How the Record Is Made

Every figure on this site and in The Myanmar Conflict Monitor is coded under the MCARI Methodology and Codebook. The codebook is versioned independently of the reports that cite it. A change takes a new version, and each version states the first issue it binds. A report is never re-made against a later version.

MCARI's quarterly and annual reports are published in full, in Burmese and in English. The underlying dataset is not published in full. It is supplied on request, and a request should state the period, the geographic scope, and the purpose it is needed for.

Two fields are never supplied. The narrative note records what MCARI's local sources and correspondents established about an event; releasing it would expose the people who did that work. The coding record that travels with each row is withheld for the same reason.

Figures on this site come from a numbered data release, cut from MCARI's dataset as it is updated. Each release is exported once and kept with the SHA-256 checksum of the data it was cut from, so any figure on the site can be reproduced.

Issues of The Myanmar Conflict Monitor are coded from frozen editions. An edition holds the year it is named for and every earlier year in the series, and its checksum is recorded when it is cut.

MCARI's dataset is a living record. A row is corrected when better information reaches it, and corrections are not held back to preserve a published figure.

Reports are therefore cited against a frozen edition. A figure published in an issue remains reproducible from the edition that issue names, even after the live dataset has moved past it. This site shows the current data release, which includes every correction made since. Where the two differ, the issue's figure is correct for its edition, and the site's figure is correct for the current release.

MCARI records what reporting establishes. What follows is what the data therefore cannot show.

• The record is built from reporting. Where reporting is thin, the record is thin. A zero in MCARI's data means nothing was recorded. It does not mean nothing happened. A change in the record over time may reflect a change in reporting as much as a change in the conflict.
• The counts are minimums. Where a source gives no number, a fixed conventional figure is recorded; where it gives a quantity in words, the least those words allow. The real figure is higher than the recorded one more often than it is lower, and by an unknown margin. Nothing is estimated, interpolated or modelled.
• What the fields do not count. There is no field for the injured. Displacement counts recorded displacement events, not distinct people.
• Location. A coordinate is recorded only where the place a source names resolves to exactly one candidate inside the row's township. Township-level analysis excludes rows not located to a township, and the count of excluded rows is published with each edition.
• Actors. A perpetrator that reporting does not identify is recorded as unidentified. Names are standardised against MCARI's registry, so the string in the data is often not the wording the source used.
• The source field. The outlets that reported an event are named, but not classified by type. The share of events resting on a conflict party's own account therefore cannot be separated out from the data as supplied.
• The series is revised. → Why an issue's figures and this site's figures can differ

Corrections found after an issue is published are recorded here and carried into the next edition. Each entry names the issue or release, the date, and what changed.

No corrections to a published issue have been recorded.