I came across this chart of data, organized by Alan Reifman, professor at Texas Tech University, looking at the weighting in the major national polls in 2006.
(For now, put aside his assertion that Rasmussen’s 37 percent Democrat, 32.3 percent Republican split was the “Gold Standard” for that year.)
The New York Times exit poll — that was of actual voters who showed up on Election Day – put the Democrats’ advantage in terms of turnout at 3 percent, 39 percent to 36 percent. So while this doesn’t capture the absentee voters and has its own small margin of error (sample size of 13,251) we can conclude it’s one of the best tools we’re going to find to measure the partisan breakdown of Election Day voters, to compare to the pollsters’ final results.
That year, Fox Opinion-Dynamics put the Democratic advantage at 4 percent, 39-35.
Pew Research also put it at 4 percent, 35-31.
The ABC/Washington Post’s registered voter split (35-32) was pretty accurate, but their likely voter split actually put the Republicans ahead, 34-33.
Time put it at 29-26. Sure, overpolling independents is probably going to throw off a result, but at this point I’m looking for pollsters who dramatically overestimated one party’s turnout advantage.
The New York Times CBS poll put it at 7 percent, 35-28.
Cook/RT Strategies was in the neighborhood for their registered voter numbers (32-30) but pretty far off in their likely voter numbers (39-29).
AP-Ipsos overstated it in both, 36-30 and 37-30.
Diageo-Hotline also had it pretty close on registered — 31-28, further off on likely — 35-28.
The most recent Newsweek poll he has in his chart put the split at 7 percent, 36 percent to 29.
Carville’s Democracy Corps poll put the split at 8 percent, 40 percent to 32 percent.
When the pollster for NBC/Wall Street Journal pushed independents to say which party they identified with more, their result was a 6 percent split, 43 percent to 37 percent.
The most recent USA Today/Gallup poll on the chart, from Oct. 6-8, puts the split at 9 percent, 38 percent to 29 percent.
So when I start looking at samples that split 37.4 percent Democrat, 29.1 percent Republican, and 33.3 percent Independent, as in the most recent CBS/New York Times poll putting Obama up by 14, I get a bit skeptical that they’re accurately reflecting the actual makeup of the voters who will show up on November 4.
Lately, I’ve had a few readers say, “Come on, Jim. These are professional pollsters, and you’re just some nobody blogger. What are the odds that they’re all making the same miscalculation about how much of each party will show up on Election Day?”
Well, I can point to two short years ago, when several pollsters were in the right neighborhood — Fox, Pew, Time — and a whole bunch were all making the same error in the same direction in their likely voter screen — Hotline, AP-Ipsos, Cook/RT Strategies, Newsweek, Democracy Corps, NBC/Wall Street Journal, USA Today/Gallup and the CBS/New York Times poll.
This isn’t to say McCain isn’t trailing, or that he’s not in a tough, tough situation. It’s just that a lot of pollsters right now think that the Democrats’ advantage in turnout is going to be four, five, six percent of the electorate more than 2006, which was the GOP’s worst off-year elections since 1974.