POLITICS
10/31/2010 12:32 pm ET Updated Dec 06, 2017

Do Generic Polls Show GOP Lead Narrowing?

WASHINGTON - Readers of The Washington Post woke this morning to the headline "GOP HOLDS EDGE AT FINISH" splashed across its front page. An above-the-fold graphic, under the heading "Republicans keep a slight advantage," illustrated results from the latest Washington Post/ABC News poll showing likely voters favoring the Republican candidate in their Congressional district over the Democrat by a four point margin (49% to 45%).

No one should quarrel with the lead of the Post story (which has a different online headline): Based on all available evidence, Republicans do "appear poised to capture control of the House and are likely to make substantial gains in the Senate," a point also stressed in the ABC News analysis of the same poll. What Post readers may miss, however, is the subtext of two important words in the headlines: "edge" and "slight."

A four-point Republican margin on what is known among pollsters as the "generic" House ballot, is closer than what most other recent national polls have been showing (at least as of this writing), particularly those from Gallup, Rasmussen Reports, Fox News and others showing outsized Republican leads. Another new CNN poll released just just this morning shows Republicans leading by seven (52% to 45%). Of course, the Post/ABC poll is just one estimate with a margin of error -- random change alone might have produced a wider or narrower Republican lead had they conducted multiple polls using the same methods.

But that said, a four-point Republican margin would suggest the Republicans winning control of the House but by narrower margins than the 50+ seat gains widely predicted in recent days. The statistical model of Emory University Professor Alan Abramowitz, presented at a recent conference of political scientists, shows that the difference between a tie on the generic congressional ballot and a ten-point Republican lead translates into a potential spread of Republican gains of between 40 and 57 seats (ignoring the margin of error for the model). If Republicans lead by 5 point in the generic ballot, the Abramowitz model predicts a 49-seat Republican gain. They need to pick up a net 39 seats to win majority control.

We will have many more new national generic ballot numbers to ponder over the next 24 to 48 hours, but for now, consider the following table, which shows the most recent results reported by 13 different pollsters over the last few weeks. They show everything from a 3-point Democratic advantage (Newsweek) to a 14-point Republican landslide (Gallup's 40% turnout scenario), and every one of these surveys is based on a sample of "likely voters." Our most recent generic House ballot trend estimate, based on all public polls, puts the Republican advantage at just under 7 points (49.9% to 42.7%):

2010-10-31-Blumenthal-Generic1031c.png

Those ready to question the entire polling enterprise have pointed, with good reason, to a recent Pew Research Center report showing that polls that limit their calling to Americans with landline phones may show a modest bias favoring Republicans. However, as indicated in the table, most national pollsters are now sampling and calling both cell and landline phones, including those producing the most widely varying results (Newsweek and Gallup). But what's most striking about the most recent generic House ballot results is a huge variability that seems unrelated to whether the pollster is interviewing over cell phones to reach those with only landline service. The differences in the likely voter "models" are much more important in determining the size of the Republican lead (or deficit).

Of course, the polls in the bottom half of the chart are dated, including those from Gallup and the Pew Research Center which are based on data that is roughly a week to two weeks old. As noted here previously, the Gallup and Pew data show evidence of a coming-down-to-earth of the enormous partisan gap in attention paid to the campaign that is contributing to some of the outsized Republican margins. So the variation among the final round of national polls may be narrower.

Or it may not. It's worth remembering that national polls at the close of the 2006 campaign also showed considerable variation that did not completely converge at the end. The following table shows how the final week polls compared to the final official national House vote:

2010-10-31-Blumenthal-Final2006Generic.png

The final polls from Gallup, ABC/Post and the Pew Research Center had the final Democratic margin about right (although all three showed much bigger Democratic leads a few weeks earlier). However, the final polls from Newsweek, Time and CNN overstated the Democratic margin considerably.

What will the final round of national polls show this time? Stay tuned to HuffPost Pollster -- we will be back with more updates as pollsters release their final national polls.