A Century of Missing Women

On the morning of 31 March 2011, C. Chandramouli, Registrar General and Census Commissioner of India, released the provisional totals of the fifteenth census. Two numbers about women went out that day. The first was that the sex ratio for the population as a whole had risen to 940 females per 1,000 males, the best figure since 1971. The second was that among children aged nought to six, the ratio had fallen to 914, lower than any count since 1961. Chandramouli called the second figure a matter of grave concern. When the final tables were published, both numbers moved up by a few points, to 943 and 919 respectively, which is why the two sets of figures circulating in Indian newspapers that decade never quite matched. Most of the coverage led with the first number.

A sex ratio is the cheapest thing a census produces. It requires no definition of poverty, no consumption basket, no decision about what counts as work. Two columns are added and one is divided by the other. That is the source of its authority: there is very little in it to argue about, and so the argument has to be about what the state did with what it found.

The column nobody read

The decline is present in the tables from the beginning. The census of 1901 returned 972 females per 1,000 males. Every subsequent count to 1941 returned a smaller figure: 964, then 955, then 950, then 945. The first census of the republic, in 1951, produced a single point of improvement, to 946, and then the fall resumed. By 1961 the figure was 941 and by 1971 it was 930.

There is an obvious objection to reading that slope as a fact about women rather than about enumerators, and it deserves a hearing. Colonial censuses were known to undercount women; households in much of northern India were reluctant to name them to a stranger with a schedule, and the schedules themselves were filled by men who had no particular reason to insist. If coverage of women was poor in 1901 and improved thereafter, then the recorded ratio should have climbed as the machinery got better at finding people it had previously missed. It did the opposite. Whatever the level of the early numbers, the direction of travel cannot be explained by an enumeration that was steadily improving.

What is remarkable about those seven decades is not that the figure was falling. It is that the figure was published, tabulated, sent to libraries and read by nobody as an accusation. The census reports of the period discuss it as a demographic curiosity, with speculation about migration and about maternal mortality. It was in the public record for seventy years before an official body treated it as a finding about the state.

The audit of 1974

That body was constituted on 22 September 1971, by a resolution of the Ministry of Education and Social Welfare, in response to a United Nations request that governments assess the position of women in advance of International Women’s Year. The Committee on the Status of Women in India was chaired by Phulrenu Guha, then a Union Minister of State. Vina Mazumdar joined later as member-secretary and did most of the intellectual work. Lotika Sarkar, Leela Dube, Maniben Kara, Neera Dogra, Urmila Haksar, Sakina A. Hasan, Savitri Shyam and Vikram Mahajan sat with them. The committee presented its report, Towards Equality, to S. Nurul Hasan, Minister for Education and Social Welfare, on 31 December 1974.¹

Its central demographic finding was the slope in the chart above: a fall from 972 to 930 across seven decades, which the committee read as evidence that conditions for women had deteriorated in absolute terms while the country modernised around them. Set beside that was a second finding, drawn from the same census volumes. Women had made up 34.44 per cent of the recorded workforce in 1911 and 17.35 per cent in 1971.² Part of that halving was real, as women were pushed out of manufacture and into unpaid household production. Part of it was definitional, the product of successive changes in what the census counted as economically active work, changes that fell hardest on the unpaid, the seasonal and the home-based.

The committee’s method was to treat the state’s own numbers as evidence about the state.

That is what makes Towards Equality a document about measurement and not only about women. Nothing in it required new data collection. The committee took what the Registrar General had been publishing since 1901, laid the columns end to end, and asked why an apparatus capable of recording a seventy-year deterioration had never been asked to explain it. Mazumdar and Sarkar went on to found the Centre for Women’s Development Studies in 1980, in part because there was no institution whose job it was to keep asking.

Putting a number on the gap

A ratio tells you the shape of a shortfall but not its size. Amartya Sen supplied the size in the New York Review of Books of 20 December 1990, in an essay whose first move was to correct a common belief: “It is often said that women make up a majority of the world’s population”, he wrote, before showing that in South Asia, West Asia and China they do not.³ His arithmetic was simple enough to check on paper. Take a benchmark ratio from populations where women and men appear to receive comparable care, apply it to the male population of a country with a skewed count, and subtract. For India the difference came to roughly 37 million women, for China roughly 44 million, and for the affected regions together more than 100 million.

The estimate has been contested ever since, and it is worth being precise about what is in dispute. Nobody quarrels with the counts of men and women; those come from censuses and are as solid as anything the discipline has. The quarrel is entirely about the benchmark, which is a counterfactual and therefore a choice. Ansley Coale, working from West model stable populations, arrived at about 60 million worldwide in 1991. Stephan Klasen and Claudia Wink, using East model populations and allowing for variation in the sex ratio at birth, put the figure near 101 million for the year 2000. John Bongaarts and Christophe Guilmoto, comparing age distributions with populations that show no evidence of discrimination, reported about 126 million for 2010.⁴ The range is wide, and the width is honest: it measures how much of the answer depends on the reference population an analyst selects. It does not measure uncertainty about whether the women are missing.

The average rises while the leading edge falls

After 1991 the all-ages ratio turns upward, from 927 to 933 in 2001 and 943 in 2011. Read alone, that line describes a country correcting itself. It is the wrong line to read alone.

The two lines measure different things and respond to different forces. The all-ages ratio is dominated by adults, and it rises when women stop dying at the rates they used to: female life expectancy in India passed male life expectancy in the 1980s and the gap has widened since. The ratio among the very young carries almost no history. It records what is happening at conception, at birth and in the first years of care, which is to say it records the present.

The present, on this evidence, got worse for four decades. The child ratio was 976 in 1961, near the biological range. It fell to 964, then 962, then 945 in 1991, then 927 in 2001, then 919 in 2011. Two developments coincide with the steepest part of that fall. Ultrasonography spread through Indian towns from the middle of the 1980s and made the sex of a foetus cheap to know. Fertility declined at the same time, which raised the stakes attached to each birth in families that wanted at least one son and no longer expected to have six children.

Parliament legislated against the first of these. The Pre-natal Diagnostic Techniques (Regulation and Prevention of Misuse) Act, Act 57 of 1994, received assent on 20 September 1994 and was brought into force on 1 January 1996. It was amended in 2003, with effect from 14 February that year, and renamed to cover pre-conception selection as well.⁵ The two censuses conducted after its commencement recorded child ratios of 927 and 919. That is not proof the statute failed, since the counterfactual is unobserved and the practice it targets is committed in private by parties who both want it concealed. What can be said is narrower and still damning: no census since the Act came into force has recorded the reversal it was passed to produce.

Sixteen years without a count

The census due in 2021 was postponed when the pandemic arrived, and stayed postponed for years after it left. A notification of the Ministry of Home Affairs, S.O. 2681(E) of 16 June 2025, sets the next enumeration for 2027, with 1 March 2027 as the reference date.⁶ The gap between counts will be sixteen years, the longest since the synchronous census began in 1881. India’s most reliable statement about its own sex composition is therefore a statement about 2011.

Other instruments have filled the space, and they do not agree. The fifth National Family Health Survey, covering 2019 to 2021, reported 1,020 women per 1,000 men, and a great deal of the press wrote that India now had more women than men. The survey’s own report advised caution, and demographers made the reason plain within days: the figure comes from de facto enumeration, counting those present in the household on the night before the interview, and male out-migration inflates it. The rural figure of 1,037 women per 1,000 men should have settled the point by itself.⁷ The same survey put the sex ratio at birth for the preceding five years at 929, against 919 in the previous round.

The Sample Registration System, which follows births and deaths continuously in a sample of villages and urban blocks, gives a third answer. Its sex ratio at birth improved from 899 in 2014 to 913 in 2021, and the report for 2023 puts it at 917 for the three years to 2023, against 914 for the three years to 2022. Within that national figure sits a spread from 868 in Uttarakhand to 974 in Chhattisgarh.⁸ The biological range for the sex ratio at birth is around 950 females per 1,000 males. India has not been inside it in the lifetime of anyone reading this.

So the country now argues about its own sex composition using a decennial count that is sixteen years stale, a health survey designed for other purposes whose headline figure is an artefact of migration, and a sample registration system whose estimates are three-year moving averages. In 2027 the census will produce both of the numbers Chandramouli read out in 2011. The first will probably be higher than 943, and it will be reported. The second will be the first census evidence on a cohort born entirely after the 2003 amendment, and it is the one the committee of 1974 would have turned to first.

1. Committee on the Status of Women in India, Towards Equality (New Delhi: Ministry of Education and Social Welfare, Government of India, 1974), presented 31 December 1974 and published in 1975; the committee was constituted by ministry resolution of 22 September 1971. ↩
2. Ibid., on women’s declining share of the recorded workforce, 34.44 per cent in 1911 against 17.35 per cent in 1971. ↩
3. Amartya Sen, ‘More Than 100 Million Women Are Missing’, The New York Review of Books 37, no. 20 (20 December 1990); and ‘Missing Women’, British Medical Journal 304 (1992), 587. ↩
4. Ansley J. Coale, ‘Excess Female Mortality and the Balance of the Sexes in the Population’, Population and Development Review 17, no. 3 (1991); Stephan Klasen and Claudia Wink, ‘Missing Women: Revisiting the Debate’, Feminist Economics 9, nos 2–3 (2003); John Bongaarts and Christophe Z. Guilmoto, ‘How Many More Missing Women?’, Population and Development Review 41, no. 2 (2015). ↩
5. Pre-natal Diagnostic Techniques (Regulation and Prevention of Misuse) Act, 1994 (Act 57 of 1994), assented 20 September 1994, brought into force on 1 January 1996 by G.S.R. 706 dated 20 December 1995; amended and renamed by Act 14 of 2003 with effect from 14 February 2003. ↩
6. Ministry of Home Affairs notification S.O. 2681(E), 16 June 2025, issued under the Census Act, 1948. ↩
7. International Institute for Population Sciences and Ministry of Health and Family Welfare, National Family Health Survey (NFHS-5), 2019–21, national factsheet; on de facto enumeration and male out-migration, Nand Lal Mishra of the International Institute for Population Sciences, quoted in Down To Earth, December 2021. ↩
8. Office of the Registrar General, Sample Registration System Statistical Report 2021 and Statistical Report 2023. Sex ratio at birth under the SRS is a three-year moving average. ↩