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.

BharatDistrictPanel

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.

One district x year panel for India that joins sixteen official sources on current Local Government Directory (LGD) district codes, with Census 2011 codes carried alongside. Every value keeps its source resource id, the way its district was resolved and any caveat flag.

Districts: 796 current LGD districts (795 with at least one value)
Indicators: 31,839 in the dictionary, 31,783 with values; long rows: 10,868,342; wide table: 19,649 district-years x 300 indicators

Quickstart

The repository is gated: accept the terms on the dataset page and log in with huggingface-cli login first. long.parquet has one row per district x year x indicator; indicators.parquet explains each indicator id.

# pip install pandas pyarrow duckdb huggingface_hub
import duckdb
import pandas as pd
from huggingface_hub import snapshot_download

root = snapshot_download("judicialmind/BharatDistrictPanel", repo_type="dataset")

ind = pd.read_parquet(f"{root}/indicators.parquet")
print(ind[ind.source == "mgnrega"][["indicator", "label", "unit", "n_districts"]].head(3).to_string(index=False))

# one indicator, every district and year, with provenance (source resource, flags, how the district was matched)
mg = pd.read_parquet(f"{root}/long.parquet", filters=[("indicator", "==", "mgnrega.avg_days_per_hh")])
print(mg[mg.district_name == "Nashik"][["period", "value", "source_resource_id", "flag", "geo_method"]].tail(3).to_string(index=False))

# join two sources on the current LGD district code
print(duckdb.sql(f"""
    SELECT e.state_name, e.district_name, e.value AS eshram_workers, p.value AS pmkisan_beneficiaries_2024
    FROM '{root}/long.parquet' e JOIN '{root}/long.parquet' p USING (district_lgd)
    WHERE e.indicator = 'eshram.workers' AND p.indicator = 'pmkisan.beneficiaries' AND p.year = 2024
    ORDER BY eshram_workers DESC LIMIT 3
""").df().to_string(index=False))

wide = pd.read_parquet(f"{root}/wide.parquet")
print(wide.shape)

Output:

First 3 MGNREGA indicators

Indicator	Label	Unit	Districts
mgnrega.active_job_cards	Active job cards	job cards	733
mgnrega.active_workers	Active workers	persons	733
mgnrega.avg_days_per_hh	Average days of employment per household	days	733

Nashik, average days of employment per household (last 3 years)

Period	Value	Source resource	Flag	District match
2024-25	35	ee03643a-ee4c-48c2-ac30-9f2ff26ab722	–	name_exact
2025-26	41	ee03643a-ee4c-48c2-ac30-9f2ff26ab722	–	name_exact
2026-27	11	ee03643a-ee4c-48c2-ac30-9f2ff26ab722	partial_year	name_exact

Top 3 districts by eShram workers, with PM-KISAN beneficiaries (2024)

State	District	eShram workers	PM-KISAN beneficiaries, 2024
West Bengal	Murshidabad	2,424,210	328,167
Uttar Pradesh	Prayagraj	2,332,328	264,555
West Bengal	South 24 Parganas	2,290,466	326,784

Wide table shape: 19,649 rows × 306 columns.

With datasets: load_dataset("judicialmind/BharatDistrictPanel", "indicators", split="train") (configs long, wide, indicators, districts).

Source data

source	what	years	rows	indicators	districts
aadhaar	UIDAI Aadhaar enrolment and update counts (data.gov.in), de-duplicated	2025-2026	10,588	7	762
census2001	Census 2001 Primary Census Abstract and B-01/B-04 workers (censusindia district files; NDAP 7097 where complete); unchanged-boundary districts or complete village re-aggregation	2001	21,576	116	302
census2011	Census 2011 Primary Census Abstract (village/town-wise via data.gov.in; NDAP 6000 where complete), B-01/B-04 workers and C-16 mother tongue	2011	96,674	259	748
eshram	eShram registrations (aggregates from Bharat-Workforce-Atlas); suppressed < 10	2026	41,152	63	744
hmis	HMIS sub-district monthly reports and item-wise district history (MoHFW via data.gov.in), re-aggregated	2009-2021	204,808	80	753
jjm	Jal Jeevan Mission habitation-level tables (Ministry of Jal Shakti, data.gov.in) rolled up to districts	2024-2026	9,762	14	707
kcc	Kisan Call Centre call logs (category counts); suppressed < 10	2009-2025	76,047	18	299
lgd	Local Government Directory local bodies (panchayats) per district	2026	3,815	7	782
mandi	Agmarknet daily mandi prices (data.gov.in); district names matched	2001-2026	400,675	63	577
mgnrega	MGNREGA district-wise data at a glance (data.gov.in)	2018-2026	139,975	22	733
ndap	NITI Aayog NDAP standardised district datasets, plus village-level NDAP sets rolled up to districts	1991-2026	9,163,428	31,042	785
nrlm	DAY-NRLM village-wise self-help groups and members (data.gov.in) summed to districts	2017-2025	56,053	9	752
pmkisan	PM-KISAN village beneficiary counts (aggregates); suppressed < 10	2018-2025	23,383	4	777
rain	IMD district-wise daily rainfall via NDAP dataset 7319, summed over the year	2021-2023	8,192	4	710
udise	UDISE+ district tables: enrolment, schools, teachers (data.gov.in); district names matched	2012-2019	222,071	50	705
udyam	UDYAM MSME registrations (aggregates); suppressed < 10	2020-2026	390,143	83	785

Original data: published by the Ministry of Agriculture and Farmers Welfare (PM-KISAN, Kisan Call Centre, Agmarknet), Department of School Education and Literacy (UDISE+), Ministry of Health and Family Welfare (HMIS), Ministry of Labour and Employment (eShram), Ministry of Micro, Small and Medium Enterprises (UDYAM), Ministry of Rural Development (MGNREGA, DAY-NRLM), Ministry of Jal Shakti (Jal Jeevan Mission), UIDAI (Aadhaar), Ministry of Panchayati Raj (Local Government Directory), India Meteorological Department, Registrar General and Census Commissioner, India, and NITI Aayog (NDAP) 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.

Files

file	content
`long.parquet`	one row per district x year x indicator (schema below)
`wide.parquet`	one row per district-year, one column per indicator (top 300 indicators by coverage; no provenance - use `long` for that)
`indicators.parquet`	indicator dictionary: `indicator, label, unit, kind, domain, source, period_type, description` and coverage (`n_districts, n_district_years, year_min, year_max, n_years, n_values, district_coverage`)
`districts.parquet`	current districts: `district_lgd, state_lgd, district_name, state_name, n_villages, origin, n_sources, n_indicators, year_min, year_max, census2011_districts, boundary_changed_since_2011`
`crosswalk_census2011.parquet`	Census 2011 district x current LGD district with population shares and relation (`one_to_one` / `split` / `merge` / `split_merge`)
`census2001_matching.parquet`	each Census 2001 district file and the decision taken for it (kept, or why it was not used)
`lineage.parquet`	retired or pre-split LGD district codes -> current codes with village-share weights (empty in this release: no retired codes needed mapping)
`coverage.parquet`, `coverage_source_year.csv`	indicator x year and source x year district counts
`unresolved.parquet`	source rows that could not be placed in a current district (`source, geo_kind, state_name, district_name, district_code, rows, value_sum`)

`long.parquet` schema

column	meaning
`district_lgd`, `state_lgd`	current LGD district / state code
`district_name`, `state_name`	current LGD names
`census2011_district`	Census 2011 district code where one applies
`year`	calendar year, or the starting year of a financial / academic / agricultural year
`period`, `period_type`	period label (`2019-20`, `2011`, ...) and its type (`calendar`, `fy`, `academic`, `agri_year`, `census`, `snapshot`)
`indicator`	indicator id, `<source>.<name>`; see `indicators.parquet`
`value`, `unit`	value (float) and unit
`source`, `source_resource_id`	source key and the data.gov.in resource id(s) (`;`-separated) or `ndap-<id>`
`flag`	`;`-separated caveat flags (see Conventions)
`suppressed`	true when a count from person-level records was below 10 and is withheld (`value` null)
`geo_method`	how the row was placed in its current district (`source_code`, `name_exact`, `name_alias_exact`, `name_prefix`, `name_fuzzy`, `census2011_code`, `village_reaggregation`, `subdistrict_reaggregation`, ...)
`weight`	share of the source value attributed to this district (< 1 when apportioned)

NDAP indicator ids are ndap.<dataset_id>.<I col>[...]; NDAP column names (I<id>_<n>) are kept because their labels were not available.

How districts are harmonised

Census 2011 is re-aggregated from villages: each Census village is located in its current district through the LGD village directory (Census 2011 village code), so districts carved out after 2011 get their own 2011 values; urban areas follow the majority district of their tehsil / CD block. data.gov.in publishes the village/town-wise PCA for only part of India, so a current district gets Census 2011 values only when >= 98% of its villages are covered (NDAP's village-level PCA fills districts where its copy is complete). Census 2011 B-01/B-04/C-16 counts for districts split after 2011 are apportioned (flag apportioned).
Census 2001 values come from the district rows of the census tables where the 2001 district maps 1:1 to a current district (name match, 1:1 with Census 2011, plausible 2001-2011 growth), or from complete NDAP village rows re-aggregated into current districts; every 2001 district's decision is in census2001_matching.parquet.
HMIS is re-aggregated from sub-districts the same way.
Sources that publish LGD codes (PM-KISAN, UDYAM, NDAP, eShram, KCC) keep them; a retired or pre-split code is mapped to its successors with weights from where its villages are today (lineage.parquet). Counts are apportioned by those weights (flag apportioned, weight < 1), rates and prices are copied (flag inherited), and unlabelled NDAP values are not split at all. Where a source gives one current code to several differently named districts (the eShram atlas does this for a few dozen districts), each name is placed by name matching instead; names that match nothing go to unresolved.parquet.
Sources with names only (mandi, UDISE+, part of KCC) are matched by name within the state (exact; alias incl. official renames, Census 2011 names and spellings used by coded sources, where a district's own LGD name always wins over another district's alias; whole-word prefix such as Agar -> Agar-Malwa; then fuzzy >= 0.88 with the same first letter), recorded in geo_method. Mandi names are resolved before aggregating, so all spellings of a district share one median price and one market count. A district that kept its name after part of it was carved out (boundary_changed_since_2011 in districts.parquet) receives such name-matched values for its old extent in years before the split.
Geography: 675,247 LGD villages, 0 retired codes mapped, 0 of them split.

Conventions

year is the calendar year, or the starting year of a financial / academic / agricultural year (period shows 2019-20; period_type says which).
Flags: partial_year (fewer than 12 months reported), lower_bound (sum includes suppressed cells), apportioned, inherited, combined (several source rows summed or averaged), reaggregated_split (Census 2011 value of a district created after 2011), complementary_suppressed (hidden so a suppressed cell cannot be recovered by subtraction), excludes_suppressed_transgender (PM-KISAN total without a suppressed transgender count), hmis_dicr_legacy (FY 2010-13 HMIS value from the legacy district sheets), hmis_itemwise_history (HMIS value from the item-wise monthly district reports, FY 2009-10 to 2020-21), definition_change_2017 (HMIS item whose 2008 and 2017 definitions were matched with medium confidence), ndapx (NDAP district dataset with extra dimensions fixed or aggregated over), ndapx_village_rollup (NDAP village-level dataset summed to districts).
Ratios such as sex ratio at birth or female share are recomputed from the harmonised counts.

Privacy

No row-level or personal data. Every count derived from person-level records (PM-KISAN, eShram, KCC, UDYAM) below 10 is null with suppressed = true, before and after harmonisation. Where a published total would reveal a single hidden category by subtraction (KCC call categories, eShram gender / age / education / occupation, PM-KISAN female / male), a second category is suppressed too (complementary_suppressed), and shares are computed only from counts that remain published.

Coverage

India, 796 current LGD districts (795 with at least one value), 1991-2026; per-source years and district counts are in the Sources table and in coverage_source_year.csv.
KCC 2025 is a partial year (flag partial_year).
Rows that could not be placed in a current district (name not matched, unknown code) are listed in unresolved.parquet.
Census values exist only for districts where they are complete: the village/town-wise PCA on data.gov.in covers part of India, and NDAP's village-level copy fills districts where it is complete.

Source data

Derived from 32,613 resources published on the Open Government Data Platform India (data.gov.in) and NITI Aayog's NDAP, snapshot May 2026, under the Government Open Data License - India. Every value names its resource in source_resource_id (a data.gov.in resource id resolves at https://api.data.gov.in/resource/<id>; ndap-<id> at https://ndap.niti.gov.in/dataset/<id>). GODL-India does not cover personal information; this dataset holds only aggregates.

Providers: Department of Agriculture and Farmers Welfare (PM-KISAN, Kisan Call Centre), Department of School Education and Literacy (UDISE+), Directorate of Marketing and Inspection (Ministry of Agriculture and Farmers Welfare), Ministry of Health and Family Welfare (HMIS), Ministry of Labour and Employment (eShram), Ministry of Micro, Small and Medium Enterprises (UDYAM), Ministry of Panchayati Raj (Local Government Directory), NITI Aayog (National Data and Analytics Platform), Registrar General and Census Commissioner, India (Ministry of Home Affairs).

Also: Ministry of Rural Development (MGNREGA, resource ee03643a-ee4c-48c2-ac30-9f2ff26ab722; DAY-NRLM, resource d4206736-a28b-4552-8900-7e0c23c707ac), Ministry of Jal Shakti (Jal Jeevan Mission habitation-level block tables; these values carry jjm:<table> in source_resource_id), Unique Identification Authority of India (Aadhaar, resources 19eac040-0b94-49fa-b239-4f2fd8677d53, ecd49b12-3084-4521-8f7e-ca8bf72069ba, 65454dab-1517-40a3-ac1d-47d4dfe6891c), Ministry of Panchayati Raj (local bodies, resource 1a6c26ed-d67c-40ea-aa20-d38d35f341a5), India Meteorological Department (district rainfall, via NDAP dataset 7319).

NDAP data: NITI Aayog, National Data and Analytics Platform (ndap.niti.gov.in), used under its open data terms; NDAP indicator column names (I<id>_<n>) are kept because their labels were not available.

Citation

@misc{judicialmind2026districtpanel,
  title={BharatDistrictPanel: A harmonised district x year panel of Indian official statistics on LGD codes},
  author={JudicialMind},
  year={2026},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/judicialmind/BharatDistrictPanel}}
}

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.