Pollster.com

Articles and Analysis

 

OH: 45% Portman, 38% Fisher (PPP 8/27-29)

Topics: Ohio , poll

Public Policy Polling (D)
8/27-29/10; 475 likely voters, 4.5% margin of error
Mode: Automated phone
(PPP release)

Ohio

2010 Senate
45% Portman (R), 38% Fisher (D) (chart)

Favorable / Unfavorable
Less Fisher: 24 / 32
Rob Portman: 29 / 28

Job Approval / Disapproval
Pres. Obama: 41 / 54 (chart)
Sen. Brown: 33 / 36 (chart)
Sen. Voinovich: 22 / 48 (chart)

 

Comments
StatyPolly:

Beats Ras again.

I bet if you combine all the various state gubernatorial and Senate races from Ras and PPP for the past few months, the difference between the two robopollsters will be indiscernible.

They almost always track each other to near perfection.

____________________

Field Marshal:

But Ras has an obvious R bias while the other is just an accurate Dem pollster.

____________________

boomer40:

It's like Fisher isn't even trying.

____________________

StatyPolly:

I probably shouldn't have said "past few months". But definitely since PPP went to LV model. They're like twins. But one is good, the other - evil:)

____________________

Von Wallenstein:

how can PPP match Ras when Ras has a GOP house effect of like +5? is PPP now part of the vast right-wing conspiracy too?

____________________

dpearl:

"I bet if you combine all the various state gubernatorial and Senate races from Ras and PPP for the past few months, the difference between the two robopollsters will be indiscernible."

That is my impression too. Of course in this case we can actually look at the data (I'll post one state at a time as I look this up). Let's do LV comparisons only for the polls since June with Ras poll closest in date to PPP polls - dates below will be for PPP closing date.

date Ras PPP Ras/PPP difference

Ohio Gubernatorial race:

6/27 +7R +2R Ras +5R
8/29 +12R +10R Ras +2R

Ohio Senate:
6/27 +4R +2D Ras +6R
8/29 +6R +7R Ras -1R

average for Ohio for four results Ras +3R compared to PPP.

____________________

dpearl:

Here's Pennsylvania Senate

date Ras PPP Ras/PPP difference
6/27 +4R +2D Ras +6R
8/16 +8 +9 Ras -1R

and Pennsylvania Gubernatorial
6/21 +10R +10R Ras +0R
8/16 +14R +13R Ras +1R

average for these three Pennsylvania results is Ras +1.5R

____________________

dpearl:

Here's Florida Senate (comparing Rubio to Crist - taking margins off of Pollster charts)

date Ras PPP Ras/PPP difference
7/18 +2R +6i Ras +8R
8/22 +10R +8R Ras +2R

Florida Gubernatorial (Scott vs Sink only)
7/18 +4R +6D Ras +10R
8/22 +3R +7D Ras +10R

average of these four polls has Ras +7.5R compared to PPP


____________________

dpearl:

Here's Washington Senate
date Ras PPP Ras/PPP difference
8/1 +2D +3D Ras +1R


4 states so far - Ras gives results more favorable to Republicans in all four (or PPP gives results more unfavorable to Republicans if you want to see it that way).

____________________

StatyPolly:

Okay okay Dpearl. Stop it Stop it I give I give.

Take a lunch.

But.. are you sure those PPP's are all LV's?

____________________

dpearl:

Here's Nevada Senate (here the PPP poll is dead center in time between two Rasmussen polls so I averaged the Rasmussen polls)
date Ras PPP Ras/PPP difference
7/18 +1.5R +2D Ras +3.5R

Gubernatorial
7/18 +15.5R +14R Ras +1.5R

average of these two results has Ras +2.5R compared to PPP

Seems like a pattern at this point so I won't go on (and Pollster.com is making it hard for me to post because of the too-many-posts-in-a-short-time-rule).

Sorry guys - I think data has to trump impressions.

____________________

dpearl:

Yes - all LV's.

____________________

Cederico:

I think its clear Rasmussen is considerably more GOP friendly in its results.

On this Ohio race, it just seems that Fisher is failing to gain traction. There are many reasons to point to...Obama's ratings, Strickland's struggles, a more GOP leaning electorate, etc. My thinking is that Fisher would fair much better if he had money to get his message out and attack Portman. But instead he faces an 8 million dollar cash on hand gap!

That last fact more than anything is why I beleive that Portman will win this race in the end. Portman has many weaknesses that could be exposed but Ohio is expensive to run ads in and I am not sure the DSCC is gonna spend much to assist Fisher after seeing poll numbers like these.

____________________

StatyPolly:

Dpearl, hate to nitpick, but the only meaningful diff is in FL, and both Gov and Senate PPP polls on 7/18 were RV's. I realize that they started doing LV's for most of their polls back in June, but those two weren't LV's for whatever reason.

And if you look at August only, just about all results are damn spanking close.

Here is perhaps an easier way to compare.

http://www.realclearpolitics.com/epolls/latest_polls/elections/

If you click on individual races in the third column..

____________________

seg:

dpearl:
Thank you for sharing the data. It does suggest a modest but fairly consistent difference. You have been careful to avoid saying that represents a bias by Ras, which is good since there is little else to compare and either one may be the more accurate or both may be wildly inaccurate.

The thing I will add is this: if two analysts tried to put together an LV filter to predict November, I would consider it remarkable if they differed by only a couple of percents. They are both using educated guesses. Whatever weighting they used would presumeably be used for all of their polls and thus could be expected to produce consistent differences.

Since LV requires judgments about voting behavior, both PPP and Ras are subject to conscious or subconscious biases and both have strong political viewpoints. For that reason, I would have the greatest confidence in the average of the two because the most likely outcome is that each has a modest but opposite unintentional bias.

That is mostly academic, however. I view your data as showing that they corroborate each other as well as can be reasonably expected given the uncertainty all observers have about which voters will show up in November. There is decreasing doubt every day that that reps will do very well; they remaining issues are HOW well and about specific races.

____________________

nick283:

Yeah, PPP even says in it's latest Ohio poll that it was the first one of likely voters. So with likely voters for Ohio it's +.5 R for Ras and for PA its 0 for ras if you use likely voters. Just looking at the senate races in those two states Ras is actually 1 point more D among likely voters.

____________________

dpearl:

"Dpearl, hate to nitpick, but the only meaningful diff is in FL, and both Gov and Senate PPP polls on 7/18 were RV's."

I did all of the above pretty fast and was just reading off of the Pollster.com charts the way Emily classified them and they were labeled LV there. Looking at the original press release from PPP - I don't see it mentioned whether it was RV or LV they just said a survey "of voters".

____________________

dpearl:

"For that reason, I would have the greatest confidence in the average of the two ..."

SEG: Your intuition here is absolutely correct. Just like averaging individual measurements gives a more reliable result - it is essentially a theorem of statistics that averaging across different models will also produce more reliable results (and often less biased results as well).

____________________

Emily Swanson:

Hi dpearl,

To clarify -- based on conversations I've had with Tom Jensen at PPP, they sample from registered voter lists, but in almost all cases they accept answers from others in the household. We classify that as a likely voter screen, however they do move toward a more active likely voter screen later in the election cycle. Unfortunately for the purposes of our charts or a relatively short blog post it's difficult to get the complexities in there, and they rarely explain it in their releases. In the case of their national polls or a few other releases, they do say that they specifically and only accept responses from registered voters.

Emily

____________________

StatyPolly:

Yeah, they have "PPP surveyed 900 Florida voters from July 16-18."

I was looking at RCP and they had both polls classed as RV's.

http://www.realclearpolitics.com/epolls/2010/governor/fl/florida_governor_scott_vs_sink_vs_chiles-1607.html

But no big deal. The two pollsters definitely got closer in Aug.

____________________

dpearl:

Thanks Emily. That's good to know. It will be interesting to see how much better things align as the election draws closer and their LV screen becomes more "active".

____________________

dpearl:

"But no big deal. The two pollsters definitely got closer in Aug."
StatyPolly: Cool. I didn't notice that at first. For the five states I looked at above: In June/July we find the eight Rasmussen polls to average +5R compared to the PPP polls. In August we find the seven Rasmussen polls to average +2.3R (median of just +1R) compared to the PPP polls.

If PPP really changed their methodology during that time period ... this fits very well with the often cited idea that shifting to a model based LV screen from a basically RV screen will move things 3 or 4 points toward the Republicans.

Nice catch.

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR