Interactive LabData Feminism & Intersectional Analysis Lab
Analyze data through gender, caste, and power — and spot the biases hiding in plain sight. A hands-on ImpactMojo lab for practitioners and researchers across South Asia.
Why Data Feminism?
Data is not neutral. It reflects who collected it, who was counted, who was left out, and what questions were asked. Data feminism asks: Whose data? Who benefits? Who is harmed?
When "neutral" data hides inequality
In 1990, economist Amartya Sen estimated that more than 100 million women were "missing" across Asia — women who would be alive under equal care, but who died from sex-selective abortion, infanticide, and systematic neglect. India's own share is estimated at roughly 40–60 million. The censuses did count individuals; the crisis only became visible once analysts disaggregated by sex and compared the ratio against a "no-discrimination" benchmark. The number is an estimate, sensitive to the benchmark chosen.
GDP conventions exclude unpaid domestic and care work — work done disproportionately by women. India's Time Use Survey 2024 found women spend about three times as much time as men on unpaid domestic work, and roughly twice as much on caregiving; economists estimate this work is worth on the order of 15–17% of GDP. Female labour-force participation was long undercounted (much unpaid/subsistence work isn't recorded as "economic"), though official estimates have risen sharply in recent years.
Apart from Scheduled Castes and Scheduled Tribes, India's decennial Census has not enumerated caste since 1931. So when a programme "works for everyone," the aggregate can mask the fact that it works well for dominant castes and badly for Dalits and Adivasis. (The Census 2027 is slated to include full caste enumeration for the first time since 1931.)
The seven principles of Data Feminism D'Ignazio & Klein, 2020
1. Examine power
Ask who has power in data collection, analysis, and use.
2. Challenge power
Use data to challenge, not reinforce, unequal hierarchies.
3. Elevate emotion & embodiment
Data represents lived experience, not just numbers.
4. Rethink binaries & hierarchies
Categories (e.g. gender) should reflect real complexity.
5. Embrace pluralism
Multiple sources, methods, and knowledges are valid.
6. Consider context
Data means little without its social and historical context.
7. Make labor visible
Credit those who collected, cleaned, and coded the data.
Intersectionality in Data
Intersectionality (Kimberlé Crenshaw, 1989) means that gender, caste, class, religion, and disability don't operate independently — they intersect to create distinct experiences of privilege and oppression.
The same program, different experiences Illustrative dataset
A government skill-training programme reports 70% female participation. Sounds good — until you disaggregate. (The table below is a teaching dataset: the numbers are invented to illustrate how aggregates hide inequality, not drawn from a real programme.)
| Group | Enrollment % | Completion % | Job placement % | Avg. salary |
|---|---|---|---|---|
| General category men | 45% | 82% | 68% | ₹18,000 |
| General category women | 35% | 75% | 45% | ₹12,000 |
| OBC men | 12% | 70% | 55% | ₹14,000 |
| OBC women | 8% | 58% | 32% | ₹9,000 |
| SC men | 5% | 65% | 48% | ₹13,000 |
| SC women | 3% | 42% | 18% | ₹7,500 |
| ST men | 2% | 60% | 40% | ₹11,000 |
| ST women | 1% | 35% | 12% | ₹6,000 |
| Overall (aggregate) | 70% female | 72% | 52% | ₹14,500 |
Spotting Bias in Data Collection & Analysis
Bias enters data at every stage: who asks, who answers, what's asked, what's recorded, what's analysed, and what's published.
Where bias creeps in
Sampling bias
Surveying only households with phones; excluding women who can't leave home; sampling only urban areas.
Question bias
Asking men (not women) about household decisions; using male-defined categories for "work."
Recording bias
Assuming the male is "household head"; coding all domestic work as "unemployed."
Analysis bias
Reporting averages without disaggregation; treating male norms as the baseline; ignoring outliers.
Reporting bias
Cherry-picking positive findings; deficit framing for marginalised groups; hiding invisible labour.
Algorithmic bias
Training data underrepresents women; facial recognition fails on darker skin; credit scoring discriminates.
Interactive: find the bias
For each scenario, pick the primary type of bias, then check your answers.
Visualizing Data Intersectionally
How you visualize data shapes what people see. A single bar can hide as much as it reveals.
From aggregate to intersectional Illustrative dataset
Scenario: school enrollment in a district. Watch what appears as we add dimensions. (Numbers are a teaching example.)
Aggregate view: "92% school enrollment"
Disaggregated by gender
Fully intersectional: gender × caste
Interactive: design your visualization
For each goal, pick the visualization approach that reveals rather than hides inequality.
Data Justice in Action
Data feminism is not just critique — it's practice. Here's how to apply it in your own work.
The data-justice checklist
1. Design for disaggregation
Collect data that can be broken down by gender, caste, religion, disability, geography, and age from the start. You can't disaggregate what you didn't collect.
2. Use intersectional analysis
Don't stop at "women vs men" or "SC vs General." Look at SC women, Muslim women, rural disabled women.
3. Name the missing
When you report an aggregate, state what it hides: "Overall enrollment is 92%, but ST girls enroll at only 68%."
4. Center marginalized voices
Include qualitative data from those most affected. Numbers tell you WHAT; stories tell you WHY. Both are data.
5. Challenge deficit framing
Document structural barriers and power imbalances, not just what marginalised groups "lack." The problem is the system, not the people.
6. Make labor visible
Credit those who collected, cleaned, and coded the data — often field workers whose labour is invisible in publications.
Interactive: apply data justice
You've collected data on a rural employment programme. Select every action that aligns with data-justice principles, then check.
Case Study: Re-analyzing a "Successful" Program
An NGO reports: "Our water program increased household water access from 45% to 78% in 20 villages." Let's re-analyse this through a data-feminism lens.
The NGO's headline data
| Indicator | Baseline | Endline | Change |
|---|---|---|---|
| Households with water access | 45% | 78% | +33pp |
| Time to fetch water (minutes) | 45 | 20 | −25 min |
| Water quality (safe %) | 32% | 65% | +33pp |
| User satisfaction | 2.8/5 | 4.1/5 | +1.3 |
What they didn't report
- Water points were built near the main road — convenient for dominant-caste households, far from Dalit hamlets.
- Women still do 100% of water-fetching, but the programme had no gender-specific design.
- The "user satisfaction" survey was administered by male field staff to male household heads.
- Three villages with the worst outcomes were dropped from the final report ("outliers").
- No data collected on: caste, gender of the water-fetcher, time-of-day constraints, or safety.
- The programme cost ₹2 crore — but no cost-per-household data disaggregated by group exists.
Interactive: your re-analysis
What would a data-justice re-analysis do? Select every appropriate action, then submit.
Lab complete
You can now analyse data through an intersectional lens, spot hidden biases, and design more just data practices.
- Data is never neutral — it reflects power and choices about who is counted.
- Aggregate data often hides the most important inequalities.
- Intersectionality reveals what single-axis (gender-only or caste-only) analysis misses.
- Bias enters at every stage: collection, recording, analysis, visualization, reporting, and algorithms.
- Visualization choices shape what people see and believe.
- Data justice needs both critique and better practice: disaggregate, name the missing, challenge deficit framing, make labour visible.
Recommended next steps
Read further: Data Feminism by Catherine D'Ignazio & Lauren F. Klein (MIT Press, open access) · Invisible Women by Caroline Criado Perez.