Addition and Subtraction

It's become clear that, as many of us predicted, Sen. John Kerry came out of the Democratic convention with only a modest bounce. Indeed, much of the bounce arguably came before the convention and was the result of Kerry selecting Sen. John Edwards as his running mate.

The argument made here and in other places was that, with President Bush already laying claim to roughly 90 percent of the Republican vote, and Kerry holding almost that high a share of the Democratic vote, there was very little room in this race for either candidate to enjoy the kind of bounce that used to occur, when partisan cohesion was not nearly as strong as it is today. As veteran Democratic pollster Peter Hart puts it, we have a "concrete trampoline," which makes it almost impossible to get much of a bounce.

For more than four months, both Bush and Kerry have been hovering at around 45 or 46 percent, give or take three percentage points, depending upon the poll, the week and what was going on around the time the survey was conducted. In recent days, most polls have Kerry up a touch over the incumbent, but still well within the surveys' margins of error. The only reason Kerry's slight lead is worth noting is that it occurs in almost all of the recent national polls, suggesting that an edge truly exists, albeit a very narrow one. The notable exception is the Gallup/CNN/USA Today numbers, where Bush leads among likely voters while Kerry leads among registered voters.

There is considerable skepticism among pollsters about the reliability of likely voter screens months before an election, and many of the biggest fans of the Gallup Poll overall, myself included, are less enthusiastic about its screens, based on past history.

In the Aug. 3-5 Associated Press/Ipsos Public Affairs survey of 798 registered voters, the Kerry/Edwards ticket had a three-point lead over the Bush/Cheney ticket, 48 to 45 percent, with the Nader/Camejo ticket rounding out the field with three percent and another three percent undecided. This marked a five-point decline for Bush since the July 6-7 AP/Ipsos survey, when the president was ahead of Kerry by four points, 50 to 46 percent, with Nader picking up two percent. The latest survey also showed Bush with a 49 percent approve, 50 percent disapprove, not much different from June or July numbers. Suffice it to say, this race has remained very, very stable since April.

Updating some figures that this column published last month with the latest AP/Ipsos survey data, the far more sobering numbers for Republicans are those of voters in the undecided column. Keep in mind two important factors: First, when an elected president is seeking re-election, the contest is a referendum on the incumbent far more than it is a competition between candidates. Second, undecided voters historically have broken heavily against well-known, well-defined incumbents. This has proven true on the congressional, senatorial, gubernatorial and presidential level. That's the origin of the phrase in politics for incumbents, "what you see is what you get" -- you get pretty much the percentage on Election Day that the last round of polls indicate that you will get, while the undecided vote goes elsewhere.

With those two premises in mind, let's take a look at the combined last five months of AP/Ipsos national polls of registered voters. Among the 3,719 registered voters in the combined April through early August surveys, 41 percent believe the country is headed in the right direction, while 56 percent say the country is off on the wrong track. While those are not good numbers for an incumbent, we have certainly seen much worse. But among the 327 registered voters who are undecided, only 19 percent say the country is headed in the right direction; 74 percent say its off on the wrong track.

While 49 percent of all registered voters approve of Bush's overall job as president, another 49 percent disapprove. Among just the undecided voters, only 25 percent approve, and 68 percent disapprove. Those are very ugly numbers for an incumbent. Not surprisingly, this pool of undecided voters tend to be disproportionately more Democrat than Republican, with 43 percent identifying themselves as Democrats, 32 percent as independents and only 25 percent as Republicans. <

The natural tendency for a Bush supporter is to say: "But wait, those numbers for undecided voters are from only 327 interviews. That has to have a higher statistical margin of error than if it were a broader sample of 500, or 800 or 1,000 voters." This is absolutely true, but given the relatively small number of undecided voters, it's hard to get larger samples of undecided voters.

It is also true that the margin of error figures so often cited in news stories and columns about polls are those for percentages close to 50-50, where the margin of error is the greatest. The closer percentages are to 90-10 (or 10-90), or 80-20 or 70-30, the margin of error is significantly less, and the numbers are much more stable and reliable.

Others might say, "but these are national polls, not polls of the 15 or 18 or 20 battleground states that really matter in the Electoral College." That is also true, but we have seen repeatedly that the polls of the combined battleground states match up, almost digit for digit, with national surveys.

Privately, Republican pollsters concede that undecided voters usually do break against incumbents, and that this pool of undecided voters is particularly ugly for their team. Nonetheless, they insist that the pool is still in play. They have to go all out for them, no matter the odds.

To the extent that Republicans could take solace that Ralph Nader is running again and could tip the election their way, as he clearly did in Florida in 2000, signs of Nader not being on the ballot in as many states in 2004 as he was last time are disheartening. Stories of Nader petition-gatherers being harassed by people because they believe Nader "effectively elected Bush" suggests that his vote total may be stripped of many liberals who supported him last time, and may be confined mostly to anti-establishment types who would never vote for any major party candidate, minimizing his impact.

Of course, it is important to emphasize that there are 84 more days to go before this election, and that American politics rarely remain static for two or three months. Much can and certainly will happen between now and Nov. 2, not the least of which are one vice presidential and three presidential debates. Other, less predictable events are possible as well, including the chance of (God forbid) another major terrorist attack.

But with those major caveats and disclaimers now safely out of the way, President Bush must have a change in the dynamics and the fundamentals of this race if he is to win a second term. The sluggishly recovering economy and renewed violence in Iraq don't seem likely to positively affect this race, but something needs to happen. It is extremely unlikely that President Bush will get much more than one-fourth of the undecided vote, and if that is the case, he will need to be walking into Election Day with a clear lead of perhaps three percentage points.

This election is certainly not over, but for me, it will be a matter of watching for events or circumstances that will fundamentally change the existing equation -- one that for now favors a challenger over an incumbent.