Can Less be Better?
An underlying assumption in this series of articles has been the notion that more democracy can only be an improvement. But could there be an unseen down-side? It is hard to imagine, but perhaps that is only because even an adequate portion of democracy is so outside of our experience.
It is certainly understandable how experiences with shortages might lead one to conclude more as necessarily better. But people seem too quickly to conclude that more of something good is necessarily better. Often it is of course, but we all know it is not always purely the case. More cars means more traffic; more construction means more congestion and more demands on our environment. One pill might cure what ails you while a dozen of them could kill you.
The mistaken notion that if something is good then more is necessarily better can even corrupt our thinking about voting. Are more choices and more freedom of expression always better? For example is it always better to have more than just two candidates in an election? It is tempting to think so, but an extra candidate can easily distort an election when plurality voting is used.
That particular voting system simply misfires much too frequently when there are even three candidates in an election. This does not mean that more candidates is what is bad. More reasonable it to conclude that plurality voting is a decidedly poor choice for elections even though we do keep using it. So long as we do keep using it, more candidates is not entirely a good thing.
We have argued in the past that balanced approval voting is a good system and perhaps even the best system of voting. In that system, each voter has the option of voting approval (+1) or disapproval (-1) for each of the candidates. But necessarily also there is the option of marking neither approval nor disapproval (0). The sum is computed for each candidate and the winner is the candidate with the largest net vote. In this system the candidate with the largest net approval wins. It is worth noting that this is a description of the simplest example of a balanced range voting. Range voting is also known as score voting.
But some people, apparently operating on the belief that surely more is better, will argue that allowing the voter more choices is undoubtedly better. For example, allowing the voter a larger range of say -2, -1, 0, 1, 2 would obviously be an improvement over balanced approval voting. Would they argue that a range of -100,...,0,...100 with 201 possible ways to vote for each candidate be an even greater improvement? Let's look at an example.
Consider an election with five candidates where the issue of the day happens to be whether more choices in the election are better. For clarity let us assume there are 1000 voters.
The five candidates are La, Lb (both of whom favor less), N (who is neutral on the issue), Ma and Mb (both of whom wish for more choices). Let us assume also that the voters split evenly with 450 voters approving of both La and Lb and disapproving of Ma and Mb while another 450 voters disapprove of both La and Lb while approving of Ma and Mb. The remaining 100 voters approve of N but remain neutral on the other candidates. In a balanced approval election, La, Lb, Ma and Mb would have a net vote of zero while N would be the winner, being a consensus candidate with a net vote 100 -- arguably the best possible outcome.
But now suppose the M-voters get what they wished for, namely ranked voting with five ratings available for each candidate. And suppose that the 450 voters who approve of both La and Lb (preferring as they do to vote only approval or disapproval) all vow to give both La and Lb a rating of +2 and to give Ma and Mb a rating of -2. These 450 L-voters provide 900 votes for both La and for Lb while providing -900 votes for both Ma and Mb.
The voters who want more choices have, let us say, an even mix of those who prefer Ma and those who approve of Mb and they likewise have a mix of opinion about La and Lb. The M-voters provide 675 votes for each of Ma and for Mb and likewise they provide -675 votes for each of La and Lb. In total for this election using 5-way ranked voting, La and Lb each get a net vote of 900-675 = 225 votes. This ties L1 and L2 for a win against the compromise candidate N with only 200 votes. Ma and Mb come in last, tied with -225 votes. In effect, the M-voters, by exercising their right to express themselves more precisely have managed to elect the candidate they most dislike. Perhaps after the election they will feel so satisfied that they voted in a way express exactly how they felt that it will not matter to them that the result for them was the worst possible; winning is not everything after all.
If balanced ranked voting were adopted with more than just the three ranks (-1, 0, +1) that are available under balanced approval voting, it seems likely that most voters would eventually understand it to be strategically best to cast only maximum, minimum or neutral votes for all candidates. The voting system would in time behave nearly the same as would balanced approval voting. But no doubt there would always be some voters who would prefer to dilute their vote, continuing to insist that no harm could possibly come from expressing their opinion more precisely. And no doubt efforts by one side would be made to convince the other side to believe this -- so long as there remained essentially two sides at least.
In earlier articles I have pointed out various problems with Instant Runoff Voting (IRV). But the fact that this alternative voting system remains so popular makes one wonder why people are so attached to it. I think that one important reason is that it has to do with the belief that more is better; in this instance, they prefer more expressive ability. Voters want to be able to express their opinions fully and they feel that ranking all of the candidates the way they can achieve this. But once again we have to consider whether more for the sake of more is really better.
Perhaps that argument can be made if you are considering only elections with three or four candidates, but what if there are seventeen? How accurately can a voter construct a list of seventeen candidates in order of preference? How hard would it be for even just eight? Could such a contrived list really express fully the voter's actual opinions of the candidates? There is no way for the voter to express likely ambivalence about a few of them.
And how, using just an ordered list, can one express which candidates seem acceptable and which simply seem unacceptable? There really is no way; two voters might well construct exactly the same list while one of them considers only one of the candidates to be acceptable while the other considers only two not to be acceptable. And just how important is it to express a very fine-grained preference for candidates? Is that more important than being able to express approval or disapproval clearly?
Anyone who has followed this series of article will know that these specific problems can be side-stepped by using IRBV instead of IRV. IRBV has much the same look and feel as IRV, but few people have ever heard of IRBV. However, the issue here is not to promote IRBV. Rather, it is to understand why IRV has such a strong following. One reason surely is that IRV does have a movement promoting it. But another reason could well be that IRV has an intrinsic appeal based on the uncritical and incorrect belief that IRV uses a best-possible sampling of a voter's opinion about the candidates.