Posted in Quick Hits on September 8th, 2016 by Ed

So I'm trying hard to collect more data on this, but I noticed an odd anomaly on the the "Poll of Polls" aggregator on (now part of Huffington Post). There are three blue dots (representing Clinton support in the respective polls) that are equal (at 37%) and clear outliers, showing much lower Clinton support than any other polls.

All three are conducted by the same agency, Rasmussen Reports. Rasmussen is a well-known partisan hack outfit that makes its money by generating results that please potential conservative donors. It tells clients what they want to hear, which in this case would be that the presidential race is either close or even favorable for Trump. The rightward lean of RR has been well documented, most publicly by Nate Silver after the 2010 midterm election. The green arrows identify the three Rasmussen results that stand out, as you can verify for yourself on Pollster's interactive site.


Two questions: What does the "poll of polls" look like without Rasmussen's data? Since they contribute only a few polls to a very large pool of data for a site like Pollster, I wouldn't expect a dramatic change. Their inclusion is, however small, exaggerating the competitiveness of the race in national polling. Second, why is data so clearly suspect not being looked at more closely, or perhaps withheld until its notable skew toward Trump relative to other polling can be explained? I tend to be suspicious of media efforts to depict the race as competitive, so although the explanation is probably innocent ("We need to be Fair and Balanced! Which means including intentionally skewed data as long as it's skewed to the right!") part of me feels like we're beginning a repeat of 2012 when all the networks insisted up until the last possible minute that Romney was OMGSOCLOSE to winning in order to prevent viewers from getting bored and wandering away.