Tuesday, November 2, 2010

Beware “leaked” exit polls

Edison Research, the largest and most widely used company for conducting exit polls during US federal elections, is conducting research in 26 states today. Per a contact of mine at the Mobile Press-Register, all exit polling data will be quarantined until 4:30 CDT.  None of the races in Alabama or Mississippi are being sampled according to my source, but the Florida Senate and Governor races are, as is the Senate Race in Louisiana.

The Louisiana decision is odd, as David Vitter will almost certainly cruise to reelection, while the Alabama Governor’s race, the AL-02 Congressional race and the MS-04 Congressional race would be interesting.

Exit polls have not been as good at predicting a race’s outcome as they have been in understanding why a particular candidate won or lost (more on that, below).  They’re supposed to show trends among demographic groups.  The breakdown in using exit polling as a predictor is that those trends may or may not be representative of the overall electorate, introducing bias.  Furthermore, those trends don’t always reflect turnout.  An exit poll may breathlessly report that white protestant females are breaking 75-25 for Candidate A, but not tell you that the white female Catholics voted in much larger numbers.

The real point of this post—if you hear anything about leaked exit polls, dismiss it out of hand.  Exit polling data is almost never leaked now, and those with alleged inside information about what the polling data contains are talking because they have an agenda, and its never an agenda that favors our side.  In the 2000 election, there were “sources” floating around exit polling data from several states showing Al Gore with big leads among key demographics.  The data were floated by political operatives attempting to suppress GOP turnout in afternoon voting. Some think it may have worked. And don’t forget that Edison’s own exit polling data predicted that John Kerry had defeated George W. Bush in 2004.

Even though I do have a good source at the paper, I won’t get any exit polling data until you see it on one of the major networks, and even then there might be an agenda behind the discussion. And, even if there is no agenda, the data are likely to be as useless as they were six years ago.