Random commentary and senseless acts of blogging.
The first Republican president once said, "While the people retain their virtue and their vigilance, no administration by any extreme of wickedness or folly can seriously injure the government in the short space of four years." If Mr. Lincoln could see what's happened in these last three-and-a-half years, he might hedge a little on that statement.
Prisoners of Azkaban
Wednesday, December 24, 2003
Crunching the Electability Numbers
What evidence is there that Wesley Clark is really more electable than Howard Dean? I ran a meta-analysis of eight recent polls (2 Newsweek, 2 CNN, Nov NBC, Fox, Gallup, and Dec Quinnipac) to find out. In seven of the polls Clark outperforms Dean, in one they are tied.
In meta-analysis, a group of polls are added together into one larger poll. In the total poll Dean loses to Bush by a little over 13 points, 52.20% to 39.18. Clark loses by just under 10, 50.62 to 40.83. So Clark does 3.23% better than Dean. In this kind of analysis the MOE is quite small; in this case N=8003, MOE=1.12%.
This may seem like a small gap, but in a close election 3% can be critical. In 2000, 7 states with 60 electoral votes were decided by 3% or less, 5 by 1% or less.
Very attentive readers will notice that I have excluded the poll most favorable (relative to Clark) to Dean, the December NBC poll taken the day that Hussein was captured. The poll was excluded because the numbers were too inconsistent with both earlier and later polls. Hussein's capture naturally caused a Bush spike, but the main change wasn't an increase in Bush voters but a decrease in votes for both Dean and Clark. The 28% for Clark recorded in that poll has a z score of -3.75, which means it is equal to the average (40.875) minus 3.75 times the standard deviation (3.44). In statistics, any measurement which varies from the average by more than three times the standard deviation is considered to be untrustworthy. Had the poll been included, Clark's advantage would have been 2.79%
Still, 3% isn't a huge number a year before the election. But there's evidence that Clark's real edge is larger. I ran through some polls in swing states, to see how Clark and Dean compared in the crucial states. In most states, there was no difference, with one odd exception. Two polls taken in New Hampshire each showed Clark six points ahead of Dean in a contest with Bush.
That doesn't seem to make sense. Dean runs far ahead of Clark in New Hampshire. He polls there better than in any other state, possibly including Vermont. His favorable/unfavorable ratings, straying into negative territory in some national polls, have hit +50 in New Hampshire. Dean comes from next door and has spent far more time and money in New Hampshire than Clark has. Here, if anywhere, he should be able to show that he can do better against Bush than Clark. Yet the results are the opposite.
The cause is probably name recognition. Voters in NH have simply tuned in to the election in a way that voters elsewhere haven't. Even back in October, 86% and 73% of New Hampshire voters knew enough about Dean and Clark respectively to form an opinion. In the most recent national poll available, those numbers were 45% and 33%.
As the rest of the nation becomes more aware of the candidates - at least if the media deigns to discuss the candidates instead of crowning Dean - both Clark and Dean are likely to close the gap on Bush. But Dean is unlikely to close Clark's electability gap, in fact it is likely to grow wider. Dean, with the advantage of being local, and with better name recognition, still runs six points (the margin of 15 states and 167 electoral votes in 2000) behind Clark in the Granite State. It may well be twice that elsewhere.
Deaniacs believe that as their man rises in the polls that gap will disappear, or perhaps already has. But the truth is that Dean's recent rally has been entirely among Democrats. Between mid November and mid December, Dean went up an impressive 14 points in Newsweeks polls of Democratic candidates. But there was no movement in the electability gap; three consecutive polls showed Clark 1% ahead of Dean against Bush. In the Gallup poll, Dean moved from a tie with Clark to 15 points ahead, yet, over about the same period, started out 6 points behind Clark in electability and actually fell back to 7.
Will Dean's impressive grass roots campaign make up the difference? Hardly likely. In New Hampshire and Iowa, Dean has worked his grass roots to great effect, and with far more intensity than will be possible to apply to every swing state in the general election. Yet in New Hampshire, Clark's 6 point edge over Dean doesn't even make him the strongest candidate. Kerry runs 10 points ahead of Dean, sinking campaign and all, as does Lieberman. A recent Iowa poll didn't test Clark, but had Gephardt 5 points ahead of Dean, Edwards 1 point, and Kerry tied. Dean will probably win Iowa and New Hampshire, but in both states he is the weakest candidate against Bush.