Mark – my words

Monthly Archives: March 2011

How much is my AV vote worth?

AV skeptics and campaigners for a “No” vote in the May 5th referendum frequently assert that because voters rank the available candidates, there must be some associated reduction in the value of a vote as it gets reallocated from a voter’s first preference, to their second, third and so on. This is wrong-headed nonsense, and it arises from a fundamental misapprehension of the purpose and operation of the alternative vote system. Reallocated votes should and must be counted at full face value, in order to preserve the principle of “one man, one vote.”

The first point that needs making is that AV is a form of run-off voting. In attempting to elect the candidate who is most acceptable to most voters, there may be several rounds in the contest, with the field of candidates being narrowed down each time by excluding the least popular choice. Each voter gets a vote in each round, and as the choices narrow, so the selection focusses onto the candidate who can command a majority of all the votes cast in the voting round. AV does the run-off “instantly”, rather than as a series of separate ballots, but the underlying principle is the same. The compromises involved in doing an instant run-off1 rather than a multiple ballot runoff make the system both quicker and more economical – clear advantages when what is wanted is a quick, simple and decisive election.

No-one would seriously suggest that in a multiple ballot run-off election, certain individual voters’ votes in successive rounds should count less if the voter had been prevented from voting for the same candidate as in an earlier round due to that candidate’s name no longer appearing on the ballot paper. That would be a nonsense. It would negate the whole point of the run-off. If a candidate isn’t standing, you can’t vote for them. Yet that devaluing of votes is exactly what is being suggested should be the procedure in an instant run-off, AV election.

Here’s another reason to reject the idea. As I’ve described elsewhere, under FPTP no-one questions how passionately people vote. One voter’s deeply sincere and committed “X” is never said to be somehow more valid than another’s hesitant, doubtful, perhaps even reluctant mark for the candidate she hopes will prevent her personal nightmare MP getting elected.

Thus it seems unreasonable to insist that the degree of enthusiasm or sincerity (or whatever) that informs a voter’s ranking of candidates in AV makes any particular position in the ranking more or less valuable as a vote. Rather, the places in the ranking merely serve to indicate which, out of a particular subset of candidates, the voter wishes their vote to count for. The ranking is more akin to an instruction to the returning officer than it is any absolute measure of the amount of favour the voter assigns to the respective candidates.

Each voter in an AV election has one and only one vote to award. At any given stage in the AV instant run-off, that vote can count for only one candidate. A lower-ranked preference counts if and only if all of that voter’s higher ranked preferences have been eliminated from the race and no overall winner has yet emerged. Clearly it would be impossible in those circumstances for their vote to count at all unless it were awarded to their next highest ranked candidate. Equally, it would be entirely inappropriate to take into account the lower preferences of voters whose first preference candidate is still in the running.

From the perspective of the voter who submits their more-or-less carefully considered ranking of the candidates, the descending order of preference clearly does indicate a result that they will find personally less and less desirable. If we then additionally penalise the voter by making their lower preference vote count less than a full vote towards the determination of the eventual winner, we add insult to injury.

There are still more reasons to insist on full value for all ranked votes. If the value of a re-allocated vote were to be factored down, according to its position in the voter’s ranking, it would cause all sorts of problems for the returning officer2, and the contest might prove unwinnable by any candidate, because the winning criterion – “more votes than all other candidates put together” – could turn out to be unattainable were some votes to count significantly less than others.

Finally, devaluing lower preference choices comes perilously close to saying that some people’s votes are less valuable than others. That’s a slippery slope that ends somewhere in the direction of the gas chambers.

1 In a multiple ballot run-off, voters have more information. They know which candidate or candidates have been eliminated. This, it is argued, may cause them to choose a different candidate to vote for in the next round. By forcing voters to rank all candidates from the outset, AV denies them this additional flexibility.

2 What should the fractional value of each transferred vote be exactly? How should the reconciliation between votes cast and votes counted for each candidate proceed when some votes are fractional? Should the turnout or shares of the vote be reported using whole or fractional votes? I’m sure there will be quite a few more such issues.


How to vote in an AV election

Faced with a ballot paper that allows you to rank any or all of the candidates in an AV election, how should you award your preferences? Which candidate should you give your 1st preference to, and which 2nd, 3rd, and so on? Should you even bother to mark preferences beyond your first?

Before going any further, there are some ground-rules that you must understand.

  1. You have only one actual vote to award. Your preferences merely tell the returning officer which candidate your vote should be counted for.
  2. The winner of the election will be the candidate who gets more votes than all other candidates put together. This means that there may need to be several rounds of counting, eliminating the lowest-scoring candidate each time, and redistributing votes that had gone to the eliminated candidate, according to those voters’ stated preferences.
  3. Your lower numbered preferences only take effect if and when the candidate(s) you gave higher rankings to have been eliminated from the contest. So if one candidate storms the contest with over half of all votes cast, your second preference will never be looked at. Similarly if your candidate has a decent percentage of first-preference votes, and thus has a good chance of winning if s/he attracts preference votes from less popular, eliminated candidates, then again your second preference may not be looked at.
  4. If all the candidates you marked a preference for get eliminated, your vote will no longer count. It will be like you never voted. There’s nothing to stop you voting for just one candidate, but bear in mind that your vote won’t influence the result if all your preferred candidates get eliminated.

So, given these basics, here’s the strategy I plan to adopt, and I humbly suggest the same strategy makes sense for everyone, irrespective of their political affiliations.

  • I will look at each of the candidates and choose the one I would most like to see winning, let’s call her “A”, and I’ll mark “1” next to her name.
  • Next I will imagine that A gets knocked out. Of the remaining candidates, which would I most like to win?  That’s “B”, and he will get my “2”.
  • Next, imagining that both A and B are eliminated, I choose my 3rd preference, C. And so on.[1]

There may come a point where I cannot choose between the remaining candidates, or I may refuse to endorse any of them in any way. At that point I stop marking preferences.

Through this entire decision-making process, I can vote entirely honestly. I know that my vote can continue to influence the election until the final outcome is decided. One of my favoured candidates may win, or I may end up having backed losers all the way. But I need not think that my vote has been wasted, because it can continue to count until the winner is elected.

1 I suspect there is never any point in marking your absolute lowest preference, the “nth” out of “n” candidates, because the contest will always stop when it comes down to two candidates, and votes for the losing 2nd place in the final round won’t get ever re-allocated.

Busting some No to AV myths

No voter in AV gets more than one vote, nor is any voter’s vote more infuential than any other.  This has been established in legal precedent in the USA.

There is no question of multiple bites of the cherry. All voters’ votes are counted in each round of the AV runoff (unless all candidates they indicated a preference for have been eliminated). If your single first-preference vote turns out to have been for the winning candidate, your vote has nevertheless been counted exactly the same number of times as that of a voter whose preferred candidates have been successively eliminated. And who knows, his final preference may have been the same as your first, and been the one vote that pushed that candidate over the winning line!

Consider a BNP supporter whose second preference is UKIP, and third preference is Conservative. When the BNP candidate is eliminated, this voter becomes to all intents and purposes a UKIP voter, indistinguishable from someone who voted UKIP as their first preference. When the UKIP candidate is eliminated, our voter is supporting the Conservative party. His vote is no less valid because his desire for minority representation has been thwarted.

Some people get very exercised about lower-preference votes being somehow less worthy than first-preference ones. But in this country we’ve never been in the least bit concerned to examine how enthusiastically a voter’s single vote is awarded to a candidate under FPTP. I  would venture to suggest that any given individual’s level of committment to his FPTP vote could be any of: highly positive; cynical, opportunist & tactical; a chore, undertaken out of a sense of duty; a coerced act; an act of despair; and much else besides.

Allowing voters to express any or all of these sentiments on a single ballot paper, from which ultimately only a single vote will count in the eventual result, seems to me to be self-evidently a superior and more empowering form of democracy.

Three arguments for AV and against FPTP

I was challenged on Twitter by @ticobas :

Hi Mark, what in your view are the best three arguments FOR #AV, aka the three biggest problems with #FPTP?

Here is my response.

  1. AV gives MPs more legitimacy by requiring a higher level of support: the AV winner requires “more votes than the other candidates put together“.   FPTP on the other hand  allows your MP to be elected with a minority of votes, when majority of voters might positively disfavour the winner. (Technically, FPTP allows the Condorcet loser to win, AV does not.)
  2. AV Encourages honest voting for truly preferred candidates: voters need have no fear that their vote will be wasted by giving it to a minority candidate; conversely, FPTP  encourages dishonest, tactical voting. (Technically, tactical voting is possible in AV but it is problematic and risky. See this LSE paper.)
  3. AV is a simple evolution of the familiar current FPTP system; it retains the constituency link, it is very similar to vote in (instead of X, you mark 1, 2, 3, …), it is very similar to count and won’t take much longer; it produces arguably equally decisive results nationally – no more likely to result in coalitions than FPTP. But it is a system that is better adapted to our multi-party electoral environment. FPTP really only works when the contest is two-way. See the IPPR report downloadable from here.

I hope this helps. Please feel free to comment – I promise to approve all non-spam comments.

More votes than all the other candidates put together

One of the arguments against the Alternative Vote (AV) system is that its claim, that the winner of an AV election must have over half the votes cast, is false. This objection hinges on the question of which body of votes the winner must get more than half of, and the linked issue of “exhausted votes”.

The actual wording which defines how an AV winner is decided is there for all to read in the Parliamentary Voting System and Constituencies Act 2011. It’s a lovely piece of plain English, and it states: “if one candidate has more votes than the other candidates put together, that candidate is elected.”

Obviously that condition may not exist when the first preference votes are counted. The first round of counting in an AV election is identical to the one and only count that happens in the present, First-Past-The-Post (FPTP) system. We know all too well that FPTP often results in a winner who has substantially fewer votes than “the other candidates put together”. In fact, fully two-thirds of MPs in the present House of Commons were elected with a clear minority of the votes cast. AV improves on FPTP by insisting on a greater level of endorsement of the winning candidate.

To that end, AV is a form of runoff voting, and may require several rounds of counting in order for a winner to emerge. So long as no candidate “has more votes than the other remaining candidates put together”, the lowest-scoring candidate — the one with the fewest votes — is effectively eliminated from the contest. As the Act says,

the candidate with the fewest votes is eliminated and that candidate’s votes shall be dealt with as follows—
(a) each vote cast by a voter who also ranked one or more of the remaining candidates shall be reallocated to that remaining candidate or (as the case may be) to the one that the voter ranked highest;
(b) any votes not reallocated shall play no further part in the counting.

Case (b) applies where all a voter’s preferred candidates have been eliminated. Their ballot “shall play no further part in the counting”. These are referred to as “exhausted votes”.

The objection I hear from No supporters says that exhausted votes should be included in the total number of votes which the winner has to have more than half of. An example may make this clearer. The table shows an AV election over 4 rounds. After each round, for the sake of simplicity, the votes of the lowest scoring candidate are divided equally among the remaining candidates and the exhausted pile.

Candidate Round 1 Round 2 Round 3 Round 4
A 1000 1020 1075 1300
B 800 820 875 1100
C 600 620 675 (eliminated)
D 200 220 (eliminated)
E 100 (eliminated)
Exhausted 0 20 75 300
Total 2700 2700 2700 2700

At round 4 only two candidates remain, and A is declared the winner, having “more votes than the other candidates put together” – because there’s only one other candidate remaining. However, candidate A, with 1300 votes, clearly does not have more than half the 2700 ballots that were turned in at the election, even though he has more than the 1100 votes that B has accrued.

Does this matter?  The rules, established in the Act, are explicit: exhausted votes “play no further part in the counting”. We may assume that voters whose votes became exhausted were unable or unwilling to express a preference between the remaining candidates. It therefore seems fair that their votes should lapse, as though they had not voted at all. Alternatively you could view it as being no different to what happens to the votes of voters in a FPTP election whose candidate does not win. Their preferred candidate or candidates simply lost. They certainly didn’t get enough votes to win.

There’s a third possibility: exhausted votes should continue to be counted, but counted for a special, virtual candidate called “none of the above”. I plan to explore that concept in a later posting, but as it’s not a possibility that’s entertained by the current AV legislation, arguing about it is a rather sterile activity in the face of the decision we’re being asked to take on the 5th of May.

Meanwhile, is the objection valid? Should the entire concept of AV succeed or fail on the basis that its claim that the winner of an AV election must get over half the votes can be shown to be false, albeit only in some circumstances and under a set of assumptions that don’t match the relevant legislation?

To my mind, the objection doesn’t invalidate AV at all, but of course you may have a different opinion.