Re: exit polls vs. election results

From: Kathy Dopp <kathy_at_uscountvotes_dot_org>
Date: Sun May 15 2005 - 23:13:47 CDT

Fred McLain wrote:

>Opinions? Note that this column has supported the USCV position in the
>past.
>
> -Fred-
>
>http://bellaciao.org/en/article.php3?id_article=6046
>
>
They quoted almost directly from our paper in the above article, and
also from Mitofsky. Clever.

Well, not that it changes our conclusions much, but USCV probably should
not have rushed our paper for the AAPOR conference because:

I've been staring all day today at my "vote shift simulator" and I had
made some mistakes in it.

a. I'd shifted votes using the existing simulated data distribution,
when a vote shift analysis (I realized this a.m.) needs first to take
the existing data distribution "back" to the pre-vote shift location,
and then apply the effect of vote shifting.

b. The vote shifted pattern must then be compared with the pre-vote
shift pattern (not the existing E/M pattern) to see what vote shifts do
to the data patterns.

Luckily we labeled our paper a "Working Draft". I've attached two
updated improved charts from it.

According to Elizabeth Liddle, who just admitted to me that she was
hired by Mitofsky (temporarily), says that there is so much variance in
the data, that it is still plausible that it can be explained by a
response bias with constant mean. However, this variance may also be
the result of vote shifts from some precincts which should rightly be
included in more pro-Kerry precinct groupings and got shifted.

What is really interesting to me, is that the slope of the "vote shift"
trend line falls exactly between the trend lines for Pre-vote shift and
E-M data.

This is all from simulations, but to me, it shows a likelihood of the
influence of both a response bias and a vote shift to explain the E-M
data. I don't think we'll be able to conclude anything definitively
without studying the election data.

We must get the Nationwide Election Archive Done, but I suppose first I
must correct and reissue our paper.

Feel free to play with the excell simulator here:

http://uscountvotes.org/ucvAnalysis/US/exit-polls/simulators/dopp/

I've attached a graph from it.

Best,

Kathy

_______________________________________________
OVC discuss mailing lists
Send requests to subscribe or unsubscribe to arthur@openvotingconsortium.org
==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================

Received on Tue May 31 23:17:38 2005

This archive was generated by hypermail 2.1.8 : Tue May 31 2005 - 23:17:52 CDT