Women and men in India, measured side by side.

Left alone, about 105 boys are born for every 100 girls, roughly 950 girls per 1,000 boys. A ratio well below that means girls are being selected out before birth; telling parents the sex of a foetus has been illegal since 1994. Each point averages three years, so a change shows up slowly.

In the first windows towns ran lower than villages; the lines have since crossed.

The Sample Registration System follows a fixed sample of households and records every birth, so its ratio is the steadier one. The Civil Registration System counts births registered with the state; where registration is incomplete, and girls go unregistered more often, its ratio can mislead.

This ratio counts everyone alive, not births. It moves with births, with who dies young (for decades more girls died in childhood than boys) and with migration. Women outlive men, so an ageing population pushes it up even when births stay unequal.

Newborn boys are biologically frailer, so where girls are cared for equally, slightly more boys die in the first year. A higher death rate for girls is a sign of neglect: less food, later treatment, fewer trips to the doctor. The crossing of these lines is one of the quieter changes in Indian family life.

Maternal deaths are rare enough that one year's sample cannot measure them, so the Registrar General pools three years. The windows overlap, and the gaps between them are uneven, which is why the points are not evenly spaced.

The ratio is per 1,00,000 births, not per woman. A state with fewer births and a high ratio may still lose fewer mothers in all than a populous state with a middling one.

Life expectancy at birth is how long a baby would live if today's death rates at every age held for her whole life. In the early 1990s women and men lived about as long as each other in India; the women's advantage seen in most of the world opened up only as maternal deaths fell.

Schooling and fewer children travel together: a woman in school marries later, knows more about contraception and has more say. The link is not only cause and effect; the women who stay in school also tend to come from richer, more urban families.

An average hides the girls married young. A state with an average of 22 can still have many brides under 18; the share married before 18, from the health survey, tells that story.

The gross enrolment ratio divides everyone enrolled at a level, whatever their age, by the children of the age for that level. It can pass 100 when older children repeat a class, and it says nothing about attendance or learning. The dropout rate is the share of a year's pupils who do not come back the next year.

Both ask whether a person can read and write a simple message with understanding, but the census asks everyone in the household and the survey asks a sample, with a different questionnaire. The gap between the 2011 census and the 2017-18 survey is not a measured change.

Mean years of schooling counts every adult, including the grandmothers who never went to school, so it moves slowly and records the past as much as the present. Enrolment shows today's children; this shows the stock of education in a state's adults.

"At work" in usual status means a person's main activity over the past year, or a subsidiary one for at least 30 days, was work for pay, profit or the family's own farm or business. Much of the rise in women's work since 2017-18 is unpaid help in a family enterprise and self-employment, not wage jobs.

The survey changed twice in this window: from agricultural to calendar years, and in 2025 to a new sample design. Change is read within a ruler, never across the marked lines.

The states with the most women in the labour force are mostly hill and north-eastern states, where women farm; the fewest are in the north-west and in cities, where paid work for women outside the home is scarcer. Men's rates vary far less from state to state.

Casual labour is daily-wage work: harvesting, construction, loading. Part of the gap is the kind of work each is given, part is a lower rate for the same work. Public works such as MGNREGA, which pay women and men the same rate, are counted separately and are not in this measure.

The survey asks each person to account for every half hour of the previous day. "Per person" averages over everyone, including those who did none of the activity; "per participant", in the downloads, averages only over those who did. Unpaid work at home is cooking, cleaning, fetching water and fuel, and looking after children and the old.

An account in a woman's name is not money she controls. Many scheme accounts were opened to receive government transfers; they show that the money reaches her name, not that she decides how it is spent. The investing figures are closer to choice, and there women are a small minority.

Parliament passed the Constitution (106th Amendment) in 2023 to reserve a third of Lok Sabha and assembly seats for women; it takes effect only after the next census and delimitation of constituencies. Women who contest win more often than men do: parties field few women, and mostly in seats they expect to win.

The 73rd and 74th Amendments (1992) reserved a third of panchayat and municipal seats for women, and many states have raised that to half. No such rule applies to the courts, the police or company boards, apart from the Companies Act's requirement of at least one woman director on larger boards.

These are crimes the police registered, not crimes committed. A state where women are more willing, and police more ready, to file a case records more. A high rate can mean more crime, more reporting, or both; a very low one can mean women do not go to the police.

Section 498A of the Indian Penal Code makes cruelty by a husband or his relatives a crime; most of these cases arise from violence and dowry demands at home. The health survey asks women directly and finds far more violence by husbands than ever reaches a police station.

This is not the usual "conviction rate", which divides convictions by the trials completed. It divides by every case due for trial, including the backlog, so it shows how few cases end in a verdict at all in a year. The counts of convictions, acquittals and cases pending are in the data downloads.

Cause of death is assigned by trained staff from a structured interview with the family after the death, a "verbal autopsy", because most Indians die without a doctor certifying the cause. A larger share of women's deaths is left ill-defined.