The indicator moved.
That is not the same as empowerment.
Naila Kabeer’s resources, agency and achievements, made clickable. Four Indian series, each drawn beside the companion series from the same survey that changes how it reads.
Lifetime experience of spousal violence fell from 37.2 per cent to 29.1 across fifteen years. Over the same three rounds, the same survey put the last-twelve-months figure at 23.0, then 23.9, then 24.0. Only one of those numbers is ever quoted.
The framework is Naila Kabeer’s, from Resources, Agency, Achievements: Reflections on the Measurement of Women’s Empowerment (1999). ImpactMojo added the Indian series and pulled each one from the source. Full credits ↓
Empowerment is the expansion of the ability to make strategic life choices, in a context where that ability was previously denied
That is Kabeer’s definition, and every word in it is load-bearing. Strategic excludes the second-order choices a person may make freely while the first-order ones are settled for them. Ability to choose requires that an alternative existed, that the person could see it, and that choosing carried consequence. Previously denied means empowerment is a change over time, so a snapshot cannot show it.
The three parts are a sequence and not a menu. Resources are the pre-conditions, agency is the process, achievements are the outcomes, and a measurement of any one alone can move for reasons the framework would not count.
The half of the paper that gets dropped
Kabeer’s 1999 article is cited constantly for the three words and rarely for what follows them, which is a long and specific warning about proxies. Her argument is that the standard measures each fail in a characteristic way: a resource can be delivered to a woman and controlled by someone else; agency asked about in a questionnaire cannot see the decisions nobody raised; and an achievement can improve because circumstances changed rather than because anyone chose.
The sharpest of her cautions is about preference. A person who has adapted to a constraint reports the constraint as a preference, so an instrument that asks what someone wants will record the adaptation as a want. That is not a technical problem to be corrected with a better question; it is a limit on what asking can establish.
This page is that warning, run against Indian data. Every indicator here is one a gender report would quote, every one moves in the direction the report would want, and every one has a companion in the same survey that changes what it means.
Pick an indicator
Each is drawn twice: as it is normally reported, and beside the series that complicates it. Both come from the same instrument.
Two readings of one survey
Both charts run 0 to 100 on the same axis. A chart scaled to its own data would make a two-point drift look like a collapse, which is the visual form of the mistake this page is about.
Whose framework this is
The framework
Naila Kabeer, Resources, Agency, Achievements: Reflections on the Measurement of Women’s Empowerment, Development and Change 30(3), 1999. The definition, the three parts, and the measurement cautions this page is built on are all in that article.
Her The Power to Choose (2000), on Bangladeshi and British garment workers, is where the method is worked out at length
The data
National Family Health Survey rounds 3, 4 and 5, retrieved from the DHS StatCompiler rather than from secondary reporting, and the Periodic Labour Force Survey, retrieved from the MoSPI API. Every value on this page came out of one of those two, and the definitions quoted are the instruments’ own.
The DHS definitions matter here: “alone or jointly” is theirs, not a gloss
What ImpactMojo added
The pairing. For each indicator, the companion series or breakdown from the same instrument that changes how the first one reads, and the note on which of Kabeer’s conditions the pair puts in doubt. The choice of which four indicators, and the reading of each pair, are ours.
ImpactMojo, 2026 · content CC BY-NC-ND 4.0 · code MIT
What this page does not do
It does not claim these indicators are worthless, or that nothing has improved. Married women reporting a say in their own health care rose from 62.2 per cent to 81.1, and women reporting no say in any of the three decisions fell from 23.5 to 11.3. Those are large movements in a consistent direction across three rounds.
The claim is about what a single series can establish, not about whether India has changed
Using this in a workshop
Bring the indicators your own programme reports on gender, and for each one ask the two questions this page asks. Which of the three parts is it measuring? And what is the companion series in the same instrument that would complicate it?
The second question is the useful one, because the answer is usually that nobody has looked. Most gender reporting quotes one series per claim, and most instruments carry a second series that bears on it directly. The twelve-month violence figure sits in the same NFHS table as the lifetime figure.
Read alongside Practical and Strategic Gender Needs, which is about what interventions do rather than what empowerment is, and the Results Chain, which is about the opposite failure: an outcome nobody measures at all.
If you reproduce the framework, credit Kabeer. If you reproduce the pairings or the readings, credit ImpactMojo, and argue with them.