Over the last week I ran a little experiment comparing First Past the Post (the current system) and the Alternative Vote (a voting system that we will have the chance to adopt instead pending a referendum on it on May 5th).
I wanted to look at how accurately the two systems allow voters to represent their views on available candidates in an election. It seems pretty clear that people are not simple-minded when it comes to making a choice. They weigh up competing factors and deliberate of which of them are more important before making a final decision.
So it would make sense to have a voting system that allows you to express your sophisticated views accurately. When presented with a list of options you have positive and negative views of them all. There might be a candidate you want to win more than anyone else, but your opinion will also view the remaining candidates differently. Some of them, while not your favourite might be reasonably satisfactory winners in your eyes, especially compared to some others which you might find particularly objectionable.
THE EXPERIMENT:
So onto the first part of my experiment: I ran a ballot where voters voted for the same outcome using FPTP and AV in parallel. Rather than use a political theme and risk a cross-over with party political issues, and to make it a bit more amusing and interesting for the participants, I picked five deeply unpopular figures and asked the voters to vote for one of them to be executed. 31 participants submitted their ballots to me by email after I sent an open appeal with a link to a zip file containing four kinds of ballot papers and detailed instructions of how to participate, appealing on twitter, facebook and a few internet message oards I frequent.
Here is the First Past the Post version of the ballot:
The results are as follows:
Rupert Murdoch 20 (65%)
Sepp Blatter 4 (13%)
George Osborne 3 (10%)
Gordon Ramsay 3 (10%)
Simon Cowell 1 (3%) n.b. percentages are rounded
OK, so it certainly looks like the vote was overwhelmingly in favour of Murdoch. Even if all the other voters desperately wanted Murdoch to live and had all voted for the same rival candidate, the outcome would have been the same. Still, it would be helpful to take the voters holistic views into account. We don’t know, for example, how fond the voters were of the other candidates, neither in absolute nor in relative terms. It would certainly help everyone if a party could know how popular their candidate was compared to all of the others and use that information to inform future campaigns. Supposing Murdoch was pulled out of the running at the last minute, how might people have voted between the other four candidates? You have no way of knowing from this.
MODIFYING FPTP TO INCLUDE A POSITIVE AND A NEGATIVE VOTE:
So I have devised a crude modification to FPTP to allow voters to show the negative end of their views as well as the positive. It is my aim to show that AV is merely a more sophisticated modification of FPTP. In my simple modification I asked the voters to mark a candidate whom they would like to be saved. I used a separate ballot, but if such a system were ever used in a real election it could be done with an alternative symbol to go along with the X on one ballot. Here is the ballot used in this experiment.
This negative vote could be subtracted from the totals of positive votes. In FPTP where the winner is the one who has more votes than any other individual, it’s possible for more voters to actively dislike the winner than the number who voted for the winner. Although in this case Murdoch has a clear majority so the outcome wouldn’t be affected, the results do show that voters’ views are more sophisticated than a straightforward FPTP vote indicates.
Here they are:
Cowell +1 -3 = -2
Blatter +4 -0 = 4
Osborne +3 – 16 = -13
Ramsey +3 -12 = -9
Murdoch + 20 – 0 =20
So comparing the modified set of results with the straight forward FPTP results shows us a few interesting things.
First of all Murdoch’s “victory” is shown to be even more convincing because voters had an opportunity to vote against his execution without detracting from their preferred victim. That they declined this opportunity, gives Murdoch an even stronger mandate to be executed. This shows that the modification benefits a strong candidate who appeals to more than just a core group of voters.
Secondly, the rankings of the candidates change rather dramatically.
Simon Cowell has gone from being the least popular execution victim to being ranked in the middle. What has happened here is that compared to the other candidates few voters have much of an opinion of Cowell at all. In First Past the Post having very little opinion of a candidate is translated as having a negative opinion. The modification shows that a negative opinion and a neutral opinion are quite different things.
Similarly, the inclusion of the negative votes shows that George Osborne is clearly the darling of the electorate in that a considerable number (although not a majority), would like him saved. Gordon Ramsay is not far behind, but then there is a gap which was not indicated by the standard FPTP vote.
Both have gone from being just behind Blatter and just ahead of Cowell to being definite NOs for execution. Had Murdoch not been a candidate, who knows what the outcome might have been? Either of Ramsey or Osborne might have been executed in a FPTP ballot, upsetting a very large number of the voters.
So one can already see that voters’ views are too sophisticated to be represented by First Past the Post with any real accuracy.
FURTHER MODIFICATION TO AV:
OK, well that’s all well and good, but how does AV fit in?
Well AV is simply another modification of First Past the Post, along the same lines. Instead of marking a ballot with one positive vote and one negative one, you rank all the candidates in your order of preference. So not only does it allow voters to convey the extreme positive and negative end of their views, it also allows them to rank all of the other candidates in between.
As well as asking the participants to submit me an FPTP ballot, and a ballot of negative votes, I also asked for two AV style ballots. One ranking in order of preference to be executed, and another ranking in order of preference to be saved.
Here is the first AV ballot used:
And the results are as follows:
Cowell | 0 |
Blatter | 5 |
Osborne | 2 |
Ramsay | 3 |
Murdoch | 21 |
This from first preferences alone, and under the rules of AV if a candidate has a majority (more than 50% of the votes) then they are declared the winner there and then. As one might expect having already seen the FPTP ballots Murdoch has a clear majority and gets executed without the need for any elimination of candidates or redistribution of votes.
It is worth noting though that the 1st preferences in the AV ballots are similar, but not identical to the FPTP votes. Murdoch and Blatter have one more vote each and Cowell and Osborne one fewer each.
The X vote that you give in FPTP doesn’t always go to your first preference, especially when you are in a constituency where there is a lot of pressure to vote tactically instead of preferentially. There are other reasons as well. Choosing one candidate and ignoring the others is often more difficult to do than to rank in order of preference.
Ok, so this result has showed that AV does not compromise things for MPs who have a majority. They will continue to win their seats under AV and rightly so, as they have backing from more than half of their constituents. In a sense this is all that matters regarding the outcome of this particular vote. Murdoch dies, and the other four live and the majority are happy with that outcome in both cases.
WITH AV WE CAN ASK OTHER QUESTIONS THAT FPTP CANNOT ANSWER:
In this case we have a clear winner and any voting system would show Rupert Murdoch as the obvious choice for execution. However, situations are not always going to be that clear. Remember, after looking at the situations I asked “Had Murdoch not been a candidate, who knows what the outcome might have been?” While it’s not possible to tell from looking at the First Past the Post ballots, not even when you factor in the negative votes, the AV ballots make it quite clear.
We can let the same AV ballot with Murdoch removed as a candidate represent a situation where the competition is somewhat closer. So I shall now eliminate Murdoch (as if he had come last in an AV vote without a majority) and transfer his votes over to the second preferences of those who voted for him.
So redistributing the 2nd preferences of the Murdoch votes and adding them to the 1st preference votes of the other candidates the results were as follows:
Cowell 0 + 7 =7
Blatter 5 + 8 = 13
Osborne 2 + 5 = 7
Ramsay 3 + 1 = 4
The first thing to notice is that no candidate has a majority. The total number of votes cast is 31 and so a majority would be 16 votes or more.
Another interesting thing is that although Simon Cowell received no 1st preference votes he was ranked very highly among the 2nd preferences of those who chose Murdoch as their 1st preference.
Since there is no winner, the lowest ranked candidate needs to be eliminated and his votes redistributed according to the preferences of those who voted for him. In this case it’s Gordon Ramsay.
Note that at this stage the redistributed vote could either be a 2nd preference or a 3rd preference, depending on whether the voter placed anyone other than Murdoch or Ramsay as their 2nd choice.
So, of the 4 votes that had been previously assigned to Gordon Ramsay, 3 go to Sepp Blatter and 1 goes to Simon Cowell with the resulting totals being:
Blatter 16
Cowell 8
Osborne 7
And so this time Sepp Blatter has a majority and is executed after the last minute withdrawal of Rupert Murdoch as a clear choice of the voters. Thus AV is able to give a clear answer to the question “How would people have voted if Murdoch hadn’t been a candidate?” where FPTP is not.
The AV result is consistent with the modified FPTP ballot putting Murdoch clearly in the lead and Blatter a good way ahead of the others. The AV ballot even yields the same ranking order of the top three candidates, Murdoch, Blatter, Cowell as the modified FPTP ballot while Osborne and Ramsay are never in any danger of being executed in either case.
But AV is a significantly further improvement on FPTP than the simple modification looked at earlier. Rather than just catering for the extreme ends of a voter’s preference spectrum it allows the voter to compare the relative merits of all candidates and express his or her full view. In the situation without Murdoch, people who voted for Ramsay, haven’t wasted their vote because they got to make a contribution to the eventual decision between executing, Blatter, Cowell or Osborne.
On the other hand the meaning of putting Blatter first isn’t compromised, as Blatter with his healthy quota of votes from the previous round had a head start on the others, so the voters for Ramsay didn’t get any more power out of transferring their votes than is their democratic right. But at the same time they weren’t robbed of their democratic right by a voting system that would have forced them to choose between a tactical vote and a wasted vote.
REVERSING THE POLARITY OF THE AV BALLOT:
The fourth and final ballot asked voters to rank the candidates in order of preference for them to be Saved (rather than executed). I did this for two reasons:
The first reason was to see how many voters would submit mirror images with their two AV ballots.
As it happened only 21 out of the 31 pairs of AV ballots submitted were mirror images of each other. This demonstrates that people tend to think slightly differently when considering a positive and a negative vote. How much you don’t want one person to win an election is just as important as how much you do want another to win, but if you are concentrating on the positive or the negative end according to the ballot’s question you may find that you contradict yourself between one and the other. The more candidates you have the more diversity there would be.
FPTP can only really cater for this when there are only 2 candidates. With just 2 candidates, a vote for one is a vote against the other and vice versa and all of the pairs of FPTP ballots would be mirrored. With 5 candidates even the AV ballots aren’t entirely mirrored which is a big indicator of how important it is to represent voters’ views as fully as possible.
The second reason was to demonstrate that AV allows voters to vote both positively and negatively on the same ballot and to a fairly sophisticated extent. If this is the case, then despite the slight variations among the non-mirrored ballots, one should be able to ignore the first AV ballot and still execute Murdoch with a convincing majority. One would then be able to ask what might have happened if Murdoch had been withdrawn and see if Blatter is still the executed candidate. Let’s have a look at what actually happened then.
Here is the 2nd AV ballot used:
It is identical to the previous one, just with the word EXECUTED replaced with the word SAVED. It is still for an election to execute one candidate and save all of the others and the participants were clearly informed of this.
To decide whom to execute from these reversed ballots, the first preferences of candidates to be saved was counted. If only one candidate has less than an equal share of the vote (20% in the first round between the five candidates) he is executed. If not, the candidate with the most 1st Preference votes to be saved is eliminated and as many further rounds as are necessary ensue.
So here are the results:
First round:
Osborne 16 (52%)
Ramsey 9 (29%)
Cowell 6 (19%)
Blatter 0 (0%)
Murdoch 0 (0%)
Already it’s looking consistent with the previous ballots.
There is no clear loser so George Osborne is Saved and Eliminated and the votes are redistributed to the 2nd preferences of those who voted to save him.
This time we’re looking for only one candidate to have less than 25% of the votes to be saved.
Second round:
Ramsey 9 + 11 = 20 (65%)
Cowell 6 + 3 = 9 (29%)
Blatter 0 + 1 = 1 (3%)
Murdoch 0 + 1 (3%)
So now Ramsey is the next favourite to be saved and because Blatter and Murdoch are still neck and neck at the bottom we must Save (eliminate) Gordon Ramsay and proceed to another round.
This time we’re looking for just one candidate to have less than a third (let’s call it 33%) of the votes to be saved and it’s the votes for Gordon Ramsay that are redistributed among the remaining three candidates.
Third Round:
Cowell 9 + 13 = 22 (71%)
Blatter 1 + 6 = 7 (23%)
Murdoch 1 + 1 = 2 (6%)
So although we can see an inevitable conclusion ahead of us, we don’t have a definite loser as both Blatter and Murdoch have failed to gain 1/3 of the vote. This means in the interest of fairness we must save Simon Cowell and distribute his votes to see out a head to head contest between Sepp Blatter and Rupert Murdoch.
Final Round:
With only 2 candidates remaining in this round, it is not necessary to worry about who voted for whom in which order. All that is necessary is to look at each ballot and see who is ranked lowest out of the two (unless the voter declined to give a preference to either of them, as is their right).
In the case where one is given a preference and the other is not, that is viewed as a vote for the ranked candidate to be saved. In other words there is no difference between leaving one candidate blank while ranking the other four, and ranking the remaining candidate 5th.
After redistributing the votes that were previously assigned to Simon Cowell the count is as follows:
Sepp Blatter 26 (84%)
Rupert Murdoch 5 (16%)
And so even by asking voter to vote negatively, AV is able to produce a result that is not only convincing in its own right but in full agreement with the positive AV ballot, and with the FPTP ballots. Murdoch is executed once more.
It is also clear without doing any recounting that Sepp Blatter would have been executed had Murdoch been withdrawn as a candidate, because Murdoch had no 1st preference votes to be redistributed and only one 2nd preference, not enough to bring Blatter anywhere close to Simon Cowell’s share of the vote.
CONCLUSION:
So each of the four Ballots have pointed towards a clear mandate for Rupert Murdoch to be executed ahead of the remaining four Candidates. This shows that the added complexity of AV does not compromise the position of a candidate with strong support who clearly deserves to win. This should alleviate the fear that a preferential system detracts from the conviction of a clear victory that one is accustomed to under First Past the Post.
We have also seen that First Past the Post on its own does allow a voter to convey any more information than which candidate they most wish to choose. From the point of view of the FPTP ballot each voter would like one candidate to win and would be equally unhappy with any other winner regardless of who it may be. This doesn’t matter so much in the case of the experiment because there was a winner with a clear majority. But when the winner does not have a majority you have a problem and you need to be able to ask questions.
One question you need to ask is are there more who are happy with the winning candidate than are unhappy to the same degree? The modification of FPTP to include one negative vote and one positive vote to cover each voter’s most extreme preferences would show this. After this modification the candidate with the highest total would be the one with the clearest mandate from the people.
However this modification is fairly crude. It isn’t perfectly equipped to deal with closer encounters where there are 2 candidates dividing the positive votes and the negative votes between them. And when there are many more than 5 candidates all the less extreme candidates get left out altogether and we simply don’t have enough information to distinguish between them.
That is where AV comes in. Ranking the candidates in order of preference from top to bottom allows us to extend the simple modification of First Past the Post to one sophisticated enough to be worthy of conveying a voter’s holistic view. We saw that AV produces the same outcome where there is a clear majority. We also saw that AV is capable of answering the question “What if Rupert Murdoch had been withdrawn as a candidate?” which also served to demonstrate how AV comfortably deals with close contests taking into account all of the voters views to find the result that suits the most people.
Finally, we saw that you can even reverse the polarity of the AV ballot asking voters to vote negatively saving a candidate from execution being analogous to keeping a parliamentary candidate from office. The result was the same. Murdoch is still executed. And even the will of the voters regarding the order of the remaining candidates remained the same. This serves to demonstrate that an AV ballot can cover the full range of a voter’s views from positive to negative because even when you run it backwards it still works out to be consistent with what we know from First Past the Post and with the modified version, and of course with the standard positive voting in AV.
I hope I have managed to demonstrate with this experiment that the Alternative Vote gives voters the most say, allowing them to express their full view, without compromising the chances of a deserving candidate to win an election.
ACKNOWLEDGEMENTS:
I'd now like to thank people who helped me with this by participating in this experiment. The reason I haven't included 31 names is that some did not give their full name and others asked to remain anonymous.
Luke Curner, Chris Read, Hywel Davies, Ben Cadwallader, Richard Shaw, Samantha Best, Laura Blyth, Kenny Aisling, Gavin McLaughlin, Amy Sellers, Angela Upton, Lydia White, Mary Wright, Rachel Speed, Kirsty Bagnall, Nicola Beckett, Katie Sharratt, Amy Nathan, Penny Homer, Manisha Gandhi, Dusty La Rue and Gareth Colwell.
The rest of you, you know who you are, thank you too!
Special thanks to Mevan Babakar for spreading the word and recruiting many of the participants for me.
Say Yes to AV!
No comments:
Post a Comment