Pollsters know it instinctively. Commentators debate its nuances, while journalists largely ignore it. Named by Nick Panagakis in 1989, the “incumbent rule” argues that most voters who are undecided late in a race against an incumbent will break toward the challenger. Thus, the important poll number is not the margin between an incumbent and a challenger, but rather the percent of the vote garnered by the incumbent.

Pollsters know it instinctively. Commentators debate its nuances, while journalists largely ignore it. Named by Nick Panagakis in 1989, the “incumbent rule” argues that most voters who are undecided late in a race against an incumbent will break toward the challenger. Thus, the important poll number is not the margin between an incumbent and a challenger, but rather the percent of the vote garnered by the incumbent.

One of the first races I did, in 1982, matched an unknown legal aid lawyer against an incumbent House member. When Congressional Quarterly listed 125 competitive races, ours was not among them. Understandably. Six weeks out, we were behind by over 20 points — about 48 percent to 25 percent. I urged the candidate to ignore the margin, noting that the central fact was that the incumbent was under 50 and that most of those undecideds would come our way. Six weeks later, after just 10 days of carefully crafted TV ads, we won by less than 1 percent and I had a career.

Some have reified this “rule” into an iron law, ignoring the mechanism through which it operates. During the last Presidential campaign, some rejoiced when we were a couple of points behind, confidently asserting since the undecideds would break against incumbent President George W. Bush, Sen. John Kerry (D-Mass.) “had it won.” Much of the optimism that permeated activist circles in the waning days was based on blog posts invoking the incumbent rule. One analyst applied the rule on a state-by-state basis and forecast a 327 electoral vote victory for Kerry.

Others of us were quite doubtful. The “incumbent rule” is not a mechanical result, but emerges from a particular underlying structure absent in the 2004 presidential contest, as in some others.

Most incumbents are much better-known than their challengers. So when a voter says they are “undecided,” what they really mean is that they know the incumbent and do not particularly want to vote for him or her. However, they do not yet know enough about the challenger to support that candidate. By Election Day, these voters pick up at least a scintilla of information about the challenger, enough to justify casting a ballot against the incumbent.

Mostly that’s the way it works. From the final polls to Election Day, in 155 Senate races involving incumbents from 1994-2004, those incumbents picked up an average of 2.8 points, while challengers gained 6.4 points. Thus, on average 70 percent of undecideds broke to the challenger, 30 percent to the incumbent.

It doesn’t always work that way, however. Last cycle Sen. Jim Bunning (R-Ky.) picked up 7 points from the last poll while his challenger added 6 points to his total. John Kerry and George Bush split the undecideds evenly, 1.1 points each, from the average of the final polls.

When undecideds don’t break to the challenger it is for a reason. In those circumstances, voters either know the challenger and do not particularly like him or learn something disqualifying in the closing hours of the campaign.

In the presidential race, those who remained undecided disliked both candidates, albeit for different reasons. They were not unfamiliar with Kerry, but had grave doubts about him and about President Bush. These were true undecideds — uncertain not out of ignorance, but out of perceived knowledge that left them with serious reservations about both candidates. Bunning unleashed a broad last-minute fusillade against his under-funded opponent that left voters unwilling to take a chance on the challenger.

So headlines heralding an incumbent’s 45-35 margin as a “big lead” betray an ignorance of what the numbers mean. But keep in mind, the “incumbent rule” is not an unbending law — its applicability depends on individual campaign dynamics.

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.