ranked.vote: Understanding Discrepancies
All of the analysis that goes into ranked.vote reports is based only on ballot-level data. This includes the IRV tabulation, which backs the “Runoff Rounds” section.
In some cases, these tabulations don't match the official tabulations. There are two main reasons for this. First, there are multiple ways to tabluate an instant-runoff election. The main decisions are:
- Whether you remove a single candidate on each round, or all remaining candidates who are mathematically unable to win.
- Whether you stop eliminations when it is clear that one candidate has won, or whether you continue eliminating until only two candidates remain.
These result in the same winner, but the count at each round may differ. For consistency and information density, we tabulate all elections the same way: all candidates who are unable to win are removed at each round, and we continue until two candidates remain.
Another source of discrepancies is the handling of undervotes/overvotes and write-ins. To the extent possible, we follow the statutes in the relevant jurisdiction in order to treat ballots the same way that the elections board does. But statues aren't code, so there are sometimes cases where there are ambiguities around edge cases. In particular, the treatment of write-in candidates can be a complicating factor, because many jurisdictions replace them with a generic “write in” dummy candidate, which does not always reflect how they are treated in the official tabulation.