Rhodes Cook, one of our most astute and least heralded political analysts, continues a venerable tradition of focusing not on polls, but on counting votes. Prior to November, he divined three patterns among big-wave House elections.

A “one-party surge” occurs when that party significantly increases its national vote, while the vote for the other party remains essentially unchanged. Such was 1994: The GOP vote ballooned by almost 9.5 million over the previous midterm, while the Democratic vote dropped by about 400,000. 1974 witnessed a “one-party collapse,” as Democrats gained 1.1 million votes from 1970, but Republicans lost nearly three times as many. Finally, there are “unequal gains,” typified by 1982, when Republicans added a substantial 3 million votes, but were outdone by Democrats, who picked up an additional 6 million votes from 1978.

The dramatic fact about 2006 is that it fit none of the prior patterns. It combined a one-party surge with a one-party collapse. Democrats added over 5 million voters, while the GOP tally declined by over 3 million from the last midterm. The Democratic victory and the repudiation of Republicans were both broad and deep.

As instructive as these vote counts can be, they do have limitations. In studying a relatively small number of wave elections, Cook identified three different patterns. Yet 2006 fell into none of the preexisting categories. It was different. While the past is always prologue, it is not always prophetic. New patterns and new relationships can emerge.

Karl Rove apparently neglected this lesson, assuming patterns he found in two previous elections would hold in this one. As Newsweek reported, “The polls … pointed to a Democratic sweep, but Rove dismissed them all . . . He believed his ‘metrics’ were far superior to plain old polls … before the elections, Rove showed Newsweek his magic numbers… graphs and bar charts that tallied early voting and voter outreach. Both were running far higher than in 2004 ... Based on his models, he forecast a loss of 12 to 14 seats in the House.”

Late-night C-SPAN watchers may have seen me impolitely berating Rhodes Cook over another limitation.

He and too many other analysts use these vote totals to evaluate turnout efforts. After 2004 they focused on the GOP “turnout machine,” arguing Bush’s gain of 11 million votes over his 2000 showing, compared to Kerry’s 8 million vote increase from Gore’s total, demonstrated the power of Republican turnout efforts. Cook took that “lesson” into 2006, where the faulty premise led to an inaccurate prediction: “A one-party surge in the Democratic vote in 2006 or a one-party collapse in the Republican vote is unlikely, given the GOP’s recent success at voter mobilization.” In fact, both “unlikely” events actually occurred.

At root the problem is conflating turnout with persuasion. Changes in aggregate vote totals result from increases in turnout, decreases in turnout and from people who switch their vote from one party to another. The aggregate vote totals cannot tell us which underlying process was responsible for the overall result.

Imagine an electorate where 100 people voted, dividing 50-50. In the next election, exactly the same 100 people vote but it’s 60-40. By the method of the vote counters, “turnout” for one party “increased” by 10 votes while “turnout” for the other “declined” by 10. But in fact turnout was completely unchanged. The difference in the vote totals was wholly attributable to “persuasion” — voters shifting from one party to another — and not at all to a differential ability of either party to turn out their voters.

While the real world is more complex, vote counts alone do not tell us everything we would like to know.

Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982, including Sen. John Kerry (D-Mass.) in 2004.