Speaker votes should record WHO people voted for
Opened this issue · 2 comments
Jeff and I spoke about this today.
Currently speaker votes reflect whether the voter voted for or "against" the eventually victorious candidate. This reflects that although votes for speaker are multinomial, our system wants them to be binary + not voting the way all other votes are.
There are three issues with this:
- People voting "Present" for speaker are not "Not Voting" the way we maybe consider them to be when it comes to other substantive votes.
- People are not voting "Yea" or "Nay", they are voting for candidates
- People voting "Nay" could be voting for one of several candidates.
We think that in the medium term, votes for speaker should have additional metadata making it clear who each person voted for. Then, the actual vote page for speaker votes should display this information in the vote table, and potentially ultimately the map.
None of this should reflect NOMINATE estimation, just WebVoteView display of votes.
We think gathering this data could easily be done by SRP type students, while implementing might require a bit of programming.
Looks like there may be general problems with the way that speaker votes are handled in Voteview. So far, we have discovered that there are missing speaker voters (many of them) for the 36th House and it looks like also in the other Houses of the 1850s in which protracted speaker fights occurred.
Additionally, although we understood the practice of Poole to be to recode speaker voters such that supporters for the winning speaker candidates received a "1" and voters for other candidates received "6"s, It appears that that is not (always) the case. For example, the 32nd House data retains the raw ICSPR vote codes in which names are mapped to values between 1 and 8 (as defined in the descriptions). Voteview wrongly converts these codes in the standard 1-3=y, 4-6=n, 7-9=nv in the displays and codings.
All of this highlights the need to identify all of the speaker votes in the data.