Tideman Criticism of Range Voting – response by Abd ul-Rahman Lomax

T.N.Tideman [in email]:
I agree that range is quite good if everyone is sincere, and not bad if everyone is strategic so that it becomes Approval. I agree that in practice you are likely to get the same amount of strategy on each side. What I don't like is the pretense that people are supposed to vote sincerely when people who understand what is going on have no qualms about voting strategically.

Tideman has succinctly stated a common argument against Range. But it's a defective argument.

Range is untried in public elections, so we really don't know how it will behave, we only have theory – and, as we all know, people can have conflicting theories. But there are some very good arguments for Range, I think good enough, by far, that Range deserves to be tried. In my opinion, it is at least as good as IRV, and, actually, should be a lot better.... of course, we need to define what "better" means. We have no consensus on that. (By "we" I mean the election methods community. Range voting advocates do have a metric for "better." But it has not been universally accepted.)

Here are the flaws in the argument:

(1) There is no "pretense that people are supposed to vote sincerely" under Range. Voters may vote as they choose, they are not constrained. Indeed, what seems to be objectionable to so many critics is that voters have choices with Range that they don't have with pure ranked methods, i.e., methods which neglect preference strength or only use an imputed strength to discriminate between members of the Smith set.

(2) There is a hidden contradiction in the argument. Suppposedly there are some people who vote "sincerely," and others who vote "strategically." However, nobody tells you how to rate candidates in Range. Under Approval, there is a strategy which is suggested only, and which is anything but insincere. Essentially it suggests that one draw a line somewhere between the two pre-election-poll frontrunners, and approve everybody above that line in quality. This is clearly sincere.

That one might not "approve" of a candidate one votes for under this strategy is moot. The name of the method is modern and not entirely accurate. Approval is merely standard FPTP voting with overvoting allowed.

Continuing the discussion of the contradiction: supposedly Range rewards "insincere" voting. In order to draw this conclusion, one must assume that "bullet" voting (voting for one only, at 100%) is insincere. However, what bullet voting amounts to, assuming that one gives the maximum rating to one's true favorite, is setting the Approval cutoff at one's favorite, and then setting what I call the magnification high enough that all other candidates are disapproved. It's not insincere. It is merely, at most, melodramatic or exaggerated.

Now, why would one do this? Why would one "exaggerate"? Remember, there is no fixed standard for rating candidates, but we can assume that ratings correspond to rankings, that is, that ratings will be in the same sequence of magnitude as would correspond with rankings. Truly insincere voting would involve rating candidates out of sequence, not compressing the scale such that all candidates are ranked either at maximum or minimum.

One would exaggerate if the outcome of the election, that one's favorite wins over all others, is important to the voter.

Now, if all voters vote in this way, Range reduces to Approval, which, as has been stated and which seems to be a rather common opinion, is not at all a bad method. Here, though, it is asserted that Range rewards "insincerity," which, in context must mean that it harms sincere voters. Sincere voters, it is assumed, have voted other than the extremes. That is, if they like B almost as much as A, their favorite, they may rank A at 100% and B, say, at 80%. Now, it is absolutely true that this could result in the loss of the election by A to B, particularly if B voters do not similarly rate A (which, if we assume that sincere votes have to do with position in issue space or affinity space, might seem likely). But this only will happen if most A voters really don't mind B winning the election. In other words, the harm to them is small, relatively speaking.

What seems offensive is that B voters have been "rewarded" for voting "insincerely." But it mattered to them. Who are we to claim that their expressed preference and ratings are "insincere"?

Range allows – but does not require – voters to express a "weak" vote for one candidate, while at the same time expressing a "strong" vote for another. Voters are not required to do so, they will only do it, I presume, if their preference of the strong over the weakly approved candidate is not strong.

Range leaves to each voter the decision of how to rate candidates. I assume that some parties and candidates will try to get their supporters to bullet vote for them. Some critics seem to assume that all candidates would do this. But consider this: I would consider it a strong negative consideration for a candidate to refuse to acknowledge that an opponent might be a good choice as well. It means to me that the candidate is either insincere or is a fanatic. Our present system requires candidates to more or less behave in this way, so it does not reflect on them individually at present. In other words, for a candidate to engage in puffery is not, at present, for me, a disqualification. However, with Range, it would be.

If you are going to have a system in which knowledgeable people use what looks like approval voting, then in my view you should invite everyone to cast approval votes.

It is simply not true. It is not true that "knowledgeable people" would vote only approval style. I wouldn't. Unless the candidates were so far apart that I could only approve one of them, or a few. Indeed, I might do this, but in many elections I would not. It depends on how important it is for my favorite to win or my most-disliked to lose. Depending on the importance, I would "set the magnification."

What Range does (like Approval) is to equate the "selection space" of all voters. Your maximum rating is the same as my maximum rating, and your minimum is the same as mine. That's a simplifying assumption and it is probably not literally true from a social welfare point of view, but it is a necessary assumption unless we get something like delegable proxy, where elections could become a deliberative process where it is possible to consider, for example, the expertise of voters. Absent something like that, which would assume general consent to such a process, we must assume that all voters have an equal right to influence the outcome.

But having a right does not mean that one is required to exercise it at all, or, in this case, to exercise it fully. Range gives me the choice. Where it is important to me, I can "exaggerate." Where it is not important, I can give graduated ratings. This, in fact, is only an expression of how I rank importance. It is not at all insincere, whether I exaggerate (bullet vote or Approval vote) or spread out my ratings.

I think that the idea that Range somehow rewards insincerity comes from a paradoxical realization that there is something good about allowing voters to express intermediate ratings. Supposedly such voters, the thinking would be, are superior to voters who only will vote for their favorite and against everyone else. And then by "allowing the inferior voters to prevail," we have somehow turned this on its head. It's a paradox. In order to avoid the alleged harm of allowing intermediate votes, which we somehow sense are good, being sincere, we prohibit them....

But there is no harm that is not accepted by the sincere Range voter.

The most harm is done, in particular, to voters who truly vote out of sequence. I.e., they favor A>B>C (A over B over C), but because they fear that B will win, they rank A highest (sincere), B at zero (which could be sincere in itself), but C at an intermediate rating, or even at full rating. If it happens that C wins, because too many voters think similarly, they will experience the full consequences of their insincerity – maximized regret.

And this will only happen if, say, other voters, say B supporters, rank C at some higher rating than zero. (I'm assuming here that the honest Condorcet winner would be B.)

But I am not happy with approval either. I think we agree that under approval the strategy that generally makes sense is to try to determine which two candidates are most likely to be ranked first and second, and then give a vote to the one you favor between these two, and to all candidates whom you prefer to that candidate.

Yes. Range strategy is quite similar; the only question is the "magnification," which most writers on the subject seem to have neglected, as if it were obvious how to "sincerely" distribute intermediate ratings, as if it is necessary that if I prefer A to B, I must rate A over B. Sure, I must, if the system allows me sufficient resolution. It won't in many cases. What must be understood is that the distribution depends on the importance that the voter places on the victory of the favorite over others and the loss of the most disliked compared to others. With strong preferences in this regard, Approval style voting is sincere. And if preferences are weak, then intermediate ratings do not harm the goal of the voter: a pleasing outcome to the election.

If I wanted to devise an evaluation of Approval that I regarded as informative, it would need to involve a statistical process that reflected this understanding of how approval works in practice. But even if that evaluation turned out well, it would make me uncomfortable to have a voting procedure in which getting the most out of one's vote required one to begin by figuring out who the two leading candidates were.

But, of course, this is what we presently have. There is no requirement that voters make this determination. They can vote "sincerely," but present systems require that they do so for one candidate only. The described Approval "strategy" is rational and allows one easily to determine how to vote under conditions close to the present. If you don't know who the frontrunners are, you have been living on some other planet for a very long time. Almost always, they are the Republican and the Democrat (in the U.S.)

If you fail to use this strategy, no harm. That is, no harm over the present method, which forcibly prevents you from using this strategy – giving you only the choice of either picking the frontrunner or almost certainly wasting your vote. (With the asterisk that your wasted vote still might be able to affect ballot position or public campaign funding next election – which actually distorts the process.)

Allowing overvoting is the number one simplest and easiest-to-implement voting reform. All that has to be done is to delete a few lines in the election code, those that require disregarding ballots where the voter votes for more candidates than there are allowed winners. I can see no reason for doing this, except for essentially preventing third parties from getting any traction. Is this the reason? Perhaps. It may also be that election process was originally designed without long experience with elections, and having everyone express first preference was just what they thought of first. I think there were other proposals before the founding fathers, though....

And it is my opinion that once voters can vote Approval, they will then want to be able to express intermediate votes. Can Mr. Tideman provide us with any reason why they should not be allowed to do so? His argument above is essentially that they would be harmed if they do it. Isn't it our prerogative, as voters, to make that determination ourselves? Indeed, most of the U.S. currently exercises the ultimate Approval voting: we vote zero for all candidates by not bothering to vote. Which, of course, is exactly equivalent in effect to voting for all of them.

Range is a ranked method, we should not forget. Even if one wants to evaluate ballots according to pure preference, perhaps to determine the Condorcet winner, voters should be free to express the kind of detailed information that a Range ballot collects. And, of course, once they are free to express it, we can then begin to understand if preference strength is important. I have no doubt, however, that it is, and thus pure preference methods which neglect preference strength are inherently flawed. It is easy to show how poor decisions can be made by systems which disregard preference strength but only consider rank order.


Return to main page