A short video about star voting recently attracted my attention to this voting system and I'd like to share some thoughts about it.
The video features an election with nine voters to choose one of three fabrics. The star election is conducted with the same ballots as suited to score voting. The ballots allow five scores numbered 1,...5, and the ballots are first counted as a score election for choosing the two top contenders (in this case, just to eliminate one candidate). The remaining candidates then face a runoff using, effectively, plurality voting. The votes in that runoff are determined by the ballot scores with the vote from each ballot indicating the candidate who is assigned the higher score. Plurality voting is a balanced system when there are only two candidates and in that case there is no concern about the spoiler effect that, in general, afflicts plurality voting.
But it does seem a troubling notion that the election outcome will somehow be improved because we ignore, in the last round, details that the voters took the trouble to express. The video shows how a favorable result might happen, but it seems quite possible this is only for some special cases.
With only nine voters, it does not seem unreasonable that they might all vote exactly as they are told, specifying a score for each and every candidate. But suppose there were 3001 or even just 2001 voters; would it still be a reasonable assumption? If we cannot still assume that not even one voter skipped over a candidate, assigning no score for some candidate, then how will star voting handle such exceptions? A distasteful possibility would be to simply throw out those ballots as being spoiled. At first glance, a more appealing possibility would be to just skip over those missing scores, but this is essentially the same as treating the missing scores as zero and that would mean that there are now six scores for voters to choose from, not just five. A third possibility would be to choose a default value between 1 and 5 and just assume that is what the voter must have meant. But which value?
In a recent article, we showed that unless a voting system is balanced, it will bias the election in favor of the most famous candidates. Given a polarized politics, these will typically be the candidates from the two dominant parties, perpetuating the two-party duopoly. If we wish to avoid such a bias we would choose 3 (the middle score) for the default value as that choice makes the system balanced. With this added restriction, star voting might seem to be an attractive system, being both balanced and evaluative. True, it is considerably more complex than BAV, but perhaps it offers some advantage (without further disadvantage) that would warrant the additional complexity. But for now, it is worth remembering that in an early article of this series we found that with more than three scores, balanced score voting invites voters to vote strategically.
A concern with star voting is that the score tallies might produce a three-way winning tie in the preliminary score election; it is not clear how star voting would deal with such a malfunction. Similarly, any tie in second place (in the initial score election) could force the runoff to be between more than two candidates.
A further complication is that perhaps a large number of ballots might have to be ignored in the final runoff round. For example, there might be two or more candidates assigned a score of 5 on a majority of the ballots. No doubt these very candidates win the score round of voting, but none of this majority of ballots show a preference between the two. So in deciding between the two finalists, none of a majority of the original ballots could be counted. We might argue this not to be unfair since these voters already had their say in selecting the finalists. But it does seem troublesome that the final selection could rest on a very small number of voters. Would it not have seemed better to have selected the winner of the single-winner score voting, assuming there is one? Perhaps the choice might even rest on a coin toss (consider how, in our very polarized elections, what would happen in the runoff with two Democrats competing against one Republican).
Just as we can re-interpret the 5-score ballots for a plurality voting, we could as easily re-interpret them to simulate a 3-score election with scores less than 3 treated as -1, scores of 3 treated as 0 and scores greater than 3 treated as 1. This seems a reasonable way to turn to balanced approval voting (BAV) for making the final decisions. The advantage would be that BAV would handle much more gracefully than could plurality voting, the unfortunate multi-way ties at the end (just as could 5-score voting). But real value in this approach seems hard to find; just as with the plurality finale much of the detail that the voters have already provided is simply discarded. Voters had to deal with extra complexity when voting, only to have their efforts ignored in the final runoff. And to what purpose?
It is possible that when two or more candidates tie under 5-score voting, that BAV might produce a single winner. But just as with any other voting scheme a tie is still a possibility with BAV. Still, for resolving ties, voters are apt to find employing a suitable alternative voting system preferable to tossing a coin or drawing straws so it seems like it is first worth a try. For use as a tie-breaker, BAV does offer the advantage of offering two reasonable alternative counts, the second being to assign the scores 2, 3 and 4 all to the middle value of the three-score system. Using BAV for breaking ties would provide the chance of a second tie-breaker should the first fail. There is still no guarantee, but the odds would appear to improve.
