Post Mortem: 27th Congressional District in Texas

TX-27 Special Elections: TapResearch/PredictWise predict winner correctly, and two-party vote share among leading candidates very well.


 

TapResearch and PredictWise ran a poll in TX-27 before the special election to replace Blake Farenthold (yes, this guy below who resigned in disgrace over an ethics investigation and allegations of sexually harassing staffers, and likes to wear, shall we say, comfortable (?) pajamas).

 

blake-farenthold-ducky-jammies-thumb-700x466-182788

 

This was a difficult election to poll, with public polling all but non-existent, and only 36,268 total votes in total, less than a third compared to 2014. Our prediction (top-two-candidate vote share) was Michael Cloud (R) 67%, Eric Holguin (Dem) 33% (As a reminder: see plot below for top-three-candidate vote including Libertarian Daniel Tinus whose showing we over-predicted slightly, ignoring "other:); The truth was Michael Cloud Michael Cloud (R) 63%, Eric Holguin (Dem) 37%.

 

Figure1

 

This poll is a both a validation of Random Device Engagement methods to collecting data, especially for districts that are otherwise hard to poll, the reach and coverage of TapResearch (our RDE partner we conducted this study with), and the methodology of PredictWise. As a reminder, this is not just one incident of successfull RDE polling PredictWise has conducted, but fits nicely into our election predictions of MT-AL 2017, PA-18 and AZ-08. In these three races, PredictWise predictions had an average error of 0.5 percentage points.