Saturday, September 16th, 2006...3:43 pm
An Answer?
I recieved an e-mail from Douglas Johnson, Vice-President of National Demographics Corporation, detailing some reasons the polling in the gubernatorial primary did not resemble the final results:
Here’s the answer to the question you pose about why CD 8 surveys were so accurate when the poll about Pima views on the Gov primary were so far off:
CD 8 surveys included from 400 to 700 respondents in CD 8, giving them minimal margins of error.
In contrast, the Behavior Research Center talked to 627 voters STATEWIDE, and only 251 of them were Republicans.The report does not give the details by county, but odds are that less than 40 of those respondents were Pima County Republicans.
With 40 or fewer respondents, the margin of error for Pima County Republican views on the primary is insanely high, at least plus or minus 15 percent.
The temptation to announce survey subtotals is always strong, but often it’s a mistake because the subsample size is simply too small to conclude anything from its results.
1 Comment
September 17th, 2006 at 12:14 am
It always come back the sample, doesn’t it?
I guess size really DOES matter…