For most of the summer, I've been doing weekly simulations of the general election using polling data from Andrew Tanenbaum's www.electoral-vote.com. And since Barack Obama clinched the Democratic nomination, he's held a commanding lead over John McCain in these simulations. Now, however, the landscape has changed, with the race currently looking like a very close. As before I'm doing two different runs, one using the 4% margin of error common to most state polls, and another with an 11.1% margin of error, taken from a regression comparing 2004 polling data against the actual results.

When you just sum electoral votes of who is ahead in each state at www.electoral-vote.com, McCain now leads, 270-268, his first lead since May. The current breakdown has Obama winning all of the states John Kerry won, plus winning Colorado and Iowa. Obama still wins a little more often in each simulation, because McCain has more electoral votes in states with very small leads, and Obama needs to flip just one more state to win. The summary:

4% Margin of Error
Obama wins 53.4%, averages 272.7 EV (low 219, median 273, high 328)
McCain wins 46.2%, averages 265.3 EV (low 210, median 265, high 319)
Electoral tie 0.4%

11.1% Margin of Error
Obama wins 52.2%, averages 270.9 EV (low 131, median 271, high 379)
McCain wins 46.1%, averages 267.1 EV (low 159, median 267, high 407)
Electoral tie 1.7%

The two runs differ mostly in spread, and the wider margin of error case results in a tie a lot more often. Most recent national polls show McCain leading by between 1 to 4
points, so at this point he might be ahead in the popular vote. Obama
is still slightly ahead in the electoral college. I think it is far more likely that Obama could win the Presidency while losing the popular vote than McCain could.

The shifts come as there are more post-convention state polls, which have tended to favor McCain, albeit less on average than his shift in national polling. Florida continutes to be a volatile state, with two polls this week showing large McCain leads (Insider Advantage 50-42, Quinippiac 50-43) but another showing a dead-heat (Rasmussen 48-48). These three average to McCain ahead by 5, a solid lead. The three prior polls had Obama leading in one, with an average edge to McCain of 1 point.

McCain has retaken a slight lead in New Mexico, up by 2 compared to two prior polls from August that had Obama up 13 and up 6.


Michigan also is tightening, with three new polls averaging to a 3 point Obama lead, including the first poll to have McCain leading the Wolverine state since late May. Similarly the last three pre-convention polls of Pennsylvania showed Obama ahead by 5-7 points; the two surveys since have shown Obama leads of just 2 and 3. So the model, which before had basically given PA and MI to Obama, now sees McCain winning those states more of the time.

Virgina was a toss-up in August (McCain and Obama each led one poll by a point, and a third poll was tied), but now it's tipping towards McCain, with three new polls giving him leads of 2, 2, and 4 points.

In addtion to making states leaning to Obama more competitive, McCain's bounce also has increased his lead in states that had previously been quite close. Two North Carolina polls give McCain large leads, +13 and +18. McCain has strengthened in the upper great plains, retaking the lead in North Dakota (now by double-digits), and widening his lead in Montana from 1 point in late July to 11 now.

Ohio has 4 polls since the convention, with results ranging from Obama +5 (Quinippiac) to McCain +7 (Rasmussen), with McCain ahead by 2 on average, esentially unchanged since August, where four polls showed one tie and three McCain leads from 1-4 points. Missouri also hasn't moved much, with two post-convention polls showing McCain ahead by 5, similar to his average pre-convention lead.


27 of the 50 states have had polls in September, with most of the
"swing" states having multiple polls. But some states that were close
in their last poll don't have any polls this month. McCain led Indiana
by 2 in late August, Obama led Wisconsin by 5 in early August, and
McCain led South Dakota by only 4 in early July, but none have more recent polls. It would be good to
get newer polls from thes states, as well as Oregon (two early August
polls had Obama up 3 and up 9). If these states all show McCain gains of around 3-5 points, then he likely would win a majority of simulations with my model. But most of the states without recent polling (such as TX, NY, UT, and IL) are safe for the leading party.

If this is near the peak of McCain's post-convention bounce, and his numbers start to fall back, Obama is still in quite a strong position. But if this represents the new equilibrium, then this could be quite a close election. The current momentum certainly is in McCain's favor. I wouldn't panic if I were Obama, but I would remain nervous.