Ranked-Choice Voting in Wisconsin: Good or Bad?

Introduction

In September of 2023, Wisconsin legislators considered implementing a Ranked-Choice Voting (RCV) scheme in the form of Instant-Runoff Voting (IRV) for U.S. House and Senate elections (AB 563, 2023; SB 528, 2023).

IRV is one of many RCV schemes and probably the most popular. Its implementation and consideration in several States suggests that the current system of Plurality Voting (PV) is undesirable and would be improved by the adoption of RCV. If true, RCV would be a good system to adopt. The question, then, is what qualifies as “good?”

To understand whether Ranked-Choice Voting (RCV) in the form of Instant-Runoff Voting (IRV) is good, we must first understand the purpose of voting. Once we understand voting’s purpose, we can determine whether IRV achieves the goal of voting.

If RCV does a better job of achieving this goal than PV, then it is preferable to the current system and ought to be implemented. Other considerations such as the purposes of those promoting Instant-Runoff RCV are relevant but beyond our scope.

The Purpose of Voting

For Wisconsin, the stated purpose of its government is to “secure [freedom’s] blessings, form a more perfect government, insure domestic tranquility and promote the general welfare”; and that these purposes are established by “the people of Wisconsin” (Wis. Const. Preamble). The State, therefore, is a creation and agent of the people whose task is to pursue these goals.

The Wisconsin constitution then immediately declares that the State derives its “just powers from the consent of the governed” (Wis. Const. art I, § 1). A State which is entirely just must have perfect representation and the unanimous consent of the governed while a tyrannical state has neither. The extent to which the people give their consent, then, determines the justness of the State’s exercise of power.

The people’s goals are then pursued by representatives—elected to office by qualified electors of the State—exercising their powers in pursuance of the Constitution, and who express the will of the people through their legislation: “The people of the state of Wisconsin, represented in senate and assembly, do enact as follows…” (Wis. Const. art IV, § 17). To the extent that representatives attain these goals with the consent of the people, then, the government is representative.

The State, therefore, can justly exercise its powers while lacking perfect representation provided it meets a minimum standard for the exercise of power and provided the people’s consent to be ruled is not rescinded; but it cannot be truly representative in the exercise of power unless it is consensual and in pursuance of the people’s goals.

Based on Wisconsin’s Constitution, the State government is constructed as a representative government. In Wisconsin, voting is how this representation is attained. The more closely the actions of the State align with the desires of the people they govern, the greater the level of consent and the better the representation. The purpose of voting is thus to elect representatives that act in accordance with the desires of the people to the greatest extent possible. The ideal voting system according to Wisconsin’s Constitution is thus that which installs the most representative candidates.

How Instant-Runoff RCV Works

In a PV election, each voter picks one candidate and the candidate with the most votes wins. But in an IR election, voters rank all the candidates in order of preference and the election proceeds through rounds until one candidate has a majority.

For example, say an election is held between Al, Barry and Carl. In the first round, Al gets 28 votes, Barry gets 33 votes and Carl gets 39 votes. Although Carl had the most votes and would’ve won in a PV election, he did not get a majority. Carl received 39/100 votes but needed 51 to win a majority. And since there isn’t a majority, the election goes to the second round.

Since Al had the least number of votes in the first round he is eliminated from the election and everyone who voted for him will have their votes for their second-ranked candidate counted instead. As it turns out, 18 people who voted for Al ranked Barry as their second choice and the other 10 ranked Carl as their second choice. As a result, Carl gets 49 votes and Barry gets 51 votes—so Barry wins the election.

Is Instant-Runoff RCV More Representative?

For IRV to be representative the results of an election must do a better job of electing the candidate who most closely resembles the wishes of the voters and has the broadest support. Yet, IRV fails to satisfy these criteria in multiple ways:

• IRV violates the principle of One Person, One Vote
• IRV leads to non-majority, non-representative outcomes
• IRV is prone to the Center-Squeeze Spoiler Effect (t=4:43-7:24)
• IRV increases the frequency of Exhausted Ballots and discourages voting

One Person, One Vote

This principle states that “each vote be given as much weight as any other vote” (Wesberry v. Sanders, 376 U.S. 1, 1964). While the judicial merits and implications of this decision are contentious, intuition says, “Yes, each person gets one vote.” But in fact, IRV gives some voters more than one vote.

In our example from before, those who voted for Barry and Carl in the first round had their choices considered once. But those who voted for Al had their choices considered twice. This gives a strategic advantage to first-round voters for Al that is not afforded to the supporters of Barry and Carl. To understand why, consider these thought experiments:

You are playing a game of chess. Your opponent moves their Knight to a square. In turn, you respond by moving a piece to counter them. This is a fair interaction between you and your opponent—they move, you move.

But now imagine that after you move your piece your opponent is allowed to take back their move and move somewhere else. You, on the other hand, must keep your piece where it is. Is this fair?

How about this? You are playing rock-paper-scissors against a friend. At the exact same time, you shoot “rock” and they shoot “scissors.” You win! Right? Wrong—they win. And they win because after you have shown “rock” they change to “paper.”

This is precisely what happens in IRV and precisely what happened in our fictional election. The votes of Al’s supporters are over-weighted. This leads to a result that is less representative of the will of the people because some voters have the privilege of potentially changing the result of an election when they do not like the initial outcome.

Non-majority, Non-Representative Outcomes

As it turns out, our fictional election from before was not so fictional after all. In fact, the results of that election are precisely the results of an election that took place in Burlington, Vermont in 2009 conducted under IRV. In the aftermath of that election, IRV was then repealed the following year. So, what happened?

Just as before, three candidates ran in the election—Al, Barry and Carl. In the first round, Al got 28% of the votes, Barry got 33% and Carl got 39%. Al was eliminated from the running and the second-round votes of the Al voters were counted. Barry then won with 51% of the votes. The issue here is that Barry was actually the least popular candidate despite his electoral majority.

According to the ranked preferences of the Burlington voters, a plurality preferred Al to Barry and a majority preferred Al to Carl. In other words, if Al were running against either Barry or Carl he’d have won the election in both cases. Since Al was the preferred candidate according to most voters, he was the most representative. Yet because only the second-round votes of Al’s supporters are counted, a suboptimal candidate is elected. Had Barry and Carl’s voters known that the Al voters would get Barry elected, the Barry voters would’ve kept their votes the same, but the Carl voters would’ve changed their vote to Al and Al would’ve won.

Furthermore, because several ballots were exhausted (ballots that are no longer viable) after the first round, Barry was able to win without receiving a true majority. He won with a majority of the viable ballots but received only 48% of the total ballots cast. Had Carl’s voters been allowed to choose their second choice, Al would’ve won with 54% of the viable ballots.

This scenario has materialized in a number of State and local IRV elections. In Alaska’s 2022 election for U.S. Representative, two Republicans—Nick Begich and Sarah Palin—faced-off against Democrat Mary Peltola. In the first round, 28% voted for Begich (R), 31% for Palin (R) and 40% for Peltola (D). After Begich was eliminated, 50% of his votes were transferred to Palin, 29% were transferred to Peltola and 21% were exhausted, i.e. not counted. Of all ballots cast in the election, 6% were thrown out and had no weight in determining the result of the election.

Furthermore, like Barry, Peltola did not get a true majority—she received only received 48.4% of the total ballots. And just as Al was the most preferred candidate while Carl was the least preferred candidate, Begich was most preferred by Alaskans while Peltola was the least preferred. This brings us to the third failure of IR RCV: the Center-Squeeze Spoiler Effect.

The Center-Squeeze Spoiler Effect

This effect occurs when the more moderate candidate, preferred by a majority of voters, is eliminated or “squeezed-out” of an election by more radical candidates. According to the ranked preferences of Alaskans, in head-to-head matchups Begich beat both Palin and Peltola with 61% of the vote and 52% of the vote, respectively. In total, 90% of Palin voters preferred Begich to Peltola while 91% of Peltola voters preferred Begich to Palin. This is because IRV confers an advantage to more radical candidates and further increases the political polarization IR RCV advocates claim to reduce (here too).

With IRV, the more representative, moderate candidate, Begich, lost because voters are told that they can vote for their preferred candidate without any downside. Rather than vote for the candidate who most closely aligns with their views and has the best chance of winning, voters split their vote. In fact, the power of a vote is inversely proportional to the number of candidates in a faction. And under IRV, the more candidates a faction has the greater the vote-splitting and the greater the advantage of the radical candidate.

More radical candidates, of course, do a worse job of representing the electorate than more moderate candidates because voters are further from the radical candidate on average than they are from the moderate candidate (see footnote for proof).[1] This proof further demonstrates that IRV becomes increasingly less desirable the more the number of candidates exceeds three. In the common five-way IRV elections these effects are thus particularly magnified and even more prone to elect radical candidates when paired with a Multi-Winner election scheme.

Finally, Kenneth Arrow showed definitively that not only does RCV produce a Spoiler Effect, but that no RCV scheme can ever eliminate it. Thus, voting your conscience remains problematic in IRV and does not eliminate strategic voting.

Exhausted Ballots

Exhausted ballots are votes that do not influence the results of an election. If a ballot becomes exhausted, it is not counted even if it could have made a difference.

Some people think that exhausted ballots are only due to voter error but a ballot can become exhausted even if a voter fills out everything correctly. Additionally, the fact that voter error persists in IRV—and to a greater extent than in traditional PV elections—despite being in place in some States and communities for several election cycles is enough to warrant its repeal on the basis of disenfranchising voters.

In one analysis by RangeVoting.org comparing IRV to traditional PV, they found that voter error was seven times higher in San Francsico, 84 times higher in Aspen, Colorado, 4-7 times higher in Minneapolis and twice as frequent in Australia when compared to the U.S.

Neely and McDaniel found that “Asking voters to do more than pick a single candidate from a list leads to an increase in disqualifying errors,” in San Fransisco’s IRV elections. One SF election even resulted in 53% of ballots becoming exhausted.

One national survey found that in four separate IRV contests, the winners not only failed to receive true majorities, but also that in each election the rate of ballot exhaustion ranged from “9.6% to as high as 27.1%” (pg. 21). The Maine Heritage Policy Center also found in their survey of 96 municipal IRV elections held nationwide that on average, IRV elections exhausted 11% of all votes. They also found that with respect to these 96 elections, 61% of the time the winner failed to receive a true majority (pg. 15).

Two Alternative Solutions to the Problem of Representation

The first solution is an intuitive and age-old alternative to voting reform measures like IRV. It was suggested during the Ratification Debates, a condition of Virginia’s ratification, demanded by George Mason, Patrick Henry and Richard Lee, argued for in Federalist #58, submitted as the first amendment in the Bill of Rights, and was the only suggestion ever made by George Washington during the Constitutional Convention: Increase the number of Representatives.

The consensus among the Founders was that the number of Representatives ought to number one for every 30,000. Today, there are 435 Representatives in the House, each representing more than 750,000 people.

This apportionment recommendation was the first amendment to the Bill of Rights—Article the First. It read: 

"After the first enumeration required by the first article of the Constitution, there shall be one Representative for every thirty thousand,” and that once a certain number of Representatives was reached, that there shall not be “more than one Representative for every fifty thousand persons."

The second solution accomplishes the goal of better representation as well: Increase the voting threshold required for acts of legislation to a 2/3 majority. This threshold mimics those required in Senatorial impeachments, the overriding of presidential vetoes and the ratification of Constitutional amendments (3/4). Its justification is the same as that given during the Convention and those later given by Calhoun in his proposal for a Concurrent Majority. This solution has also recently been adopted by the Texas GOP in their party platform (pg. 6).

Conclusion

Ranked-Choice Voting in the form of Instant Runoff Voting for Wisconsin is unlikely to be a worthwhile substitute for the current Plurality Voting system.

It fails to satisfy the primary objective of voting to establish representative government, and fails in the following way: it does not satisfy the rule of “First, do no harm.”

While IRV may result in more representative officials, this is only true for those voters whose votes are counted. The extent and persistence of disenfranchisement via ballot exhaustion is an aspect of IRV that is directly at odds with the construction of Wisconsin’s government. All eligible citizens are entitled to have their voices heard and their votes counted. IRV systematically eliminates the votes of some voters to the benefit of others. This makes some votes worth more than others.

IRV also fails to provide sufficient representation by systematically conferring advantages to more radical, and necessarily less representative candidates. If one of IRV’s goals is to reduce political polarization, a voting system that promotes polarization cannot be seriously entertained.

Fundamentally, the issue of sufficient representation comes down to the insufficient number of representatives and the power of a majority. Better representation can be attained automatically by increasing the number of representatives which involves more citizens in government. Furthermore, the stakes of political elections would be reduced dramatically if the threshold required for positive acts of legislation were increased. Such a change would make reapportionment almost entirely unnecessary by ensuring that large majorities are required to pass legislation.

IRV is unlikely to be an improvement for Wisconsin and likely to be a detriment.

Footnotes