Counting RCV Ballots (for Analysis)
Ballots that rank all candidates provide more information than incomplete ballots.
[In future posts, I analyze past Ranked Choice Voting (RCV) elections. This post explains one subordinate challenge and how I handle it. You may want to skip it until you find yourself interested in it.]
In Ranked Choice Voting (RCV) elections, many voters fail to rank all of the candidates. For instance, so-called "bullet voting" is when a voter ranks only one candidate. This is a challenge for analyzing RCV elections because it is not clear why a voter failed to rank others. I believe there are two competing explanations for why a voter might rank fewer candidates than possible:
The incomplete ballot may be an indication that the voter preferred her ranked candidates to her unranked candidates, AND that she might be indifferent to the unranked candidates, or
The incomplete ballot may be an indication that the voter excluded candidates that she thought had no chance of winning. Note that this is not an expression of preferences, but rather a strategic (or lazy) decision.
For example, in the 2021 New York City Democratic primary for mayor, suppose that a voter only ranked the three candidates who were polling the best, Adams, Garcia and Wiley, but left their fourth and fifth ranks blank. One reasonable interpretation is that the voter preferred all three to all other candidates.
Another reasonable interpretation, however, is that the voter thought that the other candidates didn't stand a chance of winning, and so she didn't bother ranking them despite liking some of them better than the top three.
For instance, where did Yang fall in the voter's preferences? Did she prefer those she ranked to him, or did she omit him because she thought he had no chance of winning despite being her favorite (or somewhere in between)?
We cannot know why the voter cast an incomplete ballot, or how to interpret the omissions.
Therefore, I believe that the only preferences that we should trust (and, therefore, analyze) are those that are fully expressed. I therefore limit analysis to complete ballots. This is a limitation, but I believe it is the appropriate, conservative approach.