So, I’m pretty sure we’re all aware of Burlington 2009, but I’m wondering if any of us knows of a real-world IRV failure where the more conservative candidate won because a more centrist candidate was eliminated in an earlier round. I’m trying to present the problems with IRV to Democrats, and am concerned that they might be less upset by that result because their side won…
The problem is that IRV is likely going to only get used in more left places where conservatives are going to usually lose…
Try Australia. See if you can find some data on their house elections. They have conservatives win frequently enough.
I think also Ireland uses IRV for something.
Ireland uses 3-5 seat STV, which is far less likely to throw that type of error than single-seat STV.
And while Australia is the likeliest candidate, there are hundreds of ridings, with scores of elections that I would have to sift through to find a Coalition Win/Condorcet Failure example, requiring a lot of time or full ballots with standardized data format, and pulling an example where the Green + Labor vote (or Labor + Lang-Labor) outnumbered the Coalition vote, yet Coalition still won…
I was hoping someone would have known of one off the top of their head
I think the best example in this regard is Egypt 2012. I realize that that was a 2-round runoff, not IRV; but it is a case of center-squeeze that arguably would have worked the same with IRV. The three factions were Islamists, reformers, and old-guard; the reformers were Condorcet winners according to pre-election polls (imperfect, but the best we have); and yet the only two candidates in the runoff were an Islamist and an old-guard. Of course Morsi, the Islamist, won, with disastrous consequences for democracy in Egypt.
The great thing about that example is that pretty much any group in the US can agree that it’s a bad outcome. Whether you’re right-wing, centrist, or left-wing in US terms, you probably have little sympathy for the Muslim Brotherhood, and no joy at seeing Egyptian democracy end with a coup.
Well, the 1952 and 1953 Parliamentary elections in British Columbia (Results here) put right-wing Social Credit into power, and the results look ridden with pathologies. 1952 isn’t a great example, though, because the transfers were more along the lines of establishment vs anti-establishment than left-right. When the far-left got eliminated, their votes often went to the far-right (and vice-versa). You also don’t have complete rank orders.
Indeed, and I may use it for the presentation to the Republicans, but my current example of Top Two failure is WA’s own 2016 State Treasurer’s race, where the Democrats split their 52% 3 ways, and the Republicans split their 48% only two ways, and we had 2 Republicans in the general.
Unfortunately, the fact that the #4 candidate and #5 candidate both covered the spread between First Place third Third Place (several times over, in fact), it isn’t any good for a “this really happened” example.
How would you rate the parties of that day on a left/right axis? Because I don’t really need a full breakdown if I can show the First and Last counts…
CCF was a socialist party (today it’s the NDP). The Liberals and PCs were center-left and center-right respectively. Before the 1952 elections, they formed a ‘free-market’ coalition to prevent the possibility of the socialists getting power on vote splits. (Actually, the reason that IRV was enacted in the first place was that they had originally agreed not to compete against each other, but tensions in the coalition were making this agreement untenable. But they weren’t willing to risk putting CCF into power, hence IRV.) Social Credit ran as a populist right-wing party. 1952 is tricky, though, since many of the voters were anti-establishment and ranked both CCF and SC over the coalition. In 1953, after Social Credit actually got power, this stopped being the case. When either CCF or SC got eliminated from a riding, most of the support went to the Liberals (the PCs had lost most of their support to SC, and so did not last long enough to receive transfers from CCF, SC or the Liberals.)
Here is a good article on the subject:
That article notes on the 1953 elections: “What is remarkable is that the Liberals won so few seats considering that every other major party offered them second-choice votes. They simply started with too few first-place votes to capitalize on their second ballot strength.”
So yeah, the 1953 elections, which gave Social Credit a majority, are probably what you are looking for.
There you go. That’s perfect. Rossland Trail, 1953:
Clearly at least some of the PC supporters preferred SC to Liberals, but because there weren’t enough of them, it went to Social Credit. It’s exactly like the “Favorite Betrayal in IRV” video, except with an extra round.
57.02 would have almost certainly preferred Labor to PC/SC, but because not enough people picked Liberal high enough…
Thank you, Marylander!
If you can’t find the list with every count (not just first & last) for each riding, it’s there, they just put it after the tables with only the first and last counts.
So it is! Thank you, that’s exactly what I was looking for.
Also as an interesting example of Center Squeeze: IRV was initially implemented by the center-based Liberal/Progressive Conservative coalition, who held 39 of the 48 seats following the 1949 election. They were worried about the rise of the CCF.
And under the first IRV election, the CCF got double their previous average, (mean: 9, median: 7, 1952: 18), the centrist coalition lost nearly 75% of their seats (10, from 39), and the right-most party (of those that ran in more than half the constituencies) went from 0 seats ever to being the primary party with 19 of the 48 seats.
The formerly dominant centrist coalition didn’t win double digit seats for the following third of a century (10 in 1952, 17 in 1991)
Well, they ditched IRV after the 1953 elections because it went terribly (see pg 109), not just in the ‘IRV gets poor results’ sense but because voters were confused, and a lot of them preferred to vote in the old style (i.e. bullet vote.)
But the BC results do suggest that Burlington was far from a freak occurrence, and would instead be a regular outcome in elections with more than 2 competitive candidates. Esquimalt, Nelson-Creston, and Saanich all look problematic, and probably would have been pathological were it not for the high bullet rates caused by voters (and candidates) not understanding the system. Also, in Prince Rupert, CCF might have won if some of the CCFers voted SC instead. In 48 elections, the number of problematic ones seems to be in the high single digits.
Start from first principles. First past the post creates a voice choice dilemma. The choice of 3+ options makes the voice of single support inadaqaute there are 2 options:
Reduce the number of choices : primaries, multiple elections (this is IRV)
Augment voice: Approval voting… Or the other gobbly goop like score and whatever.
IRV tries to sell as increased voice but it’s actually just the same voice pick 1 but now we hold a series of shitty elections instead of just 2 shitty elections. At its core instant runoff elections are about systematically reducing choice till its moot. It’s choice reduction not voice augmentation. The problem is the system for reducing choice is terrible. It lacks monotonacity. It isn’t immune to irrelevant alternatives and it’s tedious.
More voice > less choice.