# Are Indians protesting more? A research programme

Working folder for a six-part research series and a compiled report on protest in India, 2000 to 2026. The dataset comes first, the writing second. Nothing here is published until it is reviewed and moved into `content/` on `main`.

## The question

Who is protesting in India, why now, who pays and who amplifies, and how much of it is engineered?

The series tests the "fifth-generation warfare" hypothesis (that the recent protests are manufactured or amplified from outside) against the ordinary explanations: jobs, prices, exam leaks, the election calendar, copycat effects from Nepal and Bangladesh, and the dynamics of social media. Where the evidence shows foreign money or coordinated amplification, the report says so, with the documents. Where it shows ordinary grievance, it says that too. The conclusion is not written before the evidence.

Legal note: naming an organisation or person as a foreign agent without documents is defamation. Every such claim in the series must quote a court order, an MHA filing, an FCRA notice or an official statement, with the rebuttal where one exists.

## The six parts

1. **Is there really a spike? Count it.** This folder. Protests per month, by state, by type, 2000 to 2026, with the sources below.
2. **The grievance baseline.** Youth unemployment (PLFS), exam leaks and recruitment backlogs, food and fuel inflation, farm prices, the US tariff shock, the Gulf remittance hit. Where protests cluster against where the pain is.
3. **The money.** FCRA: which organisations lost licences 2020 to 2026 and why, MHA filings, court orders, foreign-funding flows by donor country. Only what the documents show.
4. **The information layer.** How protests are amplified: coordinated hashtags, bot networks, Pakistani and Chinese state-media patterns, foreign press narratives; India's own grey-zone doctrine. Network evidence, not assertion.
5. **Case studies.** Five or six protests examined the same way: trigger, grievance, organisers, funding, amplification, state response, outcome. CAA, the farmers' movement, Manipur, Ladakh, the Cockroach Janta Party movement of 2026.
6. **The counter-case and the verdict.** The strongest argument that this is ordinary democratic protest in a young, unequal country; what the evidence settles and what it does not; what the state can do without eroding the rights that make India credible; what it means for investors.

Published as Research briefs on the site, one or two a week, under a series title; then a compiled PDF with a methodology and data appendix; then a 7,000 to 8,000 word journal version (Strategic Analysis, India Quarterly).

## Part 1: the dataset

ACLED's India coverage starts in January 2016. To reach back to 2000 the dataset is built in three overlapping layers, so that a real rise shows in all of them and a perception rise shows only in the news-based one.

### Layer 1: official annual counts, 2000 to 2024
Consistent over time and conservative: they count what the state recorded.

| Series | Source | Years | File |
|---|---|---|---|
| Rioting cases, all-India and by state | NCRB Crime in India (via reclaimchennai/NCRB, every table as CSV) | 1953 to 2024 | `data/clean/ncrb_riots_*.csv` |
| Riot sub-types: communal, political, agrarian, student, industrial, caste | NCRB Crime in India | 2014 to 2024 | same |
| Internet shutdowns | SFLC.in tracker (via DW's verified 2012 to March 2020 extract, with tags for protest, state violence, communal), SFLC annual reports 2020 to 2024, Access Now 2025 | 2012 to 2025 | `data/clean/shutdowns_*.csv` |
| Strikes and lockouts, workers involved, man-days lost | Labour Bureau | 2000 to latest | to collect |
| Communal incidents | MHA figures in Lok Sabha and Rajya Sabha replies | 2004 to latest | to collect |
| Election calendar | ECI, via TCPD Lok Dhaba and datameet | 2000 to 2026 | `data/clean/elections.csv` |

Caveats: NCRB counts registered cases, not protests; many protests never produce a riot FIR; from 2014 NCRB changed its tables, and from 2024 the IPC was replaced by the BNS, so series are spliced and marked.

### Layer 2: machine-coded news events, 1995 to 2026
Event-level counts for the long window, at the price of counting news reports rather than verified events. Normalised by total India news volume (GDELT) to correct for the growth of media and phones.

- ICEWS (Integrated Crisis Early Warning System), Harvard Dataverse, 1995 to 2023.
- GDELT, protest events (CAMEO 14) for India, 2015 onward at full resolution, coarser back to 1979.
- Mass Mobilization Project (Clark and Regan), Harvard Dataverse, 1990 to March 2020, hand-coded anti-government protests with crowd sizes.

### Layer 3: ACLED, 2016 to 2026
Hand-coded, with actor, type and location. The gold standard and the calibration key: in the overlap years we compute the ratio of ACLED events to Layer 2 events and use it to back-estimate 2000 to 2015, with a stated margin of error.

### The curated catalogue
About 150 to 200 major protests, 2000 to 2026, one row each: dates, state, type, trigger, organisers, police, organiser and press crowd estimates kept separately, violence, deaths, internet shutdown, state response, outcome, whether an election fell within six months, and for 2023 onward any official or press claim of foreign funding or amplification with its rebuttal. Sources on every row. `data/clean/catalogue.csv`.

### What is blocked from this container
The build container's network policy allows GitHub and PyPI only. These hosts are denied and need either an allow-list entry on the environment or a manual download placed in `data/raw/`:

- acleddata.com (ACLED export; account registered 2 October 2026 under editor@)
- data.humdata.org (ACLED India mirror)
- dataverse.harvard.edu (ICEWS, Mass Mobilization)
- api.gdeltproject.org and data.gdeltproject.org (GDELT)
- internetshutdowns.in and sflc.in (current shutdown tracker)
- ncrb.gov.in, mospi.gov.in, labourbureau.gov.in, eci.gov.in (official sources; the NCRB tables are mirrored on GitHub so that one is not needed for Part 1)

## Outputs of Part 1
- Clean CSVs in `data/clean/`, each with a `sources` column or a sidecar note.
- Five or six charts in `charts/`: the long trend, the breakdown by type, the state map, protests against inflation and unemployment, protests against elections, reported protests against news volume.
- A methodology note (`notes/METHOD.md`).
- A 1,500 to 2,000 word Research brief: "Are Indians protesting more? What the data shows." Draft in `content/` on this branch; reviewed before merge.

## Log
- 2 Oct 2026: folder created. NCRB table archive, SFLC/DW shutdown data, Lok Dhaba and datameet election data cloned from GitHub. ACLED, HDX, Dataverse and GDELT blocked by network policy. Catalogue and published-counts searches running.
- 2 Oct 2026, later: NCRB series 1953-2024 with motives and states; shutdowns 2012-2025 (two trackers); MHA communal incidents 2004-2017; Labour Bureau disputes 2000-2023; election calendar (2025-26 verified, older rows from memory, flagged); catalogue at 166 rows, all sourced, 2000-2013 still being completed; nine charts; Part 1 draft in content/are-indians-protesting-more.md on the branch, not merged. Finding so far: official confrontation series all fall; major-protest rate roughly doubles from the late 2010s to 2024-26, led by jobs and exams. ACLED layer still blocked.
- 2 Oct 2026, evening: ACLED weekly export received from the editor; Layer 3 built (yearly, monthly, state); three charts; Part 1 draft completed with the verdict section. Remaining for Part 1: editor's review; spot check of pre-2025 election dates; 2014 rows in the catalogue.