In the New York Times, G. Elliott Morris and I discussed our American Voter Twitter bot, and what we can learn about the public’s beliefs system, ideological idiosyncrasies, and polarization from it. Find the article here, and follow the Twitter bot account here.
Last year, I blogged about bias in the CCES’s validated turnout estimates — how much they were over- or under-estimating the true turnout rate. However, I used the voting eligible population (VEP) as the denominator in the calculation of actual turnout, whereas the best comparison for rates from a survey like the CCES — that covers all adults — ought to use the voting age population (VAP) as the denominator (thanks to Shiro Kuriwaki for pointing this out). Below are recalculated state-level bias plots that use VAP turnout as the “actual turnout” comparison. I also add on validated turnout from the 2018 CCES.
One quicker (and likely better) way of summarizing these results is through a CCES vs. actual turnout by year small multiple graph, with a 45-degree line indicating over and under-estimates. I show that below, where the key takeaway — the validated CCES turnout more closely matching actual turnout over time — becomes clearer.
I recently posted a GitHub repository that contains raw data, questionnaires, and information for surveys I conducted on the Dartmouth College student body from the Spring 2016 to Spring 2018 terms, as part of my work for The Dartmouth (The D) student newspaper. You can find it here.