While 50.6% of the population of the United Kingdom was female in 2020, only 48.0% of India’s population was female in the same year (World Bank, 2021). The differences in these percentages between the West and many countries in Asia, the Middle East, and North Africa led Amartya Sen to argue in 1990 that more than 100 million women were missing, and more than a century earlier, caught the attention of census administrators in India (Report of Census of 1881; Sen, 1990).
There are many explanations of why women are missing, and each of these has different implications for both where and when women should go missing – over space, over the life cycle, and across history (see here an earlier post on the topic at this blog). To describe these patterns, then, we have assembled novel data on the percentage of the population that is female in the countries comprising modern-day Bangladesh, Burma, India, and Pakistan every decade between 1881 and 1931, by religion, by district, and by age. We outline variation in the percentage of the population that is female by region, religion, geography, and age.
Missing Women: Patterns
By region, the literature has long noted that Northern India has more male-biased sex ratios than does the South. Several reasons may account for this difference, including greater participation of women in rice cultivation elsewhere than wheat cultivation in the North, or greater prevalence of dowry and patrilocal exogamy in the North (Bardhan, 1974; Dyson and Moore, 1983; Kishor, 1993). All of these would contribute to a greater preference for sons. We confirm the North-South gradient in our data (see figure below). Further, we show that sex ratios in 1931 are still predictive of sex ratios in 2011, even looking within states, though the predictive power of colonial sex ratios has declined over the twentieth century and is stronger at younger ages than later ages. This is evidence in favor of explanations of missing women that emphasize deeply rooted variables such as agricultural practices and marital residence patterns.
By religion, the literature has noted the importance of male relatives in performing last rites as a source of son preference among Hindus, particularly high caste Hindus, while often taking Muslims as a comparison group because of the more comparatively balanced Muslim sex ratio (Jayachandran, 2015; Visaria, 2015; Chakraborty and Sukkoo, 2010; Bhalotra et al., 2021). We do record differences on average across religions – the Sikh percentage female is below 45% in all years and the Buddhist percentage female is above 48% in all years (these differences correlate with differences by region, with Sikhs concentrated in Punjab and Buddhists concentrated in the territory of modern Burma). Muslims have slightly more male-biased sex ratios than Hindus, for example at 48.5% female in 1901 versus 49.2% for Hindus (see figure below).

But we also note that intra-religious differences in sex ratios across regions are often just as large. There are substantial inter-religious correlations in sex ratios at the district level when we focus on the largest religious groups; in a district where the population of Hindus is disproportionately male, for example, the Muslim population is likely to have the same skew. This is a pattern we find across age groups, and that is also apparent in the 2011 census. This is evidence, then, that either different religious groups influenced and continue to influence each other’s sex ratios, that there are common causes of the sex ratio across religions, or both. The greater male bias among Hindus than among Muslims that exists in the present is a recent phenomenon, possibly due to the greater aversion to abortion among Muslims (Bhalotra et al., 2021).
By geography, we merge our colonial data with several sources of Geographic Information Systems (GIS) data on variables that help determine the forces that the literature has claimed shape the sex ratio. Geographic suitability for wheat and rice, cultivation, for example, are exogenous predictors of the actual cultivation of wheat and rice. The relatively low participation of women in wheat cultivation in India has been cited as a possible source of missing women in wheat-growing regions. And yet, we find that in multivariate regressions the predictive power of wheat suitability is low and only dryland, rather than wetland rice suitability correlates with greater female shares once controlling for other variables. By contrast, variables that have been comparatively neglected in the literature, like suitability for malaria, emerge as significant predictors of the sex ratio.
By age, explanations of missing women rooted in son preference and dowry should predict that women go missing due to neglect in girlhood, and that this pattern should be religion-specific. While we do find many religions have fewer girls than boys in the 10-20 age bin, it is in the age bin 10-20 that this gap becomes largest. Missing women in adolescence are, then, more indicative of both maternal mortality and of misreporting of adolescent girls, and we find this pattern across several census years and across several religions.
Conclusions
Some religious groups, notably Muslims, have colonial sex ratios that differ from their current relative position. There is evidence of either spillovers or common determinants of sex ratios across religious groups within regions. While the geographic distinctions we find are consistent with a role for kinship and marital residence patterns in explaining missing women, the drop in the percentage female in adolescence is common across several regions and religious groups, suggesting a role for maternal mortality that cross-cuts kinship and marital residence. We do not provide broad support for explanations based on the determinants of the economic role of women, though we do find female labor force participation correlates positively with the population share female. And we believe that other factors, including the roles of disease and cash crops, have been comparatively neglected.
- Bardhan, P. K. (1974). On life and death questions. Economic and Political Weekly, 1293-1304.
- Bhalotra, Sonia, Irma Clots-Figueras and Lakshmi Iyer. 2021. “Religion and abortion: The role of politician identity.” Journal of Development Economics, 153:102746.
- Chakraborty, T. and Sukkoo, K. 2010. “Kinship institutions and sex ratios in India.” Demography, 47(4):989–1012.
- Dyson, T., & Moore, M. (1983). On kinship structure, female autonomy, and demographic behavior in India. Population and development review, 35-60.
- Jayachandran, S. (2015). “The roots of gender inequality in developing countries.” Annual Review of Economics, 7(1):63–88.
- Kishor, S. (1993). “May God give sons to all”: gender and child mortality in India. American Sociological Review, 247-265.
- Sen, A. (1990). More than a hundred million women are missing. New York Review of Books, 20, 61-66.
- Visaria, A. (2015). Religion and son preference in India and Bangladesh: Three essays on comparing Hindus and Muslims on son preference and sex differentials in child health. PhD thesis University of Pennsylvania
- World Bank. World Development Indicators, https://databank.worldbank.org/ (accessed 2021.07.30).
Professor (University of Warwick)
Professor (University of Warwick)
PhD Student (University of Warwick)