In a previous post, I explained why FPTP is a very bad voting system.
So if FPTP is bad, what are better systems? A lot of voting systems that have been proposed. In this post we will consider one of them, Instant Runoff Voting.
Instant Runoff Voting (IRV) is also known as Alternative Vote (AV), Ranked Choice Voting (RCV) and Single Transferable Vote (STV) with 1 winner.
Note that AV isn't a very descriptive name as there are lots of alternatives to FPTP, and RCV isn't a good name either as there are lots of systems that allow you to rank choices. For some reason, voting systems often have crappy names.
On the other hand, STV is a good name as IRV gives each voter a single vote that is transferable between candidates (more on that later), but the name STV is normally used for elections that elect more than one winner.
Here's an example of an IRV ballot paper:
In this example, our voter's first preference has gone to Charlie carrot, their 2nd preference to Betty Banana, and their 3rd preference to Emily Eggplant. They have no preference between Apple and Date, so have not marked any further preferences.
How is IRV counted?
the number of 1st preferences for each candidate is tallied. If one has more than half of all the votes, they are declared the winner. If not, the candidate with the least number of votes is eliminated, and those votes are transferred to the voter's next preference. This continues until there is only one candidate left, who is the winner.
A worked example of an IRV election.
Going back to our example from the post explaining FPTP, we had an election result as follows:
votes
Bad Party 12000
Worse Party 11000
Because of this, a new party, the Better Party, put their hat in the ring. They attracted a few voters from the Worse Party, more form the Bad Party (who they were ideologically closer to), and some new voters who hadn't bothered to vote at all the last time as they didn't like either candidate. The result was:
votes difference
from last time
Worse Party 10500 -500
Bad Party 10000 -2000
Better Party 4500 +4500
And because of the Better Party's intervention, things have gone from Bad to Worse! This is called the spoiler effect and it's a major flaw of FPTP.
Here's how the election might play out under IRV:
IRV election -- first round
Worse Party 10500
Bad Party 10000
Better Party 4500
If this was a FPTP election, the counting would end there, and the Worse Party would be the winner. But in IRV, the Better Party has the least number of votes and gets eliminated. 2000 of their voters don't have a 2nd preference, so thier voters are exhausted. Of the Better party's other 2500 votes, 2000 votes go to the Bad Party and 500 to the Worst Party:
IRV election -- 2nd round
Bad Party 10000 +2000 => 12000
Worse Party 10500 +500 => 11000
Better Party 4500 -4500 => 0
Now Bad Party has more than half the votes, so is elected. Because the losing Better Party's votes get transferred to their voters' next preferences, there is no longer any spoiler effect. Voters can happily give the Better Party their first preference, knowing that if they don't win, their 2nd preference goes into effect.
Tactical voting is still possible in IRV
So if there is no spoiler effect, does that mean we can all agree that IRV is the perfect system? Not so fast: tactical voting is still possible in IRV.
That is to say: it is possible for some voters to get a more-preferred outcome by voting other than their true preferences.
Consider an election where the candidates are Apple, Banana and Carrot.
Banana is a centrist party. Banana supporters are equally split on their 2nd choices.
Apple supporters' preferences are all: Apple > Banana > Carrot.
And Carrot supporters all prefer the order: Carrot > Banana > Apple.
The 1st round of the election goes like this:
IRV election -- first round
Apple 8000
Banana 6000
Carrot 5900
Carrot is eliminated, and all Carrot voters 2nd preferences go to Banana:
IRV election -- 2nd round
Apple 8000 +0 => 8000
Banana 6000 +5900 => 11900
Carrot 5900 -5900 => 0
And Banana is the winner.
But what if 200 sneaky Apple voters vote for carrot on the 1st round? Then we get:
IRV election -- first round
Apple 7800
Banana 6000
Carrot 6100
Banana is eliminated, their voters split equally between Apple and Carrot on the 2nd round, and Apple wins.
Apple has won the election by some of their voters voting for their least-liked candidate on the first round. This is called a center squeeze, because it results in the center party being squeezed.
Center squeeze can happen naturally
In the previous example, a center squeeze happened because some voters voted tactically. But a center squeeze can also happen naturally. Consider if party support was roughly:
Leftist Party c.40%
Centrist Party c.20%
Rightist Party c.40%
The Centrist Party has roughly 20% support, so would be first eliminated. It would then come to a close fight between the Leftist Party and the Rightist Party.
Research questions
IRV is fairly widely used: for example it has been used in Australia since 1918. Has any research been done to see how common center squeeze is in practise?
Further reading
Electowiki focuses on electoral reform and describing alternative electoral systems. It has pages on: