The underlying model and algorithms for
The Political Machine are pretty
good. Ultimately, the model is a mishmash of polling data from dozens of sources
combined with demographic data. What, I think makes The Political Machine
special is that it makes heavy use of demographic data rather than just looking
at polls. What that means, fundamentally, is that at the end of the day,
conservatives tend to vote for Republicans and liberals tend to vote for
Democrats and once you understand that, it's really a matter of turn out on both
sides. There really aren't nearly as many "independents" as are sometimes
claimed. Independents tend to be conservative or liberal in their views and as
such, they tend, on a macro-scale, to vote one way or the other consistently
enough to model with some degree of accuracy.
And that's bad new for Bush. Because barring a miracle, Bush has already
lost. Not by much but the issue is electorally. We've run the simulation
countless times in house and we come up with different results:
First of all, there really aren't that many toss-up states. The media
has states like Michigan and Wisconsin and Washington State and Oregon as
toss-ups. The only possible way they're toss-ups is if Ralph Nader is able to
get a significant share of the vote. But even there, people who think Nader is
going to get 6% of the vote aren't realizing the infrastructure involved in
getting onto the ballots in all the states and maintaining some semblance of
momentum.
So who should be Kerry's Vice Presidential pick? If he's smart, he'll pick
Richard Gephardt. By bringing him in, he gets Missouri and probably Iowa (our
model doesn't yet handle the VP pick effect very well). So write this prediction
down so that if we're right we can be oowed and awed appropriately: 
Without Gephardt VP choice:
Kerry 279 / Bush 259
With Gephardt VP choice:
Kerry 290 / Bush 248
Now, you say to yourself, how can you be so sure? We're not. We're confident
we'll be close but there are a few real toss-ups in which the events between now
and election day will matter. Here are the real toss-up states:
- Ohio
- Missouri
- Iowa
- West Virginia
Bush has to win all of these to win. Barely.
States that COULD become toss-up states based on events:
- New Mexico
- Arkansas
- Wisconsin
Bill Richardson as a VP pick could shore up NM if something causes it to be a
toss-up, but that's unlikely. General Clark could bring Arkansas into play.
Some pollsters have said that some blue states have become "purple".
I'd like to know which ones they think are purple now. I can't think of any blue
states that have softened since 2000. Wisconsin? Possibly. But only if Nader
becomes a major factor there again. We do think Florida will be much more solid
for Bush this time, for what it's worth.
But go ahead and look at the map above, which states do you think Bush will
win that we're wrong about?
We'll keep making predictions, after all, what good are computer models for?
And you can then check the accuracy after the election to see how well we did.