The flaw in the design of IRV is that with the exception of the last runoff, each of the (virtual) preliminary elections involve more than just a pair of candidates. These preliminary runoff counts are modeled on plurality voting with more than two candidates.
The IRV concept amounts to capping off a series of potentially faulty elections with a single good election at the very end. On its face, that would not seem to offer any improvement over plurality voting whatsoever, but in practice IRV actually does seem to work better. Probably this is because the conditions for failure (using plurality voting) to eliminate a candidate are more improbable than the conditions for failure when selecting one as winner. However, as the number of candidates increases, so does the number of problematic counts, and so does the likelihood of a failure in at least one of the counts.
A failed IRV election was illustrated in an earlier article that was designed so that the very first of the preliminary runoff-elections would reject the voters' most favored candidate. The first candidate eliminated was the very one that a reasonable (democratic-thinking) person would say should have won the election. The example is hardly different from the one constructed above in which Susan deserved to win.
As a last note, I should recommend a video that also compares IRV some other voting systems (none of those systems being balanced, however). It provides alternative ways to think about these and other voting methods.
Readers of this article may be interested in a later article that explores further the problems with IRV.
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
1
