Posted: November 30, 2010
Looking Backward: Instant Runoff Voting in the 2010 Midterm Elections
In some states, the recent midterm elections are far from over. However, in those races where the winners have been determined, these election results are a veritable goldmine of data. Continuing my interest in exploring alternative voting systems, looking at this data through the lens of Instant Runoff Voting (IRV) illuminates three conclusions: (1) how IRV would affect elections if broadly implemented; (2) that there are differences in voting efficiency between IRV and plurality voting; and (3) putting economics and broad application together, it is possible to answer the ultimate question of IRV: “is it better than plurality voting?” For brevity, my previous commentary on IRV walks through the basic mechanisms of the voting system.
Organizing the data of the midterm elections, the elections coalesce around three categories of results when examining them under IRV system design:
1.Races in which no candidate received a majority of votes
2.Races in which a candidate received a slim majority of votes
3.Races in which a candidate received a large majority of votes
Working backwards through these groups, Category 3 is the easiest to understand, and plentiful examples exist: Wyoming’s gubernatorial race: Meade received 123,764 votes and the next closest challenger, Peterson, received 43,336 votes; in Hawaii’s Senate election Inouye received 276,928 votes and his closest challenger had 79,830 votes. These Category 3 races, if run under IRV, would result in the same winner without IRV affecting the results. Category 3 elections indicate that even with widespread implementation of IRV, the elections that experts and pundits expect to be landslides are in fact landslides. IRV would not cause drastic, dramatic voting shifts and change “safe” elections different from those results obtained in plurality voting.
Category 2 races are those with a clear winner, but that winner achieved a very small majority of votes. This type of election is where the differences between IRV and plurality voting start to arise, as IRV could produce different winners in these races because of differences between voting system structures. IRV produces two effects that plurality voting dampens: (1) voters do not feel pressure to vote for a candidate from either major party candidates as they might otherwise feel that their vote wasted; and (2) votes can aggregate in later runoff rounds for a candidate that survives each successive round but does not achieve any sort of numerical strength and becomes electorally strong. Looking at the result of the Colorado gubernatorial election illustrates these effects.
John Hickenlooper 856,569 50.7%
Tom Tancredo 620,626 36.8%
Dan Maes 187,998 11.1%
Jaimes Brown 12,337 0.7%
Other Candidates (2) 11,183 0.7%
First, if individuals favored Maes and disliked either Tancredo or Hickenlooper, in plurality voting these voters would feel pressured to vote for the candidate in the best position to defeat their disliked candidate and not vote for their actual first choice candidate. Thus, it seems that a candidate wins under plurality voting because a significant number of voters support the candidate, but also because voters do not like the opposing major party candidate (Delaware’s Senate race might be an illustration of this principle). IRV enables voters to choose their actual first choice because there is no pressure that voters will “waste” their vote by not voting against the disliked candidate by having to select a candidate with a higher possibility to win. Then, under IRV, an individual like Maes or Tancredo can receive a majority in later runoff voting rounds by aggregating second, third, and fourth round votes when voters do not feel pressured to support a top two candidate.
In Category 1 elections, IRV requires further runoff rounds in order to declare a winner. For example, the Florida Senate election was very close, and under a system in which a candidate must earn a majority to win, the ultimate result can be far different. In Florida Rubio received a near majority, but as votes aggregate, another voter could overtake that lead. Consider the results of Florida’s Senate election as the starting point for this next hypothetical:
Election Round 1 Votes Percentage
Marco Rubio 2,615,262 48.9%
Charlie Crist 1,588,821 29.7%
Kendrick B. Meek 1,076,028 20.1%
Other candidates (6) 39, 409 0.7% (approx.)
As there is no majority on the first ballot, further rounds are necessary. There are a number of candidates in this election who received approximately 0.1% of the total votes, so several runoff cycles are necessary to eliminate these low-vote receiving candidates. Thus, after eight rounds of this hypothetical runoff, Crist and Meek have made minor gains while Rubio stands in front.
Election Round 8 Votes Percentage
Marco Rubio 2,615,262 48.9%
Charlie Crist 1,627,773 30.6%
Kendrick B. Meek 1,090, 501 20.5%
In this round, Meek is eliminated and the ninth round is a head to head between Crist and Rubio. Voters who placed Meek first in this hypothetical placed Crist second and thus Crist earns a majority and wins the election with 51.1% of votes.
Election Round 9 Votes Percentage
Marco Rubio 2,615,262 48.9%
Charlie Crist 2,718,274 51.1%
It is possible to imagine the election turning in a different direction under slightly changed voting results. Rubio needed to aggregate approximately 44,499 votes to earn a majority, but the first eight rounds do not encompass enough available votes for Rubio to win even if he picked up every vote as the second choice candidate after the minor candidates are eliminated. Thus, in the eighth round, favoring Rubio in each runoff cycle the vote would be:
Election Round 8 (Version 2) Votes Percentage
Marco Rubio 2,653,671 49.8%
Charlie Crist 1,588,821 29.7%
Kendrick B. Meek 1,076, 028 20.1%
Once Meek is eliminated and the ninth round runoff begins, Rubio only has to pick up 6,089 voters to earn a majority and win the race.
Drawing a greater point from these three categories of elections, we see results that can best be understood through economic thinking. We have already seen that IRV allows a clear winner when the electorate votes in clear favor of a candidate and removes some obstacles tow voters being able to vote without external pressures. Additionally, IRV creates the possibility that in marginal cases, non top-two candidates can win, inducing more competition.
In purely economic terms, efficiency occurs when no individual can be made better off without harming another individual. Modeling from economic efficiency, voting efficiency then occurs when no voter or candidate can be made better off without harming another voter or candidate. Measuring voting systems in terms of comparative efficiency demonstrates the differences in externalities and encumbrances of different voting systems. Of course, a voting system that is “maximally efficient” is not necessarily a good thing—for instance, when voters have no encumbrance to voting, vote fraud might be rampant due to the ease of submitting a ballot; when candidates have no encumbrances on their actions, the ballot may be flooded with similar candidates such that choosing amongst them is near impossible.
Comparing IRV and plurality voting, both appear to hit a similar point of voting efficiency. As seen earlier, Category 1 and 2 elections demonstrate that IRV voters have less external pressures on their voting behavior and candidates have fewer barriers to enter the election market. However, plurality voting is more efficient in education, as IRV forces voters to educate themselves about a far larger group of candidates because these voters must later rank the candidates. Voters must learn enough about each candidate to rank them against the rest of the field, and if voters do not rank candidates or learn enough, the primary value of IRV is defeated. Additionally, on the candidate side, plurality system provide better screening and vetting mechanisms for candidates when considering the primary and general elections—the barriers to entry that plurality voting possesses serves to exclude possible candidates who have no business in the election market.
Tying together the election data with what we see in comparing voting efficiency sheds light on the question of “which is better, IRV or plurality voting?” The answer has become clear: there is not some quantitatively better system, but IRV and plurality voting stress different values in the composition of their systems. The question then, in considering whether to continue with plurality voting or to move to IRV, changes to “what values do we want our voting systems to reflect?” Answering this question then, we must consider whether we value our top-two candidate system (favoring plurality voting), or whether we think attempting to select through majority voting is more important (favoring IRV). The answer to this valuation, then, seems best left to voters.