Skip to content

Where the two reports differ — abstentions vs Equal Support

One line: BetterVoting and the LH engine agree on the winner and the runoff margin on every election. They can differ in one bookkeeping detail — how a "no preference" ballot is classified — which shifts the abstention count, the tally total, and the score totals, but never the result.

→ Hub: STAR Reporting · canonical write-up: BetterVoting and the LH engine — when the reports differ.

The difference, at a glance (the 461-ballot pet race)

BetterVoting LH engine (all 461)
Ballots tallied 455 461
"Abstentions" 6 (every flat ballot) 1 (the blank only)
Equal Support in runoff 92 98
Per-candidate score totals 9 lower (Dog 1798) Dog 1807
Runoff Dog / Cat 190 / 173 190 / 173
Winner Dog Dog

The whole gap is one rule: BetterVoting files a flat ballot (every candidate scored the same) as an abstention and drops it; the LH engine counts it as a cast ballot and, in the runoff, as Equal Support. 92 + 6 = 98; the dropped flat ballots included an all-5s and an all-4s vote, whose 5 + 4 = 9 stars are exactly the per-candidate score-total gap.

Why "no preference" ≠ "abstention"

A ballot that scores everyone the same is a cast vote with no preference between the finalists — that's Equal Support, which is correctly neutral in the runoff denominator but still counts in the score round. An abstention is a ballot with no score recorded at all (a blank). Folding the first into the second mislabels engaged voters and makes published score totals miss a hand count. (Terms: GLOSSARY · the denominator: Runoff percentages.)

Two views of the same election

The cleanest proof is the 3-candidate case shown both ways:

  • BetterVoting view — election ID dq2dmm (reports nAbstentions: 3).
  • LH view — the same BV JSON tabulated by LH: flat_scores_abstention_c3_b8_tabulated.txt (1 abstention; the flat 3,3,3 counted as Equal Support).

Same numbers, different words

The two reports never disagree on the runoff numbers — they just name and format them differently, because one is a visual table and the other is a text audit report.

BetterVoting (Race Details column) LH engine (the runoff line) What it is
Runoff Votes the finalist counts — Boston 2, Austin 1 (also the Automatic Runoff block) each finalist's head-to-head vote total
% Between Finalists the (67%) / (33%) in the line share of the decided voters (Equal Support excluded) — the number that decides the race
% Runoff Votes not a separate line; the _tabulated "Runoff math" funnel carries the all-ballots total share of all ballots (Equal Support left in the denominator)
(denominator: a column header) Voters with a preference: N of TOTAL (E Equal Support) LH states the denominator inline instead of in a header

Why the words differ. BetterVoting is a visual, voter-facing display: it can afford several columns and let you read whichever percentage you want, with the denominator implied by the header. The LH engine is a text audit report: it can't assume you've read a column header, so it folds the decisive count and percentage into one self-reconciling line and names the denominator (Voters with a preference) — the one thing BetterVoting's two percent columns leave you to infer.

Why there are two percentages at all. When some voters rate the two finalists equally (Equal Support), "what fraction did the winner get?" has two honest answers — out of everyone, or out of only the voters who picked between the two finalists. BetterVoting prints both; the LH line prints the decisive one (decided voters) and names it, then expands the full funnel in _tabulated. With no Equal Support the two percentages are identical; with Equal Support they diverge (e.g. the flat-scores 4 of 8 case). So: LH "Boston 2 (67%)" = BV Runoff Votes (2) + BV % Between Finalists (67%), in one line that says what it's dividing by.

Evidence & the filed issue