Pollster.com

Articles and Analysis

 

POLL: Newsweek/PSRA Iowa Caucuses

Topics: 2008 , The 2008 Race

A new Newsweek/PSRA survey (story, results, conducted 12/3 through 12/5) finds:

  • Among 395 likely Democratic caucus goers, Sen. Barack Obama leads Sen. Hillary Clinton by a statistically insignificant margin (35% to 29%) in a national primary; former Sen. John Edwards receives 18% and Gov. Bill Richardson 12%.

  • Among 275 likely Republican caucus goers former Gov. Mike Huckabee leads former Gov. Mitt Romney (39% to 17%); Sen. Fred Thompson receives 10%, former Mayor Rudy Giuliani 9%, Rep. Ron Paul 8% and Sen. John McCain 6%.

  • All other candidates receive less than five percent.

From the Newsweek release: "Likely Republican caucus-goers are 16% of all Iowa adults 18+ (weighted cases). Likely Democratic caucus-goers are 24% of all Iowa adults 18+ (weighted cases)."

Compare to comparable statistics for the October Des Moines Register poll (12% and 10% respectively) as well as the actual turnout for the 2004 Democratic caucuses (5.5%) and the Republican caucuses in 2000 (3.9%) and 1988 (5.3%).

Update: Several commenters question whether the six-point Obama lead could possibly be statistically "insignificant." I used that word above to paraphrase the conclusion from the Newsweek article:

While the Illinois senator's lead among Democratic caucus-goers in this poll is not large enough to be statistically significant, things seem to be trending his way, Hugick said.

Hugick's calculation is correct. For this sample of 395 likely caucus goers, a six point margin is not large enough to be "statistically significant" assuming the commonly accepted 95% level of confidence. Remember, the margin of sampling error (+/- 6% in this case) applies to the percentage for each candidate separately, not to the margin between candidates.

While the Obama lead is not "significant" at a 95% level of confidence, it would be significant if we were willing to relax our level of confidence to about 85%. See Gary Langer's column from earlier this week for more detail on this issue.

 

Comments
Andrew :

It's the first time in my life that I hear a 6% lead described as a "statistically insignificant margin".

____________________

Chuck Miller:

I guess it is technically insignificant with a 3% margin of error... since Obama could be as low as 32% and Hillary could be as high as 32%.

But is does help confirm the trend that Obama is becoming the front-runner in Iowa.

____________________

C Miller:

I guess it is technically insignificant with a 3% margin of error... since Obama could be as low as 32% and Hillary could be as high as 32%.

But is does help confirm the trend that Obama is becoming the front-runner in Iowa.

____________________

Natt:

The margin of error for likely democrats is actually 6%, for republicans 7%.

____________________

Paul:

Huckabee +22 and Giuliani in single digits!

____________________

Andrew:

May I challenge any pollster to investigate how often, among polls showing a candidate leading by 6% before an election, this candidate wins and how often he loses? There is no doubt in my mind that the probability of said candidate to win, based on past history, is extremely good.

In fact, searching through a database of polls immediately preceding the 2006 congressional elections, we can see that in 14 of 17 polls this candidate favored by 6% won the elections. With odds like this, I would not call this lead insignificant. Here is the totality of polls, according to a Survey USA database, where this 6% margin was present, and how the favored candidate fared in the actual results:

Arizona:
Arizona State University predicts Kyl wins by 6%.
Result: Kyl wins.

Colorado House 7th District:
Mason Dixon predicts Perlmutter wins by 6%. Result: Perlmuter wins.

FL Governor:
Scroth, Elthon poll predictgs Crist wins by 6%
Result: Crist wins.

Kentucky House 3rd district:
Majority Watc predicts Yarmouth wins by 6%.
Result: Yarmouth wins by 2%.

Kentucky House 3rd district:
Louisville Courier Journal predictsd NOrthup wins by 6%.
Result: Northup loses

Kentucky 4th House District:
Survey USA predicts Davis wins by 6%
Result: Davis wins

Maine Governor:
Survey USA predicts Baldacci wins by 6%
Result: Baldacci wins.

MD Governor:
Garin Hart Yang predicts that O'mally wins by 6%.
Results: O'mally wins.

MD US Senate:
Polimetrix Hoover predicts Cardin wins by 6%
Potomac Research predicts Cardin wins by 6%.
Result: Cardin wins

MI Governor:
Survey USA predicts Granholm wins by 6%:
Result: Granholm wins.

Michigan Senate:
Strategic Vision predicts Stanebow wins by 6%.
Result: Stanebow wins.

Minnesota Governor:
U. of Minnesota predicts Hatch wins by 6%.
Result. Hatch loses.

MO US. Senate:
Survey USA predicts McCaskill wins by 6%.
Result; McCaskill wins.

OH US Senate:
Mason Dixon predicts Brown wins by 6%.
Result: Brown wins.

TN US Senate:
Hamilton Beatie predicts Ford wins by 6%.
Ford loses.

VA ballot question #1:
Survey USA predicts yes wins by 6%.
Result: Yes wins.

From this we can infer that for some reason unbeknownst to me, when a poll with a 6% difference between candidates is off, it's almost always off in favor of the top candidate.

http://www.surveyusa.com/Scorecards/SUSA2006ElectionReportCard.htm

____________________

Anonymous:

This poll is proof positive that Romney cannot win Iowa, or the nomination, because he is a Mormon. The minute they started the push polling and press to make Iowa voters realize that Romney is in fact a Mormon, his numbers started to go down. Huckabee's a born-again preacher, he's not going to go over at all outside the bible belt. But he'll do well enough solidifying the evangelical vote to take Romney out in Iowa. They must be high-fiving each other at Guiliani HQ over these numbers. Rudy doesn't care about winning Iowa, but Romney finishes a weak second, or even third there, and he can finish him off in NH.

____________________

quixote:

The significance of a 6% difference depends on sample size. Most national polls have over a 1000, which gives them their 3% margin of error. (If they went to the trouble of calling 10,000 people, the margin would go down to less than 1%, but they don't consider the added expense worth the added accuracy.)

This poll had around 400 people. The margin of error is probably a bit larger than 6%. So, yes, 35% and 29% is (just within) statistical equivalence.

____________________

Chris S.:

"May I challenge any pollster to investigate how often, among polls showing a candidate leading by 6% before an election, this candidate wins and how often he loses? There is no doubt in my mind that the probability of said candidate to win, based on past history, is extremely good."

Except that in this particular poll, the MoE is a whopping 6% for the Democratic sample, and an even larger 7% for the Republican sample. I'm guessing that in most of the examples you site, the sample size was reasonably large and the MoE wasn't so huge.

____________________



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