BAV: What Could Go Wrong?

Being curious about the objections there might be to balanced approval voting (BAV), I asked Chat GPT for a summary. Chat GPT is reported to have read and absorbed everything on the internet so that it seemed a prime source for such information. But the objections it reported seemed pretty generic, applying at least as well to other voting systems. For example, it listed complexity and susceptibility to strategic voting as objections, topics on which BAV would seem to be comparatively immune. The only objection in the summary that appeared to have merit is that BAV is new and unfamiliar.

BAV is used in Latvia, and I once read (but not confirmed) that BAV had been used in the early days of the Soviet Union; perhaps that is where Latvia got the idea. So BAV is not entirely new, but surely it is new in the American context where it remains a too widely unfamiliar novelty. And the novilty raises reasonable concern for unforeseen troubles. Caution surely should lead us to adopt it slowly, on a small scale at first, but is there any realistic alternative to this slow approach? One state or locality might try it out for a few election cycles allowing for discovery of surprises that might arise.

With BAV, as with any other voting system, tied elections are always possible. The traditional solution for this situation is to flip a coin, and provided ties are extremely rare this seems like a satisfactory plan. But conceivably, with BAV, tied elections might become more common and it would be wise to consider how to respond.

Another concern with BAV seems mostly a concern for voter satisfaction, not something to dismissed lightly. The limited choices that BAV offers to voters could raise some objections (though, to be fair, BAV provides the voter with more choices than either plurality voting, or approval voting does).

A voter is limited to choosing between support, opposition or simply no-comment and this restriction to only three ratings could make some voters unhappy. A voter who greatly prefers one candidate above all of the others might still feel strongly about a second choice and maybe even a third choice that seems a reasonable alternative; does the voter show support all three equally or maybe just abstain with respect to the third choice; from this voter's perspective there are not enough options.

But while having additional options make these decisions easier for one voter, choosing from five or more choices may may seem to other voters like making overly detailed and even meaningless decisions. Would providing extra choices guarantee a better election outcome? It is in no way clear it would, but there can be no doubt that the extra options would complicate the ballot. And more complicate ballots might also increase voter dissatisfaction.

In an early article of this series, we argued that more choices would also encourage strategic voting. For all these reasons it now seems hard to justify providing the extra scores, but it is possible more experience with BAV could provide additional justification. We can now, at least explore ways to avoid encouraging strategic voting while providing voters with additional scores.

Assume that score ballots were used in an election using the scores -2, -1, 0, 1 and 2. We could interpret these ballots for a BAV tally by treating any negative scores as an opposition vote. And positive votes would be treated as support; of course, 0 would be understood as an abstention. This would provide voters with the satisfaction of more complete expression while not raising the temptation of strategic voting. But also, this would seem deceptive and surely, many voters would object to this apparent fraud. It seems imperative to make some use of the extra detail on the ballots.

And voters might also object to how arbitrary this is. In tallying these same ballots, we might as easily treat -1 and 1 as well as 0 abstentions -2 as opposition and 2 as support. Notice that this alternative treats voters as being less open to compromise than does the previous approach. Which version would be better? Should we opt for more compromise or less? An old suggestion of caution advises wearing both a belt and suspenders. By somehow using both alternatives we could make use of all scores while avoiding the encouragement of strategic voting.

We could first count the ballots as if voters are very open to compromise; but when it is necessary to break a tie then the alternative, less compromising, method would be used in the hope of resolving the ambiguity. And if that second BAV tally also fails to settle the tie, the ballots could be counted a third time, as if this were a score election. Even then there remains a small possibility that even the third tally might fail to break the tie, but presumably this would be an extremely rare event, probably rare enough that a coin toss could again serve for making the final decision. But instead of resorting to a coin toss, we might realize there is yet another potential tiebreaker. Before the coin-toss we could explore whether balanced Condorcet voting would break the tie.

In the words of another recent article, score voting with five or more candidates has a large nuclear family. This construction shows how, in extreme circumstances, this could prove to be a useful attribute.