Loading the map…
Source: NFHS-5, Swachh Survekshan Grameen 2022, ADR/MyNeta, NCRB Crime in India 2024, WorldPop 2020Generated with iipmaps

How the index works

Dehatism is a mindset, not a place: power exercised without substance, short horizons, and custom followed because it is custom. No survey asks about it directly, so this atlas measures its traces, things that leave a mark in household surveys and police records, and combines them into one score per district.

The headline score

Dehatism Index, 0–100

A district's position among all 783 scored districts on thirteen indicators grouped into five dimensions. It is a percentile: 80 means the district shows more of these traces than 80% of districts. It does not mean "80% dehati", and a score of 20 is not "good", only lower than most.

Higher = more. Built from NFHS-5 (2019–21), Swachh Survekshan Grameen 2022, ADR/MyNeta election affidavits (2022–26) and NCRB Crime in India 2024.

The fairer comparison

Beyond development, 0–100

Poorer, less schooled districts score higher on almost everything. This version removes what three development markers already predict (women with ten or more years of schooling, households cooking with clean fuel, and population density) and ranks what is left. A high score means more than a district's circumstances would predict; a low score means less.

Those markers explain 35% of the index. Least reliable in very sparse hill districts, where density is an outlier.

The five scored dimensions

Tradition over choice

Life decisions made by custom and family rather than by the person they affect: marrying, bearing children and which children are wanted.

IndicatorExactly what is countedWorse when
Child marriageWomen aged 20–24 who were married before 18, %higher
Teen pregnancyWomen aged 15–19 already mothers or pregnant, %higher
Sex ratio at birthGirls born per 1,000 boys in the last five yearslower

NFHS-5 district fact sheets, 2019–21. District sex ratios come from small samples and swing widely.

Civic disconnect

How far households stay outside the state's records. Registering a birth or a death is free and compulsory, so skipping it signals distance from public institutions rather than cost.

IndicatorExactly what is countedWorse when
Birth registrationChildren under five whose birth was registered with the civil authority, %lower
Death registrationDeaths in the last three years registered with the civil authority, %lower

NFHS-5 district fact sheets, 2019–21.

Dominance & violence

Power exercised over others by force, status or caste, recorded as registered crime.

IndicatorExactly what is counted, per lakh peopleWorse when
MurderMurder caseshigher
Rioting & affrayAll rioting heads (communal, caste, land, water, family, political and others) plus affrayhigher
Caste-dominance offencesAgainst Scheduled Castes: intentional insult or humiliation, grabbing their land, denying use of public places, forcing out by social boycott, and untouchability under the Protection of Civil Rights Acthigher
Abduction for marriageWomen kidnapped or abducted to compel marriage, often a family's case against a couple who elopedhigher
Dowry offencesDowry deaths plus Dowry Prohibition Act caseshigher

NCRB Crime in India 2024, district tables, divided by WorldPop 2020 population. These count cases registered, so they also reflect how willing people are to report and police to register.

Squalor & sanitation

How clean homes and shared spaces are kept: litter, stagnant waste water, unused or missing toilets. This is the everyday face of civic sense, the part of the commons that people tend or neglect together.

IndicatorExactly what is countedWorse when
Observed cleanlinessDirect-observation score out of 300. Surveyors visited sampled villages and inspected households, lanes and drains, waste-disposal assets, and public places (schools, anganwadis, panchayat offices, markets, places of worship) for visible cleanliness, waste water and toilet uselower
Improved sanitationPeople living in households that use an improved toilet (flush, pit with slab, twin-pit or composting), %lower

Swachh Survekshan Grameen 2022 (Department of Drinking Water and Sanitation, fieldwork by Ipsos), Annexure 3; NFHS-5. The survey covers rural areas only, so 22 wholly urban districts (Mumbai, Kolkata, Chennai, Hyderabad, Delhi's districts and a few UTs) are scored on sanitation alone. Districts compete in this ranking and prepare for it, so the score also reflects administrative effort, not only residents' habits. No district-level survey records littering or spitting directly.

Strongman politics

Choosing power over process: electing representatives who face serious criminal charges. Research on Indian elections (Milan Vaishnav, When Crime Pays) finds such candidates do best where people trust the state less to deliver and back someone who can get things done regardless. It is the closest district-level trace of low trust and tolerated corruption.

IndicatorExactly what is countedWorse when
MLAs facing serious chargesMLAs from the latest assembly election who declared serious criminal cases in their election affidavits, as a share of the district's MLAs. Serious means, in ADR's definition, offences such as those punishable by five years or more, non-bailable offences, assault, murder, kidnapping, crimes against women and corruption. A constituency that spans two districts counts in each by area.higher

ADR/MyNeta analysis of candidate affidavits, assembly elections 2022–2026. Charges are declared, not proven, and some come from political protests, which lifts states with agitational politics such as Kerala and West Bengal. Union territories without a legislature (Chandigarh, Ladakh, Lakshadweep, Andaman & Nicobar, Dadra & Nagar Haveli and Daman & Diu) have no value. No district-level survey measures bribery or trust directly.

Shown, but not scored

Commons & predation

Living off what belongs to everyone, and preying on others: damaging public property, unlawful possession of railway property and telegraph wire, theft of government property, and offline cheating, forgery and fraud.

It is central to the idea of dehatism, but in the data it does not move with the five scored dimensions (correlations between −0.29 and 0.19), which suggests police records capture it poorly. It has its own map but no weight in the index.

Context indicators

Available in the indicator menu for reference: Electricity Act cases, cyber-fraud victims, cruelty by husband or relatives, all crimes against Scheduled Castes, Child Marriage Act cases, and MLAs facing any criminal charge (including minor and protest cases).

Each mostly measures something else: state enforcement drives (Electricity Act, Child Marriage Act), where victims report rather than where offenders are (cyber fraud), or willingness to report (cruelty, crimes against SCs).

The formula

  1. Orient every indicator so that higher means more dehatism. Sex ratio, the two registration rates, observed cleanliness and sanitation are flipped.
  2. Tame skew: crime rates become log(1 + rate).
  3. Trim outliers at the 2nd and 98th percentiles.
  4. Standardise: z = (value − mean) ÷ standard deviation, across districts.
  5. Score each dimension as the mean of its indicators' z-scores (at least half must be present), then standardise again.
  6. Combine the five dimensions with equal weight (at least three must be present).
  7. Rank: score = percentile × 100.

Why equal weights? Weighting the dimensions by how much they share (the first principal component) gives the same ranking, so equal weights cost nothing and are easier to explain.

How well do the parts agree? Moderately. The five dimensions correlate at 0.05–0.53; squalor tracks civic disconnect most closely (0.53), and strongman politics moves least with the rest (0.05–0.20). This is a broad composite, not one tightly measured trait. Look at the dimension maps to see which part drives a district.

New districts. 148 districts created after NFHS-5 carry their parent district's survey values; each district's panel says so. The same applies to 76 districts newer than the 2022 cleanliness survey.

Four things the data says

46/50

The Gangetic east dominates the top

Of the 50 highest-scoring districts, 31 are in Bihar and 15 in Jharkhand. Their median districts score 96 and 94. Kerala's median district scores 15.

r = −0.51

Women's schooling is the strongest signal

Districts where more women have ten years of schooling score lower. Population density runs the other way (r = +0.17): dense districts score slightly higher, so this is not a village trait.

35%

Development explains about a third

Schooling, clean cooking fuel and density predict 35% of the variation. The rest is something else. Switch to Beyond development to see who over- and under-performs their circumstances.

36 cr

People living in high-scoring districts

157 districts score 80 or above. Together they hold about 36 crore people, roughly one Indian in four (WorldPop 2020 estimates).

Every district, by state

Each dot is a district, placed by the measure selected above. States are sorted by their median district, marked with a bar. Hover a dot to find it on the map; click a state name to zoom to it.

What travels with it

The score against development markers. If dehatism were simply rural, the densest districts would score lowest. They don't.

Caveats and sources

Read with care

  • This measures traces in districts, not people. A high score describes conditions recorded in a district, not the character of anyone who lives there.
  • Crime rates partly measure willingness to report and register, so states with better registration can look worse. Survey measures (NFHS) are steadier.
  • Where police jurisdictions cross district lines (Hyderabad, Warangal, Ramagundam, Mumbai, Bengaluru, Delhi, Puducherry, Dimapur), neighbouring districts share one pooled crime rate.
  • Special police units (railway police, CID, cyber cells, crime branches) are excluded because their cases are not tied to one district.
  • NFHS-5 is from 2019–21, the cleanliness survey from 2022, the elections from 2022–26 and NCRB from 2024; the parts of the index are a few years apart.

Sources