# Welcome — the denominator decides

A 45-minute session on why two people can quote the same government survey, the same year, the same table, and disagree by a factor of nearly two.

You will need: something to write on, and one indicator from your own work that involves women.

Mark each step **done** when you have finished it, or **stuck** if you want the facilitator to come to you. Nobody is timing you against anyone else.

# The claim on the table

India's Time Use Survey ran a second national round in 2024. The National Statistical Office asked a sample of households what every member did with their day, slot by slot.

Here are four numbers from it. All are for people aged 15 to 59, rural and urban combined, covering all unpaid activities.

| | women | men |
|---|---|---|
| minutes per **participant** | 389 | 120 |
| minutes per **person** | 366 | 62 |

Read them before going on. They are not a typo and they are not two different surveys.

# Work out both ratios

Do the arithmetic yourself. It takes thirty seconds and it lands better than being told.

1. Divide women's minutes per participant by men's minutes per participant.
2. Divide women's minutes per person by men's minutes per person.

Write both numbers down.

You should have roughly **3.2** and **5.9**.

Now say each one out loud as a headline. "Indian women do three times as much unpaid work as men." "Indian women do nearly six times as much unpaid work as men." Both sentences are true. Both are defensible. They come from one table.

# Where the gap comes from

One more number from the same survey: on the reference day, **94.2 per cent of women** aged 15 to 59 did some unpaid work. For men the figure was **51.5 per cent**.

Per participant averages only over the people who did the thing. It quietly removes the half of men who did none — and their zero is the most interesting fact in the dataset.

Per person averages over everybody, so those zeros stay in and pull the male average down.

Neither denominator is wrong. They answer different questions:

- *Among people who do unpaid work, how much do they do?* — per participant.
- *Across the whole population, how is this work distributed?* — per person.

For most gender arguments, the second question is the one you meant to ask.

#[quiz] Check yourself

## A colleague cites "women do 3.2 times as much unpaid work as men". What does that figure condition on?

- [ ] Rural households only
- [x] Only the people who did any unpaid work that day
- [ ] Only women who are employed
- [ ] The whole population aged 15 to 59

## Why is the per-person ratio larger than the per-participant ratio?

- [ ] Women's absolute minutes are higher on that measure
- [x] Nearly half of men did no unpaid work at all, and per-person keeps their zeros in
- [ ] The two figures come from different survey rounds
- [ ] Per-person includes childcare and per-participant does not

# Your own indicator

Take the indicator you brought. Write it out in full, then answer three questions in writing.

1. **What is the denominator?** Not "women" — which women, counted how, at what moment. If you cannot answer in one sentence, that is the finding.
2. **Who is excluded by it, and is their exclusion the thing you care about?** The zeros are where the inequality usually lives.
3. **If you halved the denominator, would the indicator move?** If yes, anyone quoting your number needs to be told what it is a fraction of.

If your indicator is a participation share — "52 per cent of participants were women" — write down what it would take for that number to change without anything changing for any woman.

# Three questions for any gender statistic

Keep these. They are the whole session compressed.

**What is this a fraction of?** Ask before you ask anything else. Most disputes about gender data are disputes about denominators that nobody has stated.

**Who is in the zero?** The people who did none, earned none, attended none. A statistic that conditions them away is answering a narrower question than it appears to.

**Would this move if the underlying relation did not?** A number that rises when enrolment rises, and for no other reason, is a coverage statistic wearing a gender label.

# Close

One thing to take back to your team: find the most-quoted number in your last report and write its denominator beside it, in the report, where a reader can see it.

That is the entire practice. It costs a clause and it stops an argument you would otherwise have six months later.

**Where this comes from.** The Time Use Survey figures are from the National Statistical Office's 2024 round, retrieved from the MoSPI data API; the two denominators are the survey's own, published side by side. The longer treatment is Module 7 of [Gender-Sensitive MEL](https://www.impactmojo.in/courses/gender-mel/), which is free.