Romney: possibly West Virginia
Yeah, no. Obama's still polling in high 30's to low 40's even when undecideds are counted. He'll at least win McDowell and Boone, and probably Jefferson since it's an extension of NOVA.
Last poll, which I think had a Democratic-lean considering the numbers it gave to Manchin and Earl, had Romney up by 17, 54-37. Undecideds typically break against the incumbent, but for arguments sake, let's say it's a pretty even split. That gives Romney around 58% of the vote, compared to Obama's 41%. That swing would give Romney every county in the state - the only county that will be close under this scenario is McDowell. Hence the word "possibly" I added.
With that pattern, incumbents would largely lose most elections.
If I have seen any pattern, it is that the undecided tend to swing inadequately toward the one behind at the time. Those near the middle tend to split roughly 50-50 most of the time. A 60-30 split tends to go 62-37 or so (and not to 67-32 or so), and a 46-44 split tends to go about 51-49. The usual explanation is that if one has a 60-30 split the undecided are on the ideological side of the candidate in the lead and are likely to 'come home' to old patterns of voting. Near the middle the ideological split among the undecided is usually about even. So it was for Goldwater in 1964, McGovern in 1972, and Mondale in 1984 when the challenger was far behind.
This does not explain collapses (Dukakis 1988) or near-collapses (Carter 1976, Bush 2004) that have better explanations. If the incumbent is at all competent as an administrator or legislator and as a politician, he usually shows why he was elected in the first place. Of course it the incumbent has an inept campaign (Ford 1976), disappoints the public (Hoover 1932, Carter 1976) or can't effectively explain why he should be re-elected (Bush 1992) he then shows that electing him was a mistake.
The Lichtmann test has been far more reliable in the past than almost any other seat-of-the-pants prediction for the Presidency. The only ones with more reliability are the ones more quantitative (such as the one by Nate Silver that estimates that those incumbent Governors and Senators with a 44% approval rating at the beginning of the campaign season have a 50% chance of winning re-election with a precipitous drop-off of a chance of winning for those with less than a 44% approval rating and a sharp gain of a chance for an incumbent with more than a 44% chance of winning.
An explanation for Silver: elected incumbents have usually shown that they are competent campaigners with good organizations behind them, which is good on the average for 6% of the total vote. Challengers can carp all that they want while incumbent politicians are obliged to make polarizing decisions, but once election time comes, those politicians who are still near or above 50% usually win if they can have a spirited campaign and an effective GOTV program. (The blatant exception might be for appointed pols who never proved that they were competent campaigners, which demonstrates itself in the poor record for appointed pols winning re-election).
