YOU’RE DOING RANKED CHOICE WRONG
( To a short summary of my proposed Condorcet election: https://americarepair.home.blog/2020/12/13/brief-summary-of-practical-condorcet-election/ )
(Pros can skip to near the bottom of this page for a link to the full method, and another link with charts and examples.)
Here is an example of a ranked choice election that I say fails.
Parallel Universe 2016 IRV election:
Romney 37% 1st-choice votes
Biden 31% – eliminated first
The procedure of “Instant Runoff Voting” uses only 1st-choice votes to eliminate the first candidate.
Biden has 45% 2nd-choice votes from Romney and Sanders voters. He would beat either opponent in the top two, which makes him the Condorcet winner. IRV deems the support of 76% of the voters irrelevant, so Biden can’t win.
But a ranking election doesn’t have to shut out the most popular candidate.
ANALYZING THE ISSUE
Condorcet winner: a candidate who wins every possible two-candidate comparison in a ranking election.
IRV (instant runoff voting): a ranked choice system that will sometimes kick out a Condorcet winner in 3rd, 4th, or 5th place.
This quirk of IRV defeats the purpose of ranking, which is to elect the candidate who is preferred over the others. It happens because IRV works like a combination of two imprecise sports competitions.
The first is a pro wrestling battle royal, where a group of contenders all fight at once. In this mess, a gang of enemies can throw out the most talented wrestler early. Serious tournaments use focused, one-on-one contests.
IRV is also similar to a single elimination tournament, where if the frontrunner is unlucky one time, they’re done. A more complex tournament, such as double elimination, would be more thorough, more insurance against dumb luck. A round robin will pit each contestant against each of the others, providing even more data that can be used to logically determine a winner.
IRV uses an unreliable process to narrow a field of candidates. By ignoring part of the information provided by voters, IRV can put the wrong candidates into the final.
THERE IS A BETTER WAY TO EVALUATE RANKINGS
Runoff: an election in which voters choose between the top two candidates from the primary.
Instant runoff (the event, not the IRV system named after it): a virtual runoff, based on rankings, to determine which of TWO candidates more voters prefer.
A real instant runoff happens only in the final round of IRV.
So let’s build on that. Let’s use “instant runoffs” for all contenders, as if they’re the final two, a “Condorcet method.” It’s also referred to as “pairwise,” or “ranked pairs.” But I prefer “matches,” or to be specific:
(Since “pair” in English usually implies a harmonious relationship, and “ranked pairs” sounds like we rank two candidates together, it’s confusing to use “pairs.” And “pairwise” is weird.)
YES WAY, CONDORCET
(The t is silent, so it sorta rhymes with Jose.)
The winner of a head-to-head match will be the candidate preferred by more voters, based on rankings. Condorcet methods compare the win-loss records of the candidates, to see which one was preferred by more voters, more times.
I propose simply counting match wins, and the candidate with the most wins will be the election winner.
WE WON’T LET CLOSE ELECTIONS STOP US
A quirk of Condorcet systems is that ties for top candidate will happen, as in, two candidates having the same number of wins. This is the easy kind of tie. The easy tiebreaker is that the match involving these two candidates should tell us which one beats the other.
A more difficult tie is when there are three or more candidates, tied for most wins, who have beaten each other (called a cycle, or Condorcet’s paradox). Ranked choice supporters might enjoy using IRV as a tiebreaker, to eliminate one. IRV is better for tiebreaking than it is for narrowing down a large field.
For a persistent tie, yet another tiebreaker should be used. This one really isn’t important enough to argue about. It probably won’t be needed, and the winner will be one of the very best candidates. Keep it simple, use first-choice votes.
I would add one more thing, to simplify, and to soothe Condorcet’s critics. Early in the evaluation process, use first-choice votes to eliminate about half of the candidates, but keeping at least five. This keeps the number of head-to-head matches manageable, while ensuring the election winner will have a significant number of first-choice votes.
So the general procedure of evaluating this ranking election will be:
1. Check for a Condorcet winner. (Winning against all other candidates wins the election.) (Includes majority winner of first-choice votes.)
2. First-choice votes narrow the field.
3. Head-to-head matches determine which candidate(s) has the most wins.
This head-to-head Condorcet system will prevent the early elimination of the most-preferred candidate. We’ll still be ranking candidates, and we’ll still have “instant runoffs,” but the results will be more precise, and more fair, than pure IRV.
A link to the details of my proposed ranking election procedure: https://americarepair.home.blog/practical-condorcet-method/
And a page with charts to aid in evaluating the election, and some detailed examples of IRV going goofy: https://americarepair.home.blog/condorcet-method-tips-for-evaluating-the-election/
[Note: A fellow named Tideman has crafted some well-known ranked pairs methods. One method, based on similar concepts to the one above, is called Tideman Condorcet-Hare, or Condorcet-IRV. It should be acceptable to the public, and the software for determining a winner is supposedly available online. I would give you a link to the official rules, but it seems we might have to pay for the copyrighted material, or visit a real-world library.]