Instant Runoff Voting

Instant Runoff Voting

Paul McClintock

Atkins, Brown and Carr are candidates for mayor, a non-partisan position. The voters rank them thus:

37% prefer Atkins, then Brown, then Carr

33% prefer Brown, then Carr, then Atkins

20% prefer Carr, then Brown, then Atkins

10% prefer Carr, then Atkins, then Brown

Simple election situations with only two candidates are readily resolved with a single ballot and a majority winner. But majority decisions are often not so simply obtainable with more than two choices. In scenarios such as the one above, it behooves us to ask, who should be elected? Is that who would be elected under current law? There is no simple or completely satisfying answer to the first question, as there are several conflicting goals in election systems.

As a result of the year 2000 presidential election problems in Florida, there have been numerous election law changes and proposals, with more on the way. California's blanket primary was ruled unconstitutional by the U.S. Supreme Court in June 2000, and the Washington blanket primary is pending before that same court as I write. San Francisco voted about a year ago to dispense with its primary and use Instant Runoff Voting (IRV) in the general election. What is IRV? What are its strengths and weaknesses? Under what circumstances should it be considered a good approach? Are there better alternatives? Worse? How so? This article attempts to answer these questions.

Let's begin with the Atkins-Brown-Carr race. A Plurality election system says that the one with the most votes wins, even if less than a majority (more than half). In this case, Atkins has the most and wins. But notice that Atkins is the LAST choice of 53%.

Some states have a runoff election between the top two vote-getters, between Atkins and Brown in this case. Assuming the exact same voters showed up at the runoff election and that their preferences didn't change, then "the 20%," (a group of Carr's supporters), would vote for Brown and "the 10%" (another group of Carr's supporters) would vote for Atkins, so that Brown would win with 53% of the vote, a clear majority.

If the election system were IRV, it would simulate the runoff system but without the runoff. At the election, ballots would allow voters to indicate their first and second choices. E.g., For example, 37% mark Atkins as their first choice and Brown as their second choice. The ballots are counted according to the first choice: Atkins gets 37%, Brown gets 33%, Carr gets 30%. No one gets a majority, so a runoff election is simulated. Carr is dropped from the simulated runoff as the last-place candidate, and the ballots cast for him are now cast for the second choice, resulting in Atkins now getting 47% and Brown 53%. Brown wins with a majority vote.

Had there been 5 five candidates, IRV would potentially simulate three runoff elections (a 4-way, then a 3-way, then a 2-way race), and voters could indicate on the ballot their first, second, third and fourth choices (the fifth choice doesn't need to be indicated, as it is the only one left). In the original five-way race, if no one got an gets a majority, the last-place candidate is dropped from the first simulated runoff, and each ballot cast for the dropped candidate is re-cast for the highest ranking candidate that remains in the race. If still there is not a majority winner, the process is repeated until there is one. This is logically fairer than a single-step runoff between the top two.

IRV has the advantage of (a) providing a majority winner rather than a mere plurality winner, (b) being fairer than a single runoff between the top two, (c) avoiding the high cost of extra election dates for runoffs, and (d) avoiding the voter participation drop-off problem of extra election dates.

But IRV is not the perfect solution (mathematician Kenneth Arrow proved there is no perfect solution). Voting purists would argue that the Borda Count is better, which in a 5-way race would also have voters indicate first through fourth choices, would use a weighted counting scheme, casting 4 votes for a voter's first choice, 3 for the second choice, 2 for the third, 1 for the fourth. After all ballots are thus processed and the totals computed, the candidate with the most votes (plurality) wins. The relative pros and cons between IRV and Borda are beyond the scope of this article, but suffice it to say that IRV is the one being most commonly proposed today for public elections.

One of the biggest problems in using IRV in public elections today is the voting equipment now in use. Many counties still use punched card machines, with only one column of punches available at a time. IRV could be implemented on a punch card system by printing the same 5-way race four times, asking voters to mark their first choice in the first "race," their second choice in the second race, etc. These races would likely span multiple pages, and would likely cause voter confusion similar to the infamous "butterfly" ballots in Florida in 2000. A three-way IRV race could be done on punched cards with a single "race" and thus with less confusion, e.g.,

Mark ONE of the following:

Atkins (no second choice)

Brown (no second choice)

Carr (no second choice)

Atkins 1st, Brown 2nd

Atkins 1st, Carr 2nd

Brown 1st, Atkins 2nd

Brown 1st, Carr 2nd

Carr 1st, Atkins 2nd

Carr 1st, Brown 2nd










A paper ballot with mark-sense ovals could be prepared with much less confusion, e.g.,

Apparently some of the existing mark-sense oval systems are limited to three, four or five columns, and none of them could handle side-by-side all the choice levels in the recent California gubernatorial recall election, which had over 100 candidates. But even if allowed to rank 100 choices, virtually no voter would want to think through that many choices. Three to five choices is likely the most number of choices practically needed.

All voting systems have an element of confusion. There are folks who over-vote (mark more than one candidate) even in today's non-IRV systems. We can expect some voter confusion with IRV as well, with some marking Atkins for 1st, 2nd and 3rd choices. That ballot could still be counted as if only the 1st choice was indicated and the voter abstained in the simulated runoff elections if Atkins is dropped. But an IRV ballot with both Atkins and Brown marked as 1st choices would have to be voided, just as in the non-IRV case in today's systems.

IRV has a special attraction in partisan elections, for both major and minor parties. Suppose you have a race with three candidates: Democrat, Republican, and a minor party (D, R, M). In the plurality system, many voters preferring M will likely vote instead for their second choice (D or R), feeling that they are "wasting" their vote for M and perhaps thereby helping the person they dislike the most to get elected. The D or R will likely be elected, but in many cases it is without a majority vote, and sometimes it is obvious that the plurality winner wouldn't have won had it been a two-way race between the D and R. The D or R wins, but without a "mandate" from the voters. About half the time the D's benefit from the plurality (vs. IRV) system, and about half the time the R's benefit, so there is no net benefit for either side, but there is a net loss of the "mandate" for both sides. Thus the major parties would benefit from IRV.

Minor parties also benefit from IRV. All those truly preferring M can vote for M without "wasting" their vote, because they can mark the D or R as their second choice, and if IRV eliminates M, their second choice is counted. This would result in the statistics finally showing the true level of support each minor party has. Instead of M getting 7% of the vote in the plurality system, he may get 15% using IRV. This helps those who do get elected to know better the strength of the views of the constituency. The R or D who is elected in IRV knows that if he doesn't accommodate the concerns of the M voters who voted him in with secondary choices, he may not get those votes next time. Thus the minor party has increased influence over those elected, even when not of their own party.

Ideally an election system would not limit the number of candidates (but would give voters a full spectrum of choices), would be held on a single election date to minimize election costs and maximize voter participation, would be simple for the voter to understand, would elect the candidate most preferred taking secondary choices into consideration, and would not be too cumbersome for election workers. Instant Runoff Voting presents an attractive alternative to consider supporting, especially in contrast to today's plurality system. Your views (for or against) and questions on IRV are welcome; contact me by at


Paul McClintock is a Professional Registered Parliamentarian ( and Chairman of the Christian Liberty Party of Washington. This is a reprint of the article, first published in The Statesman, February 2004.