Posted 7:09 PM by Robert

Undecided Voters Break Towards The Challenger? III: Hearing Secret Harmonies

What Kausfiles calls the venerable "Incumbent Rule" - which holds that in the campaign's last hours "undecided" voters tend to "break" for the challenger - supposedly garners empirical support from a 1989 study by Nick Panagakis, president of Market Shares Corporation (the firm that polls for the Chicago Tribune), a study that appeared in a "famous" article in The Polling Report. The Panagakis study found that in 82% of the cases he selected the undecideds "broke" mostly to the challenger. Claims by the Mystery Pollster, for example, that the Panagakis study was based on polls which were all conducted during the last week before an election are flatly wrong. Further, most of the cases in the Panagakis were not particularly similar to federal Presidential races. Here's how Panagakis described the cases he considered:

The 155 polls we collected and analyzed were the final polls conducted in each particular race; most were completed within two weeks of election day. They cover both general and primary elections, and Democratic and Republican incumbents. They are predominantly from statewide races, with a few U.S. House, mayoral and countywide contests thrown in. Most are from the 1986 and 1988 elections, although a few stretch back to the 1970s.

The last time I looked, "most" still meant "more than half." The difference between asserting (as Panagakis does) that "most" of the final polls analyzed in his report were completed within two weeks of election day and asserting (as the Mystery Pollster does) of those final polls that they were all conducted during the last week before an election is huge. Moreover, the Panagakis analysis simply does not say how many of its final polls were, say, one day from the election and how many were, say, two weeks from the election. Why does the Mystery Pollster choose to mention "one week" at all? For all we know from Panagakis' description, 95% of his final polls may have been taken one day before the election to which they relate.

Does it matter if the Mystery pollster's formulation of the Panagakis results is wrong? Yes, it matters. And it may help to explain why Pat Caddell is looking so upset these days - and saying such tart things. In an interview with Neil Cavuto Mr. Caddell said:

[T]he undecideds always break to the incumbent at the end of a Presidential campaign. ...[N]obody studies the history of this. Undecided voters, by the middle of October, have not decided to vote for the challenger, they go for the safe choice, the person they have. ... In other words, if a challenger cannot convince them, it's what I used to call the button problem, it's the war problem. If you're not going to vote for, unless you convince yourself that the challenger will do a better job in protecting the country, or handling particularly foreign policy than will be the incumbent or the incumbent party, then if you haven't made that decision, you stay with what's safe, you stay with what you know, that you're comfortable with. And that's been working for Bush I think coming all along, and it's part of the whole--

In other words, Mr. Caddell is saying that undecideds usually break towards the incumbent.

Or is he? Mickey Kaus seems to construe Mr. Caddell that way: Pollster Pat Caddell dissents from the so-called Incumbent Rule (which holds that undecided voters never go to the incumbent). I'm not convinced. And Mr. Caddell is clearly upset with something that's being said these days. But is it the "Incumbent Rule" - or, rather, all versions of the Incumbent Rule - that has Mr. Caddell fired up?

Probably not. Mr. Caddell seems to be upset with what he views to be a particular misstatement of the "Incumbent Rule" - the one that is now being used to justify Democratic hopes on the basis of poll results now in existence. The Mystery Pollster and Mr. Caddell agree on at least one thing: The Incumbent Rule applies only based on the final poll, which should be taken very close to election day. As the Mystery Pollster cautioned (despite his incorrect description of the Panagakis study):

That is why pollsters continue to interview voters through the final weekend.

The data needed to apply the Incumbent Rule simply do not yet exist.

The Panagakis study's own description of its included cases should give any serious person serious pause in applying it to the current presidential race. "Primary elections?" "Predominantly from statewide races,with a few U.S. House, mayoral and countywide contests thrown in?" Is "thrown in" writing that usually reflects systematic, scientific choice? What would one think of, say, an experimental drug study whose official analysis submitted to the FDA stated that "a few" trial patients of a particular type (older, female, cancer victims perhaps, or those with known heart irregularities, or those who had competed in recent Olympics Games?) were "thrown in" to the study population?

The Incumbent Rule supposedly largely turns on the effect of voters getting to know and understand the candidates. How comfortable does one feel analogizing the primary, state, House and local elections studied by Panagakis - contests in which ordinary voters usually know little about the candidates, especially the challenger, until the end of the campaign (if voters know much even then) with a presidential election in which the national media has been flooding the political marketplace with information for many months? - especially in this case, where the media have been flooding the market with opinion and information about the Democratic challenger the media overtly favor. The Mystery Pollster is almost certainly seriously wrong to conclude that the historical data and theoretical underpinnings of the incumbent rule are far stronger for presidential races - especially with respect to this presidential race, in which mainstream media coverage has left little positive information about Senator Kerry to be absorbed and much positive information about President Bush obscured.

The Cavuto show was not the only Fox News program on which Mr. Caddell appeared and described his frustrations with the current widespread misstatement and abuse of the Incumbent rule. Mr. Caddell also did so last night on Hannity & Colmes, where he elaborated on his own observations. I do not have a transcript of his statements, but I believe the following fairly summarizes his position:

There is, in fact, sometimes a small "break towards the challenger" in the final day or so before an election. That is the valid content of the "Incumbent Rule" as Mr. Caddell sees it. But in the two weeks preceding, say, the second or third day prior to a Presidential election, there is actually an ongoing "break towards the incumbent." In other words, Mr. Caddell thinks the valid application of the Incumbent Rule leads to the conclusion that Mr. Bush will "top out" no sooner than the second or third day prior to the election - after which there may be a smaller "break to the challenger."

Mr. Caddell did not mention the Panagakis study by name. But his comment on the Cavuto program that "nobody studies the history of this" looks like an implied swipe at that "famous" study. In addition, Mr. Caddell also stated on Hannity & Colmes that he, personally had gone back and researched past presidential races in this respect - research on which he was relying to dispute what he considers an invalid version of the Incumbent Rule. His observation that "nobody studies the history of this" therefore seems intended to mean that he thinks that nobody who is advancing the incorrect version of the Incumbent Rule now making the rounds has adequately studied the history of its application in presidential races.

On the whole, Mr. Caddell's skepticism of any "undecided at one week break to the challenger" or "undecided at two weeks break to the challenger" version of the Incumbent Rule seems very well founded. The Panagakis study, which relies on a lot of races not very much like a presidential race at all, and especially not like this presidential race, is very weak evidence to the contrary. Further, if most of the Panagakis final polls were taken within a day or so of the corresponding elections, there is no daylight between the Panagakis and Caddell research. In that case, the Panagakis analysis would simply be the less precise.

MORE