The published 2022 sample contains 37 reconciled contests. Nineteen have margins below five percentage points of candidate votes; this is a sample description, not a statewide estimate.
Margin divided by candidate votes; excludes NOTA and rejected ballots. Descriptive bins, not future seat ratings.
Define the denominator before the headline
A vote margin is the winner’s votes minus the runner-up’s votes. The margin percentage used here divides that difference by the sum of all candidate votes in the constituency. NOTA and rejected ballots are excluded. A margin of five percentage points on this basis is neither five percent of registered electors nor a turnout figure.
The site publishes numerical competition metrics only where the candidate list and totals reconcile. Winner-register coverage is complete for all 60 constituencies in each Assembly election, but full candidate totals currently cover one contest in 2012, 58 in 2017 and 37 in 2022. The different coverages must not be mistaken for equally complete statewide samples.
The winner and the shape of the field
In the published 2022 sample, 22 of 37 winners did not exceed half of candidate votes. A plurality election permits this. It does not establish that all other voters preferred a common alternative, or that an alliance of losing candidates would have combined their votes.
Top-two concentration measures how much of the candidate vote went to the two leading candidates. The effective-candidate count is one divided by the sum of squared candidate shares. Four equally supported candidates produce a value of four; a dominant candidate and several tiny candidacies produce a much smaller value than the raw field size.
Read the closest observations precisely
Lamlai’s reconciled 2022 contest had a margin of 121 votes, about 0.42 percentage points of candidate votes. Keishamthong’s margin was 187 votes, about 0.74 percentage points. These are the closest observations within the current published sample, not certified rankings across all sixty constituencies.
The election lab can sort by margin or effective-candidate count. Every row links to the constituency report, where the full candidate table and source locator can be checked. Sorting by a measure changes the analytical question: closeness and fragmentation describe different aspects of competition.
Compare the same places across time
Comparing a median from 58 constituencies in 2017 with a median from 37 in 2022 changes both year and sample composition. The matched-sample section first intersects the constituencies available in both selected years, then compares those same constituencies.
Even this comparison remains descriptive. A changing margin can reflect a different winning party, a different runner-up, candidate withdrawal, alliances, participation or a broader political change. A smaller historical margin is not a calibrated probability of a future change in representation.
A reproducible reading protocol
Begin with coverage, inspect the candidate list, identify the denominator, compare matched observations, and read the constituency’s party sequence alongside its vote shares. Downloaded CSV rows contain the same measures shown in the table.
Two 2017 contests remain withheld because of unresolved accounting discrepancies. Withholding a numerical comparison while retaining a separately verified winner is more informative than silently treating an unreconciled total as certain. The source register distinguishes machine reconciliation from visual review.
PRIMARY EVIDENCE & REPRODUCIBILITY
Read the sources.
- CEO Manipur · Assembly election reports ↗
2012, 2017 and 2022 winner registers; constituency Form 20 tables
- Published data and verification log ↗
winners.json and results.json; verification metadata
Calculations are descriptive. Sources retain their stated reference periods. Related internal pages link to underlying documents and review notes. Corrections and missing fields remain visible in the source register.