Honesty and Strategy in real-world voters

Evidence that score voters will be both honest and strategic

There is plenty of evidence that a lot of real voters use strategic (and "dishonest") voting. There is also plenty of evidence that a lot of real voters vote "honestly" even if they know it is unstrategic (i.e. "stupid"). They do both.

A lot of people have the idea that, in a Score Voting future, voters will "get wise" and start voting strategically, not honestly, i.e. vote in approval-voting style. Therefore (the argument continues) why not just use approval voting, not score voting? Our response: in fact, we believe that the human impulse for honesty is so powerful (especially in synergy with score voting, which appears to inspire honesty to a surprisingly large degree) that even after 100 years of future experience (and even though strategic voting is fairly simple in score voting) voters will still, in large numbers and in a large fraction, choose not to be fully strategic, and to include a large amount of honesty, in their score vote. If so, then score voting is a better idea than approval voting, because it can take advantage of that voter honesty. Whyever do we believe that? Experimental facts about human behavior, that is why. See below.

Strategic (but dishonest) voters:

  1. In the USA's 2000 presidential election, voters who thought Nader best were urged not to be strategically stupid and instead to vote (dishonestly) for either Gore or Bush, and NES polls showed that over 90% of the voters who thought Nader was best in fact did vote for somebody else. Approximately the same thing also happened (polls also showed, same election) for Buchanan.
  2. Another, different sort, of example – from p.65 of Lakeman & Lambert's book: a 1950 Gallup poll showed 38% of British voters wanted to vote Liberal but only 9% did.
  3. In our exit polls using score voting, the huge discrepancy versus the official election results with plurality voting strongly suggests that a tremendous amount of strategic distortion happens. (Nader would have got about 66 times more votes relative to Kerry & Bush with score voting.)
  4. In Spain & France, the Socialists are far more powerful than in the USA. They won control in Spain in 2004; the smart money is supposedly on them to win France in 2007 too [later note: they came second in France]; but in the USA the Socialist candidates usually cannot even get on the ballot and get about 0.003% of the vote. Obviously, the US and French voters do not differ in their inherent socialisticness, by a factor of 10,000. (Maybe 2 or 3, but not 10,000.) That is a giant effect and it is caused by strategic voting.

Honest (but "stupid") voters:

  1. About 1/3 of US voters register Independent (and 1/3 Democrat and 1/3 Republican). Why? Obviously if you do that, then you cannot vote in a primary (in most states) and hence clearly lose voting power. This is obviously an unstrategic action, yet about of 1/3 of voters take it despite well over 100 years of experience showing it is unquestionably and obviously strategically stupid.
  2. In our exit polls using score voting, the vast majority of voters do not submit an "approval+blanks style" score vote, i.e. their vote is manifestly less than 100% strategic. It is not clear just how honest and how strategic the voters in our score voting exit polls (and they were instructed to vote "as they would if this were the real election") were, but clearly a great deal of honesty was involved. Specifically in our 2004 US Presidential exit poll, roughly speaking the situation appears to have been approximated by about 50% honesty and 50% strategy. (Actually over 75% of the voters cast manifestly unstrategic votes, but probably a lot of votes contained both honesty and strategy mixed inside.) In our small exit poll in the 2006 Texas Governor race, all 36 of the pollees – 100% of them – voted nonstrategically, i.e. not in approval-voting style. Not a single one went with a fully strategic vote although two went with all-10 and -X votes, which perhaps could be called strategic. (The difference presumably was, the 2004 race was 2-party dominated to a massive degree so strategy paid off more than in the 2006 Texas race, which had several credible independent and minor-party candidates.) You can also see all these range-style polls in which again, manifestly non-strategic "votes" were massively common.

    These findings are massive. Remember, Nader 2000 showed that 90% of US plurality-voting voters chose to be strategic and dishonest when it paid to do so. Our polls show 50-100% of US score voters choose to be honest even when it does not pay. That is a huge behavior swing, certainly among the hugest ever observed in voter psychological studies. Score voting evidently stimulates (and offers the opportunity for) much more voter honesty. (And that has important consequences, e.g. the nursery effect.)
  3. And also, honesty and strategy are inherently less in conflict under score voting, so even if you do want one more than the other, you often do not need to compromise. For example, if a score voter decides, on the 5 least-likely-to-win candidates (whom, let us say, she regards as intermediate in true value between the two most-likely-to-win candidates, whom she gave 0 and 99 scores out of strategy) not to give them all 0s, but instead to give them honest scores like 37 – that costs her almost nothing in terms of lost utility from not being maximally strategic. So she might as well do it. She feels happier about being honest, and the cost is infinitesimal, not enough to outweigh that extra happiness. In other words being honest and unstrategic is actually not stupid – with score voting in many circumstances it is stupid to such an incredibly tiny degree that it is outweighed by your increase in happiness from being honest (and being happy is your goal in life)... the chance your unstrategic decision to rank some no-hope candidate honestly will hurt the election, is usually far, far tinier than 10-100 which means even a tiny subatomic fraction of one molecule of happiness-neurotransmitter, is more reward than you could expect to get from maxxing and minning out these candidates' scores.
  4. Although I am willing to concede that after a lot of experience, voters will become more strategic, these experimental findings are so massive (and the logic and arithmetic last item are valid) that I do not believe that score-voter honesty will ever mostly go away.

Conclusion: expect both honesty and strategy!

Score voting is lots better than approval voting if there is a lot of voter honesty. Score voting is lots better than, e.g. Borda and Condorcet, if there is a lot of strategy. (DH3 pathology) We expect both in the voter pool. It is unrealistic to expect only one.


More about honesty & strategy

Return to main page