Are Steele and Cardin Really Tied? No

Originally posted at MoCoPolitics.

Probably not. I mentioned yesterday that I thought the Survey USA poll showing a 46-46 tie was an outlier, and having now looked at the lovely Pollster.com website (catnip for us poll geeks), I think I was right.

Here are all the polls that have been done on the Cardin-Steele race since the beginning of August

Survey USA, 10/17/06: Cardin 46, Steele 46
Zogby Interactive, 10/16/06: Cardin 51, Steele 43
Rasmussen, 10/12/06: Cardin 53, Steele 44
Public Opinion Strategy, 10/4/06: Cardin 47, Steele 43
Zogby, 10/2/06: Cardin 45, Steele 37
Gallup, 10/1/06: Cardin 54, Steele 39
Voter-Consumer Research, 9/28: Cardin 44, Steele 39
Mason Dixon, 9/26: Cardin 47, Steele 41
Zogby Interactive, 9/25: Cardin 52, Steele 39
Survey USA, 9/19: Steele 48, Cardin 47
Potomac Inc., 9/18: Cardin 51, Steele 40
Rasmussen, 9/13: Cardin 50, Steele 43
Zogby Interactive, 9/5: Cardin 49, Steele 40
Gonzales, 8/25: Cardin 44, Steele 39
Zogby Interactive, 8/21: Cardin 50, Steele 41
Rasmussen, 8/9: Cardin 47, Steele 42
Public Opinion Strategy, 8/2: Cardin 43, Steele 35

Survey USA and Rasmussen are IVR ("interactive voice response") polls or more colloquially, "robopolls." Zogby Interactive is an Internet-based poll (Zogby polls are traditional telephone polls). The reliability of IVR and Internet polling is the subject of great debate. The remaining pollsters poll by the traditional means of the telephone.

So is there a big discrepancy across poll type? Not really, if you exclude Survey USA. In fact, there is a pretty consistent pattern, which is Cardin leading by between 7 and 10 points. Let's take a look.

The two Survey USA polls are the best for Steele, but are wildly different from the Rasmussen polls utilizing the same technology. The average of the two Survey USA polls is Steele 47, Cardin 46.5. The average of the three Rasmussen polls is Cardin 50, Steele 43, with a decided trend line in favor of Cardin (47, 50, 53). The average of the four Internet-based Zogby Interactive polls is Cardin 50.5, Steele 40.8, pretty consistent with Rasmussen.

There have been eight traditional phone-based polls, with an average as follows: Cardin 46.9, Steele 39.1. One thing we notice right off is that phone-based polling has a much higher level of undecided voters (14% versus 6.5%, 7% and 8.7% for the "new age polls). The other point that arises is that the closest of the eight phone polls was a 4 point lead for Cardin. Even if you take out the Gallup outlier (54-39), the remaining 7 polls show an average of Cardin 45.9, Steele 39.1, an average lead of 6.8 points.

So to summarize:

Survey USA: 2 polls, Steele by 0.5
Rasmussen: 3 polls, Cardin by 7.0
Zogby Interactive: 4 polls, Cardin by 9.7
Phone polls: 8 polls, Cardin by 7.8

Other than Survey USA, there is remarkable consistency across all the polls by poll type, with only an average difference of 2.7%, less than the margin of error, across three poll types and 15 different polls.

Bottom line: unless and until a non-Survey USA poll shows up with similar numbers, assume that the SUSA polls are not just outliers, they're wrong. Until demonstrated otherwise, Cardin leads by 7-10 points, plus or minus the margin of error. And when you get to averaging 15 polls, that margin of error is pretty small.