.189 0.155 (0.00546) (0.0195) (0.0187) (0.00572) (0.0208) (0.0186) State Fixed Effects No No Yes No No Yes N 3644 3637 3637 3644 3637 3637 R-squared 0.00487 0.202 0.236 0.00152 0.201 0.250 Table 6: Regressions based on 3644 clusters. Standard errors in parentheses: p < 0.10, p < 0.05, p < 0.01. 29