This dataset is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 licence (CC BY-NC-SA 4.0): research, academic, personal and other non-commercial use is permitted with attribution; adaptations must be shared under the same licence; commercial use is not permitted. The original public source data remains under the Government Open Data License - India (GODL-India). Commercial use: contact surya@judicialmind.ai or hello@judicialmind.ai.

PM-KISAN Reach Atlas

Licence: CC BY-NC-SA 4.0 — research and non-commercial use only; derivatives must be shared under the same licence. Commercial use: contact surya@judicialmind.ai or hello@judicialmind.ai

Access is approved automatically when the email you enter matches your Hugging Face account email and belongs to the organisation you name; other requests are reviewed manually.

Who PM-KISAN reaches, installment by installment: beneficiaries by village, sub-district, district and state for installments 1-22 (Dec 2018 to Mar 2026), with women's share, change from the previous installment, coverage against Agriculture Census 2015-16 operational holdings and Census 2011 cultivators, gender-gap ranks and a list of districts with sharp drops. Totals are corrected for the records missing from the public dump (est_* columns, with standard errors).

Quickstart

The repository is gated: accept the terms on the dataset page, then huggingface-cli login.

from datasets import load_dataset

districts = load_dataset("judicialmind/PMKISAN-Reach-Atlas", "district_installment", split="train").to_pandas()
latest = districts[districts.installment == 22]
print(latest[["state_name", "district_name", "est_total", "est_se", "female_share", "coverage_holdings"]]
      .sort_values("est_total", ascending=False).head())

With pandas or DuckDB on a local copy (huggingface-cli download judicialmind/PMKISAN-Reach-Atlas --repo-type dataset --local-dir PMKISAN-Reach-Atlas):

import duckdb, pandas as pd

nat = pd.read_parquet("PMKISAN-Reach-Atlas/national_installment.parquet")
print(nat[["installment", "fy", "est_total", "est_se", "female_share", "change_ratio"]].tail(4))

print(duckdb.sql("""
    SELECT state_name, district_name, installment, panel_ratio, parent_panel_ratio
    FROM 'PMKISAN-Reach-Atlas/sharp_drops.parquet'
    WHERE level = 'district' ORDER BY panel_ratio LIMIT 5
""").df())

Use est_total / est_female / est_male (with est_se) for totals; male, female, transgender, total are the counts actually held in the source dump. Use panel_ratio for change between installments.

Files

file	rows	contents
`census_sources`	291	Census 2011 village PCA resources used
`district_fy`	6,069	district x financial year
`district_installment`	16,723	district x installment (+ change, sharp_drop, coverage)
`gender_gap_ranks`	1,496	district female-share ranks (rank 1 = lowest), latest installment and latest full FY
`installments`	22	installment -> financial year, quadrimester, period, reference-village hit rate
`national_fy`	8	all-India x financial year
`national_installment`	22	all-India x installment
`sharp_drops`	129	district (vs state) and state (vs India) installments with sharp drops; null ratio = fell below 10
`state_completeness`	35	share of portal records held, per state
`state_fy`	276	state x financial year
`state_installment`	763	state x installment
`subdistrict_fy`	54,216	sub-district x financial year
`subdistrict_installment`	151,226	sub-district x installment
`village_continuity`	899	villages paid again next installment (completely held states only)
`village_installment`	5,397,304	village x installment, observed counts (villages < 10 combined)
`charts/*.csv`		small extracts for plotting: national series, latest state totals, female share by state and FY, 25 highest / lowest district female shares, top 50 sharp drops
`FINDINGS.md`		headline results and the estimator self-check
`summary.json`		table row counts, per-state completeness, census and holdings match rates, findings in machine-readable form

All .parquet tables carry an attribution column (provider, title, resource URL).

Schema

Keys and labels: state_lgd, district_lgd, subdistrict_lgd, block_lgd, village_lgd (LGD codes), village_code (source village code), *_name, installment (1-22), fy (financial year, e.g. 2025-26), fin_year_id, quadrimester, period_from, period_to (installment period).

column	tables	meaning
`male`, `female`, `transgender`, `total`	installment tables	beneficiaries held in the source dump (counts below 10 are null)
`n_villages`	installment tables	villages contributing to the row
`rolled_up`, `n_suppressed`	village / all	row combines villages with < 10 beneficiaries; number of suppressed cells in the row
`cell_c_raw`, `cell_c`, `cell_c_source`, `n_ref`, `n_hit`, `c_ref`, `c_unusable`	district	completeness of the district x installment cell from reference villages, and how it was measured (`cell`, `state`, `state_overall`, `complete`)
`completeness`	all	share of the cell's records held (1 = exact)
`est_total`, `est_female`, `est_male`, `est_var`, `est_se`	installment tables	estimated beneficiaries corrected for missing records, variance and standard error
`imputed`, `cells_imputed`, `cells`, `imputed_share`, `grid_filled`	installment tables	cells holding under 5%, interpolated from the district's own series, and their weight in totals
`female_share`	most	female / (male + female) beneficiaries
`prev_est_total`, `prev_total`, `change_ratio`, `change_se`, `change_z`	installment tables	change in the estimated total from the previous installment
`panel_villages`, `panel_prev`, `panel_curr`, `panel_ratio`, `panel_se`, `panel_z`, `panel_share_falling`	installment tables	change measured over villages held in both installments
`parent_change_ratio`, `parent_panel_ratio`, `change_vs_parent`, `panel_vs_parent`	district, state	the same ratios for the parent (state or India) and the ratio to it
`villages_moved_out`, `moved_out_share`, `boundary_change`	district	villages that sit in another district next installment
`sharp_drop`, `drop_basis`	district, state	sharp-drop flag (see Method) and the test used (`panel`, `full`)
`holdings_2015`, `coverage_holdings`	district, state, national	operational holdings 2015-16 and est_total / holdings
`est_total_matched`, `census_villages_matched`, `cultivators_2011`, `households_2011`, `coverage_cultivators`	district	coverage against Census 2011 main + marginal cultivators in matched villages
`est_total_mean`, `est_total_max`, `total_mean`, `female_sum`, `total_sum`, `n_installments`, `first_installment`, `last_installment`	`*_fy`	financial-year aggregates of the installments in that year
`gender_gap_pp`, `rank_in_state`, `n_in_state`, `rank_national`, `n_national`, `period`	gender_gap_ranks	(male share - female share) x 100 and ranks (1 = lowest female share)
`ref_hit_rate_complete`, `ref_hit_n`, `ref_hit_ceiling`, `ref_hit_rate`	installments	reference-village hit rates used to correct completeness
`villages`, `villages_paid_next`, `village_retention`, `beneficiaries`, `beneficiaries_in_retained`, `next_beneficiaries_in_retained`, `beneficiary_ratio_in_retained`, `villages_new_next`	village_continuity	villages paid again in the next installment
`raw_rows`, `distinct_rows`, `api_total`, `duplicate_rows`, `missing_records`, `total_source`	state_completeness	records fetched, distinct, on the portal, and missing
`uuid`, `title`, `district`, `state`, `in_dump`	census_sources	Census 2011 village PCA resources

Method

Source rows: one per village x installment with male, female and transgender beneficiary counts. Exact duplicate rows are removed.
Completeness: the loss is clustered by district x installment (0-90% held inside one state), so it is measured per cell: cell_c = share of reference villages (>= 10 beneficiaries in the installments before and after) that are held (cell_c_raw), divided by ref_hit_rate (installments.parquet). In the completely held states some reference villages really were not paid in installments 1, 2, 6 and 11 (ref_hit_rate_complete 81-93%), and the raw rate counts that as missing records. ref_hit_rate = max(ref_hit_rate_complete, ref_hit_ceiling = 99th percentile of raw cell rates in incomplete states), a conservative correction. est_total = observed / cell_c, est_se combines village sampling and the uncertainty in cell_c. Cells holding under 5% are interpolated from the district's own series (imputed, s.e. set to 50%); imputed_share gives their weight in state and national totals. Completely held states are exact (completeness = 1). state_completeness.parquet lists records held per state.
Change: change_ratio = est_total / previous installment's est_total. panel_ratio = beneficiaries now / before over villages held in both installments (panel_villages), with panel_z from the spread across villages; it does not depend on how much of each cell was lost. sharp_drop marks districts whose panel_ratio is <= 0.75, at least 15% below their state's, while the state itself did not fall by 25%, with z <= -3, at least half of the villages falling by 25% themselves (panel_share_falling), >= 30 villages and >= 1,000 beneficiaries before (drop_basis = panel); in completely held states the exact totals are tested the same way (full). Never flagged when 10% or more of the district's villages sit in another district next installment (boundary_change). States are flagged the same way against India.
Coverage: coverage_holdings = est_total / operational holdings 2015-16 (NDAP 7171, LGD-coded). coverage_cultivators = est_total in villages whose LGD code and name match a Census 2011 village / main + marginal cultivators (persons) in those villages. Census match: 43.4% of coded villages, names agree for 99.6% of matches; only name-agreeing matches are used. These are ratios of different units (registered farmer families vs holdings vs working cultivators) and can exceed 1.
Gender gap: gender_gap_pp = (male share - female share) x 100 of beneficiaries; rank 1 = lowest female share.
Village continuity (village_continuity.parquet) is only given for states held completely, where a village missing from the next installment really was not paid.

Results

From FINDINGS.md:

Installment 22 (2025-12-01 to 2026-03-31): estimated 96,704,192 beneficiaries (+/- 1,124,854 s.e.), 23.1% women.
Peak installment: 19 with 97,039,643.
Coverage vs 2015-16 operational holdings, latest installment, all India: 67.6%.
Lowest female share (large states): Punjab 1.0%, Haryana 16.1%, West Bengal 17.4%. Highest: Kerala 44.1%, Andhra Pradesh 33.8%, Rajasthan 30.4%.
102 district-installments flagged as sharp drops (95 districts). Largest: KHUNTI (Jharkhand) installment 12: -97% vs state -17%; KAKCHING (Manipur) installment 6: -94% vs state 53%; THOUBAL (Manipur) installment 6: -92% vs state 53%.
National installments with the largest falls: 6 (-38%); 12 (-19%); 13 (-6%).

Check against published installment totals (press-release headline figures):

installment	estimate	reference	difference
17	91,651,319	92,600,000	-1.0%
18	95,113,500	94,000,000	+1.2%
19	97,039,643	98,000,000	-1.0%
20	94,845,919	97,000,000	-2.2%

Estimator self-check (villages deleted at random within district x installment in the 10 completely held states, then re-estimated): district totals measured from their own reference villages (cell_c_source = cell) have a median error of 8.0% (98% within 2 s.e., n=174). Panel change ratio vs the true change: median absolute error 0.0075 (90th pct 0.0579). Details in FINDINGS.md.

Coverage

35 states and UTs, installments 1-22 (Dec 2018 to Mar 2026), down to village level. The source dump holds 48.2% of the portal's 12,928,182 (village, installment) records: the May 2026 refetch hit a 500,000-row cap per state and offset paging returned 2,004,030 duplicate rows in place of other records. Duplicates are removed; the gap is corrected per district x installment from reference villages (see Method). 10 states/UTs are held completely (completeness = 1).

state	completeness	API total
Uttar Pradesh	19.4%	2,131,258
Karnataka	22.9%	611,499
Madhya Pradesh	35.5%	1,127,052
Odisha	40.3%	1,004,743
Maharashtra	43.1%	948,480
Rajasthan	45.1%	941,707
Bihar	46.2%	871,581
Jammu And Kashmir	57.4%	138,964
Jharkhand	61.2%	624,603
Himachal Pradesh	62.0%	371,396
West Bengal	63.4%	569,902
Haryana	66.6%	155,623
Assam	67.6%	472,786
Tamil Nadu	67.9%	372,026
Kerala	68.1%	295,470
Meghalaya	70.4%	67,458
Chhattisgarh	70.6%	420,073
Punjab	72.9%	258,268
Andhra Pradesh	73.7%	377,011
Arunachal Pradesh	74.4%	70,016
Gujarat	74.6%	399,689
Telangana	74.9%	234,788
Uttarakhand	74.9%	323,335
Manipur	75.5%	43,604
Nagaland	97.8%	31,153

Privacy

The source is public at village level on data.gov.in and holds counts only, no names of people. Every count below 10 is null, shares are null when the male or female count is below 10, and villages with fewer than 10 beneficiaries in an installment are combined per sub-district (rolled_up).

Source data

Derived from 292 resources published on the Open Government Data Platform India (data.gov.in), snapshot May 2026, under the Government Open Data License - India. Per-resource attribution (provider, title, resource URL) is in the attribution column. GODL-India does not cover personal information; the source holds counts only.

Providers: Department of Agriculture and Farmers Welfare (Ministry of Agriculture and Farmers Welfare), Registrar General and Census Commissioner, India (Ministry of Home Affairs).

Attribution:

PM-KISAN: Department of Agriculture and Farmers Welfare (Ministry of Agriculture and Farmers Welfare). "Village and Gender - wise Beneficiaries Count under PM - KISAN Scheme". data.gov.in resource 388208c6-d82a-4190-90df-91aa2c326fec, https://api.data.gov.in/resource/388208c6-d82a-4190-90df-91aa2c326fec. Licensed under the Government Open Data License - India (https://data.gov.in/government-open-data-license-india). Snapshot: May 2026.
Census 2011 village Primary Census Abstract: Registrar General and Census Commissioner, India (Ministry of Home Affairs), 291 data.gov.in resources "Village/Town-wise Primary Census Abstract, 2011 - District of "; uuids in census_sources.parquet.
Holdings: Source: NITI Aayog, National Data and Analytics Platform (NDAP), dataset 7171: Estimated Number and Area of Operational Holdings by Size Classes and Tenancy Status. https://ndap.niti.gov.in/dataset/7171

Original data: published by Department of Agriculture and Farmers Welfare, Ministry of Agriculture and Farmers Welfare (PM-KISAN); Registrar General and Census Commissioner, India, Ministry of Home Affairs (Census 2011) on the Open Government Data Platform India (data.gov.in) and, where marked, NITI Aayog's National Data and Analytics Platform (NDAP), under the Government Open Data License - India (GODL-India). The original public data remains under GODL-India; the CC BY-NC-SA 4.0 licence of this repository covers JudicialMind's compilation, derived fields, annotations, benchmarks and models.

Citation

@misc{judicialmind2026pmkisanreach,
  title={PMKISAN-Reach-Atlas: PM-KISAN beneficiaries by village, district and state, installments 1-22},
  author={JudicialMind},
  year={2026},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/judicialmind/PMKISAN-Reach-Atlas}}
}

License

CC BY-NC-SA 4.0 (see LICENSE): research, academic, personal and other non-commercial use, with attribution to JudicialMind and to the original data publishers; adaptations must be shared under the same licence. Commercial use requires a separate licence.

Commercial use: contact surya@judicialmind.ai or hello@judicialmind.ai.

Downloads last month: 20

Total file size:

58.7 MB

Collection including judicialmind/PMKISAN-Reach-Atlas

India Open Government Data — Bharat Agriculture Datasets

Collection

Research and non-commercial use (CC BY-NC-SA 4.0). Commercial: surya@judicialmind.ai / hello@judicialmind.ai • 3 items • Updated 4 days ago