Thursday, September 9, 2010

Don't put too much stock in PPP's latest Texas poll

PPP is touting its recent Texas poll, in which it claims GOP voters prefer New Gingrich for the 2012 Presidential nomination. They show Gingrich with 23 percent, with Sarah Palin and Mike Huckabee tied for second at 19 percent each and Mitt Romney with 17 percent.

The problem is it's a PPP survey, and the Democrat polling firm is notoriously off the mark with its polling results. One reason is that the company uses ridiculously small sample sizes. In this poll, for example, it only surveyed 400 Texas Republican primary voters. With such a scarcity of data points, the poll has a margin of error (MOE) of nearly 5 percent, plus or minus. That means any of the top three potential candidates could be in the lead. So basically, it's a three-way tie.

Nevertheless, PPP's Tom Jensen tries to spin the poll metric that former speaker Gingrich beats Gov. Palin among Texas conservatives by 24 percent to 21 percent. The three-point difference, again, is well within the 4.9 percent MOE, so either one could actually be in the lead.

In grad school, Dr. Phil Taylor, who taught our statistics class, said that you really need a minimum sample size of 800 or so to reliably make inferences based upon your sample. A 3 percent MOE at a 90 percent confidence level, for example requires a sample size of about 750 data points at the bare minimum.

It gets even sloppier. PPP, in the disclaimer about its methodology, advises that "other factors, such as refusal to be interviewed and weighting, may introduce additional error that is more difficult to quantify." Yikes...

- JP

No comments:

Post a Comment