Last month, inspired by a recent blog post in the Washington Post, and more generally the sabermetric revolution in baseball, I set out to sort out who the MVP was. You know, the Most Valuable President.
We have a pretty good handle on who the great presidents were, who the lousy ones were, and who falls in between. Those assessments tend to be outcomes based, however. They ignore things like context and process. It’s obvious that some Presidents were dealt a lousy hand, and others were able to coast on good times. Put differently, some Presidents made lemonade out of lemons; others took the lemons they received, carefully planted them, lovingly watered and fertilized them, and eventually produced an entire orchard of lemons.
Baseball analytics tries to strip away such external factors, and get measures that come as close as possible to players’ “true” performance: how do they “really” hit, pitch, run, and field, independent of the influence of opponents, teammates, and surroundings alike.
I wanted to do something similar for Presidents, so I developed the concept of the EAR statistic (Executive value Above Replacement). Basically, if we account for a few basic factors (I included economic growth, the year of the presidency, and the onset of war or other significant foreign adventure), what should have happened to the President’s approval rating in a given year? The extent to which Presidents over or under-performed the expected change in approval is the EAR statistic for that year.
I know, right? Genius! (Admittedly, the entire regression fails the F-test in rather decisive fashion, arguably rendering all that follows bunk. Still, proof of concept, right? Right?)
By popular request I said I’d post once more with a bit more detail. This is that post.
Here’s the model I used:
The EAR statistic is the model’s residual, the difference year over year between predicted and actual results. The higher the EAR, the more the president outperformed expectations for the year. It’s quite limited by the data I had easily available to me, namely the changes in presidential polling compiled by Philip Bump in the WaPo story I linked to up above (coverage only from Ike to Obama) and even moreso by GDP growth (Not a US politic specialist by any means, I used what I had to hand, which was WB data, which runs from 1960 to the present.) Sadly, that means we’ll never be able to compare with some of the great duds of the 19th century. (I’m looking at you James Buchanan.)
My friend and colleague Charles Breton was kind enough to diagram out the results I calculated. Here’s how the best and worst EAR years break down since 1960 (the first year for which the World Bank provides GDP growth data.)
A few things are immediately clear–external circumstances still matter quite a lot. I don’t think even GWB’s greatest fans would claim that his work in 2001 represented the single best Presidential performance in the last 50+ years, an intuition reinforced by the fact that his turn in 2002 proved to be the third worst in the set.
Even so, there seem to be some recurring patterns. Glancing at the graphical portrayal of the results. I’m struck by how much better comparatively recent presidents do. Is it possible that the US has been in better hands during the last thirty years than it was when the Greatest Generation had its hand on the tiller?
Statistically this is borne out as well. As I mentioned in the previous post, by including a “presidential dummy”™ term in the predictive regression, I developed a sense of how well the different presidents performed on average. The full results can be seen here. Near the top are Reagan (2nd), Obama (3rd), Bush Mk II (4th), and Clinton (5th). At the bottom are the Nixon/Ford year from hell in 1974, Johnson and Kennedy(!) Again, Nixon would have come off likely at the bottom had I been able to pin him with his final fall.
Another possibility is that presidential spin has become more effective, insulating them from the worst bumps. A third is that the current state of two solitudes existing in US politics tends to provide a cushion—no matter how bad things get, Democrats and Republicans prefer to support their own POTUS than think about voting for the other team.
Or maybe the fault lies with the voters. Perhaps we live in a quiescent age, in which we love to complain, but have forgotten how to demand and expect the best from leaders.
Or perhaps, like everything else here, it’s just spurious.
One more graph from Charles, this one showing changes over time: