Skip to content

TCJ Research: A Look At Our 2010 Polling

October 28, 2012

When we posted our poll showing Romney up 3 in Ohio, a result that no one else seems to be showing, we knew that our poll would raise more than a few eyebrows. Quite frankly, I don’t blame you for being cautious; we acknowledge up-front that we’re a Republican-backed polling group.  At the same time, however, we value accurate results more than anything because it doesn’t do anyone any good if the results we’re showing are blatantly skewed to paint a rosier picture.  Without a further ado, a look at our 2010 polling:

Senatorial Polling

TCJ Poll Margin Actual Margin Error
Kentucky Senate
Rand Paul-R v. Jack Conway-D 12%- Paul 12%- Paul 0 0
Wisconsin Senate
Ron Johnson-R v. Russ Feingold-D 8%- Johnson 5%- Johnson 3 3
Florida Senate
Rubio-R v. Crist-I v. Meek-D 21%- Rubio 19%-Rubio 2 2
West Virginia Senate
Raese-R v. Manchin-D 5%-Manchin 10%- Manchin 5 5
Nevada Senate
Angle-R v. Reid-D 1%-Angle 4%-Red 5 5
Washington Senate
Rossi-R v. Murray-D 1%-Murray 4%- Murray 3 3
Illinois Senate
Kirk-R v. Giannoulias-D 3%- Kirk 2%- Kirk 1 1
Colorado Senate
Buck-R v. Bennet-D 2%- Buck 1%- Bennett 3 3
California Senate
Fiorina- R v. Boxer- D 3%- Boxer 10%- Boxer 7 7
Pennsylvannia Senate
Toomey-R v. Sestak-D 3%- Toomey 2%-Toomey 1 1
Delaware Senate
O’Donnell-R v. Coons-D 10%- Coons 16%- Coons 6 6
Alaska Senate
Murkowski-WR v. Miller- Rv. McAdams-D 3%- Murkowski 4%- Murkowski 1 -1
Connecticut Senate
McMahon-R v. Blumenthal-D 7%- Blumenthal 12%-Blumenthal 5 5

Let’s start with the bad stuff first. We acknowledged in the wake of the midterm that our Senatorial polling left much to be desired.  On average, our polls skewed from the final results, in favor of the Republican, by about 3 points.  We performed particularly poorly in California and Delaware because our LV model showed more enthusiasm on the side of the Republicans than was realized on election day.  Had the polls been reweighted to more accurately reflect turnout among Democrats and Republicans, we would have been off by 3 or 4 points in both cases.  Still, that’s kind of like guessing the correct lottery numbers the day after they’re drawn.  If we can toot our own horn for a moment, we performed particularly well gauging Senator Lisa Murkowski’s Write-In campaign in Alaska and balancing the 3-man contest in Florida.

We are by no means making excuses, but let’s be honest: Polling in 2010 was TERRIBLE. And while I do hate to link to FiveThirtyEight, please take a look at their analysis of the 2010 polling.  It’ll show you that, relatively speaking, we did quite well.

Gubernatorial Polling

TCJ Poll Margin Actual Margin
Arizona Governor
Brewer-R v. Goddard 10%- Brewer 12%- Brewer 2 -2
Texas Governor
Perry-R v. White-D 11%-Perry 13%-Perry 2 -2
Pennsylvania Governor
Corbett-R v. Onorato-D 9%-Corbett 9%-Corbett 0 0
Colorado Governor
Hickenlooper-D v. Tancredo-C v. Maes -R 9%-Hickenlooper 14%-Hickenlooper 5 5
Nevada Governor
Sandoval-R v. Reid-D 15%-Sandoval 12%-Sandoval 3 3
Rhode Island Governor
Chafee-I v. Robitaille-R v. Caprio-D 3%- Chafee 3%-Chaffee 0 0
Ohio Governor
Kasich-R v. Strickland-D 5%- Kasich 3%- Kasich 2 2
Massachusetts Governor
Patrick-D v. Baker-R v. Cahill-I 2%- Patrick 6%- Patrick 4 4
Florida Governor
Scott-R v. Sink-D 3%-Scott 1%-Scott 2 2
Connecticut Governor
Malloy-D v. Foley-R 1%-Foley 1%-Malloy 2 2
Average Error: 2.2
Average Trend to Republican 1.4

Our Gubernatorial polling came out a bit better, with an average error of 2.2 points.  Our Republican bias was even less, skewing towards Republican by 1.4 points.  Our worst performance was in two of the three-way races of 2010; with those polls taken out, our average error fell to around 1.5 points.

House Polling

TCJ Poll Margin Actual Margin
Palazzo-R v. Taylor-D 4%-Palazzo 5%-Palazzo 1 -1
Roby-R v. Bright-D 3%-Roby 2%-Roby 1 -1
Noem-R v. Herseth-Sandlin-D 3%- Noem 2%- Noem 1 1
Perriello-D v. Hurt- R 5%- Hurt 5%- Hurt 0 0
McNerney-D v. Harmer-R 3%-McNerney 1%-McNerney 2 -2
West-R v. Klein-D 7%-West 9%- West 2 -2
Djou-R v. Hanabusa-D 5%-Hanabusa 7%- Hanabusa 2 2
Labrador-R v. Minnick-D 10%-Labrador 10%-Labrador 0 0
Walsh-R v. Bean-D 1%-Walsh 0%-Walsh 1 1
Dold-R v. Seals-D 1%- Dold 2%-Dold 1 -1
Buerkle-R v. Maffei-D 2%-Buerkle 0%-Buerkle 2 -2
Average Error 1.181818
Aberage Trend to Republicans -0.45455

Our House polling was particularly strong in 2010.  Our polls showed an average error of just 1.2 points, and our “bias” actual trended toward the Democrats by 0.5 points.  I challenge anyone to find a pollster that showed less error (or bias, for that matter) in their House Polling. Taking FiveThirtyEight at face value, the next most reliable polling group for House polling was SurveyUSA which showed more than 5-points of error across its polls.

Sure, our solid performance in 2010 could have been beginner’s luck. I guess we’ll find out in about 9 days.

TCJ Research Team


From → Uncategorized

  1. Federalist Patriot permalink

    Nice analysis. You look like you are doing a great job. Do you have plans for more state polls this week leading up to Election Day? Thanks for your work.

    • TCJ Research permalink

      Thanks for the kind words. We’ll be releasing polls in Ohio, Wisconsin, and Nationally later today. We view Ohio and Wisconsin as the true battlegrounds this year, so we’ll have one more poll out in each of those states before election day. Additionally, we’ll conduct final polls in Virginia, Florida, Pennsylvania, Minnesota, Michigan, Nevada, Connecticut, Oregon, Iowa, New Mexico, Colorado, North Carolina, Indiana, and Missouri. If another state looks competitive by week’s end, we’ll throw that in as well!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: