Suppose that you are playing a game in a hilly landscape, and the object of the game is reach a higher altitude than any other player, and all you've got is your own feet & eyes (no GPS, etc.) What do you do? Well, presumably you try to find the highest hill you can spot, and head straight for the top. But what if you're playing in a thick fog, and can only see a few dozen feet in any direction? In such conditions, you can determine what your local slope looks like, and climb up it. And you can reach the top of whatever hill you're on that way, just by always climbing higher and higher until you run out of up. But... what if that hill is not, in fact, the highest hill? You'll be at a top, but not the tallest top. What then?
If you're on top of a hill, but it isn't the highest hill, then you are at what AI researchers (among many others) call a local maximum. Many problem-solving tasks can be modeled as playing this sort of foggy hill-climbing game. And one always runs into this basic dilemma: make your problem-solver too risk-seeking, and it'll trade away a perfectly good hilltop for bupkis; but make your problem-solver too risk-averse, and it'll trap itself in a local maximum, unwilling to climb a little bit down in order to have a chance at finding an even higher hill to reach the top of.
From the conversations that have been going on on this site, it seems to me that the right today is in a real danger of being caught in a local maximum. We know that the coalition that has worked at least since Reagan is no longer working. After Bush and Rove and Iraq and Katrina and... well, the independents just don't seem to trust the right anymore. To the extent that they are interested in McCain, it's because they accept that he's not really a right-winger (as many around here are happy to agree.) And the demographics are against the right, too, in terms of the youth vote, Latinos, etc. Trying to just keep pushing forward with this coalition -- staying put on top of this particular hill -- is a recipe for short-term electoral suicide. Maybe the left will burn itself out in a few cycles, but I wouldn't count on it; they've learned from their losses in the mid-90s, and from watching the GOP meltdown in this century.
And it's worse than that, even, because the current coalition is fraying. There are too many tensions that have become too clear in the last decade. We can't have both lower taxes and an invasion of Iran. We can't have a smaller, less intrusive government and unlimited domestic security powers. We can't have a laissez-faire government and a government that continually helps out agribusiness, or the airline industry, or the banking industry. Yet all of these are elements that have been crucial to GOP electoral success thus far.
The right needs to climb down a little, before it can climb higher. And what that means, in cashing out the metaphor, is that some elements of the current coalition are probably going to get jettisoned in order to better pursue new voters that haven't been reached before, or who have recently been lost. Some might think that "just return to our conservative roots!" is a solution that avoids this difficulty, but that's an illusion -- much of the reason that GOP legislators vote for government- and deficit-expanding measures is that there are some voters that these actions keep in the GOP fold. Ditto with, e.g., Medicare D. To propose more small-government conservatism is to propose to lose those voters. Maybe the true-blood conservatives whose votes will be gained will more than compensate for the kitchen-counter voters who will be lost. But it's a hard answer to determine one way or another, in the current fog.
Let me be clear that I am not advocating here any particular rejiggering of the conservative coalition, nor am I making any suggestions as to which groups of voters are those that are not currently in the coalition but which should be pursued. But I think it's important to face up, as clearly as possible, to the difficulties being faced standing here at the top of this no-longer-dominant hilltop.