Democrats in special elections and horseshoes

I appear to have skipped a post on that Arizona special election to replace Trent Franks.  We're all going to miss Trent, aren't we?  Lesko won.  It was a Republican district.  Tipirneni outperformed the partisan baseline.  So, yay Democrats!  You lost by less than expected, because politics are like horseshoes!  "Close" counts, right?  Or... not.

This is a pattern.  We have had a bunch of special elections, and at the beginning of the series, I warned about them being "shiny things."  Don't get distracted by them.  They are tempting because outside of an election season, they look like fun, but they don't mean much, individually anyway.  At what point, though, do we have a pattern?  What about Conor Lamb?  You know, the one who won?  (Doug Jones being a special case...)

There are two basic points here.  The more observations we have, the more evidence there is of Democratic overperformance (hold your horses, though...).  One data point means very little.  Two?  Eh...  Ten?  Now you're telling me something.  Social science is about looking for patterns.  Tipirneni lost.  Hence, Democrats have nothing to celebrate regarding that race.  The fact that she overperformed means nothing.  The fact that Democrats have been consistently overperforming in special elections may tell us something.

Note that I am still being cautious here, which brings me to the second point.  Special elections still occur under different circumstances from the general election in November.  Political attention is focused on one district rather than spread across the country.  That changes the dynamic, or at least, it can.  Why might that matter?  Consider the storyline.  Democrats are looking at races where they expect to lose, and celebrating for overperforming.  From the Republican perspective, who cares?  They still win, except against Conor Lamb, or when their candidate is a child rapist.  Get to November, when flippable seats come into play, and the Democratic argument is that they overperform by the same levels, but Republican responses change too, and that's the inference problem.  Turnout levels change non-uniformly,* lots of things change.  What does this mean for November?  We still don't know.  Democrats are likely to pick up seats.  Betting says they'll take the House.

Still, don't read too much into one, or even a few special elections.  My standard rules about focusing on shiny things apply.

Ooooh!  James Comey did a town hall!  Lemme watch that!

(No!  Bad political scientist!  Off the couch!)



*At some point, I'll do a post debunking the currently fashionable line of bullshit that midterm elections have an intrinsically Republican bias because of turnout.  This one became fashionable as an explanation for the 2010 election among people who didn't have a long enough political memory to see the problem of, oh, say, 2006.  The previous midterm.  Midterms run against the party of the president.  When the president is a Republican, the bias is towards the Democrats (unless you have something really weird, like 9/11 counteracting that).  Don't fall for this bullshit.

Tweet

Subscribe to receive free email updates: