In a post from a few days ago, Patri Friedman points to a little back-and-forth between Alex Tabarrok and Brian Caplan about “freaky economics”. Therein, Tabarrok references a paper by Hess and Orphanides that suggests that presidents go to war to get re-elected:

If the economy is doing well, a sitting president is up on one score and without evidence can be assumed to be as good as the challenger in war-making ability. Thus, the president gets reelected. But if the economy is doing badly then an incumbent who cannot present evidence that he has superior war-making ability will lose for certain. Crucially, an incumbent can’t demonstrate war-making ability without a war – so when the economy is doing poorly and the President is up for reelection the model predicts more wars.

Hess and Orphanides define a war as “an international crisis in which the United States is involved in direct military activity that results in violence.” Using data from the International Crisis Behavior Project, they compare the onset of wars in first terms when there is a recession with (a) the onset of wars in first terms with no recession and (b) second terms. Stunningly, however, they find that in the 1953-1988 period wars are about twice as likely in first terms with a recession than in first terms with no recession and second terms (60% to 30%). The probability of this result occurring by chance is low.

Need I mention that the Hess and Orphanides model has proven to have predictive power?

Now, based on psychological intuition, I don’t necessarily disagree with Hess and Orphanides’ conclusions. Nor do I necessarily disagree with the Mencken line Patri summarizes with: “Democracy is the theory that the common people know what they want and deserve to get it good and hard.” However, I do have an issue with the justification that H&O give for their theory (at least as summarized by Tabarrok; I can’t access JSTOR from home). First of all, why only use 1953-1988? The American presidency’s been around for over 200 years, so why only use 36 of the more recent years? My guess: because, given that those are the years at the heart of the Cold War, armed, international conflicts were far more prevalent and, therefore, indicative of the desired conclusion. Also, before I get into the above more deeply, note that the H&O model only predicts war if the economy is bad…but the 36 years being discussed were probably (I’m more or less just making this up, but I suspect the numbers would pretty much back me up here) the 36 years of greatest sustained large-scale growth in the history of the world and that, furthermore, arguably the worst period, economically speaking, occurred during the first (and only) term of Jimmy Carter, who was (again, arguably) the least war-like president of at least the last 7 decades or so.

Anyway, let’s be honest, here: those 36 years cover 9 election-cycles (which, depending on how you slice it, gives anywhere from 9 to 11 “presidential terms,” depending on how you classify 1964 and Ford’s time in office), which is a pretty damned small sample size to be drawing “predictive” conclusions from. This becomes more clear if we break it down more explicitly. I’m assuming that Johnson’s first year in office (after JFK’s death) isn’t really being counted as a first term, since he was still eligible to be re-elected in 1968. On the other hand, I’m assuming that Ford’s time in office does qualify, since, even if he had won in ’76, he wouldn’t have been eligible in 1980. Hence, we have the following terms and associated wars (as defined in the H&O paper and as I recall them; that being said, I’m sure I’ve missed some):

First Terms

• Ike, 1953-56: Korea
• JFK, 1961-64 (died 1963): Bay of Pigs, start/escalation (depending on how you look at it) of conflict in Southeast Asia
• LBJ, 1965-68: Vietnam
• Nixon, 1969-72: Vietnam
• Ford, 1974-76: Vietnam (still)
• Carter, 1977-80: None
• Reagan, 1981-84: Lebanon, Grenada

Second Terms

• Ike, 1957-60: None
• Nixon, 1973-76 (resigned 1974): Vietnam
• Reagan, 1985-88: None

So, by this count, we have “war” in 6 out of 7, or 71% of first terms (I’m guessing, in the paper, JFK gets put in the “no war” category, which is misleading, but, I suppose, technically correct, given their definition; anyway, that gives 5 of 7 or 57%, which rounds off to the cited 60%) and 1 of 3 or 33% of second terms.

Anyway, the point is, yeah, the percentages look pretty significantly different, but when you only have 3 items in the second category, it’s sorta hard to take those percentages seriously in terms of statistical significance.

Also note that if the paper had taken into account the years since 1988, things would look somewhat different. Each of the last five terms has included “an international crisis in which the United States is involved in direct military activity that results in violence”: Bush I had Iraq, Clinton had Somalia in his first term and Kosovo and Afghanistan in his second term (betcha forgot about Afghanistan, didn’t you?), and, obviously, Bush II has had Iraq and Afghanistan in both terms. If we include this period, all of a sudden the percentages become 8 of 10 or 80% for first-term presidents (or 90% if we include JFK) and 3 of 5 or 60% for second-term presidents going to war.

Viewed in that light, the apparent difference between hawks trying to get re-elected and dove-like sitting ducks (sitting doves?) begins to vanish. In fact, the truly cynical might look at those numbers and wonder whether the primary reason presidents get involved in violent international conflict isn’t simply that they have the capacity to do so. The cynic who holds such views might also note that none of the 12 aforementioned “wars” were waged under the auspices of a Congressional declaration of war. And the truly, incurably committed of such cynics might point this little factoid out not in the wistful if-only-we-would-go-back-to-the-Constitution way of knuckleheaded conservatives/Libertarians, but rather as a nod to someone with the initials JTK.