If you want a good pessimistic trend, though, note that GOP incumbents have lost in the 20th century only in the following cases:

1. GOP fractures into two parties (1912, 1992).

2. GOP is discredited by monstrous scandal (1976).

3. GOP is discredited by total economic collapse (1932).

So Kerry's doing a lot better than that pattern would suggest.

I don't quite get how he arrives at his conclusion from his data, but let that pass. Instead I'd like to point you to another analysis of presidential elections, this one by Larry Sabato of the Center for Politics of the University of Virgina, which includes some of Matthew's observations but systematizes them in the process, and goes farther as well. Sabato claims that there are three "keys" to the presidency, three predictive factors which can help project who the winner of a presidential election will be.

The factors are The Economy, Peace and Scandal, and they are summarized in a chart that I, being a cheapskate user of the free (non-graphical) Blogger service, cannot post here, so I'll describe it:

For each of these factors, pluses and minuses are awarded the administration in power. For the Economy, 2 pluses are given for an excellent economy, 1 for a good one, 1 minus for a fair or poor economy, and 4 minuses for a very poor one. (The idea is that "voters punish incumbents harshly for bad times, while rewarding incumbents for good times less emphatically. This represents the intersection of American history with human nature.") In the Peace category, a minus is given for a unpopular war, and a plus for either peace or a popular war. Under Scandal, the administration gets 2 minuses for a big scandal, 1 minus for a small one, and a single plus for no scandal.

The administration's pluses and minuses are added up, and the claim is that:

The incumbent party won all 13 of the contests where it had amassed three to four net pluses, with 8 of these 13 elections producing landslide victories.

In addition, there were four elections in the 26 total where the incumbent president or his party had one or two net "pluses", and the incumbent party won three of these four. (Yes, Al Gore is the only exception, a fact that explains why just about every senior Democratic official sighed with relief when Gore bowed out of the 2004 race on December 15, 2002.) That leaves 9 elections where the incumbent president or his party's nominee ran with one to four net "minuses"; the incumbent party lost all nine of these contests. In four of these cases, the incumbent party accumulated four net minuses, and in all four cases, the challenger party won in a landslide.

So, assuming for a moment the accuracy of Sabato's judgment calls in awarding the points, and the predictive power of the model, what does this tell us about the upcoming election?

Well, it rather depends.

What, for instance, is the state of the economy? My own opinion is that the economy continues to be bad, despite the supposed signs of an upturn, because the jobs situation is still bleak, but there are those who argue that it's getting better. Certainly the very best possible rating Bush can be given on that predictor is a single plus for a "good" economy, and two minuses, meaning a fair to poor economy would, I think, be more accurate.

Then, what about the popularity of the war? Well, it's falling, and continues to decline, but not yet so much as to rate being definitively called "unpopular". I think it's most likely that it can, and will, fall far enough to warrant that rating, if only because the Bush administration has demonstrated over and over that they really don't know what they're doing there, and that given two choices of action, they will invariably choose the worst one. (Fallujah being perhaps an exception.)

Finally, there's the "scandal" predictor, and once again I'm somewhat flummoxed about how to rate the situation. Certainly the malfeasance and misdeeds of the Bush administration should rate a "big scandal" double minus, but there's no indication that the media is planning on picking up on any of these stories. (When is a scandal not a scandal? When the press doesn't report on it.) We can be heartened, though, by the fact that there are something like a dozen major investigations underway at this moment, a small number of which have the potential to blossom into a real, solid scandal. Still, it looks as if Bush has the advantage here.

So, if everything breaks right for us, Bush would have 2 minuses for the economy, 1 minus for the war, and two minuses for a big scandal, for a total of 4 minuses and, by this prediction scheme, a certain Kerry presidency.

More likely, though, Bush's results will be a plus for the economy, a minus on the war issue, and a plus for no scandal, making a total of 1 plus. By this scheme, as described by Sabato, this is an ambiguous case. Sabato lumps together the results for one plus and two pluses, but gives no real reason for doing so. I rather think that the data support the idea that at least two pluses are needed for an incumbent to be viable winner, and an incumbent who can only provide a measly one plus's worth of reasons to keep him in office is going down for the count. (This comports well with the historical surveys of approval ratings as predictors -- there's a grey area where the result is uncertain -- 45%-51% -- but an approval rating of under 45% in mid-June is the kiss of death for an incumbent.)

So my take is that the scoring system should really require at least 2 pluses, and preferably more, for the incumbent to win, so with only a single plus on hand, Bush will either have to improve the economy dramatically (unlikely, since to do so he'd have to rollback some of his sacred tax cuts), or make the war in Iraq popular (or, rather, reverse the current direction of opinion of stop it from becoming definitely unpopular, as it is now headed).

Regardless of one's political persuasion, those sound like very tough tasks to pull off, which means, at least by this particular crystal ball, that Kerry has a real chance of winning. That jibes with my own perception, which is that Kerry's positives are basically irrelevant to the end result, since the election will basically be a referendum on Bush. As long as Bush's numbers continue to go south, Kerry doesn't really need to define himself very specifically, or present strong plans for what he will do as President. To some extent, he's better off being undefined and amorphous (all things to all people) as long as he can avoid having his public persona created out of whole cloth by Bush's smear attacks -- so Kerry's efforts should be limited to preventing that. In other words, just the kind of positive defense posture that Kerry is in right now.

Note that in Sabato's predictive system, the public status of the challenger is not factored in at all, reinforcement for the notion that presidential elections are essentially about the incumbent (or the party of the incumbent) and not really at all about the challenger. To get more points to shore up his chances of a win, Bush has to fix the economy or fix the war, and no amount of driving down Kerry's approval rating with negative advertisement or smears from the right-wing attack machine is going to change that. If Bush can't hack it, he'll be out, says this crystal ball, and the only man who can then be in is John F. Kerry.

Update: Link to Sabato's site corrected.