Full Preferential Voting (or How I Learned To Stop Worrying And Vote For People I Don’t Like)

This post is about why full preferential voting is good thing, why in an optional preferential voting system you should preference as many candidates as you reasonably can, and why preferencing someone you don’t like is not always a bad thing.

Mostly I am talking here about the system that applies to the lower house in Australian federal and state elections, in which one candidate is elected to represent an electorate. I’ll comment on upper house elections, which involve proportional representation, towards the end.

Basics – how does preferential voting work?

All of the first preferences are counted. If someone had over 50% of the vote they win. If not, the candidate with the least number of votes is excluded and the second preferences on their ballots are counted and their votes reassigned to those candidates. Once again if some one has 50% they win, if not the next candidate is excluded. At each stage the redistribution is based on what voters have filled in on their ballots, the candidates (or parties) have no say. If the next preference on a ballot is for someone already excluded then it just moves on to the next preference for a candidate who is still in the running.

An example: Suppose an election has five candidates A,B,C,D,E and 100 voters. A candidate requires 51 votes to be declared the winner. On first preferences the votes are

A: 39, B:35, C:16, D: 7, E:3

Nobody has 51 so the 3 votes from E are redistributed.

A:39, B: 35+2=37, C: 16+1=17, D:7

Next D is eliminated, and their preferences distributed, any with a second preference for E just move on to the next one

A:39, B:37+7=44, C:17

Now C is eliminated

A: 39+10=49, B: 44+7=51

Note that even though it looks like it is C that pushes B over the line to be the winner, some of the 7 votes B gets at this stage could well be votes that were initially for D and E, so you couldn’t just say that B won because of C’s preferences without looking at the details.

In Federal elections a valid ballot must have preference for all candidates, in some states there is optional preferential which means that votes can exhaust if no further preferences are given, at which point a winner requires 50%+1 of the remaining votes.

Why have preferences?

A first past the post system, where you just take the candidate with the largest vote, is easily seen to be unfair. Two similar candidates can split the vote, for example if A and B have policies that are generally more popular with the electorate than those of C, then with A and B both running the vote might be A:30 B:30 C:40, but if only A and C ran then it might be A:60 C:40. This election is too easily open to manipulation and does not give a fair outcome. It also tends to result in strategic voting, meaning that voters are forced to speculate on who is likely to win and perhaps vote for someone who is not their preferred candidate to get a better outcome. There are numerous alternatives to first past the post, and none are perfect (a perfect system is mathematically impossible), but preferential systems are in general much fairer.

The Two Party Preferred Vote

Often the outcome of an election is expressed in a two party preferred vote, either for a particular electorate or the overall result for all electorates. This is just the result of distributing all preferences until two major party candidates are left. It is a way of viewing the overall outcome, but is not part of the count that determines the winner. In the example above, the winner was not decided until only two candidates were left however this doesn’t have to be the case. In an election with 10 candidates where one of them gets 51% of first preferences they are declared winner with no preferences being distributed. To determine the two party preferred vote you would distribute preferences until only two major party candidates remain but this has no effect on the outcome, and just gives an overall view of the popularity of the two major parties given that they have most of the seats. If people voted in more candidates not from the major parties, then this would not be done as it would no longer be a useful measure of anything.

There is also the two candidate preferred count, used to compare the popularity of the top two candidates in an electorate , once again this isn’t the official count to determine the winner, it just gives a away of measuring the margin of the victor over the next most popular candidate. It does not mean it was predetermined that one of those candidates must win.  There is a common misconception that somehow the two party/candidate preferred count gives a built in bias towards the major parties in the electoral system but it is not part of the system, just a way of viewing the outcome, and it mostly involves the major parties because they get most of the votes. In the Federal electorate of Melbourne the two candidate count is between The Greens and Labor, and in the recent Vasse by-election in WA it was between the Liberals and Nationals.

[edited to clarify party/candidate distinction]


Preferences don’t make Labor and Liberal win

Often people complain that their preferences end of with Labor and Liberal and so they always win. Nothing about preferential voting favours Labor and Liberal. There are other factors such as single member electorates which do have an effect, but the major parties do not win because of the preferences of people who vote for other parties – they win because most people give first preferences to one of them. In many cases one of them gets over 50% of first preferences so no preferences are distributed. In other cases the preferences of people who voted for others are used to decide between the two major parties, but in this situation the winner is guaranteed to be one of the major because most people voted for them, when your preference gets distributed to them they have already won, your vote isn’t making that happen, but it is helping to decide which one wins. Once again they haven’t already won because of the system, it was because most people voted for them.

For example, consider my electorate of Hindmarsh. At the last Federal election, the results were (in percentages)

Greens – 8.84
Liberal – 46.17
PUP      – 2.47
DLP     – 0.88
Katter  – 0.63
Labor   – 37.95
Family First – 3.05

The distribution of preferences can be seen here http://results.aec.gov.au/17496/Website/HouseDivisionDop-17496-185.htm but it isn’t hard to see that one of Labor or Liberal is going to win and preferences only serve to decide which one because neither has over 50% on first count. As it turns out most people who didn’t vote for a major preferred Labor, but not enough to get them over 50%, in this case all preferences had to be distributed before the Liberal candidate gets over the line.

Why you should allocate as many preferences as possible

In a Federal election, and some states, you have to allocate preferences to all candidates, however in some states (and other elections such as local council) there is optional preferential voting, which means you can number as many as you like. In such a system you should still preference all candidates if possible, even the ones you don’t like. The reason is simple, not allocating a preference to someone does nothing to stop them winning, only a preference for another candidate can actively work against them. If at least one of the candidates you like is popular with others as well, then they won’t be eliminated and your vote will stay with them until either they win or lose out to the eventual winner.

Suppose though that none of your favoured candidates survive to the decisive count. This has happened because of how everyone else has voted, there is nothing you can do about it, this is democracy. However if your vote exhausts you have no further say over the election, so this only makes sense if you really genuinely don’t care who wins out of the remaining candidates. Let’s say an election has a great candidate, a mediocre candidate, a bad candidate and an awful candidate. If you vote only for the great candidate, you are potentially letting a bad or awful candidate beat the mediocre one. Many people instead would vote 1 great, 2 mediocre and that’s it. This is still not a good idea though, because this means that if they get eliminated then you have no say between bad and awful. If you really think the awful candidate is much worse than the bad one then you must vote 3 bad.

It might not feel right giving a preference to someone you don’t like but the only way your vote can count for them is if all your preferred candidates have been eliminated (because not enough other people liked them, which you can do nothing about), and you are helping them to beat someone you like even less.

There is no need to guess in advance which candidates will be popular, your vote only transfers when someone is eliminated, if the great or mediocre candidates are popular enough (or if the awful one gets eliminated early) then your vote will never transfer to the bad candidate.
On the other hand what is the impact of not preferencing, i.e. let’s say you vote 1 great only, instead of allocating 3 preferences?

  • If Great wins – then it makes no difference, your extra preferences would never have been counted.
  • If Mediocre wins – then it makes no difference, either your vote would have stayed with Great until they lose the final stage, or would have transferred to Mediocre and helped them win. Either way your vote couldn’t have changed the outcome.
  • If Bad wins – If they beat Great or Awful at the last count, there is no difference, but if they beat Mediocre at the last count then your vote could have helped stop them. That is, they are more likely to win in the scenario when you use fewer preferences.
  • If Awful wins – Then if at the last stage they beat either Mediocre or Bad then you may have been able to stop them, but you can’t by voting only 1. So they are more likely to win if you don’t preference.

So failing to allocate preferences can only improve the chances of candidates you don’t like, whereas preferencing candidate you particularly want to win can only help them to win against candidates you like even less, which is a good thing.

Thus the rational choice is to preference as many candidates as possible, so you should only leave the very least favoured one blank, or any which you know nothing about (but even better is to make sure you do find out who they are). Some of your preferences will be for candidates that seem highly unlikely to win, but that doesn’t matter, if eliminated your vote just moves on to the next candidate.

Just remember that in a full preferential election you should not leave any blank.

A Disclaimer
I have glossed over the implications of Arrow’s Theorem, but with good reason. A perfect voting system is mathematically impossible. This means that it is possible to devise situations where ordering your preferences in a way which does not reflect your actual preferences can achieve a more desirable outcome. For practical purposes there is not much you can do, since manipulating the outcome requires a very specific set of circumstances and good information on how everybody else will vote. It only really comes into play when there are 3 or more candidates with a reasonable chance at winning, so it doesn’t come up so often in the sort of elections I’m talking about. As such I would say that preferencing as many candidates as you reasonably can in the order which you like them is an optimal strategy, but not perfect, in that for most possible configurations of votes it will achieve the best possible outcome however it is not always guaranteed.

The Senate and other Upper Houses
For proportional representation elections with preferential voting the principal is basically the same despite the extra complications. Some states (and quite possibly in the Senate in the future) have optional preferencing. Once again it is best to preference as many candidates as possible, including ones you don’t like if it means you get to put them ahead of ones you like even less. The main difference is that there could be a very large number of candidates so it may well be the case that there are many you know nothing about so it is reasonable to not allocate preferences to them.

The basic message is that by preferencing as many candidates as possible then you equip your vote with as much information as possible and it has the potential to achieve more for you. Putting a preference next to someone isn’t saying you approve of them, it only says you prefer them to whoever is lower in your preferences. Letting a ballot exhaust cannot achieve a better outcome than if you had kept preferencing, only the same or worse, and only makes sense if you genuinely have no preference at all between the remaining candidates. Most of all preferential voting is good thing and should be defended against those who take simplistic views like “it helps the major parties”, “it’s a scam”, “it should be one person one vote” because they are demonstrably wrong.