[MLUG - DISCUSSION] my final prediction

[Jonathan King <EMAIL:PROTECTED>]

OK, so I have to post this and get on with my life. :-)  Here's my
best guess as to the popular vote in tomorrow's election, based solely
on current and previous polls from 2000 and 2004.

049.3 Kerry
048.7 Bush
000.9 Nader
001.0 All others combined
--------
100.0 Total

Method: last time out, the top 15 polls gave an average reading of
Gore 44.2%, Bush 46.5% Nader 4.5%;
Bush ended up with 47.9%, Gore with 48.4% and Nader with 2.7% with 1%
straying elsewhere, as always.  Averaging the top 10 polls this time
we have: Kerry 46.9%, Bush 47.6%, Nader 0.9%.  Last time, Bush gained
1.4% on his last average number, while Gore gained 4.2%, but Nader
lost 1.8%.  If we give Bush his 1.4%, allow 1% for "Other" (always
close to this) and give the rest to Kerry, we get:

049.1 Kerry
049.0 Bush
000.9 Nader
001.0 Other
------- 
100.0% Total

The difficulty with this approach is that the Nader vote collapsed 
last time, which bumped Gore up smartly; this time, we're more in a
position of allocating undecideds.  If we take the 3.6% truly
undecided implied by "other" getting 1% and the average final polls,
and allocate it 2:1 for Kerry (almost exactly the historical pattern)
we get:

049.3 Kerry
048.8 Bush
000.9 Nader
001.0 Other
---------------
100.0% Total

A third approach is to look at the 3 polls that got Gore's total right
within 2% last time and that are polling this time (Harris apparently
isn't?)  Their average last time was:

047.0 Gore
047.6 Bush
004.3 Nader
001.1 Other
-------
100.0% Total

The big error was in the Nader + other category; only 3.8% total went
for these guys, and Bush gained 0.3% while Gore gained 1.4% (the
rest).  This time, the same 3 polls (Gallup, Zogby and Princeton/Pew)
have:

047.0 Kerry
047.4 Bush
001.0 Nader
001.0 other*
003.6 undecided
-------
100.0%

I asterisk "other" since they don't appear in the polls, but they
monotonously get about 1% of the vote, so nobody doubts they are
there.  Dividing "undecided" 2:1 against the incumbent is pretty much
historical, so we get as a final tally:

049.4 Kerry
048.6 Bush
001.0 Nader
001.0 other*
-------
100.0%

Taking the average of my two models, I get (with some not very
creative rounding) the 49.3/48.7/0.9/1.0 split that I predict at the
top of this.  In other words, I have Bush losing the popular vote by
500K to 1 million.  But I *don't* have Kerry breaking 50%. That could
happen, but would require an undecided break pretty massively against
Bush which I just don't see.

So have it, guys.

jking
_______________________________________________
discussion mailing list
EMAIL:PROTECTED
http://mlug.missouri.edu/mailman/listinfo/discussion