Since last week's voting flowchart I've been pondering whether tactical voting is sensible under AV. I don't think it is.

Under FPTP, tactical voting is not at all uncommon. If your favourite candidate isn't going to be one of the top two, your vote could be used to support whichever of the top two candidates you prefer.

Under AV, however, a tactical vote would be sensible only under the following circumstances:

  1. It's a close three-way contest.
  2. These three candidates lie along a clear spectrum (call it left, centre and right).
  3. Your favourite candidate is either the left or the right candidate.
  4. In the penultimate round the outcome will be as follows: the opposite candidate in the lead, then your candidate second, and the centre candidate third. And the next-preference votes of the centre candidate won't cause your candidate to win in the final round (in other words, the centre candidate is not much closer to your candidate than he/she is to the opposite candidate).

Now, none of this is particularly unrealistic, except point 4, which assumes a totally unrealistic level of knowledge. Having been in a three-way marginal constituency in the 2010 election, I can confidently predict that it will be pretty much impossible to predict the order of the candidates. All three of the main candidates were predicting different orders for the final outcome, based on opinion polls, historical voting patterns in the constituency, and the national situation (I think all three were wrong in their predictions!) and the more impartial onlookers were similarly divided. Tactical voting requires you to guess how everyone else will vote, and if that is not possible then neither is effective tactical voting.

But what if (for some reason) I think it is likely that all of the above conditions will be satisfied? I could either cast a "sincere" vote (remember that under AV you get to rank the candidates in order of preference):

  1. My candidate
  2. The centre candidate
  3. The opposite candidate

or I could cast a "tactical" vote, effectively deserting my candidate, and hoping that the centre candidate will defeat the opposite candidate:

  1. The centre candidate
  2. My candidate
  3. The opposite candidate

Then what might happen?

  • Sincere vote, my candidate wins. Hoorah!
  • Sincere vote, the opposite candidate wins, but a tactical vote would have given victory to the centre candidate. D'oh!
  • Sincere vote, the opposite candidate wins, but a tactical vote would have made no difference.
  • Sincere vote, the centre candidate wins.
  • Tactical vote, the centre candidate wins against the opposite candidate. Hoorah!
  • Tactical vote, the centre candidate wins against my candidate. D'oh!
  • Tactical vote, the opposite candidate wins. Slight D'oh! (people looking at the counts of the votes will assume my candidate had less support than he/she really had)
  • Tactical vote, my candidate wins. Hoorah! (but slight D'oh! as above)

So really it would be your call, based on how confident you are in your predictions for the order of the three candidates in the penultimate round, based on whether other supporters of your candidate will also vote tactically (your vote alone is unlikely to swing it), and based on how much weight you give to the various "Hoorah!" and "D'oh!" exclamations above. But once again, I seriously doubt that it would be possible to predict the outcomes sufficiently well to be able to cast a confident tactical vote.

So I stand by the flowchart: in AV you need simply to rank the candidates in order of preference then write these preferences on the ballot paper.