Methodology

A reaction is a public, attributable statement by a party organ — a spokesperson, a state unit, or the government itself — that takes a position on a specific issue. Each recorded reaction carries four things: the stance classification, a direct quote, a link to the source, and the time it was made.

Statements are classified with a fixed, tier-specific vocabulary. The government is classified by what it did (action taken, announced, acknowledged, deflected); ruling alliance partners by their distance from the government (backed govt, distanced, dissented); opposition parties by their position on the issue (supported, opposed, neutral, deflected). The vocabularies are never mixed, because the fair question differs by role: the government is asked what did you do, an ally is asked did you defend this, and the opposition is asked where do you stand.

Silence is never entered by hand. Each issue lists the parties scoped on it — the parties from which a response is reasonably expected. A scoped party with no recorded reaction is silent, and its silence duration is the time since the issue was published. When a reaction is recorded, the silence counter for that party ends permanently.

Silence is flagged — shown in amber across the site — once it crosses the tier benchmark below. A late response is still recorded, with its reaction time visible; the flag applies only while the silence is ongoing.

The government carries one extra metric: the statement-to-action gap. When its stance on an issue is announced and no follow-up action taken is recorded within 14 days of the announcement, the issue is flagged “Announced, not delivered” with a running day count. The flag clears only when action is recorded.

No stance appears on the record without a quote and a source link. Quotes are attributed to party organs (“Party spokesperson”, “State unit statement”, “Government statement”), not to individuals. Where a party has made multiple statements, the latest is shown as its current stance and earlier ones remain part of the record.

Corrections: a wrongly classified stance is corrected in place; silence computed from a missed statement is corrected by recording the statement with its original timestamp, which retroactively ends the silence.

This build runs on demonstration data. Issues are fictional but plausible; every quote is an illustrative sample marked “(sample)”; source links point to placeholders. The methodology above is exactly what the production site will apply to real statements.