When this series began in 2014, its articles focused primarily on the study of the many different balanced voting systems. Quickly, one particular system, Balanced Approval Voting stood out as particularly promising from the standpoint of improving elections and even politics.
But the first example of balanced voting, Balanced Plurality Voting (BPV), remained interesting for its primitive simplicity. It seemed so simple, but perhaps because the familiar always seems simple and natural. With plurality voting, a voter is asked to choose just one candidate to vote for. With BPV, a voter still has to choose one candidate, but the voter is given the option of voting against or for that one chosen candidate. This system is quite expressive when there are just three candidates, but less so as more candidates join in.
It was not until 2020 that an article in this series explained that the vital ingredient, missing from many of these other balanced voting systems, was the important property of being evaluative. Evaluative voting systems strictly avoid requiring or even hinting that voters should make any selection from among candidates; these systems simply ask voters to evaluate any or all candidates individually, based only on appraising the merit of that candidate. The flaw in so many other voting systems is in requiring voters to compare one candidate against another; this often drives voters to base their vote on differences that are unrelated to the candidate's perceived merit and sometimes even to choose arbitrarily between two equally qualified candidates.
The evaluative voting system, Approval Voting, relates to BAV much the way plurality voting relates to BAV. In a number of articles since 2020, I have repeated the observation that evaluative voting systems have a sure-fire way to avoid any hint of the spoiler effect. Voters are in no way motivated to consider extraneous issues like projections of electability or pundit advice because, with an evaluative voting system, voters can freely assign the very same rating to candidates that are judged equal.
Though BPV is not evaluative, it does steer clear of the common example of the spoiler effect (an example that involves exactly three candidates). But might the spoiler problem somehow persist as a worry in elections with more candidates?
A difficulty in answering this question is that the spoiler effect is commonly explained through that example with three candidates. In the absence of a widely accepted but more general definition, we have to resort to simply the hope that we will no doubt recognize the spoiler effect when we see it.
With that abject preamble, let us consider an election using BPV with five candidates: A, B, C, D and E. We should think of A, B and C as favored equally by 600 voters who also will likely oppose both D and E. But D and E are supported more or less equally by 500 voters who oppose A, B and C.
Prior to considering any particular voting system, let us think about which candidate should win election, based only on democratic principles. If A, B or C win, 500 voters will be disappointed, but if D or E win then 600 voters will be disappointed. If there is to be a single winner it seems only reasonable to expect that one of A, B and C would be the winner.
With that in mind, what might happen in a BPV election where a voter can express an opinion about only a single candidate? Casting only positive votes will probably appeal to most voters, if only because it is particularly easy to appreciate the intent and effect of such votes. With this assumption, A, B and C will split nearly evenly the 600 votes three ways; let us say, 210 votes for A but with B and C each getting 195 votes. On the other hand, D and E will split the 500 supporters with perhaps 249 votes for D and 251 votes for E. So E will be declared the winner. Unfortunately this makes 600 voters unhappy with some surely calling foul on a bad election. Does this qualify as an instance of the spoiler effect? It seems a reasonable characterization but you be the judge.
In contrast, what happens if this election uses BAV? From A, B and C, they will likely get 600 votes of support and from D and E, 500 votes of opposition, so a net-vote of 100 each; because of having a close friend among the otherwise likely opposition, perhaps B will get a net-vote of 101 instead. Meanwhile, D and E will each get a net-vote of -100. So with BAV, as democracy would prefer, the majority of voters will be satisfied with the outcome. The spoiler-like effect exhibited by the apparently simpler BPV election is, as might be expected, avoided using BAV. It is worth appreciating how, with BAV, once the positions of voters is known it is fairly clear how they will vote. There will surely be some gray areas where a voter may be forced to ponder whether to support or abstain or whether to oppose and abstain, but as noted in an earlier article this internal debate actually serves a useful purpose (through voting statistics) in taking a more refined measurement of net support for candidates.
In contrast, using BPV, voters are more limited in expressing their opinion so their voting is not as predictable. Returning to the BPV example, what if all 600 voters who support A, B and C were somehow convinced to cast their only vote against D or E. While it might be difficult to convince them to do this, there is an appealing rational that this would split the power of their vote only two ways instead of three. Indeed, while that would reduce the net-vote for A, B and C to zero, it would also reduce the net-vote for D and E to around -50. Although there is a possibility that would hand the election to a widely ignored minor party candidate, F (not previously mentioned here). But in the absence of such a sixth candidate, this negative voting would result in A or B winning the election. So (just as with the three-candidate version of the spoiler effect), a spoiled election may not happen even when conditions are ripe for that possibility.
An argument sometimes made in favor of BPV is that it is simpler than BAV. Surely the ballots are simpler and the tallying of ballots is simpler. But as compensation, BAV actually simplifies voters' decisions about how to vote. There is, with BAV, much less opportunity or motivation for strategic voting and there is no encouragement to make judgments about the relative merits of any two candidates or to make guesses about which candidate is more electable. All the voter needs is to learn enough about a candidate to establish an opinion of whether to support or oppose candidates. Of course, a voter can also decide to neither support or oppose a given candidate (abstain) and as noted before this is a remaining, though helpful, gray area.
