Skip to content

RCV-IRV in North Carolina — a short-lived experiment (2006–2013)

North Carolina ran one of the earliest and boldest US trials of instant-runoff voting — including the first statewide general election ever decided by IRV — then repealed it. A real case of IRV adopted, tested, and reversed, and (unusually) one where the loudest objections were administrative and election-integrity concerns rather than the social-choice pathologies.

→ Broader context: origins & spread · the method: RCV-IRV (Hare) · another real case: Alaska 2022

Timeline

  • 2006 — enacted. The General Assembly authorized IRV for certain judicial vacancies and set up a pilot program (up to 10 cities in 2007, up to 10 counties in 2008) to try ranked ballots in real elections.
  • 9 Oct 2007 — Cary pilot. The Cary City Council District B race was decided by IRV — the first real IRV contest in the state.
  • 2 Nov 2010 — statewide Court of Appeals. A 13-candidate, nonpartisan, statewide race for the "Wynn seat" — believed to be the first time IRV decided any statewide general election in the United States.
  • 2013 — repealed. The pilot program was ended by the General Assembly, tucked into a large omnibus bill (H589, the "monster bill," best known for voter-ID). Special judicial elections with more than two candidates reverted to plurality.

Note the small footprint: only two NC contests ever actually reached an instant-runoff round — Cary 2007 (District B) and the 2010 statewide seat. Hendersonville also used IRV but its races were decided in the first round, so the later choices were never tallied.

The 2010 statewide election — the first-round leader lost

  • First round: Cressie Thigpen — the sitting judge, favored by Democrats — led with about 395,000 of ~2 million votes, roughly 100,000 ahead of Doug McCullough (favored by Republicans). With 13 candidates, Thigpen was far short of a majority.
  • After the IRV rounds: as low candidates were eliminated and their voters' next choices distributed, McCullough overtook Thigpen and won the third round 543,980 to 537,325 — a margin of about 0.62%. It took roughly 48 days and a recount to certify.

So the candidate who led the first count lost after the transfers — and the winner still held only a plurality of the ballots cast, not a majority of everyone who voted (the runoff produces a majority only among continuing ballots). Supporters called that the method working; critics called it an experimental tally overturning the front-runner and producing a drawn-out, near-tie result. Either way it's a clean example that the first-choice leader is not always the IRV winner — the same lesson as Alaska 2022, here without a clear center-squeeze.

The critics' case (in their own terms)

The organized opposition — the NC Coalition for Verified Voting (Joyce McCloy) and allied writers at ncvoter.net / irvbad4nc — focused less on social choice and more on how the ballots get counted under NC law and equipment. Their headline list, "Instant Runoff — what they don't tell you," argued IRV:

  1. hurts election integrity and raises cost and labor, making audits and recounts harder;
  2. doesn't give voters a real "second chance";
  3. leaves not all votes counted (2nd/3rd choices are never tabulated unless a runoff round is reached);
  4. demands an unusually informed electorate;
  5. can let a voter hurt their own candidate by ranking them (non-monotonicity);
  6. can't be tabulated by NC's certified voting machines;
  7. disadvantages third parties;
  8. complicates campaigning.

The machine-certification problem was the core NC-specific objection. NC's optical-scan system (ES&S Unity) had no certified capability to tabulate IRV, and there were (and are) no federally certified voting systems compatible with IRV. For the touchscreen county (Henderson), the State Board set up an uncertified "workaround" that exported vote data out of the ES&S system into Notepad/Excel to compute the rounds — a process critics said had no audit trail (audit data erased as steps progressed), over 100 manual steps where a single keystroke could flip the result, and no federal/state certification, in tension with NC's own equipment-certification law (NC GS 163-165.7). Officials voiced the same worry: State Board of Elections chair Larry Leake said he was "leery" of the software count, and BoE voting-systems manager Keith Long wrote that "there are no provisions on ES&S equipment to tabulate IRV."

The Cary 2007 count is their exhibit A. Because the machines couldn't do it, workers hand-sorted and stacked paper ballots to count the second round of District B, the election-night counts didn't match the candidates' informal tally, and a non-public audit led to a recount and a "correction" of the results. Critics listed process objections rooted in NC statute: ballots were counted away from the precinct rather than immediately on-site (they cite NC GS 163-182.2), there was no election-night reporting of 2nd/3rd choices, incomplete raw data was released (so you can't even check for non-monotonicity), and provisional/absentee ballots had to be folded in after the transfer rounds even though IRV isn't additive. Notably, Don Frantz — the one person ever elected by IRV in NC — later opposed it, saying his top concern was election integrity: "If we can't trust the election process we've got, I don't care what it costs. I don't care what the turnout is."

What's solid, what's contested (staying fair)

  • Solid / on-point. IRV genuinely is not precinct-summable — who is eliminated depends on the whole electorate, so ballots must be centralized and a recount re-runs the rounds; that makes audits harder. And in 2007–2013 NC's certified equipment truly could not tabulate IRV, forcing hand counts or an uncertified workaround. Those were real, concrete barriers, not just rhetoric.
  • General IRV critiques the repo agrees with elsewhere. "Not a true majority" (it's a majority of continuing ballots), the monotonicity risk, and Duverger-style pressure on third parties are legitimate and covered in this repo's concept pages — though they apply to IRV everywhere, not just NC.
  • Advocacy framing to flag. Phrases like "not all votes are counted" describe a normal IRV feature (later choices aren't tallied unless a round needs them) in the most alarming terms; and cost claims (e.g. Minneapolis RCV ~$244,000) come from opponents' framing. IRV's proponents (FairVote) counter that the pilots ran smoothly where machines allowed, saved the cost of a separate runoff, and delivered majority-tested winners — and that most of the machine problems were fixable equipment gaps, not inherent flaws.

The lasting significance either way: NC remains the only US state to have decided a statewide general election by IRV, and its arc — adopt, pilot, scale up, repeal — is a frequently-cited data point in US ranked-voting debates.

Sources: Electoral reform in North Carolina — Wikipedia, 2010 North Carolina judicial elections — Wikipedia, The Short Life and Quiet Death of Ranked-Choice Voting in NC — The Assembly, NC Court of Appeals IRV Results — FairVote. Critics' case (reader-supplied): Instant-Runoff Voting regrets in NC — ncvoter.net and the NC Coalition for Verified Voting (Joyce McCloy) writings.

file: RCV_IRV_north_carolina.md