[EM] Real Ballot Data & Analysis

[Eric Gorr]

At 11:31 PM +0000 3/5/04, James Gilmour wrote:
>Eric Gorr wrote:
>  > I am sure others are aware of this data, but I just got my
>>  hands on it and spent some time analyzing it with single winner
>>  methods. Apparently in Dr. Tideman's research into voting
>>  methods, he was able to obtain many (86) ranked ballots from
>>  elections held in England.
>
>I don't know how it will affect your analyses (but I think it
>will), but you should be aware that some of the ballots in the
>dataset Tideman obtained from the UK (not just elections in
>England) may have been for STV-PR multi-member elections, not
>single-winner elections.

Yes, this is true.

However, there are some datasets where the goal was to select a
single winner. There were 15 such cases.

In one of those cases (A26), IRV computed a different result then
RP(d-wv). 1 out of 15 cases makes the percentage of IRV failures
at ~6.67%.

>It MAY not affect the conclusions of the single-winner analyses
>you wish to do, but voters do vote differently in single-member
>elections and multi-member elections where some parties put up
>more than one candidate.

I do not see how it would affect any conclusions, but I admit
that it might.

We should still be able to view the ranked ballots as an
expression of preferences among the various candidates.

>If these ballots come from the dataset I think they do, much of
>those data were machine-generated (reconstructed) from election
>result sheets, ie they are not actual voters' ballots.

What would be the actual difference?



-- 
== Eric Gorr ========= http://www.ericgorr.net ========= ICQ:9293199 ===
"Therefore the considerations of the intelligent always include both
benefit and harm." - Sun Tzu
== Insults, like violence, are the last refuge of the incompetent... ===