Interpreting a BAV Election

You may recall that the previous article compared Balanced Approval Voting (BAV) with two similar voting systems and that one of those two systems was balanced Plurality Voting (BPV). To continue with that comparison, look at the BPV election with the results shown in the first column of Table 1. Looking only at that single column, we can conclude that Candidate B is the winner and that Candidate C, the only one of the three with a negative vote count, came in last. We might conclude that Candidate C is a decidedly unpopular candidate, since she has the only negative tally. And should her party take this to heart, Candidate C will not have much of a political future.

But now we compare these results with those of the corresponding BAV election shown in the last three columns of Table 1. Candidate B still wins the election, and C still comes in last, but we see

that, far from being unpopular, candidate C was the only candidate that experienced no votes of opposition. The apparent votes of opposition (in the BPV election) did not represent animus towards C but was more likely due to voters not being unfamiliar her. The candidate with the most serious unpopularity problem is not candidate C but rather candidate A. This becomes particularly clear if we alter the presentation of the BAV election to the format displayed in Table 2.

The Support column in Table 2 shows, for each candidate, the percentage of support votes among voters who did not abstain. The Awareness column shows percentage of voters who did not abstain. This presentation of the election should make it clear to Cs party that the problem is not with Candidate C but rather that Candidate C needs wider exposure to the voters.

In earlier articles, I have argued that BAV collects more accurate and more appropriate information that other, more widely familiar voting systems. To the extent this is true, the results of a BAV election should make it relatively easy for voters and analysts to interpret the data from the ballots. To pursue that line of thought we will return to the example election from the last article as summarized in Table 3.

Table 3 does present extra information than what might be natural to collect in a BAV election. But the additional information could be provided by keeping track of which party a voter belongs to. That would make it possible to know how party registration affected the election results. When using voting machines, keeping track the votes in this way would not seem difficult; no doubt it would even be possible for the machines to keep track of which voter submitted each ballot. As concerning as that may be, merely identifying ballots by party affiliation (perhaps only for the largest parties) would not seem like an objectionable invasion of privacy. A computer-readable (but human unintelligible) code on paper ballots could identify the ballot by party affiliation when they are tallied.

Table 4 shows the same information but tabulated much like Table 2. For each party, the percent of non-abstentions for each candidate is shown in the Knows column. And the percentage of non-voters who voted for a given candidate is shown in the Approve column. This table does not show much that is surprising, but it does make it clear that the Socialist Party needs to do some promotion within the Democratic Party and inside the Progressive party.

In Table 3, the only two candidates with positive net votes are the Democrat and the Progressive. And in Table 4, these are the only two candidates with net approval over 50%. Table 4 presents the same information as does Table 3, but in a different light, and different lighting can reveal things not previously noticed. In this connection, it is worth noticing that it is the Balance property that makes the Approve Column at all interesting. With Approval Voting, the only alternative to supporting a candidate is to abstain, so the Approve Column could only report 100% for each candidate.