The state of Muslim India, in data

Definition. Muslim population as a share of total population at the given geography level.

Methodology. Population shares from the primary RGI Census religion volumes, 1961 to 2011. Two all-India figures exclude a state where the Census could not be held: 1981 excludes Assam and 1991 excludes Jammu & Kashmir. The 2021 Census is delayed.

Where this data comes from. Census 2011 · C-series · Census 1961 · C-VII Religion · Census 1971 · Paper 2 of 1972 · Census 1981 · HH-15 (Paper 3 of 1984) · Census 1991 · C-9 Religion · Census 2001 · C-series. Download CSV: pop-share.csv. You can open any of these to check the numbers on this chart yourself.

Computed Reproduce this. This figure is computed from the linked published tables. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Percentage of each community living in urban areas. Census Table C-01 provides religion × residence (rural/urban) × sex breakdowns at national, state, district, and town levels; series covers Census 2001 and 2011.

Methodology. Urban share = urban_persons / total_persons * 100, per Census C-01. Two-census series (2001 and 2011) at national and state level; the 2021 Census is delayed. Muslims (~40%) and Christians (~40%) are markedly more urbanised than the all-India average (~31%), a Sachar-era fact that conditions outcomes downstream (urban-poor labour markets, segregation, housing access).

Where this data comes from. Census 2011 · C-series · Census 2001 · C-series. Download CSV: urban-share.csv. You can open any of these to check the numbers on this chart yourself.

Computed Reproduce this. This figure is computed from the linked published tables. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Number of females per 1000 males in the population. A value below 1000 means fewer women than men. Census series across six censuses, 1961-2011.

Methodology. Females / males * 1000 from Census religion tables across six censuses (1961-2011): 1961 Social & Cultural Tables C-VII, the 1971 and 1981 census religion papers, 1991 C-09, 2001 C-01 and 2011 C-15. National for all six years; state-level for 2001 and 2011. Two all-India figures exclude a state where the Census could not be held: 1981 excludes Assam and 1991 excludes Jammu & Kashmir.

The 2021 Census is delayed.

Where this data comes from. Census 2011 · C-series · Census 1961 · C-VII Religion · Census 1971 · Paper 2 of 1972 · Census 1981 · HH-15 (Paper 3 of 1984) · Census 1991 · C-9 Religion · Census 2001 · C-series. Download CSV: sex-ratio.csv. You can open any of these to check the numbers on this chart yourself.

Computed Reproduce this. This figure is computed from the linked published tables. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Percentage of live births in the years preceding each survey round that took place in a health facility (public or private), among women age 15-49.

Methodology. Time series across four NFHS rounds: NFHS-2 (1998-99), NFHS-3 (2005-06), NFHS-4 (2015-16) and NFHS-5 (2019-21); years 1998/2005/2015/2020 are the round labels. For NFHS-5, Table 8.13 (page 324); the "all" value is a sample-size-weighted average of the 7 religion rates (the published Total row was hard to extract cleanly from the dual-column PDF layout) and matches NFHS-5's published 88.6% within rounding. Earlier rounds come from each report's facility-delivery religion table; every canonical row's methodology_note records its exact source.

Where this data comes from. NFHS-5 (2019-21) · NFHS-4 (2015-16) · NFHS-3 (2005-06) · NFHS-2 (1998-99). Download CSV: inst-delivery.csv. You can open any of these to check the numbers on this chart yourself.

Read Reproduce this. This figure is published directly in the linked source table. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Probability of dying before the first birthday, expressed per 1000 live births, for the period preceding each survey round.

Methodology. Time series across four NFHS rounds: NFHS-2 (1998-99), NFHS-3 (2005-06), NFHS-4 (2015-16) and NFHS-5 (2019-21); years 1998/2005/2015/2020 are the round labels. For NFHS-5, Table 7.2 publishes URBAN and RURAL IMR by religion separately with no total-residence-by-religion column, so the national total is a population-weighted average using Census 2011 urban/rural population by religion as weights (<1% drift; urban/rural shares by religion are stable). Earlier rounds come from each report's religion table; every canonical row's methodology_note records its exact source.

Where this data comes from. NFHS-5 (2019-21) · NFHS-4 (2015-16) · NFHS-3 (2005-06) · NFHS-2 (1998-99) · Census 2011 · C-series. Download CSV: imr.csv. You can open any of these to check the numbers on this chart yourself.

Computed Reproduce this. This figure is computed from the linked published tables. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Percentage of children under age 5 classified as stunted (height-for-age more than 2 SD below the WHO Child Growth Standards median). Stunting indicates chronic malnutrition in early life.

Methodology. NFHS-5 (2019-21) Table 10.1 (Nutritional status of children by background characteristics). Religion column = % below -2 SD HFA; all-India value is the sample-size-weighted average of the 6 religion rows (reproduces the published national rate of 35.5%). Single round in canonical for now; NFHS-2/3/4 extension possible (same pattern as IMR / inst-delivery / anaemia trend extractors).

Where this data comes from. NFHS-5 (2019-21). Download CSV: stunting-u5.csv. You can open any of these to check the numbers on this chart yourself.

Read Reproduce this. This figure is published directly in the linked source table. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Percentage of women age 15-49 with any anaemia, defined as haemoglobin <12.0 g/dl (non-pregnant) or <11.0 g/dl (pregnant), altitude-adjusted.

Methodology. Time series across four NFHS rounds: NFHS-2 (1998-99), NFHS-3 (2005-06), NFHS-4 (2015-16) and NFHS-5 (2019-21); years 1998/2005/2015/2020 are the round labels. For NFHS-5, Table 10.23.1 reports total-residence women's anaemia by religion directly. NFHS changed anaemia measurement between NFHS-4 and NFHS-5 (capillary vs venous blood at the same Hb <12.0 g/dl cut-off), which breaks cross-round comparability, so the three pre-NFHS-5 rounds carry break_flag=true: do not read the 2015 -> 2020 step as a pure trend.

Where this data comes from. NFHS-5 (2019-21) · NFHS-4 (2015-16) · NFHS-3 (2005-06) · NFHS-2 (1998-99). Download CSV: women-anemia.csv. You can open any of these to check the numbers on this chart yourself.

Read Reproduce this. This figure is published directly in the linked source table. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. What a household pays from its own pocket for one hospital admission, after any insurance or employer reimbursement: doctor's fees, medicines, diagnostics, bed and package charges. Childbirth stays are excluded. A lower figure does not mean cheaper care: it mostly shows who uses government hospitals and who can afford private ones.

Methodology. Computed from the unit-level microdata of two NSS Household Social Consumption: Health rounds: the 75th (Sch 25.0, July 2017 - June 2018, 1.14 lakh households, via MoSPI's original fixed-width TXT distribution; the NADA copy of that round is a proprietary binary, so the build reads the surviving mirror recorded in the repo's provenance notes) and the 80th (January - December 2025, 1.40 lakh households, via the NADA CSV distribution). Out-of-pocket medical expenditure per hospitalisation case excluding childbirth is the 2025 release's own headline construct, computed identically in both rounds: medical expenditure (fees, medicines, diagnostics, bed and package charges; transport and food excluded) minus any insurance or employer reimbursement, averaged over cases of the last 365 days. The 2017-18 estimator reproduces all nine published cells of NSS Report 586 Statement 3.15 within 0.01% and the published reimbursement shares exactly; the 2025 estimator reproduces all six press-note OOPME cells within 0.01%.

Neither publication breaks expenditure down by religion. A Muslim hospital stay cost INR 30,104 against the all-India INR 34,064 in 2025 (Hindu INR 34,323), and INR 14,827 against 18,088 in 2017-18: about 12-19% less per stay. Muslims lean more on government hospitals (41.5% of Muslim hospitalisation cases in 2025 against 36.4% of Hindu cases, down from 47.6% and 41.8% in 2017-18), where the average bill runs about an eighth of a private one, and lower spending capacity does the rest; the survey cannot tell cheaper care apart from care that was needed but never bought.

Values are nominal rupees of each round; health insurance coverage roughly tripled between the rounds, so the netted reimbursement matters more in 2025. NSO unit-data rider: religion is self-reported and the survey is stratified for states, not religions, so the split is indicative with no sub-state estimates.

Where this data comes from. NSS Health 2025 (NSO unit-level data) · NSS 75th 25.0 Health (2017-18). Download CSV: hospital-oop-spend.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Share of households with access to a toilet of any type.

Methodology. NFHS-5 Table 2.4 (page 74). Total residence, national. Year=2020 = NFHS-5 fieldwork midpoint.

Muslim toilet access (90.3%) runs *above* Hindu (80.7%), partly because urban-Muslim share is higher than urban-Hindu share, and urban access is uniformly higher. Not strictly "improved sanitation" per the WHO/UNICEF JMP definition: the published NFHS-5 table aggregates flush, pit and other toilet types.

Where this data comes from. NFHS-5 (2019-21). Download CSV: improved-sanitation.csv. You can open any of these to check the numbers on this chart yourself.

Read Reproduce this. This figure is published directly in the linked source table. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Share of households whose main source of drinking water is an improved source (piped, tubewell, protected well, bottled and similar) located within the household's own premises, so no one has to fetch water from outside.

Methodology. Computed from the unit-level microdata of NSS 76th round Schedule 1.2 (Drinking Water, Sanitation, Hygiene and Housing Condition, Jul-Dec 2018; 1.07 lakh households): improved principal source located within premises, by religion of household head, weighted. Report 584 publishes this by state and social group, never religion. The extraction reproduces the published all-India values across 24 validation cells to a worst gap of 0.24 points (improved-within-premises all-India: 63.6 computed vs 63.8 published, the strictest blank-distance treatment).

Muslim households are AHEAD of Hindu households here (69.3% vs 62.5%), driven by a rural advantage (65.9 vs 54.3); urban Muslims trail urban Hindus on piped-into-dwelling water (36.5% vs 41.4%, in the data file). Unit data: the original mospi.gov.in TXT distribution (the current NADA catalog ships only a proprietary binary; re-fetch URLs + sha256 in sources/nss76/). NSO unit-data rider: religion is self-reported; the split is indicative; no sub-state estimates.

Year=2018 = survey period.

Where this data comes from. NSS 76th 1.2 (2018). Download CSV: improved-water-premises.csv. You can open any of these to check the numbers on this chart yourself.

Definition. Share of households living in a pucca structure: walls and roof both made of permanent materials (brick, concrete, stone, metal and similar), the standard Indian measure of housing quality.

Methodology. Computed from NSS 76th round Schedule 1.2 unit-level microdata (Jul-Dec 2018), by religion of household head, weighted; reproduces the published all-India values (76.7 rural / 96.0 urban / 83.3 all) exactly. Muslim 83.6% vs Hindu 83.3%: parity overall, with rural Muslims slightly behind rural Hindus (75.9 vs 76.8) offset by the community's higher urban share. Same source, validation and caveats as the water card (see sources/nss76/); religion is self-reported per the NSO rider; year=2018 = survey period.

The "Electricity" tab folds in the companion measure from the same Schedule 1.2 microdata (the decarded household-electricity metric, reproducing the published 93.9 rural / 99.1 urban / 95.7 all exactly): Muslim 96.6% vs Hindu 95.4%, near-parity by 2018 with the small Muslim edge from the community's higher urban share.

Where this data comes from. NSS 76th 1.2 (2018). Download CSV: pucca-house.csv. You can open any of these to check the numbers on this chart yourself.

Definition. Share of population aged 7 and above who can both read and write with understanding in any language.

Methodology. Census defines literacy as the ability to read and write with understanding in any language. Two-census series (2001 and 2011, Census C-09 education-by-religion tables); both rounds share the same Census definition, so the 2001 -> 2011 change is a like-for-like comparison. PLFS reports a broadly similar literacy measure on a slightly different definition and is a possible future cross-check; no PLFS series is plotted here today.

Where this data comes from. Census 2011 · C-series · Census 2001 · C-series. Download CSV: lit-7plus.csv. You can open any of these to check the numbers on this chart yourself.

Computed Reproduce this. This figure is computed from the linked published tables. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. What a household spends on one school student in an academic year: course fees, textbooks and stationery, uniform, transport and other items. Private coaching is counted separately and left out of this figure. Muslim and Hindu families use free government schools at almost the same rate, so the gap is not about school choice: Muslim households spend less on fees, books and transport in whichever type of school their children attend.

Methodology. Computed from the unit-level microdata of two NSS education rounds: the 75th (Sch 25.2, July 2017 - June 2018, 1.14 lakh households, via MoSPI's original fixed-width TXT distribution; the NADA copy of that round is a proprietary binary, so the build reads the surviving mirror recorded in the repo's provenance notes) and the Comprehensive Modular Survey: Education (CMS:E, NSS 80th round, April - June 2025, 52,085 households, via the NADA CSV distribution). Average expenditure per student currently enrolled at school levels (pre-primary to higher secondary, incl. diploma/certificate up to higher-secondary equivalent) during the current academic year, excluding private coaching in both rounds (the CMS:E published headline construct; the 75th collects coaching as a separate item, so the build drops it for comparability). The 2017-18 estimator reproduces all 27 published anchor cells of NSS Report 585 Statements 19 and 21 within 0.01%; the 2025 estimator reproduces all seven published Report 595 per-student cells within 0.01%.

Neither publication breaks expenditure down by religion. A Muslim school student's year cost INR 9,249 against the all-India INR 12,616 in 2025 (Hindu INR 12,941), down as a share from 2017-18 (INR 5,532 against 7,041 all-India, Hindu 7,050): the Muslim household spends about 72% of what the Hindu household spends per student, where it was 78%. Unlike hospital costs, school choice is not the driver: 55.4% of Muslim students attend government schools against 56.5% of Hindu students (2025), so the gap comes from how much families spend in whichever school they use (fees, books, transport), not from heavier use of free government schools.

The trend line renders dashed: Report 595's own comparability section lists concept revisions vs the 75th round (school-only coverage, anganwadi counted as pre-primary, age universe 3+, itemised coaching), so the step between rounds is indicative. Values are nominal rupees of each round. NSO unit-data rider: religion is self-reported and the survey is stratified for states, not religions, so the split is indicative with no sub-state estimates.

Where this data comes from. CMS:E 2025 (NSO unit-level data) · NSS 75th 25.2 Education (2017-18). Download CSV: school-edu-spend.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Of every 100 young people in the college-going age band (18 to 23), how many are attending higher education: a graduate or postgraduate degree course (after Class 12). Measured from a household survey that asks each family directly, so Muslims, Hindus and the all-India average are counted the same way.

Methodology. NSS Gross Attendance Ratio for "post higher secondary": persons currently attending graduate or postgraduate study (any age; NSS "level of current enrolment" codes 15 and 16) over the population aged 18-23, x100, from NSS 75th round Schedule 25.2 (Household Social Consumption: Education, July 2017-June 2018) unit-level data. Code 14 (diploma/certificate at graduation level) is excluded; including it overshoots the published cells by ~6%, and codes 15+16 reproduce all nine NSS Report 585 Statement 8 cells (rural/urban/all x male/female/person) to within 0.05pp. The statement does not split by religion, so the religion split is that validated pipeline applied to the religion-of-head variable.

This REPLACES the former AISHE administrative GER, which tagged only ~7% of all enrolment to a religion and so undercounted every community (its grouped "Other Minority Community" read ~13%, impossible for the highly-educated Christian/Sikh/Jain communities it pools) and published no Hindu figure. On the household-survey basis Muslims attend at 14.5%, Hindus 24.2%, all-India 22.8% (matching the published figure exactly): Muslims are at about 60% of the Hindu rate, the lowest of the major communities, with Muslim women (12.1%) below Muslim men (16.8%). Single validated round, so a snapshot: PLFS publishes no attendance ratio and runs ~8pp high for the same year, and the 71st (2014) education round is Nesstar-locked.

Where this data comes from. NSS 75th 25.2 Education (2017-18). Download CSV: ger-higher-ed.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Of everyone aged 15 and older, the share either working or actively looking for work.

Methodology. The 2023-24 point is PLFS 2023-24 Table 48 (pages 396-400), total residence, person-level; reference period Jul 2023 - Jun 2024 (year=2023 in the canonical row is the start of this period). The 2017-18 to 2022-23 points are computed from each PLFS round's unit-level microdata, pulled via the MoSPI NADA API: the report PDFs publish the 15+ religion split only for their own year, but the microdata carries household religion in every round. The computation reproduces each round's published all-India LFPR to 0.1 points and the 2023-24 religion table within 0.2, so the published point extends the same series.

The 2004/2009/2011 points reach back through the three quinquennial NSS Employment & Unemployment Survey rounds (61st/66th/68th), computed the same way from their unit data and reproducing the published by-religion statements of NSS Reports 568/552 cell-for-cell; the EUS and PLFS designs are not strictly comparable, so the line is dashed at the 2017 break. Two decades tell one story: participation fell economy-wide between 2004-05 and 2017-18 (Muslim 15+ LFPR 55.1% to 45.0%, all-India 63.7% to 49.8%) and recovered after, but the Muslim rate has trailed the all-India rate by 5-9 points in every round since 2004. Religion is self-reported (NSO unit-data rider: indicative, no sub-state estimates).

The "Working vs looking" tab sets this rate against the worker population ratio (share actually working, wpr-15plus.csv) from the same source; the gap between them is unemployment.

Where this data comes from. PLFS 2023-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · EUS microdata 2004-12 (NSS 61st/66th/68th) · EUS microdata 2004-12 (NSS 61st/66th/68th) · EUS microdata 2004-12 (NSS 61st/66th/68th). Download CSV: lfpr-15plus.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Of the labour force aged 15 and older (everyone working or looking for work), the share that cannot find work. A low rate is not automatically good news here: people who cannot afford to be unemployed take any informal work, so read this beside the salaried-jobs card.

Methodology. The 2017-18 to 2023-24 points are computed from each round's PLFS unit-level microdata, pulled via the MoSPI NADA API: the published PLFS annual reports never break the unemployment rate down by religion, so the microdata is the only source for that stretch. The 2004/2009/2011 points come from the three quinquennial EUS rounds (NSS 61st/66th/68th) the same way, and there the gate is direct: the computation reproduces the published by-religion unemployment rates of NSS Reports 568/552. The EUS era shows the same pattern in miniature: Muslim unemployment (2.3-3.0%) sat above the all-India rate (2.0-2.3%) in all three rounds, before both spiked in 2017-18.

EUS and PLFS designs are not strictly comparable; the line is dashed at the 2017 break. Unemployed = usual principal status 81 (seeking/available for work); labour force = workers (ps+ss) plus unemployed; weights per the official estimation rule. The computation reproduces each round's published all-India unemployment rate to 0.1 points.

Muslim unemployment fell from 7.2% (2017-18, a point above the national 6.0%) to 3.2% (2023-24, at par with the national average), but over the same rounds the salaried share of Muslim workers also fell, so part of the convergence is absorption into self-employment and casual work rather than formal hiring. Religion is self-reported (NSO unit-data rider: indicative, no sub-state estimates). Year=Y is the start of the Jul-Y to Jun-(Y+1) reference period.

Where this data comes from. PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · EUS microdata 2004-12 (NSS 61st/66th/68th) · EUS microdata 2004-12 (NSS 61st/66th/68th) · EUS microdata 2004-12 (NSS 61st/66th/68th). Download CSV: unemployment-rate-15plus.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Of all workers, the share in regular salaried jobs rather than self-employment or casual labour.

Methodology. The 2023-24 point is PLFS 2023-24 Table 49 (pages 401-402), rural+urban person-level. Muslims are disproportionately self-employed (62.2% vs Hindu 58.3%) and under-represented in formal salaried work, the classic Sachar-era finding on labour-market segregation. The 2017-18 to 2022-23 points are computed from each PLFS round's unit-level microdata via the MoSPI NADA API (the PDFs publish this religion split only for their own year; the computation reproduces the 2023-24 published table within 0.2 points); the 2004/2009/2011 points come from the three quinquennial EUS rounds (NSS 61st/66th/68th), reproducing the published status-of-employment statements of NSS Reports 568/552.

The long view sharpens the finding: the Muslim salaried share has trailed the all-India share in every round for two decades (12.8% vs 14.4% in 2004-05, 16.2% vs 17.9% in 2011-12, 18.0% vs 21.7% in 2023-24), and within the PLFS years it FELL from its 2018-19 peak (22.2%) while participation rose, so the participation gains came in self-employment and casual work, not regular salaried jobs. EUS and PLFS designs are not strictly comparable; the line is dashed at the 2017 break. Religion is self-reported (NSO unit-data rider).

The "What it pays" tab folds in the pay counterpart computed from the same microdata rounds (the decarded salaried-earnings metric, anchored to the published all-ages earnings tables of the 2021-22 to 2023-24 reports across all 27 residence-by-sex cells within 0.03%): Muslim salaried workers earn 16-23% below the all-India average in every round (2023-24: INR 17,747 vs INR 21,067), so Muslims are both less likely to hold a regular salaried job and paid less when they do. Earnings rupees are nominal, so the comparable story is the gap, not the level.

Where this data comes from. PLFS 2023-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · PLFS microdata 2017-24 · EUS microdata 2004-12 (NSS 61st/66th/68th) · EUS microdata 2004-12 (NSS 61st/66th/68th) · EUS microdata 2004-12 (NSS 61st/66th/68th). Download CSV: salaried-share.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Average monthly spending per person (Monthly Per Capita Consumption Expenditure, MPCE). In India this is the standard measure of economic well-being, used in place of income, for which no reliable figure by religion exists.

Methodology. The 2023-24 figure is computed from the unit-level microdata of HCES 2023-24 (NSO, 2.61 lakh households), pulled via the MoSPI NADA API: weighted monthly per capita consumption expenditure (food + consumables & services + durables, 365-day items annualised), cross-classified by the religion of the household head. The method reproduces the published national MPCE (rural INR 4,122 / urban INR 6,996) within about 2% before the religion split is taken. NSO's note for unit-level users cautions that religion is self-reported and unverified and that the survey is designed to estimate MPCE with the State/UT as the basic stratum, so the religion figures are best read as indicative, no sub-state estimates are made, and the survey's internal Muslim population share (13%) sits a little below the Census.

The earlier benchmark is the Sachar Committee (2006) from the NSS 61st round (2004-05): Muslims INR 635 vs the all-India INR 712, near SC/ST (INR 520) and far below upper-caste 'Hindu General' households (INR 1,023). The two rounds use different consumption methods (2004-05 URP vs 2023-24 MMRP) and rupees are nominal, so the level change over time is mostly inflation; the comparable story is the relative gap, which is stable at about 10% below the national average across the 20 years and is widest in towns (urban Muslims INR 5,467 vs urban Hindus INR 7,064). The By state tab shows the 2023-24 per-state Muslim MPCE computed from the same microdata, urban and rural separately (State/UT is the survey's basic stratum, so this is the finest permitted cut; cells with fewer than 30 sampled Muslim households are suppressed).

The Sachar 2004-05 state figures remain in the downloadable data file. The "Top spending fifth" tab folds in the distribution counterpart computed from the same microdata with the same validated MPCE machinery (the decarded top-quintile-share metric): person-weighted national MPCE quintile cuts (INR 2,903 / 3,677 / 4,632 / 6,302 per person per month), with every all-India quintile share equal to 20.0 by construction as the built-in validation. Muslims sit at par in the poorest fifth (20.5%) but only 13.7% reach the richest fifth (Hindu 20.2%, Christian 30.0%, Sikh 40.6%), and the squeeze is urban: 26.3% of urban Muslims reach the national top fifth against 45.3% of urban Hindus.

Where this data comes from. HCES 2023-24 (NSO unit-level data) · Sachar Committee (NSS 2004-05). Download CSV: mpce.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. What an average household owns (land, buildings, livestock, machinery, vehicles, financial assets) minus what it owes, measured by the national debt and investment survey as on mid-2012 and mid-2018. Wealth is the stock that spending cannot show, and wealth gaps compound across generations, so they run far wider than spending gaps.

Methodology. Computed from the unit-level microdata of AIDIS (All-India Debt & Investment Survey) Visit 1, two rounds: NSS 70th (1.11 lakh households, assets and debt as on 30.06.2012, via the MoSPI NADA API) and NSS 77th (1.16 lakh households, as on 30.06.2018, via MoSPI's original fixed-width TXT distribution; the NADA copy of that round is a proprietary binary, so the build reads the surviving mirror recorded in the repo's provenance notes). The published reports break wealth down by social group but never by religion. The 2018 estimator reproduces the published all-India average assets and average debt EXACTLY to the rupee; 2012 reproduces debt and indebtedness exactly and assets within 0.75% (a documented empty-column defect in MoSPI's own CSV conversion of the non-farm-equipment block).

Muslim households held INR 15.0 lakh in 2018 against the all-India INR 18.9 lakh (79%) and Hindu INR 18.8 lakh (80%), up from 66% and 67% in 2012; the urban gap narrowed most, with urban Muslim households moving from 51% to 67% of urban Hindu net worth (INR 18.0 vs 26.7 lakh in 2018) while rural moved from 83% to 87%. Values are nominal rupees of each round's reference date. Gold and ornaments sit outside the published asset concept (roughly 3-4% on top); land is valued at guideline rather than market rates.

Values rounded to the nearest INR 1,000. NSO unit-data rider: religion is self-reported and the survey is stratified for states, so the split is indicative with no sub-state estimates. The "Borrowing sources" tab folds in the companion measure from the same microdata (the decarded institutional-credit-share metric): the institutional share of outstanding cash debt among indebted households, reproducing the published all-India shares in both rounds (2012 rural 56.0 / urban 84.5 exactly; 2018 rural 66 / urban 87).

There the trend cuts the other way: the Muslim share stood still (68.3% to 68.7%) while the Hindu share rose (72.4% to 77.1%), widening the urban gap to 17 points (71.3 vs 88.4). Muslims also borrow the least overall (26.8% of households indebted vs Hindu 31.4% in 2018). Credit-agency code 09 "others" is non-institutional despite its position in the code list in both rounds (the layout trap is documented in the repo).

Where this data comes from. AIDIS 2019 (NSS 77th) · AIDIS 2013 (NSS 70th). Download CSV: household-net-worth.csv. You can open any of these to check the numbers on this chart yourself.

From microdata Reproduce this. This figure is recomputed from the unit-level survey microdata, the raw NSS or PLFS records behind the linked source. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. How many of the 543 seats in India's national parliament (Lok Sabha) are held by Muslim MPs, counted at every general election since 1952. At the community's 14.2% population share, parity would be 77 seats; the count peaked at 49 in 1980 and stands at 24 today.

Methodology. Religion is not tabulated by either ECI or PRS Legislative Research in their official candidate-profile publications. Underlying data is ECI candidate affidavits classified post-election by journalists and researchers. Figures cross-verified across multiple sources: Maktoob Media, FACTLY, The India Forum, Statista.

Documented as "manual entry" canonical with citations in each row's methodology note. The series covers all 18 general elections, 1952-2024: the 1952-2004 figures are from historical compilations (The India Forum, Statista), 2009 onward cross-verified across the sources above. The card displays the absolute number of Muslim MPs per house (user call, Commit FW: counts read more directly than shares); the percentage rides in the canonical data, where the denominator record carries each house's N-of-M seat count, and house sizes ranged from 489 (1952) to 543 (since 1996).

Representation peaked at 49 MPs (9%) in 1980 and has sat between 22 and 27 since 2014: 2009=28/543, 2014=22/543, 2019=27/543, 2024=24/543. At the Muslim population share of ~14.2%, parity in the current house would be 77 seats, a chronic shortfall of about 50 seats. The state-assembly layer of the same affidavit-classified sourcing is the separate "Muslim share of state assemblies" card (about 6% of seats across the 31 covered assemblies).

Where this data comes from. PRS / ECI affidavits. Download CSV: ls-share.csv. You can open any of these to check the numbers on this chart yourself.

Hand-compiled Reproduce this. These counts are hand-compiled from election-wise ECI candidate affidavits and cross-checked across multiple journalistic sources, since no single official source tabulates members by religion. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Percentage of state legislative assembly seats held by Muslim MLAs. National aggregate (~6%) plus per-assembly values for all 31 state + UT assemblies that hold elections (28 states + Delhi, Puducherry, J&K, most-recent election per assembly, spanning 2020-2026). Each state pairs two bars: the maroon share of assembly seats against the grey share of population (Census 2011), so each state's seats can be compared at a glance with its share of the population.

Methodology. Manual-entry metric, like ls-share: religion is not officially tabulated by ECI or PRS. Underlying data is candidate-affidavit religion classified post-election by journalists (Maktoob, Clarion, India Forum, The Wire, Deccan Herald, Outlook, Radiance Weekly, Free Press Journal, ummid.com, thenewzradar). National aggregate ~6% across the 31 covered assemblies.

Highest: J&K 60% (2024, first election since Article 370 reorg); Kerala 25%, Assam 24.6%. Lowest: HP / Goa / Mizoram / Nagaland / Sikkim / Arunachal / Chhattisgarh / Tripura / Meghalaya, all 0% (most have never elected a Muslim MLA in state history). Each row's methodology_note cites the source compilation(s) used.

The per-state population context (the grey bars and the By-assembly table's population column) is the Census 2011 Muslim share: the J&K figure includes pre-2019 Ladakh, and Telangana's figure (12.7%) is aggregated from the ten undivided-Andhra-Pradesh census districts that formed the state in 2014, Hyderabad included.

Where this data comes from. PRS / ECI affidavits. Download CSV: mla-share.csv. You can open any of these to check the numbers on this chart yourself.

Hand-compiled Reproduce this. These counts are hand-compiled from election-wise ECI candidate affidavits and cross-checked across multiple journalistic sources, since no single official source tabulates members by religion. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Incarceration rate: prisoners of religion R per 100,000 of religion R's total population. More direct measure of disproportion than "share of prisoners", comparable across religions without needing to know population baselines.

Methodology. Cross-source: NCRB PSI 2022 total prisoner count (convicts + undertrials + detenues + other prisoners; STATES + UTs subtotals = ALL-INDIA by construction) divided by Census 2011 national religious population times 100,000. The 'denominator' column on each canonical row records the exact (count, population) pair used. Caveat: Maharashtra non-reporting of religion for ~33k undertrials/detenues, so the religion-reported numerator excludes those; full national population is the denominator.

The "Undertrials" tab folds in the sister metric (the decarded undertrial-rate-per-100k, NCRB PSI 2022 Table 2.11C over the same Census 2011 denominators): 48.8 Muslim undertrials per 100,000 against 29.3 Hindu and 33.2 all-India, a sharper over-representation than for prisoners overall, so the disproportion is largest before any conviction. Same Maharashtra caveat applies.

Where this data comes from. NCRB PSI (2018-2023) · Census 2011 · C-series. Download CSV: prison-rate-per-100k.csv. You can open any of these to check the numbers on this chart yourself.

Computed Reproduce this. This figure is computed from the linked published tables. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.

Definition. Number of communal/religious rioting incidents recorded by NCRB under IPC Sec.147-151 (rioting), specifically the "Communal/Religious" sub-classification. National annual totals (2015-2023) plus state-level breakdown for 2022 and 2023.

Methodology. NCRB Crime in India, Table 1.2 row 23.1 'Communal/Religious' rioting. National series 2015-2023, compiled across multiple Crime in India volumes (CII 2017 through 2023, each carrying a multi-year national panel); state-level breakdown for 2022 and 2023. Incident counts only, with no religion of victim/perpetrator in the published table.

Caveat: several states stopped recording 'communal' as a separate crime category since ~2017, which deflates the national total over time. Civil-society compilations typically report higher counts.

Where this data comes from. NCRB CII (2015-2023) · NCRB CII (2015-2023) · NCRB CII (2015-2023) · NCRB CII (2015-2023) · NCRB CII (2015-2023). Download CSV: communal-incidents-govt.csv. You can open any of these to check the numbers on this chart yourself.

Read Reproduce this. This figure is published directly in the linked source table. Every figure on this card, in every tab, is computed from the data file above by an open script: transform code. Each data-file row also records its own source and method.