Maine election results


Can someone break this down into something more digestible, such as pairwise match ups, a table showing the number of voters that voted under each possible vote permutation, some analysis on strategic voting, who should of won, etc?


They should display it using graphs like this: 2017 Minneapolis Election Results: Mayor
Ranked-Choice Voting Tabulation Summary
(Apparently a form of Sankey diagram)


Those diagrams omit important information. You can’t check to see if an actual majority winner (aka Condorcet winner) won, nor can you check for what forms of strategic voting would have been effective.


I processed the data into my usual url format and ran my software on it.
Condorcet and IRV agree, so this result should be relatively uncontroversial.

(data and Python code at )


I would of assumed that one of the independent candidates would have been the Condorcet winner. Nether the Dem or Rep were ranked 1st by more then 50% of voters and it seems like the Dem voters would be likely to rank the independent candidates above the Rep and vise versa for the Rep voters.

Judging from the number of first ranks many candidates got in comparison to the 2nd ranks, it looks a lot of voters only ranked one candidate. Thus I speculate that the reason why the Dem was the Condorcet winner among the pairwise match-ups was because not enough voters ranked the independent candidates.

I wonder if you can approximate who the Condorcet winner would of been had more voters ranked more candidates. Maine allows voters to rank up to 5 candidates, so for every ballot that only ranked 4 candidates, determine among the portion of ballots that ranked the same 4 candidates in the same order while still ranking a 5th candidate, what % of them ranked each 5th candidate. Then make the ballot that only ranked the 4 candidates give a fraction of a rank to each possible 5th candidate they could of ranked where those fractions are proportional to the proportions found earlier among ballots with the same first 4 ranks.

Ex: if among the A>B>C>D>X ballots, if for 60% of these ballots X was E and for the other 40% X was F, then for every ballot that just ranks A>B>C>D, switch it out with 60% of an A>B>C>D>E ballot and 40% of an A>B>C>D>F ballot.

Now repeat this process among the ballots that only ranked 3 candidates, and then among the ballots that just ranked 2 candidates, and finally for ballots that only ranked 1 candidate.

After that, re-compute the pairwise match-ups.


Well, here is a list of all the ballot combinations cast. It’s not especially readable, though.

(): 6034, (Ballots that didn’t count for any round)

(‘Poliquin’,): 87940,
(‘Poliquin’, ‘Golden’): 3242,
(‘Poliquin’, ‘Golden’, ‘Bond’): 558,
(‘Poliquin’, ‘Golden’, ‘Bond’, ‘Hoar’): 3406,
(‘Poliquin’, ‘Golden’, ‘Hoar’): 222,
(‘Poliquin’, ‘Golden’, ‘Hoar’, ‘Bond’): 2181,
(‘Poliquin’, ‘Bond’): 2733,
(‘Poliquin’, ‘Bond’, ‘Golden’): 600,
(‘Poliquin’, ‘Bond’, ‘Golden’, ‘Hoar’): 1731,
(‘Poliquin’, ‘Bond’, ‘Hoar’): 1950,
(‘Poliquin’, ‘Bond’, ‘Hoar’, ‘Golden’): 8051,
(‘Poliquin’, ‘Hoar’): 2759,
(‘Poliquin’, ‘Hoar’, ‘Golden’): 337,
(‘Poliquin’, ‘Hoar’, ‘Golden’, ‘Bond’): 1954,
(‘Poliquin’, ‘Hoar’, ‘Bond’): 2696,
(‘Poliquin’, ‘Hoar’, ‘Bond’, ‘Golden’): 11400,

(‘Golden’,): 50567,
(‘Golden’, ‘Poliquin’): 3433,
(‘Golden’, ‘Poliquin’, ‘Bond’): 524,
(‘Golden’, ‘Poliquin’, ‘Bond’, ‘Hoar’): 2721,
(‘Golden’, ‘Poliquin’, ‘Hoar’): 147,
(‘Golden’, ‘Poliquin’, ‘Hoar’, ‘Bond’): 1718,
(‘Golden’, ‘Bond’): 10302,
(‘Golden’, ‘Bond’, ‘Poliquin’): 719,
(‘Golden’, ‘Bond’, ‘Poliquin’, ‘Hoar’): 1259,
(‘Golden’, ‘Bond’, ‘Hoar’): 17031,
(‘Golden’, ‘Bond’, ‘Hoar’, ‘Poliquin’): 22702,
(‘Golden’, ‘Hoar’): 3159,
(‘Golden’, ‘Hoar’, ‘Poliquin’): 509,
(‘Golden’, ‘Hoar’, ‘Poliquin’, ‘Bond’): 697,
(‘Golden’, ‘Hoar’, ‘Bond’): 5649,
(‘Golden’, ‘Hoar’, ‘Bond’, ‘Poliquin’): 7938,

(‘Bond’,): 4318,
(‘Bond’, ‘Poliquin’): 482,
(‘Bond’, ‘Poliquin’, ‘Golden’): 93,
(‘Bond’, ‘Poliquin’, ‘Golden’, ‘Hoar’): 338,
(‘Bond’, ‘Poliquin’, ‘Hoar’): 116
(‘Bond’, ‘Poliquin’, ‘Hoar’, ‘Golden’): 617,
(‘Bond’, ‘Golden’): 1101,
(‘Bond’, ‘Golden’, ‘Poliquin’): 170,
(‘Bond’, ‘Golden’, ‘Poliquin’, ‘Hoar’): 415,
(‘Bond’, ‘Golden’, ‘Hoar’): 924,
(‘Bond’, ‘Golden’, ‘Hoar’, ‘Poliquin’): 2148,
(‘Bond’, ‘Hoar’): 1110,
(‘Bond’, ‘Hoar’, ‘Poliquin’): 279,
(‘Bond’, ‘Hoar’, ‘Poliquin’, ‘Golden’): 1199,
(‘Bond’, ‘Hoar’, ‘Golden’): 633,
(‘Bond’, ‘Hoar’, ‘Golden’, ‘Poliquin’): 2346,

(‘Hoar’,): 2139,
(‘Hoar’, ‘Poliquin’): 253,
(‘Hoar’, ‘Poliquin’, ‘Golden’): 43,
(‘Hoar’, ‘Poliquin’, ‘Golden’, ‘Bond’): 197,
(‘Hoar’, ‘Poliquin’, ‘Bond’): 88,
(‘Hoar’, ‘Poliquin’, ‘Bond’, ‘Golden’): 303,
(‘Hoar’, ‘Golden’): 273,
(‘Hoar’, ‘Golden’, ‘Poliquin’): 56,
(‘Hoar’, ‘Golden’, ‘Poliquin’, ‘Bond’): 120,
(‘Hoar’, ‘Golden’, ‘Bond’): 233,
(‘Hoar’, ‘Golden’, ‘Bond’, ‘Poliquin’): 509,
(‘Hoar’, ‘Bond’): 499,
(‘Hoar’, ‘Bond’, ‘Poliquin’): 165,
(‘Hoar’, ‘Bond’, ‘Poliquin’, ‘Golden’): 573,
(‘Hoar’, ‘Bond’, ‘Golden’): 276,
(‘Hoar’, ‘Bond’, ‘Golden’, ‘Poliquin’): 1067,
Note: the pairwise table resulting from these rankings differs from @Brian_Olson 's by a few hundred votes, and the Golden-Poliquin matchup differs from Maine’s official final count by a similar magnitude. I think the main difference between my methodology and Olson’s was that my code ignored all rankings below an overvote, but his didn’t.

Golden Poliquin Bond Hoar
Golden X 139346* 142406m 146204M
Poliquin 136506 X 142569m 144512m
Bond 48421 85211 X 90576*
Hoar 42241 73118 48160 X

Total votes: 289952; Majority: 144977
Votes with at least one valid ranking: 283918; Majority: 141959
‘*’ denotes pairwise winner
m denotes pairwise majority on all ballots with at least one valid ranking
M denotes pairwise majority on all ballots
Both major party candidates received a pairwise majority in all pairwise matchups against the others among valid ballots, so I think Golden still beats everyone pairwise even without so many bullet votes. I especially doubt one of the 3rd party candidates would be the CW.


So what percentage of that is bullet votes?


About 51% of the valid votes were bullets.


And they criticize Score Voting for being susceptible to bullet votes. Wow.

EDIT 20181123: Of course the IRV propagandists will say that the voters here “did not need” to bullet vote and stuff. Which is fallacious in its own right, because there are elections (albeit somewhat contrived ones) where bullet voting is better strategy than an honest IRV vote. (Just take any IRV election with participation failure, and imagine that the involved voters instead favored another guy that everyone else hates or does not know.)

D is eliminated, then C, and A wins.
If the five DBCA’s bullet vote for D, then B is eliminated and C wins, which is better for those voters.

Of course, bullet voting will never help D win, but it can help C win instead of A, and those people really hate A.