Redstate

It's almost as good as coffee.

By Leon H Wolf Posted in 2006 — Comments (2) / Email this page » / Leave a comment »

As I'm sure you all know, the polls (with a few notable exceptions) are not looking good for the Republicans thus far this election cycle. Cause for despair, right? Maybe not. Dan Riehl took a look at 8 polls from 2002 which were from two days before the elections that drastically deviated from the actual result in the favor of the Democrats - including several polls which had Republicans comfortably behind in races in which they won. Most of these polls, however, are Zogby, so I decided to take a look into some other results to check for consistency.

Read below...

First of all, consider the Rasmussen daily snapshot from the day of the election in 2004. It's important to realize that polling during an election year presents a different dynamic for a lot of different reasons - namely, the difficulty of assessing the true pull of the guy at the top of the ticket, and the fact that voters tend to be more tuned in and likely to vote, but it's instructive nonetheless. Rasmussen's numbers are in italics, actual results in bold:

AK Senate:
Knowles (D) 47 46
Murkowski (R) 44 49

CO Senate:
Coors (R) 49 47
Salazar (D) 48 51

FL Senate:
Martinez (R) 49 50
Castor (D) 47 48

GA Senate:
Isakson (R) 54 58
Majette (D) 42 40

LA Senate:
Vitter (R) 47 51
Others (irrelevant)

NC Senate:
Burr (R) 49 52
Bowles (D) 46 47

OK Senate:
Coburn (R) 48 51
Carson (D) 43 41

SC Senate:
Demint (R) 50 54
Tennenbaum (D) 44 44

SD Senate:
Thune (R) 49 51
Daschle (D) 46 49

So, in two of the competitive races (SD and CO), the polls erred in favor of Republicans, by margins of 5 points and 1 point. In the other 8 races, the polls erred in favor of the Democrats by margins of 6, 0, 6, 8, 2, 5, and 4. On average, then, Republicans outperformed Rasmussen polls by 2.5 points. Significantly, however, Rasmussen only "Missed" three races - two for the Republicans (AK Senate and Vitter's majority in LA), and one for the Democrats (CO Senate).

Perhaps more compelling, SurveyUSA has an interesting document here (warning: .xls) which analyzes the accuracy of the 2004 pollsters. Of the 13 polling organizations that covered the 2004 elections, three of them had a "pro-Republican bias," in that their polilng results showed Republicans ahead of where they actually finished. One of those is the Republican polling company, Strategic Vision. (The other two, Quinnipiac University and American Research Group, showed "bias" of 3.9% and 0.4% respectively). The other ten pollsters showed Democratic bias - albeit very small bias in most cases:

Global Strategy Group: 6.9%
Market Shares: 3.2%
Zogby International: 3.0%
L.A. Times: 2.7%
Rasmussen: 2.2%
Research 2000: 1.1%
Opinion Dynamics: 0.8%
Mason-Dixon: 0.5%
Gallup: 0.3%
SurveyUSA: 0.1%

Here, though, is the final point that should be made. this spreadsheet (warning: .pdf) looks at all of the competitive Senate races in 2002, as done by all the major polling companies. While there is some striking variation between the result (the small black number underneath the first column) and the polling margins (the black numbers beneath the other columns, there are very few negative numbers (indicating that the polling organizations picked the wrong winner altogether. The CO Senate race that year was the only race to have more than one poll get the wrong result. Meanwhile, the polling organizations systematically underestimated the Republicans in the Senate races in CO, OK, TX, and Democrats were regularly underestimated (with the exception of SUSA, who got the result entirely wrong) in MO. All others are fairly ambiguous.

In the end, Riehl's basic point is sound - staying home because of adverse polling leading to despair is very foolish, especially in any race that is polling relatively close. And, in places where the GOP has an on-the-ground partisan advantage (like MO, TN, MT and OH), expect even the poll numbers to shore up in the next two and a half weeks. However, we shouldn't kid ourselves into thinking that there isn't a whole lot of work to do in this race.