RESEARCH INTEGRITY / UPDATED 20 OCTOBER 2026
Know what the evidence covers.
A public account of completeness, source tiers and unresolved discrepancies. A missing value stays missing; a conflicting source stays visible.
| Dataset | Coverage | What was checked | Remaining limit |
|---|---|---|---|
| Assembly 2012 | 60/60 contests · 279 candidacies | Source totals, ballot accounting, independent winner party | Selected pages visually reviewed; not an all-page manual audit |
| Assembly 2017 | 59/60 contests · 260 candidacies | Source totals, ballot accounting, independent winner party | AC 21 full contest withheld |
| Assembly 2022 | 60/60 contests · 265 candidacies | Source totals, ballot accounting, independent winner party | Selected pages visually reviewed; not an all-page manual audit |
| Parliament 2024 | 2 constituency results · 60 ordinary segments | Candidate sums, ordinary-segment ballot totals, constituency residuals | Two special-station groups reviewed separately; remaining residual not labelled as verified postal votes |
| Disclosures 2022 | 265/265 official candidacies matched | ADR/MyNeta summary overlap, identity within election, explicit Nil values | Secondary transcription; all underlying affidavits not manually reviewed |
| Identity 2024 | 10 candidates indexed; 2 community fields and 1 religion field sourced | Published self-descriptions with attribution and date | Other fields explicitly unverified; no demographic inference |
| 2027 assessment | Historical evidence notes for all 60 seats | 2022 Assembly field and 2024 ordinary-segment lead | No complete current nominations, local opinion evidence or calibrated forecast |
The discrepancy ledger.
2017 · Naoriya Pakhanglakpa
The native Form 20 prints 30,181 valid candidate votes, one rejected ballot and 153 NOTA votes, but a total of 30,334. Those components sum to 30,335. The full numerical contest stays withheld pending clarification. Its winner is independently indexed.
Inspect the original result sheet ↗2012 · unusual party codes
AC 11 prints IVD for one candidate and AC 56 prints JMBP. Original pages were visually checked. These source codes remain literal; the platform does not silently convert them to another party or invent a party name.
2022 · declared case-count difference
ADR's candidate summaries identify 53 candidates declaring cases; the official compiled statistical summary reports 56. Scope or transcription differences remain unresolved. These counts are separately labelled, and neither establishes guilt or current case status.
2024 · votes outside ordinary segments
Constituency totals exceed the sum of ordinary geographic segments. The difference is published candidate by candidate, with two visually reviewed special-station groups separately identified. The remainder is not assigned to local segments or labelled wholly as postal votes.
Inspect the reconciliation ↗Inspect, reuse, challenge.
Source locators and SHA-256 hashes accompany new native Form 20 rows. Arithmetic validation and selected-page visual inspection are separate checks. A passing automated build does not certify every source page or settle a source disagreement.
Assembly results · CSV ↗2024 segments and residuals · JSON ↗All 60 local 2024 leads and margins · CSV ↗2022 disclosure summaries · JSON ↗240 archived official Form 20 PDFs ↗Archived native extracts and 42 ADR summary pages ↗Native-source audit and aliases ↗
↗ Archive stored in GitHub Releases; all source, research data, builds and tests remain in GitHub. Original publications retain their rights.