By Mark Mellman - 09/21/10 11:25 PM EDT
Hold on just a minute!
Based on the data coming across my desk, most of which never sees the light of day, I am willing to stipulate that this will not be a good year for Democrats. Just how badly things will turn out, however, remains murky at best.
A month ago, headlines across the country trumpeted Republicans’ record 10-point advantage on Gallup’s generic vote question. Gallup itself reported, “The Republican leads … are all higher than any previous midterm Republican advantage in Gallup’s history of tracking the generic ballot, which dates to 1942. Prior to this year, the highest such gap was five points, measured in June 2002 and July 1994.”
Respected forecasters and academic modelers, many of whom incorporate the generic vote into their projections, began talking about Democrats losing 50 and 60 seats — not out of hysteria, but based on hard analysis. Emory professor Alan Abramowitz, an eminent political scientist, revealed that his model predicted a 62-seat loss if the final Gallup margin mirrored the 10-point Democratic deficit. Nate Silver of 538, whose model includes a variety of variables in addition to the generic vote, was forecasting a 45- to 50-seat loss.
At the time, I warned a few journalists that too much was being made of Gallup’s “record result.” A statistical phenomenon called “regression to the mean” would result in the next reading being much improved. It was.
But Monday, Gallup dropped another bombshell — Democrats now enjoy a one-point lead in the generic vote! Not great, but quite a difference. Were that to be Gallup’s final number, professor Abramowitz would project a 29-seat loss — a result leaving the House under Democratic control.
So is it -10 or +1? Gallup tried to reduce the difference by aggregating all of its data for August and September to show that last month the GOP actually sported a six-point margin, which declined to just one point this month. (By professor Abramowitz’s calculations, though, even a one-point GOP edge would, just barely, leave Speaker Pelosi still in her chair.) Gallup also explains that among likely voters, the numbers are likely to skew in Republicans’ direction.
However, some other polls of likely voters also show a roughly even division. The Battleground Poll showed an exact tie, while the CBS/New York Times survey put the Republican lead at two. Still other surveys, like the AP’s and the ABC/Washington Post poll, are replicating Gallup’s earlier finding, granting the GOP a 10- to 13-point advantage.
So while the generic vote may be an accurate indicator, as Gallup claims, it is certainly not a stable indicator. It bounces around more than Lawrence Welk’s musical ball. And if the final poll is the only one that counts, it seems awfully hard to know what that will look like based on readings before then.
Moreover, while modelers can make the translation of generic votes into seats sound rather precise, in fact, it is not. Professor Abramowitz’s model has a margin of error of about 20 seats. In fact, all of these models come with great uncertainty. Another, developed by two equally distinguished academics, comes with a margin of error of 60 seats.
What’s it all mean? Well, Mr. Boehner should stop fondling the gavel, while Democrats should stop polishing their résumés and get back to work. A lot of this election remains to play out.
Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the majority leaders of both the House and Senate.