So who is the MVP (Most Valuable President)?
Last week, Philip Bump of the Washington Post published an interesting little post outlining what he thought were the best and the worst years to have been President in the last 70-odd years. The method was simple: compare across years how presidential approval, as measured by Gallup polls, changed over the course of the year. It’s just intended for fun, and not in any way scientific. The results are interesting, however. The best year? GW Bush, year 1. The worst? His father’s annus horribilis in 1991, when his approval dropped a stomach churning 33%.
Of course, it immediately struck me that I could do something similarly unscientific, yet WAAAAY more needlessly sophisticated. In part, I am inspired by the sabermetrics revolution here in its pursuit of WAR, or Wins Above Replacement. It’s a master statistic capturing a player’s total contribution to the team.
What if we could do something similar for politics? We political scientists do go on about the fundamentals, and indeed, basic factors like the decision to go to war, or changes in economic growth have consistently been associated with presidential election performance, even if they are not always decisive.
Accordingly, they provide a good starting point for what I am determined to call EAR—Executive Above Replacement. It takes into account things like year over year changes in economic performance, the start of military adventures abroad, and the presidential year. (i.e. 1 through 8. Years 5 and 6 are always a bloodbath, while years 4 and 8 tend to turn out well for the sitting president’s approval.)
Basically, using OLS regression estimates and a bit of subtraction, it tells us how much we ought to blame or praise the president for changes in his approval rating after controlling for war initiation and changes in the economy, both key fundamentals. There are lots of methodological notes I could make about how exactly I did this, and all the ways in which it is a bit sketchy here and there, but it’s nearly a stone-cold lock that no one will read them anyways, so I’m going to wait until asked about them.
And now, without further ado, I give you, the EAR of the President of the United States.
A couple of results jump out. Some Presidents, when handed lemons, made lemonade (Carter in ’79). Others took those lemons, planted in the ground, and ended up with many many more lemons (Carter again, in ’80). Conversely, some made good times seem better (Clinton in ’96), while others made even good times seem pretty lousy (Kennedy in ’62). President Bush the elder really did mess up in 1991. That’s almost all on him. Likewise, even in comparison with other foreign adventure initiation years, 2001 brought the US behind the US like nothing before or since. It didn’t last long however, as 2002 was one of the worst fundamentals-adjusted performances on record. Carter was really, really popular back in ’79, when he had little right to be. And there really was something about Ronnie and Bill. They both posted multiple top-ten years. Folks love sunshine and folksy empathy, I guess, regardless of just about anything else.
As for Obama, he’s mostly been middle-of-the-pack, but his performance in 2012 was notable. Even accounting for the fact it was a year 4, he still put up the 9th best president-year since 1961 (the first year I could get data for).
Here’s the full list, ranked from best EAR year, to worst.
|Year||President||change in growth||change in polls||EAR of the President|
One final note: the same approach allows me to isolate what each president’s average contribution is, year after year (specifically by inserting a presidential-specific dummy into my initial regression). These are in some ways the most surprising results of all. Ford actually comes off very well, though this is in part because his terrible decline in popularity coincides the year he and Nixon both were office. Reagan comes in number two, and look who’s in number three? There are two years left, but so far, the numbers suggest Obama’s been doing more than all right.
The only downside of this approach? None of it comes close to being statistically significant. It’s all bunk, and I’ve been deliberately wasting your time.