Monday, September 11, 2017

The myth of candidate agency

As discussion of Clinton's book and self-reflection (or, according to critics, lack thereof) continues, it is worth taking the time to consider the perspective of quantitative political science-- i.e., the perspective nobody wants to hear or read.  (Hi, my very few readers!)  I will acknowledge that I have only read the snippets of the book that have made the rounds in the commentariat.  I have no intention of reading the whole thing.  I have more important things to read, and more fun things to read.  I'm more interested in the discussion surrounding the book.  So, here goes...

Everybody likes to think that we have some agency over our lives.  In many senses, we do.  If you want a prosperous life, the standard advice applies.  Those who study, go to college, save and invest, etc., do OK.  Those who major in fields with a real prospect do even better.  Sorry, but a degree in art history just doesn't get you as much as a degree in math, and if you expect otherwise, there is something wrong with your expectations, not the world.  (Political science is somewhere in the middle...)  The worse your starting point, the harder it is.  If you come from a poor background with a broken family in which nobody had ever attended college, the harder it is to pull yourself up, but if you start out from a rich family, you have to try to wind up poor.  Still, even if you start out poor, you have at least some agency.  Having agency doesn't mean your life is awesome or that the world is fair.  It means you have the ability to take action to improve things.

The Equifax hack is another of those events that challenges our sense of agency.  You can find yourself victimized, having done nothing wrong.  Maybe you shred every document, properly monitor your credit, etc., but those fuckwits at Equifax can't manage their computer system.  Still, you have some options.  You can place a freeze on your credit to prevent future identity theft, and watch your credit card statements closely.  If they already got to your social security number and opened an account in your name, well, that fucking sucks, but the sooner you act, the sooner you can get things back on track.  You have agency.  You may be in a shitty position thanks to the toxic combination of incompetence and some subhuman* pieces of fucking shit around the world, but you can act to pull yourself out of that shitty situation.  It will take time, and be miserable, and maybe costly, but you can take action.  Having agency doesn't mean your life is good.  It just means you can take actions to improve your situation.

And then there are candidates for public office.  Do they have any agency at all?  I mean that seriously.  I reference the Alan Abramowitz "Time for a Change" forecasting model, and others like it on a regular basis here, and those models predict election results on the basis of factors that have nothing whatsoever to do with anything that any candidate actually does on the campaign trail.  The Abramowitz model uses GDP growth in the second quarter of the election year, the sitting president's popularity (that would have been Obama in 2016), and a variable for whether or not one party has already won two terms in a row.  Abramowitz has also been playing around with incorporating partisan polarization to reduce the range of vote shares we might observe, but the point is that nowhere in the model does it factor in things like whether or not Clinton should have told Trump to "back off, creep" during that debate, nor whether or not she should have handled any other specific situation differently.

Candidates tell themselves that they have the capacity to influence election outcomes through their tactics, self-presentation, advertisements, debate performances, etc.  Journalists play along because they need to tell narratives.  The narrative of a simple statistical model is something that only a political scientist could find fascinating.  Yet, we have remarkably little evidence of the relevance of the day-to-day shit about which candidates, journalists and, well, Politico-readers obsess.

Who did have agency during the 2016 campaign?  I'm still skeptical that the Russian interference actually moved votes directly, even though campaign collusion is a serious issue, but the poll numbers were pretty clear that Comey's late announcement did move votes.  Comey had agency.  So to speak.  Comey took action, and those actions influenced the course of the election, more than any other single actor.  And yet, in the end, the Abramowitz model predicted a GOP victory because there was tepid economic growth after two Democratic victories.  That leaned GOP.

Clinton is going back over everything in the campaign, asking herself whether or not there was something she could have done differently.  Psychologically, that is understandable.  Democrats are pissed at her.  That is understandable, but misplaced.  Journalists took the bait.  Of course they did.  That's what they do.

But, I might as well do this again.


Sequence A is the actual sequence of presidential election results from 1944 through 2016.  Sequence B is what would be necessary for DDRRDDRR all the way through.  How many letters are different between Sequence A and Sequence B?  One.  1980.  Reagan won, making it an R.  Had it been a D, leaving everything else the same, the sequences would be identical.

One year is the difference between Sequence A and Sequence B.  What could Clinton have done differently to win in 2016?  That's the wrong question.  It presumes that candidates have agency.  You have agency over your own life.  Candidates have very little agency within the electoral process, at least when it comes to presidential general elections.

*Considering how low my opinion of humanity is, this is quite a statement.


  1. That's a bit flippant, though.

    The question is whether the sequence of DDRR is the thing or whether the operating mechanism of "Time for Change" is the thing. The former is more of a Macro Polity kind of argument, which sorta hinges on voters not liking the policy produced by those 8 years. The latter is more consistent with simple concepts like incumbency advantage and coalition of minorities. But, it matters for the argument, because of two things.
    A) agency. You can possibly fuck up and not win reelection. That nobody really does so is because they have all had a base level of competence.
    B) The sequencing gets totally fucked up. 1968. 1976 (though you might put a pin in that one). 1980. 1992. That's 4 incumbents not reelected out of 10 in the post-war period. That's a thing. So, that implies that Abramowitz's model might not be as good as a simple Macro Polity version, which I don't like either. It might not be as predictable as it seems.

    Also, the DDRR thing: why does that magically appear in 1946? Wasn't there before then. Can't argue it's because we reached the modern equilibrium of roughly tied parties...that only dates to really the 1980s (and, if we're being honest, the late 1980s/early 1990s). So, you add in "well, Ike could've beaten ANYONE" and "Vietnam would've doomed ANYONE" and now we're just fucking historians.

    So, if we're really coming back to the incumbency + economy/war explanation, I think half of that has agency. Obama/McCain: no agency there. WAY overdetermined election. 2016? Models predicted a tie. Agency was possible. Small problem: bad candidates on both sides. One liked grabbing pussies, the other had one.

    I'm REALLY uncomfortable making the "candidate effects cancelled each other out in 2016" claim, because we're REALLY into elephant-repellant glasses territory here. Because candidate effects ALSO cancel each other out for, well, almost all of the other elections ('64 and '72 being the prime counterexamples). I really don't like hanging my hat on candidate effects. But, I have to admit that I'm not sure that predictability is currently the same as explainability. Which is not good. (And, by the way, the same conclusion I have come to about most aggregate congressional elections models. But "this shit sucks but my operationalization of your shit doesn't work either" is hard to publish.

    1. A) When am I NOT flippant?

      B) If you want to argue Macro Polity, which you don't buy any more than I do, that doesn't give any more agency to Clinton, specifically, than Abramowitz, given that she was Sec. State four years prior. Incumbents? Sure, but she wasn't an incumbent.

      C) At the end of the day, you can go year by year and ask about year-specific effects, but the pattern is REALLY close, and making year-specific explanations risks ad-hockery.

      D) Why '44? Same reason we don't go back before FDR for any other modern thing.

      E) So, I come back to the pattern. The only third term since the FDR/Truman thing was Poppy Bush, who had second quarter GDP growth over 5%. Overall, though, DDRRDDRRDDRRDDRR.... Why? Fuck if I know, but it works.