Monday, 21 February 2011

Arguments for Alternative Vote

Having covered the arguments against Alternative Vote, I should probably also cover its advantages. Again, because it's topical, I'll also compare Alternative Vote (AV) directly against the First Past The Post (FPTP) system, and list other voting systems that share these advantages.

There are no ideal voting systems - every voting system is good for some things and bad for others. When choosing a voting system, you need to consider the factors that are important to you, and pick a system that exemplifies those factors, while only being bad at things that don't matter so much to you.

There are also some common false arguments for Alternative Vote

AV balances core support and broad appeal

The candidate elected under AV requires sufficient broad appeal to gain transferred votes from eliminated opponents, but also sufficient core support to avoid being eliminated in an early round.

Voting systems which only require broad appeal have a tendency to massively benefit 'centrist' candidates even if they are no-one's first choice, which can harm the standard of political debate. Conversely, voting systems which only require core support can be extremely divisive and exacerbate tensions between communities.

Candidates with strong core support and very little broad appeal will find it very difficult to win AV elections. An "anyone but X" trend in the electorate is therefore magnified in AV, making it easier to remove unpopular politicians and parties from office.

Depending on your point of view, this may not be a good thing, which is why there's a very similar entry in the "arguments against" list.

Other voting systems

Borda or Range Voting, and to a lesser extent Approval Voting, also share this property, although the increased possibilities for tactical voting in those systems may make AV stronger in this regard.

Proportional Representation systems such as Single Transferable Vote balance support levels in a different way. Nevertheless, Single Transferable Vote (unlike the List systems) does provide some bonus to candidates who can appeal to a broad electorate while also electing candidates in rough proportion to their core support.

If extreme tension between disparate communities is a problem, using Proportional Representation where possible is strongly recommended. Single-seat systems such as AV are not sufficient for this problem (they may be suitable for electing a single President, but not for a Parliament).

Compared with FPTP

FPTP formally elects candidates based solely on their core support. There may be some element of broad appeal considered due to tactical voting from the voters in the core support of third-party candidates, but this is considerably weaker than in AV.

This is essentially the major point of using Alternative Vote, so it's unsurprising that it does better than other voting systems. If balancing core and broad support is your aim (and a proportional representation is either unwanted, not an option, or inappropriate for other reasons) then AV is ideal.

AV allows independent candidates

AV, like most voting systems, does not excessively penalise independent candidates. Party affiliations are not required to stand, and a lack of party affiliation only harms a candidate in terms of the campaigning resources and volunteers they have access to, not in terms of a penalty built in to the electoral system.

Other voting systems

List systems of proportional representation heavily penalise independent candidates as a consequence of the way the votes are counted. Most other voting systems treat independent candidates and party candidates identically.

AV does have a particular advantage beyond most other voting systems in that a successful independent candidate may find it easier to beat an incumbent, or maintain their position, provided their core support is strong enough, by getting transfers from sympathetic supporters of party candidates.

Compared with FPTP

AV, by allowing the collection of transfer votes, is generally better for independent candidates who amass a strong core vote - thousands instead of the tens or hundreds most independents get - and are well-placed to pick up transfers from the party candidates they get ahead of.

Such independents are very rare, under any electoral system, but have an even harder task under FPTP than under AV. The same applies to mainstream minor party candidates (e.g. UKIP or Greens)

AV is easy to count

AV is one of the most straightforward systems to count. Each vote is only being counted in a single position at a time, and only when votes are transferred do the preferences need to be re-examined.

This straightforwardness makes it possible to count AV by hand, and to use it for elections at meetings even without pre-prepared ballot papers.

Other voting systems like this

First Past The Post, List systems and Approval Voting are also straightforward to count. Pairwise voting is straightforward to count in the context of shows of hands at a meeting, but is less straightforward when using ballot papers for a secret ballot.

Single Transferable Vote is no harder to count than AV in terms of the mechanics of counting, but depending on the variant of STV in use, may require complicated calculations.

Other voting systems tend to require repeated examination of ballot papers, and so are difficult to count by hand.

Compared with FPTP

AV is marginally slower to count than FPTP, but not significantly more difficult. For an election with 4-5 credible candidates, in the worst case, an AV election will probably take around twice as long to count as an FPTP election.

Numerous non-credible candidates with very low vote totals add significantly to the number of rounds in the AV count, but provided the vote bundles were left sorted after the first counting round, this will not significantly increase the count time.

AV discourages tactical voting

AV strongly discourages tactical voting. While there are theoretical situations in which tactical voting is possible and practical, it is extremely difficult in general to identify these situations in advance.

The rare situations in which it is possible are covered in the 'arguments against' list.

Other voting systems

Condorcet is the only other single-seat system that discourages tactical voting, and it does this considerably more effectively than AV.

Proportional Representation systems such as Single Transferable Vote or List systems also largely avoid tactical voting.

Compared with FPTP

FPTP strongly encourages voters for third party candidates to vote tactically for one of the two major candidates in a constituency instead, and also gives an incentive to the two major candidates to actively encourage tactical voting.

Conversely, AV makes tactical voting generally impractical, and the forms that exist are not generally ones that political parties would encourage.

False arguments for AV

There are some arguments for AV which are either entirely untrue, or depend on the context in which AV is being used.

AV ensures that a candidate has the support of the majority of voters

Not really. This is usually implied from the fact that a candidate needs 50% (plus one) of the votes remaining at the current stage of the count to be elected.

This is not usually the same as the number of votes cast, unless the rules require every voter to use all their preferences. In optional preference AV (the form proposed for the UK), voters can use as many or as few preferences as they like. This means that in the late stages of the count, a vote without a complete set of preferences might not have any remaining candidates to be transferred to, because the voter couldn't decide between what were to them all unsatisfactory options.

If this happens, the vote plays no further part in the count, and the target number of votes (also called the quota) is reduced slightly to compensate.

In England, between 60% and 90% of constituencies - depending on how the election went - would probably have the winner receiving at least 50% of the original votes in the final round. The vast majority of the remainder would be in the 48%-50% range, with below 45% being exceptionally rare. (I don't have sufficiently reliable polling data for UK countries other than England to make a guess at those, but having four or more major parties rather than just three would generally make wins below 50% of the original votes a little more likely)

[Edit 11 April 2011: The Swingometer now gives basic statistics on the proportion of seats where fewer than 50% of the original votes are held by the winner, and the "worst-case" percentage needed for that scenario.]

However, a couple of percentage points here or there is not the main reason this argument is false. The major reason is that it relies on the interpretation "giving someone a preference - any preference - means support".

This is not necessarily true, of course. In a six candidate election, an unlucky voter who sees their four favourite candidates knocked out may still have an opinion over which of the other two is least bad, but to call their choice 'support' for their fifth may be excessive.

Another possible definition of "support" is "preferred to the alternatives", but while the elected candidate is guaranteed to be preferred by the electorate to any remaining candidates, it is possible that there is another candidate, eliminated at an earlier round, whom a majority of the electorate would rather have seen win.

A small election to demonstrate
Number of ballot papersFirst preferenceSecond preferenceThird preference
10 votersAliceCathBob
9 votersBobCathAlice
2 votersCathAliceBob

In this election, no candidate has half the votes, so Cath, with the fewest votes, is eliminated from the election, and her two votes redistributed to the remaining candidates. They both go to Alice, so Alice wins. But 11 of the 21 voters - more than half - would in a straight choice between Alice and Cath have picked Cath.

If you want a voting system where the winner is guaranteed to have majority support over any other candidate (should such a winner exist, anyway - it's not guaranteed), then the Condorcet system should be used, rather than AV.

On the other hand, if you want - as above - to balance both core support and broad appeal, this property of AV is an advantage. Should Cath really be elected in the election above when hardly anyone places her as a first preference? (The answer is "maybe": it depends on what you consider important for an election system)

AV generally requires the support of a larger proportion of the electorate than FPTP. But it doesn't always require majority support. (Proportional Representation advocates, at this point, would point out that requiring majority support is not a universal good thing anyway - allowing a party with 10% support to get 10% of the seats is the whole point of PR systems such as List or Single Transferable Vote)

AV prevents "safe seats" compared with FPTP

It's really quite hard to assess how much impact AV has on this. Any AV election result could, after all, also be achieved under FPTP in theory given sufficient tactical voting (or candidates withdrawing - former MP Martin Bell's win in Tatton, for instance). In the UK, the third of MPs who currently get half the FPTP votes already are extremely likely to keep their seats under AV.

While the power of AV to avoid splitting an "anyone but X" vote is extremely strong and important, the majority of safe seats in the UK are safe because the incumbent has massive levels of local support, not because the opposition is split.

Both Labour and Conservative supporters have worried that the supporters of the other party would join forces with Lib Dem supporters under AV to lock them out of government. In practice, this is very unlikely to actually happen, except where an "anyone but X" sentiment is widespread across the country (and in that case, that party is probably doomed under FPTP as well).

AV will make some seats less safe - semi-marginal seats are likely to become fully marginal, and semi-safe seats might become semi-marginal and vulnerable to an exceptional election campaign or extreme unpopularity of the incumbent, but the effect will be fairly subtle. Conversely, depending on how votes tend to transfer between parties, some currently marginal seats may become relatively safe.

There is also the possibility in the long term of the lack of need to vote tactically under AV opening some very safe seats up to a challenge from minor parties on the same 'side' of the political spectrum as the incumbent (for instance UKIP in a safe Conservative seat), but this is likely to take several elections to become apparent.

So, while AV will reduce the safety of some seats compared with FPTP, the effect is not likely to be spectacular.

To largely avoid safe seats, a proportional representation system is needed - either Single Transferable Vote, or an open List system. Even then, successful parties can usually get a few seats which are safe against all but the worst national landslide victories.

[Edit, 14 April: More detail in the new post Alternative vote and safe seats. The swingometer has also been updated so you can look at scenarios yourself.