June 23, 2010 05:24 PM UTC

Norton's Poll, its methodology and what it all means

•  
• by: H-man

Since Buck won the top line at the Assembly with 77% of the vote, which Norton skipped, there have been a total of three polls taken tracking the US Senate race among Republicans.  The methodology of the first two polls are fully  disclosed and relatively fair.  The results are consistent and public.  

The third (Norton) poll is not consistent with the first two and has not been make public.  When you drill down on it, you can see why the methodolgy is not fully disclosed and why the results are not a fair depiction of what it purports to survey.

The first poll was performed my Magellan, the same polling company that Buck and most of the Republican candidates in Colorado use.  It was conducted on June 8th.  It shows Buck leading 42% to 32% with 26% undecided.  Buck leads Norton in all sex and age demographics.  He does best among males and younger people.  

The Methodology employed in the Magellan poll is reported as follows:

Survey Methodology

This survey was conducted using automated telephone touch tone technology. The sample was randomly drawn from a Colorado voter file among households containing at least one registered Republican voter. The response data was weighted to reflect past Republican primary voting demographics from the 2008, 2006, 2004 and 2002 election cycles. The 1,026 interviews of likely Republican primary voters were conducted on Tuesday night, June 8th between 5:30pm and 8:00pm. This survey has a margin of error of 3.06% at the 95 percent confidence interval.

The second survey was performed by SurveyUSA and commissioned by the Denver Post/ Channel 9 and the Greeley Tribune.  It was conducted between June 15th and June 17th and released on June 20th.  It shows Buck leading Norton 53% to 37% with 10% undecided.  It shows Buck leading in all age, sex and ethinic demographics.

The methodology of the second survey was described by the Greeley Tribune as follows:

Roughly 3,000 voters participated in automated telephone survey conducted by Survey USA June 15-June 17. Independent Denver-based pollster Floyd Ciruli said while automated polls do sometimes have problems – they don’t include cell phone numbers, which can skew the results, for example – Survey USA has a strong record for accuracy.

For questions involving likely Republican primary voters, the margin of error is 3.7 percentage points. The margin for questions involving likely Democratic primary voters is 4.4 percentage points.

All registered voters were randomly assigned into one of two groups to hear questions about certain November contests and issues in the news.Questions asked of these “split samples” have a margin of error of 2.5 percentage points or 2.6 percentage points.

The third survey was conducted as a paid poll for the Norton campaign.  It was started the same day the Denver Post poll came out.  The poll has not been made public but certain results have been released.  The Norton poll claims Norton is leading Buck by 39% to 33% with 28% undecided.  The poll indicates she is leading significantly among women, and what was personnally, most unbelievable, she is leading in people who identify as Tea Partiers.

The Tea Partier Norton lead was so far over the top that I had to dig further.  How could that be?  She is not even invited to speak or attend many of their events.  It caused me to look at the description of the methodology used in the survey.  Here is what it says:

Methodology

Public Opinion Strategies conducted a statewide telephone survey among 400 likely Republican primary voters on June 20-21, 2010. The survey has a margin of error of +4.9%. Voters were contacted using a Colorado voter file of registered Republicans.

This year marks the first year Colorado will use vote-by-mail as a primary voting mechanism.  In accordance with vote-by-mail procedures, registered Republican voters who cast a ballot in the 2008 general election, or are active voters who did not participate in 2008, will receive a “live” ballot in the mail for the primary election in most of Colorado counties. New registrants or voters who have switched

their party registration since the 2008 election will also receive a live ballot by mail in most Colorado counties. Voters in counties with polling centers or polls may request a mail ballot. Our sample universe reflects this expanded potential universe in the primary election. Voters are then screened for their

likelihood to vote in the upcoming primary election

The Norton poll does not claim that it was done as a random survey.  All we know is that 1) it was a statewide telephone survey; 2)of 400 likely Republican primary voters; and 3)contacts were made using a voter file of registered Republicans.

If you get to choose the 400 likely Repblican voters based on criteria like, being members of your campaign staff, and don’t even pretend that the selection of voters was random, I guess I should not be surprised that the poll shows Norton leading with the Tea Party crowd.

Garbage in, garbage out.