Approval Voting versus IRV
Approval voting is categorically better than instant runoff voting (IRV) in a number of ways. This article outlines those ways. Note that ranked-choice voting is the same as IRV, but limits voters to ranking only three candidates. We use the names interchangeably in this article. Limiting voters to ranking only three is very common because of IRV's complexity.
Approval Voting can be done with the same ballots voters are used to; you just remove the rule that says “vote for only one”:
IRV, on the other hand, requires a new ballot; because ranking all candidates would take up too much space, the ballots are typically limited to ranking three choices:
Approval voting requires only addition. The candidate with the most votes wins, and results from multi-candidate elections are easily shown in bar graph form.
IRV, however, uses a more complicated algorithm:
Total all the first-choice votes. If a candidate has greater than half these votes among the valid ballots, then elect that candidate. If not, then eliminate the candidate with the least first-choice votes. Look at the ballots from that eliminated candidate. Transfer that candidate’ next-choice votes to those candidates and treat those votes as first-choice votes. Again, look to see if a candidate has greater than half these first-choice votes among the remaining valid ballots. If there is again no winner, then repeat this process until a candidate has greater than half the total votes among the remaining valid ballots.
For an example of how this calculation process looks in practice, see an example from a real Oakland, CA mayoral IRV election. Note that this is a simplified version in that it omits the first five rounds.
IRV results in about seven times as many spoiled ballots as plurality voting, on average. Approval voting, however, experimentally results in about one fifth as many spoiled ballots as plurality. See analysis here.
Voting Machine Cost
Whether you're doing a hand count or using a voting machine, approval voting works fine either way. You're just adding candidate selections and removing the rule saying you can only choose one candidate.
IRV, on the other hand, is much more daunting to calculate by hand. Further, voting machines require significant and costly software upgrades to run IRV elections. And most machines don't allow for this software upgrade. This means buying entirely new (and expensive) voting machines. This can quickly toll in the millions of dollars.
Precinct Summability & Delay
Approval voting lets you do tallies at multiple locations or precincts. Those separate tallies can then be aggregated to achieve a result.
IRV, however, cannot be counted in precincts or separate locations. It must be counted in a centralized area.
Requiring a central tally location can also cause delays in getting final results. For instance, this message appeared on the San Francisco city government website for several weeks after their 2008 IRV elections:
“Due to the requirement that all ballots must be centrally tallied in City Hall and not at the polling places, the Department of Elections has not set a date for releasing any preliminary results using the ranked-choice voting method."
Approval voting completely eliminates the “spoiler” problem, whereas IRV merely mitigates it. To the right is a video explanation by CES co-founder and math Ph.D. Andrew Jennings on how IRV lets in spoilers.
Risk of Ties
Approval voting, like plurality, only has one layer of calculation. This creates only one opportunity for a tie. IRV, because it has multiple rounds (more with more candidates), creates many opportunities for ties.
Approval voting is sometimes attacked on issues dealing with majority outcomes. We go more into that here to explain how this is a nonissue.
All voting methods can exhibit problems, but their frequency and severity is particularly pronounced with IRV. Here we go into how IRV makes it so you can sometimes harm a candidate by ranking them better and help a candidate by ranking them worse. Again, there's the spoiler issue. There's also IRV's sqeezing out of moderates even when they're clearly the best winner.
Approval voting performs rather well in the face of tactical voting. More on approval voting tactics here and the "threshold strategy" here. Under basic assumptions with tactical voting, approval voting elects beat-all winners (Condorcet winners) when they exist. Computer simulations using Bayesian regret calculations (shown below) demonstrate better utility outcomes in elections using approval voting versus IRV even if all approval voters were tactical and all IRV voters were honest.
IRV is susceptible to tactical exaggeration. This is so much so that when voters are tactical, IRV degenerates approximately into ordinary plurality voting. Note how approval does not degenerate into plurality. IRV's tactical vulnerability can also mean voters do not rank their favorite candidate as first.
Third Party Fairness
Approval voting always lets voters choose their favorite candidates regardless of viability. That's great news for third parties and independents because that's exactly what they need in order to grow. Approval voting's favortism towards moderates is even more generous to moderate third parties and independents.
Three of CES's co-founders wrote a lengthy piece on approval voting and third parties here.
IRV has maintained massive two-party domination everywhere it has seen long-term widespread use. The most noteworthy example is Australia, where IRV has been used in their House of Representatives since 1918.
This seems to result from voters insincerely exaggerating their rankings. For instance, an IRV voter who favors a minor party will tend to insincerely rank his favorite major party candidate in first place. Why? Two reasons:
That actually is the best strategy with IRV.
Most voters don’t understand the mathematics of their voting system, so when faced with a ranked ballot, they will tend to use the “naive exaggeration strategy“. This entails looking at frontrunners and accordingly ranking one as best and the other worst, regardless of preference to other candidates.
Third Party Support Accuracy
On Election Day, Tuesday November 6, 2012, the Political and Electoral Reform of Occupy Wall Street conducted an alternative voting experiment at polling places in Manhattan’s 69th State Assembly District. The group was granted credentials from the NYC Board of Elections to conduct the exit-poll style experiment inside the city’s official polling places. The experiment compared Plurality Voting (traditional “vote for one” method) with Approval Voting, Score Voting and Instant Runoff Voting. You can also find our extended interview with the organizer here.
Below are graphs revealing the totals for plurality, IRV, and approval voting:
While this district was clearly not representative of the overall American electorate, note the relative strength of the minor party candidates compared to the major party candidates. For instance, Green Party candidate Jill Stein received one vote for about every 27 votes for Obama. (The study authors note that their results were consistent with this princinct's official election results.)
Now let’s look at Instant Runoff Voting.
Because Obama won in the first round of voting, we don’t see the depth of support for the minor party candidates. These results actually look almost identical to the Plurality Voting results! And even when IRV elections require several rounds of elimination, the news headlines rarely shed light on the details. And how would they do so if they wanted to? There is no good way to summarize the results of an IRV election into a simple figure that says, “here’s how many votes each candidate got.” This gives the false impression that the minor party candidates had far less support than they really did.
Now compare to Approval Voting.
The Green Party now receives 58% as much support as the Democratic Party. This is over a 15x improvement in the Green Party’s strength relative to the Democratic Party, compared to where they were with Plurality Voting. The other minor party and write-in candidates also fared dramatically better.
Computer simulation studies show that approval voting is superior to IRV as measured by “Bayesian regret“, an objective “economic” measure of average voter satisfaction. The following graph, taken from page 239 of William Poundstone’s book Gaming the Vote displays Bayesian regret values for several different voting methods, as a function of the amount of tactical voting. From the figure below, you can see the complete lack of overlap between approval voting and IRV in utility, even when allowing for tactical voting