4-wk sprint: How Accurate Are Clarity Scores This Year?
A quick update on what the November 4 electorate is looking like so far: there were less than 1,000 new ballots returned in today’s update. Not surprisingly, the Dem support numbers didn’t move much: just a hair downward to 54.3%. And the ongoing point of comparison is that, among this group of voters, Bill White earned only 41.9% of votes in 2010 while President Obama earned a comparable 41.8% in 2012.
I also take some time to run a quick check on how accurate the individual scores provided by Clarity seem to be this year. Comparing it against 2010 voters, the difference was substantial. Clarity scores overstated the actual end result by a typical point or two. But when looking at only the Mail Ballot subset, it was more dramatic: over 4 percentage points. Of course, Clarity scores weren’t in use at the state level in 2010. I’d expect them to be clunkier as newer and more data found their way into the scoring system.
Looking at how well the 2012 electorate matched up against the predicted Clarity scores, the results were much tighter, with Clarity overstating actual Dem support (in this case, for Obama) by less than half a point among Mail Ballot voters. There was still enough variation among House Districts to give some cause for concern. But nothing wildly out of whack and nothing that doesn’t look un-explainable once you see the small sample size in many of the voter subsets.
All in all, the result is a good deal better than I expected to see. My assumption during 2012 Early Vote was that Clarity overstated Dem support by about 1-2 points. Here’s hoping that it’s as accurate as it would appear for this election.
Obviously, among the first things I’ll be looking for on Election Night is a comparison of how well the Clarity scores track with the actual results showing up online. In hte meantime, ProPublica’s post-2012 election story on Clarity’s scoring system is worthwhile reading.