Our exit poll showed that of those who answered the question on the survey, 48.8% reported voting for Bevin, 47.1% reported voting for Conway, and 4.2% reported voting for Curtis. According to the Secretary of State’s website, 54.7% of Boyle County voters voted for Bevin, 40.1% for Conway, and 4.6% for Curtis. That’s a difference of 5.9%, 7%, and 0.4%, respectively, and in a direction that over-states support for Conway and understates support for Bevin.
There are a few possible (and not mutually exclusive) explanations for this:
- It is possible that the 1.9% of voters who did not complete the question on the survey were mostly Bevin voters.
- It is possible that some Bevin voters indicated on the survey that they voted for Conway, despite voting for Bevin in actuality.
- It is possible that Bevin voters were less likely to agree to take the survey than Conway voters.
My strong hunch is that #3 is the most likely explanation, especially given that the 2014 Exit Poll had a 3-6% bias in favor of Democratic candidates (but not non-partisan candidates), although it is not possible to definitely prove this.
We attempt to correct for these types of effects by using a sample weighting procedure, which is a standard procedure used to correct for differences in how different demographic groups respond to the initial invitation to take the survey.
It should also be noted that a 6-7% difference is still not terribly far off from the standard accepted margin of error in the polling industry of 3% for most national surveys and professional polling firms. And I’ll also note that our survey was ultimately more accurate than the plethora of professional and partisan telephone surveys taken statewide in the lead up to the election.
This is an interesting puzzle to consider going forward: why are Democrats slightly more willing to take the Exit Poll survey than Republicans in Boyle County, Kentucky elections? I welcome ideas from interested parties.