[extropy-chat] looks like Bush will win

[Hal Finney]

For those who are interested in following the American election race,
here are some useful links.  http://pollingreport2.com/wh2004a.htm shows
the recent national polls.  Among likely voters, recent results show a
tie or are within the margin of error.  This is a slight reduction of a
Bush advantage last week.

http://www.electoral-vote.com/ shows the state by state results.  This is
the most-watched election site on the web.  American presidents are
elected by a system where states have a certain number of electoral votes,
and most states assign their votes on a winner-take-all rule based on the
popular vote in that state.  So the national poll numbers above don't mean
that much.  More important is how it shakes out on a state by state basis.

electoral-vote.com is run by an openly partisan supporter of Kerry,
but in my judgement he is neutral and fair in reporting the state by
state polling results.  Apparently because of some pro-Kerry material
on his site, it is under frequent attack, so there are mirrors at
electoral-vote2.com through electoral-vote7.com.  Personally I've never
had any trouble getting through to the main site.

You can click on a link to see the previous day's results and track the
data back in that way.  The data is extremely noisy.  Kerry is ahead
today; Bush yesterday; Kerry the 5 previous days;  Bush for quite a
while before that.  You can also click on a state to see how it has been
trending in graphical form, a wealth of information.

Below the main map is a link to Predicted Final Results, visible at
http://www.electoral-vote.com/pred/.  This assumes that "decided"
voters stick to their choice, and undecided voters break 2 to 1 for
the challenger.  The rationale for this assumption is explained at
http://www.electoral-vote.com/oct/oct20.html ; it turns out that
historically this is what has happened.  Incumbents seldom draw more
votes than their last poll numbers before the election.

This prediction is showing a Kerry win.  Several of the states that are
tied or very close will go to Kerry on the 2-to-1-undecided assumption.

So we see national polls showing a statistical tie, with Bush losing some
ground from last week; day-to-day electoral vote counting seesawing and
too close to call; and this prediction model pointing to a Kerry win.

In opposition to this are the betting markets.  This is a marvelous
innovation and moves us closer to a world where Robin Hanson's ideas are
given a real trial.  http://128.255.244.60/graphs/graph_Pres04_WTA.cfm
is the Iowa Electronic Market chart showing the market odds of winning.
Bush has led consistently since the Republican Convention, reaching
his peak just before the first debate.  Since then he has lost some
ground but is still the predicted winner.  Keep in mind that these
are not electoral vote count predictions, merely the odds of winning.
Bush is currently predicted to have a 59% chance of winning.

http://www.tradesports.com is potentially more accurate because it
allows larger bets, unlike the IEM which puts a $500 cap on traders.
At this point it also predicts a 59% chance of Bush winning.

Of course we should keep in mind that a 59% chance is not that high.
It's not the same as a candidate having a 59-41 lead in the polls,
which would mean he is sure to win.  If you had a biased coin which came
up heads 59% of the time, you'd have to toss it quite a few times to
distinguish it from a fair coin.  So to some extent these markets are
telling us that it is still an open race and could go either way.

Based on these results I'd say it is premature to forecast a Bush win.
The national polls do not show momentum in his favor, and the state
by state model gives Kerry the advantage.  Against this, the betting
markets slightly predict Bush, but they're not that far from a coin flip.

One final, personal prediction.  I don't think the election will be
that close, not as close as in 2000.  You may hit the bullseye with your
first arrow, but you're not likely to split the arrow with your second.
Odds are against the election results being flippable if only a few
hundred votes changed.  On the other hand, now that the parties know it's
possible, their lawyers will be out in earnest looking for any results
they can challenge or contest.  So it's possible that we won't see an
immediate concession.  But I suspect that it will be pretty obvious
which guy won by the end of the evening.

Hal