Earlier in this series of articles we discussed Instant Runoff Voting as well as Approval Voting and we showed how each of these voting systems could be improved by making them balanced. But only very superficially did we try to compare the two different approaches to voting. Each has its proponents and we are not apt to settle the dispute here, but we will suggest how the dispute might be settled. The Center for Election Science favors AV whereas The Center for Voting and Democracy prefers IRV.
The AV supporters observe (in comparing AV to FCV) that collecting more information from voters will result in better election outcomes. It might seem odd that that this claim would come from the AV camp because it seems to give an edge to IRV over AV; IRV does gather more information from voters.
However, more information is beneficial only if it is accurate. IRV asks voters to construct an ordered list of the candidates they favor and the order is critical for how the votes are counted. A moments reflection should reveal that voters will not always find that order easy to construct and some will perhaps resort to what amounts to a flip of a coin to determine who, for example, should be in third and who should be in fourth place. The extra information that is collected for IRV would seem to include more than a bit of noise and that noise has a potential for producing erroneous outcomes.
The information collected for AV is, except for the order details, exactly what is collected for IRV and this suggests an avenue for comparing the two voting systems. Some elections are conducted using IRV and it would be quite possible to re-count these ballots to determine how an AV election would have turned out. If the results differ, an after-election poll could measure how well satisfied the voters might be with the two different outcomes. Of course this experiment would have to be repeated over many elections for results to be meaningful.
My own view is that BAV (described in an earlier article) is clearly an improvement over AV and that IRBV (described in another earlier article) is clearly an improvement over IRV. Moreover, IRBV collects some potentially noisy data in addition to what BAV collects. If IRBV elections were conducted somewhere, the same kind of analysis of the results could be used to help determine whether IRBV or BAV might be the one to prefer.
