Although our votes are required to carry equal weight, they don't. It's time for real democracy. Join us!

Strategic Voting is any time you vote in a way that is not in line with your sincere preferences, in order to obtain a better outcome. Strategic voting can come in several forms.

Favorite Betrayal is when you vote against the candidate you like the most. Dude. You betrayed your favorite candidate? That sucks! Some voting systems, including the one we use today, as well as IRV, incentivize some voters to vote against their true favorite in order to prevent the election of a worst possible candidate (the greater evil). In Plurality voting, this known as the Spoiler Effect. The video below shows how this problem manifests in IRV:

In Score Runoff Voting, unlike IRV, for your honest vote of maximum favorite support to create your worst outcome, your vote would have to knock a ‘more viable’ candidate out of both of the top two scoring spots. SRV's runoff acts as a safety check on the outcome: It is vanishingly unlikely that the third place scoring candidate would actually beat both the first and second place top scorers in head-to-head matchups. It's actually very likely the top scoring candidate overall wins the runoff step.

Supporting the Weak Opponent is where you help out some weak candidate you don't like in order to disadvantage a candidate you think can beat your favorite. Some voting systems can incentivize voters to support a weak opponent, in order that their preferred candidate win the general election. In an open partisan primary election, for example, this is a viable strategy. If your party candidate is a shoe-in to the general election, you may cast a vote for the weaker opponent candidate so that your candidate has an easier time winning.

This is not a safe voting strategy in SRV. It is only viable if you have a very high degree of confidence that your favorite candidate will out-poll the strong opponent in the scoring round. Voting insincerely does not change the calculus between the your favorite and your strong opponent, but actually increases the likelihood that your favorite will get squeezed out.

Tactical Minimization is where you decrease your support for candidates other than your favorite. Score voting systems are said to encourage tactical "bullet voting" - supporting just one's favorite candidate and dishonestly voting 0 for all the others in order to give the favored candidate the best chance of winning. SRV's runoff step corrects for this strategic distortion in two ways: first, by incentivizing voters to differentiate scores of multiple candidates to have a meaningful say in the runoff, and likewise amplify, to parity, the preferences of those who do not bullet vote in the runoff step.

Tactical Maximization is where you increase support for candidates other than your favorite because you think your favorite is weak or you want to hedge your bet. In addition to the folks who say Score Voting systems are vulnerable to strategic "bullet voting", there are those who say the optimal Score vote is pretty much the opposite: that voters will gain strategic advantage by maximizing support for one perceived front runner and all the candidates the voter likes more than that one. As above, SRV's runoff step corrects for this strategic distortion in two ways: first, by incentivizing voters to differentiate scores of multiple candidates to have a meaningful say in the runoff, and likewise amplify, to parity, the preferences of those who do not maximally vote the top two in the runoff step.

In Summary, while no voting system entirely eliminates the tendency for voters to try to "game the system", Score Runoff Voting both minimizes tactical voting and corrects for its distortion in the runoff step. Since Score Voting with honest voters is tops for maximizing Voter Satisfaction Efficiency, adding a strategic incentive for honest voting is a huge win.

Showing 11 reactions

Please check your e-mail for a link to activate your account.
  • commented 2017-10-21 11:29:27 -0700
    Thanks for pointing me to Loomio, is this the one? I’d never heard of it before, is there a link on this site for it that i missed? Feel free to delete these to maintain the cohesiveness of the thread.
  • commented 2017-10-20 20:21:30 -0700
    Something like Loomio or other forum would be better than tangents showing up on this non-threaded reply section to this static page.

    Anyway, Ted: the issue with the presentation isn’t about describing the overall idea clearly. The issue is removing ANY element of “choose … the best candidate from the top two factions… via some math you trust experts to have right or that you can spend a while thinking about to understand”.

    Plain Star/SRV can say: “between the top two highest scoring candidates, we elect the candidate favored by more total voters” and that has no extra details left out, but even that is already pushing it for people to get the idea. Even the tiniest bit of extra qualifications beyond that goes over the edge of being too much.
  • commented 2017-10-20 19:09:46 -0700
    Just signed up, what’s the current status of getting this on a ballot? And are there any organized events we can volunteer for to help that happen?
  • commented 2017-10-20 15:50:41 -0700
    In terms of presenting the method in a simple way:

    Choose between the best candidate from those with the top two scores, and the best candidate from the top two factions.
  • commented 2017-10-20 15:36:18 -0700
    Aaron, thanks for correcting my typo.

    I agree that SRV/STAR is a good (and possibly nearly the best) starting point, and it could theoretically be amended later. However, I think that it is lacking in some respects, best described by replying to Mark’s comment.

    Mark, I don’t think that having two candidates running off from roughly the same portion of the spectrum is a feature. I live in Seattle, where we have a politically left-leaning monoculture. Even though I also lean to the left, I would appreciate a strong voice from other portions of the spectrum. Challenge makes you stronger.

    If STAR is run as a single election, media would tend to concentrate on the top two polling contenders of the majority or plurality faction, mostly ignoring other voices.

    Again, as an example in Seattle, we had 20 candidates in our top two primary. If you think that simplicity is a desirable characteristic for the voting system, you also have an implied judgment of the general population as not able to handle too much complexity. 20 candidates is way too complex to digest anyway, and I say that as one with a doctorate in applied mathematics.

    So before a single election STAR could be implemented, the voters might want to use it as a primary round, and that is where I see the problem. Including only those top two scoring candidates allows the majority faction to avoid contrarian voices, almost the same problem we have now in Washington with Top two single-vote runoff.

    And note that I’m not proposing eliminating either of the candidates STAR would use, just augmenting them with at least one more candidate.

    Finally (addressing Aaron’s points again), if you started with STAR and tended to get only the top two scoring candidates from the same plurality faction, you would eventually entrench representatives who benefit from that system, so it would be harder to amend it later.
  • commented 2017-10-20 14:51:43 -0700
    Whoah, Aaron and I posted that at literally the exact same second.
  • commented 2017-10-20 14:50:43 -0700
    Ted, this really seems like a non-issue, and the voting system you describe to fix the problem is vastly more complex and thus more difficult to describe to lay voters.

    “Clones” – i.e. actual differentiated individuals that are running to get an actual job in front of the people will differentiate themselves with one another. Fundamentally, allowing multiple viable candidates to compete from the same basic platform is a feature, not a bug.
  • commented 2017-10-20 14:50:42 -0700
    Ted’s Score-TTAPPR makes sense theoretically (aside from his typo of 10-9-1 that was meant to be 10-9=1).

    However, this sort of complexity has a big cost in explaining the concept to voters. Everything can seem like “magic math solutions by smart people” undermines the general public sense of trust and acceptance of a system. It’s probably better to be vulnerable to the crowding effect (running candidates is itself complex and costly, and real elections involve far more than the issues of the ballot) than to risk a situation where voters can’t easily understand and explain how the ballot tallying works. We don’t know yet whether the crowding issue will really be exploited or enough of a problem to worry about. It might be hard to amend the system later, but running with Star/SRV first and later adding the anti-crowding function isn’t impossible.

    If there were a way to get the crowding vulnerability addressed without the complexity and presentation costs, I’d support it.
  • commented 2017-10-20 14:20:27 -0700
    Score Runoff Voting is vulnerable to the crowding (clones) effect — the strongest plurality faction can dominate the runoff by running several similar candidates.

    This could be handled by adding an automatic primary supplementation:
    • Let the top score winner be called A0.
    • Let the runner up score winner be called A1.
    • Add an automatic primary: multiply the scores of all ballots by the complement of their A0 rating. For example, with a zero to 10 scale, scores on a ballot that rates candiate A0 at 9 would be multiplied by 10 – 1 = 9. Scores on a ballot that rates candidate A0 at 1 would be multiplied by 10 – 9 – 1.
    • The top scoring candidate using the A0-deweighted ballots is called B1. This candidate may be the same as A1 if there is no crowding, but could be different.
    • The pairwise winner between A0 and A1 is called A*. The pairwise winner between A0 and B1 is called B*.
    • The winner is the pairwise winner between A* and B*.

    This method is summable, BTW. All counts can be done on a single pass through the ballots.

    It avoids Pushover.

    By adding cloneproofing with the automatic primary, clone incentive is reduced. It could still happen, but there is no benefit to doing so.

    I call this variation of SRV “Top-Two Automatic Primary, Pairwise Runoff (Score-TTAPPR)”.

    It can be adapted to a primary … if the sum of the scores of A0 and B1 (A0-deweighted) are less than some quota (say, 2/3) of the total maximum score count, you could do further rounds of B1-deweighting, B2-deweighting, etc., until that quota of the electorate is represented, then do a runoff election with A0, A1, B1, B2, etc. This would help winnow down a crowded field. Score-TTAPPR would then be run in the final election.
  • commented 2016-12-09 12:49:13 -0800
    On bullet voting in score, although this is possible, I find the arguments not entirely convincing. It’s just as reasonably strategic to vote one candidate top score and all others zero as it is to do all you can to stop a bad candidate by voting them zero and everyone else top score. Likewise, there’s a perfectly sensible strategy to giving mixed scores to middle candidates. After all, you have an interest in both your favorite(s) winning and in the worst candidate(s) losing. So, it’s likely for strategic score voting to include multiple high and low scores and some middle scores.

    Nevertheless, it’s true that score runoff provides an extra balance against strategic voting, at the cost of a just little extra complexity that is easy enough to understand and helps satisfy those who like rank voting. SRV seems the optimal balance.
  • commented 2016-12-09 12:38:32 -0800
    Well put. The primary thing for everyone to realize is that honest score voting is far and away the best possible voting system, there’s just no way to enforce honesty. But if we minimize the impact of dishonesty, then we’ve got about as good as can realistically be achieved.

    We need more real-world experience to say for sure, but there’s lots of reasons many voters will be honest. People like passing judgment. Score voting, unlike any form of rank, allows people to register their disapproval as well as approval, not only their relative preferences. I’m sure there’s a self-selection where those interested in studying voting are also those most likely to consider strategic voting.