Josh Green's fascinating Atlantic Monthly article on Karl Rove highlights the outgoing presidential adviser's obsession with creating an electoral realignment. But as Matthew Yglesias points out, realignment theory is an antiquated approach that is no longer taken seriously in modern political science:
I will, however, note that what I found most fascinating about it was Josh's evidence that Rove's talk of masterminding an electoral realignment wasn't just bluster, but played an actual causal role in his thinking about the administration's political and policy choices. Maybe, then, Rove will be able to take advantage of his new, more relaxed schedule, to sit down and digest David Mayhew's Electoral Realignments: A Critique of an American Genre which argues convincingly that so-called realignments are a product of statistical naivete and the human penchant for hyperactive pattern detection rather than a real phenomenon of American politics.
Indeed, Green (an acquaintance of mine from when I lived in Washington) is arguably too respectful of realignment theory in this passage, especially given that he cites Mayhew's book later in the article:
Fifty years ago, political scientists developed what is known as realignment theory—the idea that a handful of elections in the nation’s history mattered more than the others because they created “sharp and durable” changes in the polity that lasted for decades. Roosevelt’s election in 1932, which brought on the New Deal and three decades of Democratic dominance in Washington, is often held up as the classic example. Modern American historians generally see five elections as realigning: 1800, when Thomas Jefferson’s victory all but finished off the Federalist Party and reoriented power from the North to the agrarian South; 1828, when Andrew Jackson’s victory gave rise to the modern two-party system and two decades of Jacksonian influence; 1860, when Abraham Lincoln’s election marked the ascendance of the Republican Party and of the secessionist impulse that led to the Civil War; 1896, when the effects of industrialization affirmed an increasingly urban political order that brought William McKinley to power; and Roosevelt’s election in 1932, during the Great Depression.
Academics debate many aspects of this theory, such as whether realignment comes in regular cycles, and whether it is driven by voter intensity or disillusionment. But historians have shown that two major preconditions typically must be in place for realignment to occur. First, party loyalty must be sufficiently weak to allow for a major shift—the electorate, as the political scientist Paul Allen Beck has put it, must be “ripe for realignment.” The other condition is that the nation must undergo some sort of triggering event, often what Beck calls a “societal trauma”—the ravaging depressions of the 1890s and 1930s, for instance, or the North-South conflict of the 1850s and ’60s that ended in civil war. It’s important to have both. Depressions and wars throughout American history have had no realigning consequence because the electorate wasn’t primed for one, just as periods of electoral unrest have passed without a realignment for lack of a catalyzing event.
Actually, the debate over realignment among political scientists is essentially over, and historians haven't "shown" anything.
For instance, the most recent examination of the supposedly critical 1896 election -- the key to Rove's theory of politics and the model for Bush -- finds no indication of a discontinuous break from the past (PDF, sub. req.):
There is no evidence of a significant shift in 1896 in the national vote for Democratic presidential and House candidates. There is no evidence that there was an abrupt decline in support for Democrats in House districts with higher density and higher percentages of the foreign born. The evidence about the coalition strategy of Democrats and the general terms of political discourse, although by no means exhaustive, indicates more continuity pre–1896 and post–1896 than change.
In the conclusion to the journal article that led to his book (sub. req.), Mayhew sums up the problems with the realignment literature:
The claims of the realignments genre do not hold up well, and the genre's illuminative power has not proven to be great. At an analytic level, the genre has proven vulnerable to at least three counterposing ideas: contingency, strategy, and valence issues.
Electoral politics is strongly influenced by the contingencies of unfolding events. To the extent that this is true, elections and their underlying causes are not usefully sortable into generation-long spans. Furthermore, victory-oriented strategy is plied by candidates and parties, both of which tend to cater to the electorate as well as emanate from it... To the degree that parties and candidates seek election victories above all else, courting the median voter, they will often accommodate major impulses from the electorate without telltale signs of realignment appearing in elections. Thus, the size of voter realignments cannot index the importance, innovativeness, and consequentiality of elections, nor the level of voter concern that underpins them.
If we consider the combination of contingency and victory-oriented strategy, certain results reported by Gans (1985) become understandable. In the sequence of presidential elections from 1856 through 1980, the distribution of victory "runs" by party (Carter, for example, was a run of one for the Democrats; Reagan's two victories and Bush's one made a run of three for the Republicans) did not differ significantly from the runs of heads and tails that would be expected from coin flips (Gans 1985:228–30). Also, in the absence of repeat major-party candidates (such as Reagan in 1984 or Bryan in 1900), a presidential election four years ago holds virtually zero predictive value for this year's election—either in predicting this year's victorious party or this year's party shares of the vote (Gans 1985:230–33).
The third idea that poses problems for the realignments genre is that of valence issues. The concept was introduced by Stokes (1966) and given major play, at least implicitly, in the sizable econometrics literature gauging the effects of ups and downs in the economy on elections. Unlike position issues, in which one party favors policy X and the other party favors policy Y—the staple kind of cleavage in the realignments genre—valence issues hinge chiefly on government management. Can another party manage the economy or the war, for example, better than the incumbent party has been doing? The more one examines American electoral history, the more it seems to tilt toward valence-issue as opposed to position-issue junctures. More than it did a generation ago, for example, the electoral turmoil of the 1890s seems to implicate the depression of 1893 as much as Bryan's insurgency. A poor economy figured in the critical 1874 midterm, and, according to one recent analysis, a quick nationwide economic downturn was a central ingredient in the Whigs' great victory of 1840 (Holt 1985). Valence issues, which also exemplify contingency and often bring into play opportunistic candidate or party strategies, are not friendly territory for the realignments genre.
The moral of the story is that Rove spent too much time talking to and reading historians and not enough time with political scientists. It's not hard to figure out why: historians tell stories, and humans love stories. Indeed, as Yglesias notes above and in a followup post, we're apt to see patterns in (possible) randomness. That's why we need quantitative political science to sort out the stories that are backed by evidence from the ones that are not. Unfortunately, most people don't enjoy reading quantitative academic research, especially if they don't have a strong background in math and statistics, so consultants and journalists tend to ignore our work and read the David McCulloughs of the world instead.