At the beginning of the primary season, the DNC defined which polls would be valid for the debate qualification criterion. Approved polls have to be associated with or conducted by the following: the Associated Press, ABC News, CBS News, CNN, The Des Moines Register, Fox News, Monmouth University, NBC News, The New York Times, NPR, Quinnipiac, the University of New Hampshire, The Wall Street Journal, USA Today, The Washington Post and Winthrop University.
As an engineer, this list immediately raises questions. Not because I don’t trust those organizations, but because there is no statistical reason as to why those polls are selected over others.
If the DNC was really interested in a transparent process, it would have defined the approved-polls based on a minimum number of samples, a certain margin of error, and a verifiable sampling methodology.
But instead the DNC pre-selected the “trustworthy” or “reliable” polls.
Well, let us see if indeed the DNC selected the most reliable sources. The figure below show the last “reliable” polls. Some were conducted via web (Like The Economist/YouGov), or others exclusively via cell phones (like the Gravis poll). Interestingly these two polls which do not rely in any measure in landlines have Tulsi Gabbard polling at 3 and 5% respectively.
An interesting trend can be seen from the busy graph above, it would seem as if the polls with a lower margin of error were actually reporting higher polling numbers for Tulsi Gabbard. Any first year student of statistics would tell you the more you sample a population, the confidence on the numbers increases. This can be easily seen in the figure below: As the sample size increases, the poll error decreases.
Four polls are highlighted, those where Tulsi is polling at >2% and are not considered valid by the DNC. The Post and Courier Poll, and The YouGov/Economist polls are among the most reliable given the sample size and methodology, and yet, they are excluded in favor or less reliable polls, some with as few as 400 samples.
What is even more concerning, is that in the latest Monmouth University poll, Question 15 asked participants about several (not all) presidential candidates.
Where is senator Gillibrand? Representative Tulsi Gabbard? One could argue low polling candidates would not make it, but if so, what are Joe Sestak or Tom Steyer doing in the poll?
Anybody is welcome to perform their own analysis with the available data, and anybody is welcomed to provide a valid alternative interpretation of these results. To me, the data is indicative of foul play. The fact that polls were selected based on organization source, and not statistical significance, is extremely troubling.
August 6th (GRAVIS)
August 4th (Monmouth University)
August 5th (Qunnipiac)
August 12th (The Post and Courier)
August 13th (Fox News)
August 16th (GRAVIS)
August 18th (SSRS/CNN)
August 20th (The Economist/YouGov)
August 20th (Monmouth University)
August 26th (Emmerson)