Baker (so thoroughly political a man that he ended up marrying another Senator) has a pendular theory of how the American Bipolarchy works. He's no believers in great ideological shifts, but argues that voters shift loyalties back and forth based on an appraisal of each party's competence in government. If Democrats screw up, people will vote Republican, and vice versa. As he puts it.
Things change because things change, not because of any ideological primacy or purity on a particular end of the political spectrum. The American people are, for the most part, highly practical and pragmatic. They like what works, and in a properly functioning political system, two broad-based national parties will offer them reasonable alternatives for what is likely to work best.
Reasonableness depends on a broad base, if I understand Baker, but it would seem that the two dominant parties don't need to be particularly reasonable so long as each is perceived as the only "reasonable" alternative to the other. According to his reading of post-1964 politics, Americans embraced an increasingly conservative Republican party not because they had become increasingly conservative themselves, but because they had decided that the Democratic party under LBJ had failed at some important task. This analysis suggests that a major factor in the Bipolarchy's staying power is Americans' short collective memory.
A voter may think that the Republicans failed in the mid-2000s, for instance, but will not recall that he quite probably felt that the Democrats had failed as of 1994. By a certain point we might expect a large number of voters to conclude that both parties had consistently failed, and to search more extensively for alternatives. They don't, in part, because they can convince themselves that Obama, for instance, will do things differently from Clinton, or that the Republican nominee for 2012 or 2016 will do things differently from George W. Bush. Personalities disguise the persistent institutional incompetence (measured by electoral defeats) or the major parties and allow people to convince themselves that they're voting for people, not parties.
Baker himself goes on to suggest that there's more to elections, after all, than mere appraisals of competence. His imagined voter seems to swing not just between parties but between two conflicting sets of "core" values. Eight years of Bush, Baker admits, have momentarily discredited the "core Republican beliefs in less government, lower taxes, more liberty and greater security in a dangerous world." But "if the American people perceive overreaching or underachieving in the Obama administration and among its allies in Congress, the Republican way may prove very attractive again in very short order." This, too, would seem to depend on short collective memories, unless Baker wants to argue (from the safety of retirement) that each party's core values may be more effective at different points in history. It also helps the Bipolarchy that each party's core values are not as consistent as Baker insists. The ones he lists might not have been recognized as "core Republican beliefs" 100 years ago, for instance. Nevertheless, Baker seems satisfied that, between them, the two major parties will provide Americans with all the reasonable alternatives they need at any critical moment -- and that strikes me as a purposefully unexamined premise. Unfortunately, that doesn't mean that history will prove him wrong.