Four reasons you can ignore that PPP poll

A Democratic polling firm’s survey of the 2014 gubernatorial race is garnering a moderate amount of attention in Maine. The Bangor Daily News has three blogs up talking about the poll and the Michaud for Governor campaign is touting it on their website.

But there are some good reasons why this survey, conducted by Public Policy Polling of Raleigh, N.C., tells us absolutely nothing about the gubernatorial race.

PPP’s automated methodology is cost-effective, but less accurate.

According to the PPP data provided, the survey was automated, meaning respondents did not actually communicate with a trained questioner. PPP has no way of knowing whether a middle school kid picked up the phone and pushed buttons until the survey concluded. Apart from that somewhat unlikely scenario, automated surveys, though cheap and useful, are widely regarded as less reliable than actual polls.

PPP oversamples women.

According to the most recent U.S. Census Bureau figures, 51.1 percent of Maine residents are female. Yet women comprise 54 percent of PPP’s survey size. The effect of this oversampling is to skew the poll results in Michaud’s favor.

According to the cross tabs PPP provides, 34 percent of woman and 49 percent of men approve of LePage. By oversampling women, PPP includes a disproportionate number of likely voters who do not approve of LePage.

Similarly, 58 percent of female and 42 percent of male poll respondents view Michuad favorably. Thus, including too many women in the sample not only depresses LePage’s approval rating, but also inflates Michaud’s favorability. (The same effect applies to Cutler’s ratings.)

PPP oversamples minority voters.

According Census data, 95.3 percent of Maine residents are white. In PPP’s survey sample, 9 percent of respondents said they were not white. This oversampling skews the poll results in much the same way as the oversampling of women.

When you look at the cross tabs, 41 percent of whites approve of LePage. Of those who pushed a button for “other,” only 38 percent approve of the governor. For Michaud, the survey indicates that 51 percent of white’s and 49 percent of non-whites view him favorably. By including a disproportionate number of non-whites in the survey sample, PPP further inflates Michaud’s favorability while deflating LePage’s approval.

The skew created as a result of the oversampling of women and non-whites extends into election-specific questions as well, as both groups report being more likely to vote for Michaud than for Cutler or LePage.

PPP’s question regarding Obama’s approval rating produced some extremely strange results which cast doubt on the accuracy of the entire poll.

According to the cross tabs, 14 percent of likely voters who identify as “somewhat conservative” and 6 percent who identify as “very conservative” said they approve of Obama’s job performance. That should strike even casual political observers as odd.

PPP is expected to release additional survey data for Maine’s 2^nd Congressional District Race Wednesday. But unless the polling firm uses a more representative sample, the results will tell little about that state of that race.