If you’re hoping for a Bush victory in November, you couldn’t have been happier when first Time magazine, and then Newsweek, uncorked polls this week showing the President with an 11-point lead over challenger John Kerry. In the history of polling, no Presidential candidate with a double-digit lead on Labor Day has gone on to lose in November. Elections typically get tighter near the end, but they don’t tighten that much.
The plot thickens, however. Rasmussen Reports publishes a tracking poll every day. Since Kerry clinched the nomination in the spring, neither candidate been more than four points away from the other. Yesterday, Rasmussen showed Bush with just a 1.1% lead, 47.6% to 46.5% over Kerry. Rounding out the post-convention polls to date, Gallup shows Bush ahead by 7 points, 52% to 45%, among likely voters.
What is going on here? Rasmussen was assailed with complaints after he failed to find the Bush bounce that Time/Newsweek did. All year long, his poll has been within the margin-of-error of all the other major national polls, and a 10-point difference surely indicates that something is wrong — the polls shouldn’t be that far apart.
In a revealing article that should be required reading for anybody who interprets polls, Scott Rasmussen cleared it up. The short answer is that Gallup has it about right: a Bush lead of 5-7 points. You can follow the link, but here is a brief explanation:
In both the Time and Newsweek polls, a plurality of voters surveyed identified themselves as Republicans. In the polling era, there has never been a presidential election in which Republican voters outnumbered Democrats. Republicans win only by getting a sufficient number of Democrats to cross over, which (luckily for Republicans) is a pretty easy thing to convince them to do. Nevertheless, party affiliation has remained pro-Democrat from one election cycle to the next.
Now, however enthused you were about the Republican convention, do you believe these four days were enough to turn a plurality of the country into self-identifying Republicans, when it has been the opposite for generations? Or is it just possible that Newsweek and Time conducted their polls during the convention itself, when a high proportion of Republicans were likely to be home with their TV sets tuned in? Rasmussen concludes the latter.
On the other hand, Rasmussen concedes that his three-day tracking sample included an extraordinarily good day for Kerry on Saturday, which explains why he shows just a 1.1% lead for the President. Excluding Saturday, Rasmussen shows a 4-point spread, which is in the zip code of Gallup’s 7-point margin. The strong likelihood is that Bush’s actual lead is somewhere in the 4 to 8-point range — not fatal to Kerry, but clearly not where he’d hoped to be. Gallup, incidentally, gives Bush a 2-point “bounce” out of his convention, which is right where the pundits predicted it would be, and comparable to the bounce that most pollsters gave Kerry after his convention.
Reading all of the polling analysis on the web reminded me that modern poll numbers are “cooked” a lot more than people realize. Most pollsters, for instance, report the views of “likely voters.” This means that the poll is not reporting the “raw” results, but the results after eliminating those judged unlikely to vote. This is a reasonable methodology, for polls show that many more people state an intention to vote than actually do. The no-shows tend to be predominantly Democratic, and a poll that failed to exclude them would consistently predict Democrat victories that fail to materialize on election day. But predicting “likely voters” is not an exact science. If turnout is higher than historical norms, it will favor Kerry.
Although Bush does not have an 11-point lead, by any measure he does have a very real lead that is right at, or perhaps slightly outside of, the margin of polling error. That lead will most likely subside a bit — that’s why they call the post-convention surge a “bounce” — but Kerry still has some ground to make-up. In addition, although the Time and Newsweek polls were clearly erroneous, Bush gets the benefit of the perception, however inaccurate, that he enjoys a potentially insurmountable lead. Kerry, on the other hand, has suffered through a 2 or 3-week period in which he has largely been responding to news (most of it unfavorable to him), rather than shaping it himself. Comparisons to the lead Michael Dukakis squandered in August 1988 are apposite.
If this election is going to be a real race, Kerry is going to need to make it so. And soon.