Forecasting elections, with my boss Bob Novak, used to be my full-time job. We did pretty well. In 2002, we correctly called a 51-49 U.S. Senate (we got two races wrong, and they cancelled each other out), and in 2004, we also did very well on House and Senate races, plus we called the Electoral College, I recall, within one state. In 2008, I thought Obama would get 52% of the vote, and 282 electoral votes (he got 53% and 365). I also called the Senate races correctly that year.
I don’t have full time to dedicate to examining polls, calling politicos in every state and visiting every swing state, but I’ve done a modest amount of this sort of reporting an analysis, and I wanted to share with you my thinking and my conclusions.
My wife, who is very good at forecasting elections — she’s won election-predicting pools against me and my friends — began a spreadsheet last night, and we finalized it after running out of candy. I’ve tweaked our joint predictions, and explained them below. Let me walk through the basics.
First, we used RealClearPolitics map as a base. Romney has basically 191 electoral votes locked up, and Obama has 201.
Then, as a starting point for each state, we went through the recent polling.The RCP average is a good starting point, but you need to look at other factors.
1) Some polls deserve more weight
2) Is there a trend?
3) Is the incumbent significantly below 50%
4) What extraordinary circumstance, or additional reporting reveal or suggest factors that wouldn’t show up in the polls?
So here we go:
Wisconsin (10): The RCP average has Obama up 3.7. But I put special emphasis on the Marquette Poll, because it’s a state-level poll, and it has the largest sample size of any poll. Marquette has Obama up by a whopping 8 points. Another 1000-likely-voter poll, Marist, also has Obama ahead. Also, importantly, Obama is at 49 or above in every poll but one going back a month (more on this principle later). My wife adds that Wisconsin always votes D in presidential races. So, Wisconsin is Leaning Obama.
Florida (29): Polls nearly show a tie, with the average leaning barely Romney’s way. The only recent local poll, which sampled more than 1,000 likely voters, has the biggest Romney lead. Three of the most recent polls also show Obama below 48 percent. Any time an incumbent is at 47 or below, it’s very concerning, because it shows deep discontentment, and undecideds often break to the challenger, in my experience.
This incumbent rule is mitigated if the challenger is either very well known or has high negatives. I could see Romney’s negatives being high among seniors, worried about him stealing their Medicare as we know it, or however Democrats are putting it these days. But ultimately, the weight of the data suggests to me that Romney is more likely than Obama to carry the state. Leaning Romney.
North Carolina (15): Obama carried North Carolina four years ago and expected to play very strong here this year — as signalled by his holding the convention here and some of his subsidies. But then a series of unfortunate events befell him: he was pushed by Joe Biden into endorsing gay marriage just as No. Carolina voted down gay marriage, the state party was racked with scandal, and the Democratic Governor imploded. His campaign has basically withdrawn from here.
N.C. polls lean Romney’s way. The poll with the largest sample size has the largest Romney lead. A local poll with a four-figure sample size shows a tie, but it has Obama 5 points down. Considering Obama won this state, and has spent much more time here than Romney has, this is a very bad sign for Obama. Leaning Romney.
Colorado (9): This is one of the toughest to call. Colorado has been a nearly 50-50 state lately, but with a definite Democratic streak. In 2002, Colorado voters re-elected Gov. Bill Owens and Sen. Wayne Allard. Those are the last top-level statewide races won by the GOP here. Today, Democrats hold both U.S. Senate seats and have their second consecutive governor. Bush did win Colorado in 2004, but if you look at Presidential, senate, and governor races Democrats are winning 6-2 over the past decade.
The polls show a tie. In some polls, Obama is down below 48 percent, which is troubling to him in a state that knows him well. But a closer look at the polls shows that the bigger the sample size, the higher Obama polls. Leaning Obama.
Pennsylvania (20): Obama is significantly up in all the polls, and consistently around 50 percent. Leaning Obama.
Iowa (6): Unemployment in Iowa is low, which is good for an incumbent. These guys have been raking in green-energy subsidies like there’s no tomorrow. Obama is whacking the GOP for holding up farm subsidy legislation, too. If Romney has a chance, it would be from a hidden evangelical surge.
Iowa polls show a small Obama lead. The only poll that shows Obama significantly below 50 is the poll with the smallest sample size. Leaning Obama.
Michigan (16): I think Obama has to worry about Michigan more than he has to worry about Pennsylvania. Sure he can tout his bailout of failed Michigan-based megacorporations General Motors and Chrysler, but this state ranks 4th in highest unemployment.
Polls show a slight Obama lead, but notably, Obama has cleared the 50 percent mark in only one poll since September. But the “incumbent factor” here needs a serious modification, because Romney is a tiny bit of an incumbent, too. He was born in Michigan. He campaigned his butt off to win the primary here this year — and the same thing four years ago. So it matters that Romney has only beaten 46 percent in one poll since the convention. Leaning Obama.
Nevada (6): Obama is at 50 percent or just below in every recent poll here. The 2010 elections show that Democrats have a, well, sophisticated turnout machine in the home state of Dirty Harry Reid. Leaning Obama.
New Hampshire (4): Obama leads in the polls here, hitting 49 percent in the three most recent and three biggest polls, including the only poll to hit 1,000 likely voters. As in Michigan, the “Incumbent Factor” applies Romney nearly as much as it applies to Obama. N.H. is next door to Massachusetts. Romney won the primaries here, and basically lived here in 2008. Leaning Obama.
Ohio (18): Yeah. Ohio.
Obama has a slim lead in Ohio polls. The largest poll, a local poll, shows him up 2. Only one poll, Rasmussen, shows Romney ahead. I just got back from Ohio, and I saw a clear path to a Romney victory — winning back white suburban voters who chose Obama in 2008 — but I also have reasons to suspect he won’t win back enough of that vote to carry the state. In short, if Obama gets even 45 percent of the suburban vote, I think he wins the state. Rasmussen’s poll showed Obama tied among six-figure earners.
Obama’s done a good job in the Toledo area making the case his auto bailouts saved the economy there. Unions will have good turnout machines around there. On the other hand, the Ohio GOP is one of the more functional state parties in the country.
Still, there’s a bit more weight on the Obama side of the scale. Leaning Obama.
Virginia (13): One of the hardest fought states in the country, this is also one of the toughest to call. Polls show a tie. But Obama has only hit 50% in one recent poll. And in four of the five polls last week, he was below 49%. In Virginia’s `burbs, I think you might see a significant swing back to the GOP, as we saw with the 2009 governor race. If black turnout ticks down a bit, and if Obama loses a bit of the youth vote, it becomes hard for him to hold on here. Leaning Romney.
Total: Obama 290, Romney 248
Here’s my caveat: When I predicted elections for Novak, I did everything on a discrete, local level. That’s why I was very right in 2002 and 2004. It’s part of why I was wrong in 2008. Some years, there is a big national tide that can overwhelm some of the local dynamics.
Potential national trends: Romney could have a silent-majority evangelical surge, as Dan Henninger suggests in today’s Wall Street Journal. Obama could get great unemployment numbers tomorrow. Obama’s well-funded, well-oiled turnout machine could swamp Romney and flip Virginia, and maybe even Florida. There could be a “Bradley Effect” — white people wanting to tell pollsters they’ll vote for the black guy, even if they won’t
But in the prediction game, we’re always just playing the odds — saying what we think is most likely. The 292-246 Obama victory is most likely, but I wouldn’t be surprised at all by anything up to 331 for Obama, or up to 269 (a tie) for Romney.
CORRECTIONS: I had misstated Florida’s and Pa’s electoral votes previously.