Saturday, September 13, 2014

Lies, Damn Lies, and Statistics

Mark Twain wrote in his Autobiography, "There are three kinds of lies: lies, damned lies and statistics."  This quotation, a favorite of mine, might sum up all this talk about polling in the Senate race between Weiland, Rounds, Pressler, and other (uhh, I mean Howie).  

Pat Powers at South Dakota War College attacks a recent poll by SurveyUSA because it doesn't look good for his man, Mike Rounds, and shows other Democratic candidates polling better than to be expected.  He does so by going after the methodology used for the poll.
Part of the problem is, according to its methodology, it’s a push button poll combined with Internet polling.
Cory Heidelberger at Madville Times, points out that just a little while earlier that Powers was raving about the same poll.  He then goes on to defend SurveyUSA's methodology:
Survey USAVoter Reg
Survey USA over-represented Republicans, not Democrats. Survey USA under-represented Independents. A more representative sample would have put Rounds even lower and Weiland and Pressler even closer to pulling off the upset of the decade.
Powers in another entry crows about another poll ( showing Wismer far behind Daugaard, but doesn't share that poll's methodology, which also uses online sampling to get to its conclusions.  The website explains its methodology in part:
The Battleground Tracker is a four-wave panel study conducted by YouGov for CBS News and the New York Times Upshot during the 2014 midterm elections. The survey provides estimates for voting on every race for U.S. Senate, Governor, and the House of Representatives. Interviews are conducted with samples of online survey panelists that have been selected and weighted to be representative of registered voters. This document describes in detail the statistical methodology used in the Battleground Tracker panel.
It targets registered voters that are available to go online.  The SurveyUSA poll actually calls people, mostly registered and some not registered.
SurveyUSA interviewed 775 South Dakota adults 09/03/14 through 09/07/14. Of the adults, 674 were registered to vote. Of the registered voters, 510 were determined by SurveyUSA to be likely to vote in the 11/04/14 general election. This research was conducted using blended sample, mixed mode. Respondents reachable on a home (landline) telephone (88% of likely voters) were interviewed on their home telephone in the recorded voice of a professional announcer. Respondents not reachable on a home telephone (12% of likely voters) were shown a questionnaire on their tablet, smartphone, or other electronic device. 
I am not an expert in statistical analysis.  I took a class in it during some graduate studies, but it is definitely not my strong suit.  These statistics are interesting to look at.  We all love polls, but we must remember that in the end, it is the real people that show up to vote in November.  If many Republicans think that Mike Rounds is going to be the hands down winner and decide not to show up, then Weiland/Pressler will pull off the upset.  If Democrats give up and don't bother to vote, then Rounds will beat Weiland by 30% or more and maybe even a third place finish for the Dems behind Pressler.

As Democrat, we should feel good that Rounds simple can't break 50% constantly because that shows people really don't like Mike that much.  However, we can't think this is solved.  The focus must be on voter turnout.  Get the early voting done.  Remind everyone that they need to get out and vote on November 5.  

As we get closer to November, the only thing that we can count on is that we will be hearing a lot of lies, damn lies, and statistics.

No comments:

Post a Comment