4460 charts, 4380 live

Civilian deaths are the figure the state has least incentive to overstate and the one most sensitive to who is counted as a civilian.

Where an encounter is disputed, this count inherits the dispute. Read it as the state's tally, not as an agreed number.

Rebuilt by summing SATP's eighteen state sheets, because its national page refuses the pipeline. Checked against that page where it could be read: two sampled years matched exactly and two were short by under half a per cent, which is consistent with incidents SATP does not assign to a state. Treat it as within a percentage point of SATP's national figure, not identical to it.

Author-constructed. Every input is published, but the choice of dimensions and neutral points is a judgement — see the methodology page before using this for anything.

Rebuilt by summing SATP's eighteen state sheets, because its national page refuses the pipeline. Checked against that page where it could be read: two sampled years matched exactly and two were short by under half a per cent, which is consistent with incidents SATP does not assign to a state. Treat it as within a percentage point of SATP's national figure, not identical to it.

Author-constructed on the same basis as the left-wing extremism index, and carrying the same caveats.

SATP counts incidents of killing: incidents in which someone died, not all attacks. An attack with no fatality is not counted, so this understates activity and is not comparable with MHA incident totals.

Author-constructed on the same basis as the LWE score, and carrying the same caveats.

SATP's Punjab table begins in 2000, roughly a decade after the insurgency it belongs to had ended. The peak — the early and mid-1980s through 1993, with fatalities in the thousands each year — is on SATP's page only as a single 1981-2000 aggregate with no annual rows, so it is outside this series entirely. Read a flat line near zero as the aftermath of a conflict that was over, not as evidence there was never one.

SATP's Punjab table begins in 2000, roughly a decade after the insurgency it belongs to had ended. The peak — the early and mid-1980s through 1993, with fatalities in the thousands each year — is on SATP's page only as a single 1981-2000 aggregate with no annual rows, so it is outside this series entirely. Read a flat line near zero as the aftermath of a conflict that was over, not as evidence there was never one.

SATP's Punjab table begins in 2000, roughly a decade after the insurgency it belongs to had ended. The peak — the early and mid-1980s through 1993, with fatalities in the thousands each year — is on SATP's page only as a single 1981-2000 aggregate with no annual rows, so it is outside this series entirely. Read a flat line near zero as the aftermath of a conflict that was over, not as evidence there was never one. Identification is made at the scene and is the most contested figure in any SATP sheet; read it as 'recorded as', not 'were'. The direction of good is left unset here for the same reason it is unset on every other side-of-conflict count on this site.

SATP's Punjab table begins in 2000, roughly a decade after the insurgency it belongs to had ended. The peak — the early and mid-1980s through 1993, with fatalities in the thousands each year — is on SATP's page only as a single 1981-2000 aggregate with no annual rows, so it is outside this series entirely. Read a flat line near zero as the aftermath of a conflict that was over, not as evidence there was never one.

SATP's Punjab table begins in 2000, roughly a decade after the insurgency it belongs to had ended. The peak — the early and mid-1980s through 1993, with fatalities in the thousands each year — is on SATP's page only as a single 1981-2000 aggregate with no annual rows, so it is outside this series entirely. Read a flat line near zero as the aftermath of a conflict that was over, not as evidence there was never one. Incidents, not attacks: an attack in which nobody died does not appear, so this tracks lethality rather than the volume of violence.

Per data *user*, not per subscriber: people who own a phone but buy no data are outside the denominator, which lifts the average above what an ordinary reading of ‘per subscriber’ would give. The Survey publishes a much longer comparison elsewhere — 0.06 GB a month in 2014 against 25 GB in 2025 — but only as those two endpoints, so it is not spliced onto this series.

VAHAN excludes Telangana and Lakshadweep, which do not report to it, so the national figure is a slight undercount. Two- and three-wheelers dominate the count; reading it as a car number overstates the passenger-vehicle transition by an order of magnitude.

Passengers handled at domestic terminals. The sheet also lists April-to-October part-years for FY24 and FY25; both are excluded, and the FY24 part-year in particular would otherwise contradict the full FY24 row directly above it.

A recognition count, not a count of companies trading. Recognition is an application a founder makes for tax and procurement benefits, so the series measures uptake of a scheme as much as formation of firms, and it never falls because recognitions are not withdrawn when a startup closes.

Visits, not visitors: one person touring four states counts four times, and the total exceeds India's population in most years for that reason.

Read from the Economic Survey workbook tabchart6, sheet "Chart VI.4". These are the tables behind the Survey's charts, so the series is only as long as the chart needed.

Overtook foodgrains by volume in the mid-2010s, which is the single clearest sign of the shift out of cereal monoculture and is rarely the number quoted.