I really enjoy Jonathan Rauch's work, but I often disagree with him, and I certainly think he's wrong about divided and unified government.
Yes, it's true that people will tell pollsters that they like divided government. They'll also tell pollsters that they don't like either party, that they are independent voters, and that they vote the candidate, not the party.
All bunk. Most voters are either hard or soft partisans, as John Sides will tell you. (By the way, easy short cut? On the surface, the electorate is about one third Republicans, one third Democrats, one third independents. Ask the right way, and you find out that of those independents one third are basically Republicans, one third Democrats, and one third real-life true independents. So one third of one third, or about 10%, are real independents).
As for voters actually doing what Rauch suggests and constantly trying to achieve divided government, the evidence that any of that is happening is very slim. One might note that for all of 2010 and 2006 yielding divided government, in 2008 (and 1992) voters opted for unified government. An electorate that actually acted on stated claims of preferring divided government would consistently try to achieve that goal in presidential election years, but that's not the case -- the party winning the presidency usually adds, not loses, seats in Congress.
Rauch claims that "partisan bomb-throwing escalates" during periods of unified government. Certainly, there was plenty of angry rhetoric in 2009-2010. But more than in 2007-2008? I don't know; angry partisan rhetoric from the Republicans was up, but I'd say that it decreased among Democrats. Surely, the recent peak of angry partisan rhetoric was during the run-up to impeachment in 1998. Under, of course, divided government.
At any rate, disgust with angry partisan rhetoric doesn't account for the 2010 elections; that would be the recession, clobbering Democratic incumbents in 2010 just as it hurt Republican incumbents in 2008.
Rauch also claims that "meaningful laws stand a likelier chance of passage" under divided government. As it happens, there's a major academic debate about whether divided government does affect the chances of passing major legislation, but the debate is over whether it's less likely or equally likely; no one who studies this finds any evidence for Rauch's claim in the empirical record. Surely, Rauch doesn't expect the 112th Congress to be as productive as the 111th, does he? I'd take that bet. Nor is there any empirical support for the idea that a GOP House will help Obama be more popular.
Bottom line: voters are partisans, whatever they say, and partisans want their party to control as many offices as possible, whatever they say.