[EM] UC davis STV election data - not very useful, actually

Warren Smith wds at math.temple.edu
Fri Nov 11 13:09:14 PST 2005


So far there have been 4 elections.
Each is a multiwinner election (e.g. the latest was 8 winners and 23 candidates)
apparently run with reweighted STV voting.

Each has a report about it with lots of charts and graphs.
Each has had between 2447 and 4068 voters.

I do not understand why these elections are of interest to single-winner
voting researchers.  Two reasons for my non-interest:
 1. they are multiwiner elections.
 2. Assuming one of these is an interesting election, we
are unable to apply other voting methods or to see if the
election exhibited non-monotonicity, etc, because UC Davis does NOT
post the actual votes, only a statistical summary of them.

This is the usual policy in IRV/STV elections, and it is apparently done
intentionally to prevent anybody from ever knowing that the election
was non-monotonic.  I.e. secrecy ==> nobody can prove there was a problem ==>
everybody is "happy".

A less-cynical interpretation is simply that they do not want to post the votes,
too voluminous data.  But Debian isn't afraid to show the world their votes.
And certainly it is no big deal memory-wise to post 4000 votes.  UC Davis's
posting of just one photographic image far exceeds the memory needed to post
all the votes in all these elections combined.
So I am afraid the cynical interpretation is the one I must adopt.

Also, the cynic will note that Davis refuses to report the number of invalid ballots,
presumably also an attempt to cover up problems.  (In San Francisco 2004, the invalid
"spoiled ballot" rate for IRV voting was 7 times their rate for plurality voting.)

Anyhow, thanks to this non-posting of their votes, the UC Davis data is NOT
useful to election researchers, but it might be if they actually did post the votes.
Also, there is some interesting data in the summary reports they do post
(e.g. about voter behavior).


More information about the Election-Methods mailing list