Instant runoff voting (here abbreviated IRV, and also known as ranked-choice voting, or RCV) offers what seems, at least, to be a plausible alternative to our current electoral system. It has become a popular cause, particularly in progressive circles, where people see election methods as reinforcing existing power structures. It comes as little surprise, then, to see the IRV debate resurface in Duluth this year: first, when the city council failed to use it properly in an attempt to fill a vacant council seat, and later, when it voted down a proposal to explore its implementation in local elections. I’m also about to head down and spend two years in Minneapolis, a city that has adopted it, so I’ll be curious to see what it looks like from the inside, and hear from people with direct experience. For now, though, I’ll lay out why I remain a skeptic. While I certainly don’t think the system we have is great, I don’t see any dramatic benefits from a change.

Thanks to fortuitous seating arrangements at the council discussion on IRV, I got myself a solid crash course in election algorithms and the critiques of IRV over the past month and a half. My guide throughout was UMD professor emerita Kathryn Lenz Peckham, who was joined by Prof. Barry James and instructor Rachel Breckenridge, all of the math and statistics department, at our final meeting. (Full disclosure: Prof. Lenz’s son was a high school friend of mine.) My teachers obviously had clear issues with IRV, but they’d been genuinely curious when they first heard about its application for political elections, and came at it with the detached, academic curiosity that I appreciate.

Most elections in the U.S. use a system known as plurality voting: each voter casts one vote and  the candidate with the most votes wins, sometimes with the help of a plurality primary in which the top two candidates advance to a general election run-off. IRV dispenses with the primary, asks voters to rank the candidates in order of preference, and conducts an instant runoff. If no candidate receives 50% of the vote, the lowest vote-getter is eliminated, and that candidate’s supporters’ second-place votes are then counted. This process continues until someone clears the 50% bar.

The most immediate critique of IRV is its confusion: it’s different, and requires a bit more thought than just casting a vote for one’s favorite candidate. IRV defenders, however, can mount a strong case in response, relying on good voter education. The evidence on voter turnout and understanding of the method is a bit mixed, though it isn’t completely damning, either. The statisticians’ greatest concerns were not really with the act of voting itself, but the manner in which those votes are counted.

One of their greatest concerns was the possibility of a tie. With so many rounds, IRV greatly enhances the possibility of a tie; there could be an election-determining statistical tie at each round.  A tie at any round could potentially be enough to trigger a recount or other such election intrigue as the order in which candidates are eliminated affects the way second- and third-place votes come into play, and could theoretically lead to a completely different result.

Transparency in reporting election results is also difficult. Since IRV elections often feature 10 or more candidates, it is practically impossible to publish a ballots table that lays out the rank-order voting data. With just 5 candidates, there are 85 distinct ways in which ballots could be marked; with 10, there would be 820. This creates challenges for accurate reporting and analysis, and data from each precinct in the voting district would have to be aggregated before the runoff process can begin.

This makes things very hard to track, and has led to delays in the counting of ballots in some cases. Even if unjustified, people may grow more suspicious of such complex procedures. This all says nothing of the fact that most standard vote-counting machines are not designed to handle IRV ballots. This would probably be a one-time cost, but it can still be substantial—Minneapolis’s first IRV election ran significantly over budget—and the machines would require new software if election officials continue to tweak the parameters, as has happened in numerous situations.

There are a few other claims by IRV advocates that seem logical enough, but struggle under further scrutiny:

IRV always elects a majority winner. This is only true if one takes a rather contorted view of what constitutes a majority. In the Minneapolis election, Betsy Hodges, despite being the clear leader from start to finish, never actually got a majority of the votes. She only cleared the 50% threshold in the 33rd round of counting, when 15,573 people (19.6% of the electorate) had seen their ballots entirely eliminated from consideration. Her final IRV total was 38,870 votes—almost 10,000 more than she’d started with, but still less than 50% of the initial votes cast. And this was in an election that was not at all controversial.

Eliminating the primary reduces costs. Primary elections are not especially expensive, and cost savings in running elections would be counteracted by the amount of money candidates would need to spend to stay in the race all the way to election day. It just spreads the cost around to different places. Moreover, a longer election season inherently favors candidates with deeper pockets. Primaries have their shortcomings, but they also channel party funds toward the most popular internal candidate. Some in the IRV crowd have argued that this just means some of the candidates were never serious anyway, and that perhaps there should be higher barriers to entry, through fees or petition requirements. This seems a rather curious way to go about making things “more democratic.”

Primary elections certainly have their issues, with small turnout and extra emphasis of activist fringe groups, though there are plausible counterarguments to these points, as Councilor Julsrud explained when the Duluth city council considered IRV. If we are to replace it, though, we should make sure the new version is certifiably better.

IRV improves behavior and minimizes mudslinging. This may be true at first, as people figure out how IRV works. In the long run, however, it’s a probably a placebo effect, to the extent that it can even be measured. No one really campaigns to get second place votes; they’re in it to win. Voters learn to play along and still vote strategically. Cynicism is just a fact of life is modern electoral politics, and it will take a lot more than a new voting method to change that.

IRV empowers minorities and candidates from underrepresented groups. Australia’s experience with IRV does not bear this out; instead of empowering minority candidates, its two major parties are more rigid than ever. If you want do give minority groups more power, you probably should consider proportional representation, but that’s a debate for a different day. The only way this argument works is by relying on the contention that minority groups are less likely to vote in primaries or runoffs, but this does nothing to guarantee that such candidates will actually run.

IRV’s greatest troubles, however, tend to come when there are more than two serious candidates. A favorite example of this trouble is that of the Burlington, Vermont 2009 mayoral election, in which the winner of the election did not have the most votes in the first round, and when matched up against the other two major candidates head-to-head in a ballot table (a method used to determine the Condorcet winner, another measure of electoral effectiveness), he lost to both. But, thanks to the round structure of IRV, he won. Burlington voters, baffled, did away with IRV shortly thereafter. Several other cities have also dumped it, and its implementation never ends the debate: Minneapolis is discussing tweaks after its first run with IRV, and San Francisco seems to have a constant string of amendment and repeal attempts underway. IRV’s adoption often leads to a protracted struggle over voting methods when there are probably more pressing issues in front of local legislators.

IRV might work well to counteract cases in which a small minority candidate saps just enough support away from a major one to flip an election (see Ralph Nader and Al Gore, 2000), but once things become more complicated, it’s not much better than plurality, and perhaps even worse. Depending on who gets eliminated first, there are actually scenarios in which it makes more sense to vote up one’s least favorite candidate over one whom the voter might prefer.

IRV advocates will retort that many of the complicated scenarios don’t happen all that often, and therefore this is all nitpicking. We could say the exact same thing about plurality voting, though, and while there are merits to IRV, it ultimately seems like much ado about very little. It also probably just distracts from some of the issues its proponents are probably most passionate about, such as voter turnout (where it cannot match the impact of a serious get-out-to-vote campaign) and money in politics (where voting methods are ants compared to the Citizens United gorilla). It seems an odd battle to choose.

Is there a better voting method? Perhaps, the professors tentatively suggested: score voting. In simple score voting, voters simply vote “yes” or “no” on every single candidate on the ballot. They can vote for everyone, no one, or any combination of them all. More complex forms allow voters to score each candidate on a scale of zero to four or five (or conceivably any number). The idea here is that the candidate with the broadest support would generate the most votes. It abandons the obsession with majority, which is increasingly difficult to come by in a pluralistic society with many candidates running for office, and settles for the broadest possible consensus. The method is compatible with standard voting machines, leaves a clear paper trail, and could work with or without a primary.

The main objection to score voting is that it appears to violate the “one person, one vote” principle, but this really isn’t the case: everyone gets one vote on every candidate, and a blank is equal to a zero. Every voter has the opportunity to cast judgment on each candidate one time. The algorithm is also mercifully easy, sparing us trouble with endless rounds and eliminations and ties. Just add the votes.

It also resonates with younger voters who are used to a world in which everything gets a rating on the internet. (While it’s obviously far too small a sample size to say anything definitive, Rachel Breckenridge’s Contemporary Mathematics classes—Math 1024, a non-major math course at UMD—named it their top choice after experimenting with several different voting methods.) Don’t expect to see it before the Duluth City Council anytime soon, though; that’s not how the professors want to push this through. They’re not in any rush to implement it, and think it ought to be tested on much smaller scales before a city goes ahead and imposes it, but it’s worth a shot somewhere. With some careful experimentation first, a new method would come across as being far less politically motivated, and might generate more organic support—hopefully from across the spectrum, and not just one wing of one political movement.

One last note: at the last meeting, we figured that the method the City Council used to elect Joel Sipress was, in fact, something akin to the Bucklin Vote, in which 2^nd- and 3^rd-place votes are added when no candidate receives a majority of 1^st-place votes. (Under both plurality voting and IRV, the Council was deadlocked 4-4, and everyone who cast a 2^nd-place vote gave it to the third candidate who got zero first place votes. However, two of Sipress’s supporters did not fill in 2^nd or 3^rd choices, and the Council erroneously used this to give him the IRV victory.)  While doing some profound research on the Bucklin Vote (aka reading its Wikipedia page), I learned from the Minnesota Bar that Duluth actually experimented with the Bucklin Vote about a century ago, only to have it declared unconstitutional. As the Council’s experience showed, a variation on the Bucklin method applied to IRV could conceivably resolve some ties, though it does nothing to relieve the problems of transparency and paradoxical results that emerge in all ranked-ballot methods that use rounds. There is no quick and easy fix that will quiet all dissent and make IRV instantly palatable.