Analytics Methodology & Data Integrity
Mathematical definitions, Form 20 reconciliation protocols, clustering confidence tiers, and ecological fallacy guardrails for Polldata.
Evidence-First Analytics Standards
Public-interest election statistics designed to answer analytical questions without speculation.
Polldata presents official polling-station data sourced from published Election Commission of India (ECI) Form 20 Final Result Sheets. All analytical indicators, distribution curves, and locality groupings are computed deterministically using standard mathematical formulations.
Core Methodological Principles
1. Evidence Before Interpretation
Every narrative insight or ranking is tied directly to underlying Form 20 counts. No speculative scores, persuasion models, or machine-learning voter estimates are used.
2. Context Before Ranking
No booth or locality is presented as "strongest" or "worst" without its total denominator (N booths) and the constituency baseline median.
3. Exact vs. Estimated Separation
Form 20 vote counts are exact aggregate counts. Geographic boundary approximations and locality clusters are labeled with explicit confidence tiers.
4. Non-Attribution of Demographics
Electoral Roll demographic data (age bands, gender) are kept strictly separate from voting returns. Ballot secrecy is absolute under Indian law.
Formula Reference
Margin = Share(Winner) - Share(RunnerUp)Calculated in percentage points (pp) rather than raw percentages to ensure comparability across booths with varying turnout.
Percentile = (Count(Booths ≤ BoothShare) / N) × 100Measures where a candidate's vote share falls within the constituency's distribution. A 90th percentile result indicates the party outperformed 90% of polling stations in that AC.
Median(Margins(Booth 1 .. Booth N))The 50th percentile of victory margins across all stations in an AC. Preferred over the arithmetic mean because it is resilient to extreme blowout outlier booths.
Count(Booths where Margin ≤ 5.0 pp)Identifies localized swing zones where a minor shift of fewer than 25 to 50 votes would have altered the station outcome.
The ecological fallacy occurs when inferences about the behavior of individual voters are deduced solely from aggregate statistics of the polling station.
Form 20 Provenance & Verification
Form 20 is the statutory return prescribed by Rule 56C(2) of the Conduct of Elections Rules, 1961, recording votes cast at each polling station in an Assembly Constituency.
Figures are presented for public information and may change as official records are corrected or updated.