Nate summarized our discussion about the Republicans and the Latino vote by pointing out which states the Republicans need to win to be competitive in the electoral college. I'd just like to go one step further and say that most of the story is the national swing. If the Republicans can get a 4% national swing, that will probably do it for them. (McCain got 46% of the vote, so adding 4% would bring the Republicans to 50/50.) The electoral math matters a little bit as, for example, we've discussed regarding the Latino vote in Florida. But, pretty much, what you need is that national swing. It's not really that much different in practice from trying to win the national popular vote. Our calculations from previous elections showed that if either party won 51% of the popular vote, they'd have about a 90% chance of winning the electoral college. (This calculation is based on research that I published a few years ago with Jonathan Katz and Gary King.) Granted, the vote might be close, and if it's a 50.5%-49.5% margin, the electoral math will matter. And, during the campaign, the parties have to decide where to spend their resources. But overall I think the state-by-state electoral college counting is a bit of a distraction.
P.S. I added a couple of clarifications above to address some of the commenters who were questioning our numbers.
P.P.S. This post is not tautological. The context is that people discuss electoral strategies, which states the Republicans need to target, etc. What I'm saying is, sure, targeting key states is important during the campaign. But that's all minor compared to the larger goal of national popularity. For the Republicans (or, for that matter, the Democrats) to improve their chances for 2012 and 2016, right now they have to be thinking about what will swing voters at the national level. There's not a lot of evidence that you can easily push buttons and swing particular voting blocs or states.
I could imagine a world in which candidates could win elections by targeting particular states. That's just not the world we live in. We live in a world of approximate uniform swing. Recall these graphs:
The swing from 2004 to 2008:
The swing from 1980 to 1984:
The swing from 1952 to 1956:
Not exactly a random scatterplot but, again, more variation than we saw from 2004 to 2008. Actually, the variation from 1952 to 1956 and from 1980 to 1984 is more comparable to the variation in two recent elections, say from 2000 to 2008:
So, the data say that swings are more national than they used to be.