« Pre-Mortem Post-Mortem | Main | Party or Person? »

ShareThis

Last-Minute Polls: Somebody's Got To Be Wrong

While there's no doubt that Republicans are going to make net gains in congressional and gubernatorial contests today (as the "out" party almost always does in the first midterm after a new presidential administration takes office), there is some pretty serious mystery about how big those gains wll be, particularly in the House. That's mainly because of an unusual degree of disagreement among the major polling outlets about the shape of the midterm electorate and the size of the GOP advantage.

Mark Blumenthal of pollster.com has a nifty write-up this morning of the problem, which includes a chart of final likely-voter generic congressional ballot polls showing a remarkable range of findings, from Marist/McClatchey's dead even to the "low-turnout"-based Gallup prediction of an astonishing 15 point Republican margin.

Complicating the issue (which would normally dictate just averaging the polls and not worrying about it) is the fact that Gallup has historically been very accurate in its final generic ballot poll. Yet today their findings look like an outlier, off there nearer to the Rasmussen numbers (a 12-point GOP margin) than to such equally well-established and sober outfits like Pew (a 6-point margin) and ABC/WaPo (4 points). TDS contributor and advisory board member Alan Abramowitz of Emory University has been saying for a good while that something's screwy with Gallup's methodology this year. One theory is that Gallup is placing too much stock in subjective "enthusiasm" indicators that may over-estimate the marginal tendency of Republicans to vote. With other apparent outlier polls like Rasmussen's, there is a suspicion that the firm has failed to sufficiently adjust for its inability to reach cell-phone-only households.

We'll know the truth soon enough, but it's comforting to know that the poll-assessment industry seems to be growing as fast as the public polling industry itself. Certainly Democrats hope that Gallup is making a mistake this year that rivals its famous prediction in 1948 that Dewey would crush Truman.