Version classiqueVersion mobile

Gender discriminations among young children in Asia

Isabelle Attané
Jacques Véron

Part I - Gender discriminations among young children in India

2. Persistent Daughter Disadvantage in India: What Do Estimated Sex Ratios at Birth and Sex Ratios of Child Mortality Risk Reveal?1

S. Irudaya Rajan et S. Sudha


Dans cet article, les auteurs étudient les niveaux et les tendances des rapports de masculinité infantiles et de la mortalité juvénile en Inde, de même que les facteurs socioéconomiques qui leur sont liés, en utilisant les données des recensements de 1981 et de 1991. La question des “femmes manquantes” a émergé de longue date en Inde, puisqu’un déficit de femmes a été pour la première fois relevé au recensement de 1871. Depuis lors, le ratio hommes/femmes dans la population a évolué en faveur des hommes, en dépit de quelques retournements ponctuels aux recensements de 1981 et de 2001. Bien que le recensement de 2001 montre un accroissement de 6 points du rapport de masculinité global par rapport au recensement de 1991 (933 et 927 femmes pour 1 000 hommes respectivement), cela ne remet pas en question le déficit de femmes. Les recherches récentes soulignent en outre la persistance d’une surmortalité féminine aux jeunes âges.
Cette recherche comporte deux volets. Ses auteurs ont tout d’abord cartographié, à l’échelle des districts, les rapports de masculinité des naissances estimés et les quotients de mortalité entre les âges de 0 et 5 ans (0q5) aux recensements de 1981 et de 1991. Ils ont ensuite conduit une analyse statistique multivariée des déterminants socioéconomiques de la surmortalité féminine pour ces mêmes années.

Texte intégral


  • 1 This paper is a revised version of the paper presented at the PAA Annual Meeting 2001. We thank Mon (...)

1In this paper we examine levels and trends of estimated sex ratios at birth (SRB) and sex ratios of under 5-mortality risk in India, as well as the socioeconomic correlates of the latter, using 1981 and 1991 census data. The issue of India's “missing women” has raised concern since the abnormal female-deficit population sex ratio was first noted in the 1871 census. Since then, the ratio has grown almost steadily more masculine, despite small upswings in the female proportion in 1981 and 2001. Though India’s 2001 census shows a 6-point increase in the population sex ratio to 933 females per 1000 males (as opposed to 927 females per 1,000 males in 1991), this still shows a substantial female deficit.

2Alarmingly, the female-to-male sex ratio among children aged under 6 years decreased from 945 in 1991 to 927 in 2001 (Registrar General of India, 2001). Earlier research debated the accuracy of census coverage as an explanation, e.g. double counting migrant men or undercounting women (review in Krishnaji 2000). Recent research, however, shows that the phenomenon is mostly due to persistent female mortality disadvantage in infancy and childhood, a view also supported by projections and simulations (Griffiths et al. 2000).

3Interpreted as a restriction of girl children’s right to live, female disadvantage in child mortality is rightly taken as one of the most significant indicators of gender bias in India. This phenomenon is not due to a greater natural frailty of girl children, but results from the practices of parents that would discourage the life chances of unwanted daughters through selective neglect or infanticide. Despite socio-economic development, fertility decline, and falling mortality for both sexes, the male-to-female child mortality gap did not shrink during 1981-1991 (Das Gupta and Bhat 1997), but also spread into hitherto egalitarian parts of the country (Basu 1999; Rajan et al. 2000). Much research investigates socio-economic and cultural correlates of excess female child mortality in India up to 1981 (e.g. Kishor 1993; Murthi et al. 1996).

4Related research enquires as to whether pre-natal sex selection techniques are spreading in India to weed out unwanted daughters before birth, rather than the age-old postnatal methods (infanticide and neglect) hitherto used. New technology and the spread of medical facilities in India since the early 1980s enable of this possibility. Das Gupta and Bhat (1997) conclude that the juvenile sex ratio (ages 0-6) in India grew more masculine during 1981-1991 due to the rise of sex-selective abortion. Reliable data on period SRB, an indicator of sex-selective abortion, are not available for all-India analyses. We attempt to fill this data gap by means of indirect estimation methods. Our previous paper on state and rural/urban patterns (Sudha and Rajan 1999), notes an increase in ‘masculine’ indirectly estimated SRBs 1981-1991 in urban North/North-West India, suggesting increasing prenatal sex selection there. This interpretation is supported by small-scale, local-level research and NGO reports.

5This research has two analytic aims. First, we map district-level data on estimated SRB and on observed child mortality (0q5) risk sex ratios from the Indian censuses of 1981 and 1991. Most prior district-level studies of gender bias in India analysed 1981 data only. We include 1991. Second, we conduct multivariate statistical analyses of socio-economic correlates of female disadvantage in death (female dominant child mortality sex ratio), for 1981 and 1991 (similar analyses for 2001 are not possible at this time as the relevant data have not yet been released). Most prior analyses have examined correlates of gender bias in infant/child mortality only up to 1981. Our discussion focuses on women’s education and economic activity, as these important characteristics are emphasized in the literature as affecting gender bias, and are underscored in policy recommendations to promote women’s empowerment.

6Prior discussion of gender bias in India distinguishes between conventional markers of economic development (urbanization, poverty, access to medical facilities, and male literacy and work), and indicators of women's societal position (female education, labour force participation and measures of different marriage systems) (Kishor 1993; Murthi et al. 1996). The findings broadly suggest that conventional economic development does not automatically reduce gender bias. Conversely, many indicators of women's status, conceptually associated with female empowerment, are related to improved life chances of girl children.

Factors influencing gender bias

Literacy and education

7There is a difference between the impact of male and female education or literacy on gender bias in child survival. Education or literacy of men had no impact on gender bias according to Kishor (1993), but increased female disadvantage in the analyses of Murthi et al. (1996). Female education has long been advocated to promote declines in fertility and child mortality (Dreze 1999; Dreze and Murthi 2001). However, the relationship between maternal education and gender bias can be positive or negative depending on region, parity or contextual details. A positive relationship may exist either because educated women more efficiently withhold high-quality care from less desired children (Das Gupta 1987), or because they have lower fertility, which is accompanied by greater gender bias since parents have a narrower latitude to ensure the desired number of sons (Das Gupta and Bhat 1997). For a negative relationship, Murthi et al. (1996) associate education with greater female agency and decreased gender bias. Further examination of trends over time is clearly needed.

8The idea that women's education promotes female agency and reduces gender inequality is reasonable. Education is a fundamental human right and a necessary pre-requisite for socio-economic development, particularly to improve the status of women. Nor are the findings that women’s schooling is sometimes linked with increased gender bias to be interpreted as suggesting limiting female education as a solution. However, to understand the at times contradictory findings, we need to consider the content and context of literacy/education in India, which few prior studies have done. Regarding content, education may assist men and women to function as rational actors in a modernizing economy, enabling them to understand and negotiate the system. However, it may not automatically lead them to transform the system. Longwe (1998) distinguishes between the concepts of education for empowerment and. schooling for subordination. Swaminathan (1991) states that Indian education in general is directed toward the privileged classes and reinforces rather than transforms traditional prejudices. This is as true of gender inequality as of any other social inequality.

9Regarding context, literacy and education in India reflect not only the provision of schooling, but also familial decisions on distributing scarce resources among children, i.e. whose education is to have priority over that of others. In other words, the educational investment of parents in children reflects gender bias, and children schooled under these conditions may not be led to transform gender inequality. In India, even basic literacy for females lags behind that of males. The all-India 2001 male literacy rate stood at about 75.9% and the female rate at 54.2%, a 22 percentage-point gap (Government of India, ND). Though this gap varies according to age, region, and class, Datta and Sinha (1997) describe women as the ‘depressed class’ as far as literacy is concerned.

10Even when girls achieve literacy, their schooling attainment illustrates familial gender bias. For example, parents in rural Maharashtra educate daughters to help them be independent and provide support during economic hardship, but also limit girls’schooling to not conflict with rural life practices and the need for early marriage (Vlassof 1994). Parents invest more in the education of children from whom they expect more returns, i.e. sons are withdrawn from household labour and sent to school, while their sisters focus on housework. This holds for smaller families, too, as fewer children mean more labour shortfalls (Jejeebhoy 1992). In asset-poor rural Gujarat households, girls perform large amounts of housework, resulting in lower school attendance and higher dropout rates, particularly after puberty (Unni 1998).

11Thus, though female literacy is rising in India, gender gaps in levels, content and context of education remain entrenched. Little is known whether girls educated under these conditions can achieve egalitarian aspirations for their own daughters and sons. Female education is necessary for women's advancement, but the content and context of schooling in India do not automatically enable of the attainment of this goal. Thus, the relationship between women's literacy / education and gender bias, including changes over time, need more exploration.

Economic activity

12Economic activity, specifically paid work participation, is another important factor considered by studies on gender bias. The literature again distinguishes between the economic activity of men and women. Although mortality in general may be higher among children of working mothers, women’s paid work is associated with lower gender differences in child mortality. Explanations suggest that women in the paid work force contribute economically to the household, raise their own value and that of their daughters, are less dependent on (male) family members for survival and are able to make and implement decisions autonomously. Clearly, paid work is distinct from the household labour performed by girls who are kept at home from school, as mentioned in the previous section.

13However, reviews suggest gender biases in the Indian labour market. In the rural sector, women constitute about 26-34% of the workforce. In the urban sector, the share of women remained at 20-21% from the late 1970s to the mid-1990s (Visaria 1996, cited in Gothoskar 2000). The share of women in the total workforce peaked during the mid-1980s, dropped by about 1 percentage point in the late 1980s and remained there until the mid-1990s. The share of the service sector in urban employment has risen for both sexes, but especially women. Though women’s share of employment in the organized sector increased in the 1980s, this was mainly in public-owned enterprises, in a few traditionally female occupations. Outside agriculture, over 88% of women were working in the unorganized sector (a much greater proportion than that for men) (Gothoskar 2000)

14Though opportunities for educated, skilled, urban women are growing, these are marked by wage, occupational and “pre-entry human capital discrimination” (Datta and Sinha 1997: 57). Deshpande and Deshpande (1992) note feminization in the urban manufacturing sector. Employers prefer women workers because they earn at least 25% less than men for similar jobs, are less easily unionized, hold semi-skilled occupations, and are thought to have the patience and manual dexterity for repetitive work. In 1961-1981, even women in top-tier professional and scientific occupations faced a 21% male-female wage gap, 67-70% of which was attributed to discrimination, i.e. gender differences in productivity were ruled out as an explanation (Duraisamy and Duraisamy 1998). Women's work participation appears subordinated to the needs of the family, even among the uppermost classes (Desai 1996).

15Thus, the economic contributions of women to the household are of great importance to their status, but societal gender inequality keeps them from being viewed or paid as equal participants. Working women’s status may improve, but patriarchal societal ideology remains firmly in place (Varma 1993). Nevertheless, economically active women are still in the best position to overcome bias against daughters. We examine this association in 1981 and 1991.

Other factors

16Additional factors considered in the literature include socio-economic variables indicating the general level of development: urbanization, infant mortality rate, male literacy rate and male work participation rate. Though development has been progressing in India since Independence, the trajectory has been slow and uneven, and the population sex ratio has generally grown more masculine. We examine how the variables indicating development might affect gender bias (female excess in deaths). Until 1981, urbanization and male literacy were associated with excess female deaths (Murthi et al. 1996).

17Gender bias also varies across socio-cultural groupings in Indian society. Although scheduled (lower) castes were often characterized by greater gender egalitarianism, they do not exhibit any less gender bias in child outcomes. Scheduled tribes, on the other hand, whose cultures lie outside mainstream Hinduism, appear to have less gender-biased child outcomes, at least up to the 1980s. Although Hindu culture is plural and has certain traditions supporting respect for women, specific customs such as hypergamy, female seclusion and dowry have been associated with gender bias. These customs seem to be spreading across all caste and religious groups and might be associated with the persistence of gender bias. While Islam in India is associated with greater conservatism regarding women’s education, work participation and seclusion, there are also Koranic injunctions enjoining good treatment of daughters along with sons, and prohibiting female infanticide. We investigate how varying proportions of different cultural groups (Hindus, Muslims, Scheduled Castes and Tribes) are associated with female excess deaths in 1981 and 1991.

18We also examine the impact of region. While it has been long established that the southern region is more gender egalitarian than the North in marriage systems, status of women and child outcomes, evidence suggests that in the late 1980s and 1990s, gender bias also penetrated South India (Basu 1999). Social and economic changes in South India have produced many factors associated with gender bias, namely the spread of dowry and high female unemployment. We investigate how region is associated with gender bias in child deaths in 1981 and 1991.

19We control for the per cent of couples using contraception, total fertility rate, F/M population sex ratio, and household size. Dreze and Murthi (2001) argue that fertility decline in India is retarded by son preference, and that gender bias in child mortality is related to higher rather than lower fertility.

Data and Methods

20We use district-level data from the Indian censuses of 1981 and 1991. The data include social and economic indicators for districts of India released by the Office of the Registrar General of India. We supplement this with district-level 0q5 mortality risk estimates for 1981 and 1991 (Registrar General of India, 1998). We compiled district counts of male and female children aged 0-4 from various reports of the 1981 and 1991 censuses (Registrar General of India, 1991 and 1998).

21As a result, we have information for each district on the following:

221. Demographic indicators:

    1. Per cent of couples using family planning methods

    2. Total fertility rate

    3. Population density

    4. Population sex ratio F/M

    5. Household size

232. Socio-cultural indicators

    1. Per cent Hindu

    2. Per cent Muslim

      • 2 These religious categories are not exhaustive and mutually exclusive. That is, they do not amount t (...)

      Per cent Christian2

    3. Per cent Scheduled Caste

    4. Per cent Scheduled Tribe

243. Development indicators

    1. Infant mortality rate

    2. Per cent urban

    3. Male literacy rate

    4. Male work participation rate

254. Women's status indicators

    1. Female literacy rate

    2. Female work participation rate

    3. Women's average age at marriage

265. Region

27a. Districts are grouped to indicate North/North-West India (including Delhi, Haryana, Himachal Pradesh, Jammu and Kashmir 1981 only, Punjab, Rajasthan, and Uttar Pradesh), South (Andaman and Nicobar Islands, Andhra Pradesh, Karnataka, Kerala, Lakshadweep Islands, Pondicherry, and Tamil Nadu), North-East (Arunachal Pradesh, Assam 1991 only, Manipur, Meghalaya, Mizoram, Tripura, and West Bengal), Central and Western India (Bihar, Madhya Pradesh, Maharashtra, Orissa, Goa, Daman and Diu).

286. Patterns of birth and death

    1. Numbers of male and female children aged 0-4, recorded by the censuses, which we use (along with the male and female child 0q 5 mortality risk rates recorded in the censuses) to estimate the sex ratio at birth for 1981 and 1991 by reverse survival methods.

    2. Male and female child 0q5 rates, which we use to construct our measure of female disadvantage in infant/child mortality risk by district for 1981-1991.

Estimating the sex ratio at birth

  • 3 We take district-level counts of boys and girls aged 0-4 and male and female q5 mortality probabili (...)

29Though the period SRB is a strong indicator of prenatal sex selection and is thus emerging as a significant data need for India, there is no systematic effort to release national-level data in this regard that would enable of direct analysis. We therefore attempt to fill this gap by applying the reverse survival technique (UN Manual X, 1983, Chapter VIII), as was also done in our earlier paper (Sudha and Rajan 1999). This technique is based on the notion that children aged x are the survivors of births that occurred x years ago. Therefore, we can take the numbers of children recorded at age x, as well as the observed mortality risk for children in that population and, using a suitable model life table for the population in question, “resurrect” the numbers who had died3. From the result we take the ratio of male to female children, and estimate the SRB that we present in maps. Because this measure is estimated, and constructed on the district level (and is thus based on smaller numbers than the state level and consequently less robust), we only present these data in the form of maps and do not conduct multivariate analysis.

30SRB values more masculine than 107 males per 100 females (M/F) are treated as an indicator of “weeding out” of girls, either through prenatal sex selection, or the under-reporting of female births; both mechanisms are varying forms of bias against girls, denying them physical or social existence. In our maps we draw attention to regions of India where the SRB is greater than the “normal” range (SRB>=107 M/F).

Calculating 0q5 sex ratios

31Infant/child 0q5 mortality risk sex ratios are straightforward, based on the recorded 0q5 mortality risk for males and females reported in the censuses of 1981 and 1991. Again, the global norm in societies where gender bias is not marked shows that male infants and children are more likely to die. We therefore present 0q5 sex ratios, drawing specific attention to areas where female deaths outnumber male deaths (0 q 5 SR <100 F/M).

Analytic strategies

32First, we present district level maps showing the pattern of estimated M/F SRB and M/F 0q5 SR for 1981 and 1991. In an associated descriptive table we present the range of values in 1981 and 1991.

33Second, we conduct multivariate analyses with F/M 0q5 SR 1981 and 1991 as dependent variables. To correct for spatial correlation in district-level analyses of India, we use the statistical package SAS (PROC MIXED). This package/procedure uses information on the geographical x and y coordinates for the centroid of each district of India in 1981 and 1991 in order to take spatial correlation into account and compute unbiased estimates.

34For the multivariate analyses, we appropriately transform the q5SR to highlight female disadvantage. We performed a logarithmic transformation of the q5f/q5m ratio; thus, higher values indicate female disadvantage. We conduct separate analyses for 1981 and 1991.

35We also note that the number of districts in India increased between 1981 and 1991. Based on areas where the census was successfully completed, we have data from 398 districts in 1981 and 452 in 1991. We do not discard any extra districts in 1991, and note that there is thus more variability in the data in 1991.


36Figure 1 (a) and 1 (b) show district-level maps of M/F estimated SRB for 1981 and 1991 respectively. Each dot on the map represents one district. For all maps, darker dots indicate female disadvantage. For 1981, in Figure 1 (a), the estimated SRB ranges from 91 to 112 males for 100 females. The underlying data show that 29 districts have a value of estimated SRB >= 107. Ten of these are clustered in Punjab alone and another five in adjacent Haryana state. These are areas with among the greatest historical gender bias, and the most evidence of prenatal sex selection technology by 1981. We also see many districts with a SRB of less than 102 males per 100 females. Figure 1 (b) shows in 1991 that there are many more districts with an estimated SRB of >= 107, with values now as high as 117, blanketing most of Punjab and Haryana, and also more than before in Rajasthan and Gujarat, western UP, MP, Himachal Pradesh and Maharashtra. We also see traces of this phenomenon in southern India.

37Figure 2 (a) and 2 (b) show district figures for M/F 0q5 SR. In 1981 (Figure 2a), the values range from 67 to 128 M per 100 F. Female disadvantage, i.e. values of <100, are mostly concentrated in North India, stretching from the western boundary states through UP, MP, Bihar, to West Bengal (with traces in the North-East and the Peninsula). In 1991 (Figure 2b), the values range wider than in 1981, as low as 58 and as high as 165. Female disadvantage remains strong in the northern belt and also deeply penetrates all four major southern states. We also see more such points in the North-East. For values of > 100, Agnihotri (2000, Chapter 3) relates high male infant/child mortality to harsh health environment, low levels of health infrastructure, and general malnutrition. High male child mortality is worthy of future analyses, but is beyond our current scope.

38The maps show that by 1991, the same areas in North and North-West India show masculine estimated SRBs concurrent with excess female child mortality. That is, gender bias in birth and death patterns appear to operate simultaneously in these areas. These foreshadow the findings of the 2001 census of India, in which especially the states of Punjab, Haryana and the Delhi UT have the least proportion of females to males age 0-6. We further elaborate this point below.

39Table 1 describes the underlying data on estimated M/F SRB and F/M 0q 5 SR. The average all-India estimated M/F SRB grew a little more male-dominant (within the normal range) over the decade, and the F/M 0q5 SR barely changed. The proportion of districts with masculine estimated SRB tripled from 1981 (about 6%) to 1991 (about 18%). The proportion of districts with female disadvantage 0q5 SR remained at around 58% and grew more dispersed in location and range. Thus, the gender gap in child mortality did not diminish in this decade.

40Table 2 describes the socio-economic covariates, indicating general changes in 1981 and 1991. We see a decline in fertility and child mortality, a rise of 1 year in female mean age at marriage, growth in male and female literacy and labour force participation, and an increasing urbanization and population density of the country. We note the well-known decline in the 1991 F/M population sex ratio after the small upswing seen in 1981 following a century's decline.

41We now turn to the multivariate statistical analysis. Because higher values of the dependent variables indicate female disadvantage, positive coefficients suggest covariates associated with more gender bias. Table 3 presents the parameter estimates for the impact of social and economic variables on the likelihood of female disadvantage 0q5 SR, corrected for spatial correlation, for 1981 and 1991. We discard the variables on population density in the district, and per cent Christian, as exploratory analyses showed them to have no significant effect. Model statistics (not presented) indicate adequate fit to the data.

42Findings for 1981 suggest that women's literacy and work participation are associated with less female child mortality disadvantage. Per cent urban is also associated with lower risk. As found in prior studies, the male work participation rate is associated with heightened risk. Sociocultural variables such as per cent scheduled tribe and scheduled caste have negative coefficients, in line with prior studies. Per cent Muslim also has a negative sign. The north-western region shows higher female disadvantage. The southern region has no significant net negative effect (though it did in bivariate associations). Male literacy, as prior studies found, is marginally associated with increased female child mortality disadvantage. The total fertility rate is associated with increased, but F/M population sex ratio with decreased, female disadvantage. In 1991, the scheduled tribe proportion is still associated with decreased female disadvantage, but the scheduled caste proportion is no longer so. Female literacy also no longer has an impact. But male and female work participation are both associated with lower female disadvantage. The urban proportion is still associated with less female disadvantage. The southern region has no significant effect (which had been found in a bivariate analysis). The north-western region is now associated with significantly less female mortality disadvantage, though bivariate associations had shown no significant difference from the rest of India, and the maps suggested a female disadvantage in 0q5 mortality ratio. The population sex ratio is still associated with less female disadvantage.

43Female literacy appears to have lessened in its negative association with disadvantage for daughters. Women's work participation lowers female mortality disadvantage.

  • 4 We derive our interpretation from the fact that throughout India sex ratios are growing more mascul (...)

44The changing impact of modernization factors, specifically per cent urban and male literacy, that are associated with decreased female mortality risk in 1991, suggest that modernization is associated with a change in strategies; families might be turning to prenatal elimination of unwanted daughters rather than postnatal4. This point is elaborated below.


45The levels and trends of female disadvantage in birth and death suggest that development in India is not accompanied by declining gender bias. Despite declines in fertility and overall child mortality, female disadvantage in 0q5 mortality risk has persisted, spread to more areas, and shows more negative values in 1991 than in 1981. Also, over the same time, the values of masculine estimated SRB and the number and proportion of districts showing very masculine values has gone up.

Is elimination of unwanted daughters occurring before or after birth?

46In countries such as China and South Korea, with different development trajectories and standards of living than India, but nevertheless sharing a strong cultural son preference, prenatal sex selection techniques seem to be replacing postnatal methods (Zeng Yi et al. 1999; Goodkind 1996). That is, SRBs are growing abnormally masculine, while sex ratios of infant and child mortality are growing more egalitarian. Fewer girls are being born, but those born are more wanted and tend to survive.

47Is India experiencing this kind of substitution, or are girls in India facing a “double jeopardy” of increasing prenatal elimination concurrent with persistent postnatal risk? On the one hand, the average estimated SRB in India grew 3 percentage points more masculine between 1981 and 1991. At the same time, the average gender gap in child mortality still shows female disadvantage. That is, in the aggregate, in India pre-and postnatal risk to daughters may run concurrently, a scenario that suggests intensifying gender bias (though we caution that our suggestion is based on estimated SRB and needs further verification).

48Some evidence hints that both risks may be concentrated in the same social groups. Sureender et al. (1997) show that in Punjab, daughters of women who approve of female foeticide survive at half the rate of daughters of women who do not, though data on actual abortions were not analysed. Moreover, 31% of daughters of the former are not immunized, compared to 23% of those of the latter.

  • 5 On page xvii.

49Further, the 2001 census also showed that the sex ratio (number of F per 1,000 M) in the age group 0-6 has fallen greatly from 945 in 1991 to 927 in 2001. In Punjab specifically, the provisional 2001 census results show only 793 girls per 1,000 boys age 0-6. Our 1991 data showed estimated M/F SRBs of 117 and 118 for rural and urban Punjab respectively, and an M/F 0q5 sex ratio of 92 for both locations (Sudha and Rajan 1999, Appendix Table 1 and 2). The steep fall in the age 0-6 sex ratio in the 2001 census for Punjab, Haryana and Delhi may be an extension of the “double jeopardy” pattern noted 10 years earlier. This idea is further supported by the Haryana State report of NFHS-2 (1998-99) that notes5 that prenatal selection and female disadvantage in child mortality are both responsible for the very low 0-6 sex ratio of the state. NFHS-2 reports from Punjab, Delhi and other North Indian states also make this point.

50The bivariate correlation between estimated SRB and 0q5 sex ratio (higher values of both indicating female disadvantage) in 1981 was. 38 (p <. 0001) and in 1991 was. 25 (p < .0001). This suggests that districts with higher female deficit in births also have greater female excess in deaths, though the strength of the association seems to diminish over time. We repeat that our SRB statistics are based on estimates, thus the conclusions need additional confirmation.

51On the other hand, local-area or household-level substitution patterns may be obscured in the aggregate, i.e. the same communities are not eliminating unwanted daughters both pre-and postnatally. One study shows that urbanizing Jat families in the outskirts of Delhi form the clientele of flourishing sex determination/abortion facilities in the area (Khanna 1997). These clinics are widely used to realize the reproductive goals of Jat families of about 3 children, with at least 2 sons. However, a gender gap in infant/child mortality is no longer seen, suggesting substitution of pre-for postnatal methods where facilities for the former exist. Moreover, our multivariate analysis showed that the north-western region now was associated with lower gender bias in child mortality. Thus, at the district level, a situation of substituting prenatal methods may be occurring.

52However, in India, distinguishing between pre-and postnatal methods of eliminating unwanted daughters is not the central issue. Although some scholars suggest that abortion is psychologically easier to bear than neglecting a living child (Das Gupta and Bhat 1997), and others examine the ethical issues surrounding the restriction of prenatal sex selection (Goodkind 1999), gender bias against daughters seems so firmly entrenched in India that the parents’ choice of methods depends on availability and convenience rather than on conscience. The central issue is, therefore, the persistence of bias against daughters in modern India.

Modernization and persistent gender bias

53Scholars link the persistence of gender bias, including older or newer methods of eliminating unwanted daughters, to modernization processes in India (Harriss-White 1998; Wadley 1993). This link suggests the continuation of the familial strategies noted in 19th-century India (Clark 1983) of parental manipulation of family size and sex composition and the arrangement of children's marriage to maximize the social and economic status of the family. These strategies lead to disfavour against daughters because of the deeply gendered and interlinked nature of productive and reproductive processes. In modernizing India, too, efforts for socioeconomic advancement are still strategized by families along kinship network ties, of which children’s marriage forms a part. Despite increasing female education and work participation, familial socio-economic advancement is still viewed as largely achieved through males, and family’s value sons accordingly. Sons also increasingly bring in dowry. The participation of daughters in modernization is still subordinated to family and marriage needs, and seen as benefiting marital rather than natal families.

54The interlinking of productive and reproductive domains (represented by work and marriage) lead to women's work and education being subject to the dictates of marriage necessities. India is still a society of early and near universal marriage for women. Kinship networks and social identity define, and are largely defined by, marriage and motherhood for adult women. Unmarried women (including widows, divorcees, separated, etc.) face greater social and economic vulnerability than married women, particularly if they have no male kin willing and able to support them. This adds further to the pressure to bear sons. Khanna (1997) and Gold (2001) describe how education and work decisions for females are always made with an eye to guarding female sexuality and the inevitability of marriage, in urbanizing Delhi and rural Rajasthan respectively. Our findings suggest that female education does not entirely transform this scenario for women nor lead them to eradicate gender bias for their daughters, though female work participation has some protective effects in this regard.

  • 6 The authors are currently conducting research in Kerala on socio-economic change, marriage system c (...)
  • 7 Indian activists point out the nexus between dowry customs and daughter disadvantage. However, ther (...)

55Our study has no measure of marriage systems other than female age at marriage, which showed limited association with female disadvantage. However, the general lack of association of women’s education and the partial lack of association of women's work participation with reduced gender bias, the increased association of per cent Hindu with female disadvantage and decreased protective effect of per cent Scheduled Caste, can also be partly understood in the light of changing marriage systems in India. Across the Indian caste and regional spectrum, marriage arrangements are fundamentally undergoing transformation: dowry and nuclear family customs have deeply penetrated communities where they were never the norm, including formerly matrilineal and matrilocal communities in peninsular southern India. Though Caldwell et al. (1982) noted the beginning of the societal shift toward dowry in peninsular India, little research on this topic considers its reciprocal links with the changing status of women6. Economic analyses view the dowry trend as demographically determined by a young age structure, such that marriage-age women are competing for a smaller pool of men in an older age group (Rao 1993a, 1993b), or that the supply of widower grooms has declined due to improvements in female mortality (Bhat and Halli 1999). However, these studies do not address the critical question of why the response to demographic pressure is the spread of the dowry custom, despite the unaffordable sums demanded and the deadly violence that often enforces payment. Options such as women marrying men younger than themselves or remaining unmarried and independent are considered unthinkable for most Indian women7.

56Thus, in modernizing India where efforts to advance socioeconomically still depend on family and kinship ties and where males are seen as the main vehicles of these efforts, the multiple intersecting domains which permit only limited educational and occupational avenues for most women, and make marriage and dowry transactions inevitable for them, coalesce to make sons valuable and multiple daughters unaffordable for many parents.

57One point bears repetition. It is sometimes suggested that growing masculine sex ratios will ultimately lead to a shortage in the “supply” of women and, since “demand” for marriageable women remains, the “value” of women will rise. This argument is not tenable. The sex ratio of India has been growing more masculine since 1871, but girls continue to be eliminated. Wives are universally desired, but few want to raise daughters. Shortfalls in the “supply” of women will lead to their being subject to greater restrictions, control and violence, as in China, where the shortage of marriageable women in some areas has led to kidnapping and sale of women from other regions (Das Gupta and Li Shuzhuo 1999).


58This paper presented evidence from the Census of India to show that abnormally masculine estimated SRB grew more prevalent between 1981 and 1991, while female disadvantage in child mortality risks persisted. Statistical analyses suggested that women's status indicators are associated with partial protection to daughters, women’s work participation more so than female literacy. Male work participation and urbanization over the decade are associated with decreasing female disadvantage in child mortality, raising the question as to whether modernization may be associated with substitution of pre-for postnatal sex selection. Changes in cultural factors also suggest links to growing female disadvantage. That is, gender bias remains over the decade.

59Our literature review illustrated how gender inequality operates in the domains of education, employment and marriage, allowing advancement within the system for some Indian women, but not transforming gender bias for the majority. In the short run, therefore, while these biases remain, development alone does not promise to transform patriarchy or lessen female demographic disadvantage in India. As Banerjee (1998: 261) puts it: “In the Indian patriarchal ideology, women are regarded more as a highly flexible resource of the household rather than fully-fledged members of it”. The evidence in this paper may be best understood in this light.

60Maithreyi Krishnaraj (2000) states that the more than 50 years since Indian Independence is not a short period for realizing progress. The progressive provisions of the Indian constitution that promise much for women remain largely unrealized. Dreze and Sen (1995) assert that persistent gender inequality and the deprivation of females are among India's most serious social failures. Though women's groups in India work toward progressive change, more efforts from public and private sectors are needed to bring about fundamental social change to reduce gender inequality.

Table 1: Patterns in estimated sex ratios at birth (m/f) and 0q5 sex ratios (f/m) in India, 1981-1991

Table 1: Patterns in estimated sex ratios at birth (m/f) and 0q5 sex ratios (f/m) in India, 1981-1991

Table 2: Descriptive statistics of the social and economic information for each district of India, 1981-1991

Table 2: Descriptive statistics of the social and economic information for each district of India, 1981-1991

Table 3: Parameter estimates of social and economic indicators affecting masculine estimated sex ratio at birth and female disadvantage in child mortality risk, 1981-1991

Table 3: Parameter estimates of social and economic indicators affecting masculine estimated sex ratio at birth and female disadvantage in child mortality risk, 1981-1991

* significant at the. 10 level. ** significant at the. 05 level or better



Agnihotri, S.B. (2000), Sex ratio patterns in the Indian population: a fresh exploration, New Delhi, Thousand Oaks (Calif.): Sage Publications.

Banerjee, N. (1998), ‘Household dynamics and women in a changing economy’ in Maithreyi Krishnaraj, Ratna M. Sudarshan and Abusaleh Shariff (Eds.), Gender, Population and Development, Delhi: OUP.

Bhat, P.N., Halli, S.S. (1999), ‘Demography of brideprice and dowry: causes and consequences of the Indian marriage squeeze’, Population Studies, July 1999, Vol. 53, No. 2, pp. 129-48.

Caldwell, Caldwell and Reddy (1982), ‘The causes of demographic change in rural South India: a micro approach’, Population and Development Review, Vol. 8, No. 4, pp. 689-728.

Census of India (1981), ‘Child mortality estimates of India’, Occasional Paper No. 5 of 1988, Demography Division, Office of the Registrar General of India, Ministry of Home Affairs, New Delhi.

Clark, A.W. (1983), ‘Limitations on female life chances in rural Central Gujarat’, The Indian Economic and Social History Review, Vol. 20, pp. 1-25.

Coale, A. and Demeny, P. (1966), Model Life Tables and Stable Populations, New Jersey: Princeton University Press.

Das Gupta, M. (1987), ‘Selective discrimination against female children in rural Punjab, North India’, Population and Development Review, Vol. 13, pp. 77-100.

Das Gupta and Li Shuzhuo (1999), ‘Gender bias in China, South Korea and India 1920-1990: effects of war, famine and fertility decline’, Development and Change special issue on Gender, Poverty and Well-being, July 1999, Vol. 30, No. 3, pp. 619-652.

Das Gupta and Mari Bhat, P.N. (1997), ‘Fertility decline and increased manifestation of sex bias in India’, Population Studies, Vol. 51, pp. 307-315.

Datta, A. and Sinha, S. (1997), ‘Gender disparities in well-being, an overview’, Indian Journal of Gender Studies, Vol. 4, No. 1, pp. 51-65.

Desai, N. (1996), ‘Women's employment and their familial role in India’, in Shah A.M., Ramaswamy E.A. and Baviskar B.S. (Eds.), Social Structure and Change: Vol. 2, Women in Indian Society, New Delhi: Sage Publications, pp. 98-112.

Deshpande, S. and Deshpande, L.K. (1992), ‘New Economic Policy and Female Employment’, Economic and Political Weekly, Vol. 27, No. 41, pp. 2248-2252.

Dreze, J. (1999), ‘A surprising exception: Himachal's success in promoting female education’, Manushi: A Journal about Indian women and Society, No. 112, May-June 1999, Available on-line at:

Dreze, J. and Manta Murthi (2001), ‘Fertility, education and development: evidence from India’, Population and Development Review, Vol. 27, No. 1, pp. 33-64.

Dreze, J., Sen, A. (1995), India: economic development and social opportunity, Delhi, Oxford and New York: Oxford University Press, Clarendon Press.

Duraisamy Malathi and Duraisamy, P. (1998), ‘Sex segregation and discrimination among scientific personnel’in Maithreyi Krishnaraj, Ratna M. Sudarshan and Abusaleh Shariff (Eds.), Gender, Population and Development, Delhi: OUP.

Dyson and Moore, M. (1983), ‘On kinship structure, female autonomy, and demographic behavior in India’, Population and Development Review, March 1983, Vol. 9, No. 1, pp. 35-60.03+

Gold, A.M. (2001), ‘Education: New Light in New Times Women's Songs on Schooling Girls in Rural Rajasthan’, Manushi: a Journal about Women and Society, March-April 2001, No. 123, available on line at:

Goodkind, D. (1999), ‘Should prenatal sex selection be restricted? Ethical questions and their implications for research and policy’, Population Studies, March 1999, Vol. 53, No. 1, pp. 49-61.

Goodkind, D. (1996), ‘On substituting sex preference strategies in East Asia: does prenatal sex selection reduce postnatal determination’, Population and Development Review, Vol. 22, No. 1, pp. 111-126.

Gothoskar, S. (2000), ‘Teleworking and Gender’, Economic and Political Weekly, June 24th 2000, Special Issue on Teleworking and Teletrade in India, pp. 2293-2298.

Government of India, ‘Comprehensive Information on Indian Education on the Occasion of the Celebrations of the 50th Year of Indian Independence’, Overview statistics on sexwise rural-urban literacy, Ministry of Human Resource Development with National Informatics Centre (NIC), Available on-line at:

Griffiths, P., Matthews, Z. and Hinde, A. (2000), ‘Understanding the sex ratio in India: a simulation approach’, Demography, Vol. 37, No. 4, pp. 477-488.

Harriss-White, B. (1998), ‘Gender-cleansing: the paradox of development and deteriorating female life chances in Tamil Nadu’, in Sundar Rajan R. and Butalia U. (Eds.), Signposts: Gender Issues in Post-Independence India, New Delhi: Kali for Women.

Jejeebhoy, S.J. (1992), ‘Family size outcomes for children and gender disparities: the case of rural Maharashtra’, in Lloyd C. (Ed.), Fertility, family size and structure: consequences for families and children, Proceedings of a Population Council Seminar, 9-10 June 1992, New York: New York Population Council.

Khanna, S.K. (1997), ‘Traditions and reproductive technology in an urbanizing North Indian village’, Social Science and Medicine, Vol. 44, No. 2, pp. 171-180.


1 This paper is a revised version of the paper presented at the PAA Annual Meeting 2001. We thank Monica Das Gupta for comments; Gary Gaddy, Rick Homan and C. Suchindran for statistical advice, and Christopher Z. Guilmoto for assistance in preparing the maps.

2 These religious categories are not exhaustive and mutually exclusive. That is, they do not amount to 100%. Each religion variable stands alone, i.e. in contrast to all other religions in the district.

3 We take district-level counts of boys and girls aged 0-4 and male and female q5 mortality probabilities recorded in the 1981 and 1991 censuses, fitting a South Model Coale and Demeny Life Table (Coale and Demeny 1966). We thus “resurrect” the numbers of boys and girls under age 5 who died prior to the census enumeration, add them to the numbers of reported males and females aged under 5, and take the ratio of male to female children in the resulting total, to estimate the sex ratio at birth.
Clearly, no indirect estimation technique can produce results as good as accurate direct records. Additionally, it may be argued that reverse survival methods would not include daughters eliminated through infanticide, where neither the birth nor death is ever reported or recorded. This type of omission would in turn bias the estimated sex ratio at birth toward masculinity. However, female infanticide is mostly a localized practice, in rural pockets of some states such as Bihar, Madhya Pradesh and Tamil Nadu. This practice may not swamp the aggregate numbers. We also note the most masculine estimated SRB in urban areas of North-West India, where infanticide practices (though historically recorded) have not been recently reported, medical facilities are widely available, and prenatal sex selection centres have been reported to flourish. Also, our estimated SRBs at the state, rural and urban levels do not vary substantially from the observed sex ratio among infants aged 0 + 1 (Sudha and Rajan 1999, Appendix Table 1). Finally, our estimated SRB is almost identical with that estimated by the SRS,
e.g. 117 for Punjab (SRS, RGI 1999).
This gives us confidence in our estimate. We also acknowledge that the district-level estimations, based on smaller numbers than the state-level estimates, are less robust. But, in the absence of reliable records on period SRB, we argue that the estimated SRB can illustrate the impact of continuing son preference in India, during social change, economic development, declining fertility and mortality, and spread of new medical technologies. We therefore cautiously proceed with reporting of estimated SRB, and draw conclusions with these cautions in mind.

4 We derive our interpretation from the fact that throughout India sex ratios are growing more masculine, although male and female child mortality are declining. Moreover, exploratory multivariate analyses using very masculine values (i.e. >=107 M/F estimated SRB) show that in 1991, the same modernization variables that are less associated with female child mortality disadvantage are associated with increased masculine SRB.

5 On page xvii.

6 The authors are currently conducting research in Kerala on socio-economic change, marriage system change, dowry penetration, women’s status, and consequent valuation of sons as opposed to daughters. We hope our findings will illustrate some of these issues.

7 Indian activists point out the nexus between dowry customs and daughter disadvantage. However, there are multiple opinions regarding the dowry issue within the Indian women's movement. One view argues that dowry provides women with the only feasible avenue to claiming their share of parental property in a social climate that makes it difficult for them to enforce legal claims to other forms of inheritance (e.g. Kishwar 1999). Therefore, until other avenues to inheritance are ensured, dowry should not be entirely discouraged, but familial support structures that limit exorbitant demands and provide some security for women should be put into place. At the same time, families should begin to view daughters as capable of providing as effective old-age support as sons, and value them accordingly.

Table des illustrations

Titre Table 1: Patterns in estimated sex ratios at birth (m/f) and 0q5 sex ratios (f/m) in India, 1981-1991
Fichier image/jpeg, 84k
Titre Table 2: Descriptive statistics of the social and economic information for each district of India, 1981-1991
Fichier image/jpeg, 208k
Titre Table 3: Parameter estimates of social and economic indicators affecting masculine estimated sex ratio at birth and female disadvantage in child mortality risk, 1981-1991
Légende * significant at the. 10 level. ** significant at the. 05 level or better
Fichier image/jpeg, 260k


Centre for Development Studies
Ulloor, Prashanth Nagar
Trivandrum 695 011, Kerala, INDIA

Assistant Professor
Department of Human Development and Family Studies
University of North Carolina at Greensboro
228 Stone Bldg.
Greensboro NC 27402-6170, USA

© Institut Français de Pondichéry, 2005

Conditions d’utilisation :


Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search