Should we switch to Instant Runoff Voting?

The question in the title comes up from time to time in the political discourse, usually when there are three-way races in which two groups with majority opinion split the vote. In my post yesterday on three-way races, I noted that some of these political contests might provide a more acceptable result for the population and stave off strategic voting, if IRV (Instant Runoff Voting) was implemented.

The Florida electorate is 23% Unaffiliated, 41% Democratic, and 36% Republican. Most Republicans and some Unaffiliateds are backing Marco Rubio, the very conservative Tea Party-backed Republican nominee, but Rubio's campaign has stalled out in its numbers, consistently failing to reach above about 45% of the vote. The assumption, therefore, is that 55% of the electorate does NOT want Rubio elected. These people are a few disaffected Republicans, many Unaffiliateds, and almost all Democrats. And yet Rubio is projected to win the election because Charlie Crist, the Independent, and Kendrick Meek, the Democrat are splitting the "anti-Rubio" vote.Why should Rubio win if most people don't want him elected?

For that matter, why should any candidate win when the MAJORITY of the electorate does NOT want him or her elected? This seems anti-democratic! And yet this is one of the major problems with the "plurality" voting method in the United States. You could win with 5% of the vote as long as no one else receives a higher number.

Normally, the plurality method penalizes the moderate in the election. But because of strategic voting, many people decide to vote for a candidate whom they do NOT prefer in order to prevent a candidate they prefer even less from winning. This is why Crist could win in Florida; there aren't enough Unaffiliateds to win him the election so he needs major defections from Democrats. But this is "dishonest" voting!

In Instant Runoff Voting, voters rank candidates first choice through third choice on the ballot. Let's simulate IRV for the Florida race:

Meek gets 20% of the vote, Crist receives 35% and Rubio receives 45% in this simulation. In plurality, Rubio wins. In IRV, Meek gets dropped from the ballot. Then all the second choices of the Meek voters are counted. It's almost certain that all Meek voters chose Crist as their second choice, since Crist is more moderate. Thus Crist is awarded 55% of the total vote and wins against Rubio who still has 45%! Perfect!

This allowed Meek voters to speak their mind, but also prevented Rubio from winning at the same time. It allowed Crist to be recognized as the overall favorite, even if he wasn't the individual favorite of all voters. This kind of method would also allow other third party candidates to be recognized by voters otherwise afraid of "wasting" a vote on them, like Libertarians or Green Partiers who poll at around 5% in some elections.

Of course, it's not that easy. Despite the proponents of IRV, there are some risks with implementing an IRV method in US elections. I found several sites that spell out the problems of IRV, but the most fascinating and disturbing from a common-sense standpoint is that under certain conditions IRV has the tendency to show chaotic patterns that are nonmonotonic. According to Ka-Ping Yee, who has published some extensive voting system research online, in the mathematics of elections, nonmonotonicity essentially means that "a shift of public opinion toward a candidate can cause the candidate to lose, and a shift of public opinion away from a candidate can cause the candidate to win". That's right, there are scenarios in 3-way IRV where Candidate X could actually cause him/herself to lose... by GAINING support. What?

That doesn't make logical sense for most of us, and rather than trying to explain it myself, I would strongly encourage you to play with this really really cool election simulation feature by Ka-Ping Yee. Click on "Hare (IRV)" at the bottom, and it'll highlight the areas where each candidate will win. Drag out three or four candidates (triangles) and play with their positions until you notice the nonmonotonicity. It's unmistakable because you might suddenly find that an electorate which most sympathizes with "yellow candidate" will consistently elect "green candidate" in IRV. The stats are noted above if you don't understand the graph. And if that's not working or you're still confused, that's fine, please read this nice visual explanation of the "pathological" behavior of IRV. I want to thank Ka-Ping Yee for his work on this, the visualizations are helpful!

If you were playing with those pages, I'm sure you noticed that five different voting systems are profiled. The ones I find most satisfactory are the Condorcet method, the Approval method, and the Range method (which is not mentioned but is similar in some ways to approval voting). Every system seems to have its form of strategic voting, but some have more problems than others. Please read about some of these systems. What is the best? What do you think we should switch to? Range certainly seems to be the most fun to me.


So, is IRV better than Plurality overall? Yes I'd have to say for the majority of situations it is. But I think it would be easiest to switch to one of the three above mentioned methods, particularly Approval voting, since it's not difficult to understand. Or Disapproval voting might be an interesting switch, though not in the interest of toning down the vicious nature of the political arena. Regardless of what form we choose, it seems clear that we need a serious overhaul of the voting system to become a major political priority in the United States.