As always, Ellis makes an excellent point. Actions that seem crazy, even fool hardy, may afterwards prove to have been strokes of genius that distinguish the truly brilliant decision maker from his or her run-of-the-mill peers. Koizumi's daring 2005 campaign in which he in effect ran against his own party seemed astounding at the time, but proved highly successful.
He tapped into a ground swell of public dissatisfaction with the status quo, and his ejection from the party of anti-reform rebels, followed by his use of the assassins (shikaku) proved such a great story line that he commanded media attention for months on end and won one of the most resounding victories in Japanese politics. As the old phrase goes, he was crazy like a fox.

Unfortunately, Ellis' argument raises interesting questions that undermine to some extent his subsequent point that social science is much better at postdiction (or retrodiction) than at prediction. It is possible to have quite a debate about why Koizumi went down this path.
The standard explanation is that Koizumi's political instincts allowed him to read the popular mood far better than his rivals. An adherent to rational choice theory would then ask, however, is why did he have a better read? If you assume that we live in a world of perfect information and that all actors have more or less equal capacity for rational calculation, this shouldn't happen. Someone made a mistake, and it clearly wasn't Koizumi. So perhaps Koizumi had access to information that his opponents didn't. Or alternatively, his opponents'
room for maneuver was constrained in a way that his wasn't, perhaps because of the interests to which they were beholden to, and/or because of the structure of decision-making that they had to work with. Of course, people can switch patrons, or change institutional structures, but doing so takes time and energy (again, to use rat-c jargon, there are transaction costs). Or, there is yet another alternative, that there was such large element of uncertainty ("imperfect information") and people were essentially placing bets. Koizumi guessed right.
"Rucky" - peace symbol and smile.

In other words, even within rational choice framework, there are many plausible explanations for this past event. Political scientists can - and no doubt profitably will - spend ages debating them. Historians are no different. Having read quite a bit of history, I have come to appreciate how many great historical debates remain open. Was the dropping of the atomic bomb on Japan necessary to end the war, or could the same result have been achieved without doing so (as a result, as Tsuyoshi Hasegawa has recently argued so brilliantly, of the Soviet invasion of Manchuria)? Was Japanese Imperialism best understood as a reaction to Western Imperialism, or was it driven by home-grown factors? Or what combination of these factors best explains the event?

To make matters worse, if you were to ask Koizumi, he would undoubtedly say that he acted selflessly out a deep rooted sense of duty to help the country, without any consideration of whether he personally benefited.
Is that a true statement? Or is there a Machiavellian calculation that his selfish personal interests are best served by a stance of selflessness? Or - again - is it a mixture of the two? No way to tell - perhaps even for Koizumi himself, since - as Hitler is reputed to have said - the most effective liar is one who believes his own lies.

There are of course facts. No one can say that Koizumi lost the election in 2005. But trying to figure out why certain events in the social world happened is a daunting task, given that human beings are motivated actors and there is no way of directly measuring their motivations, despite the best efforts of generations of pollsters and political psychologists. All social scientists can do - regardless of whether they are rationalists and culturalists, or adherents of various other paradigms that emphasize psychology or the impact of impersonal structures such as class or the environment - is try to make probabilistic statements about how the variables that they emphasize tend to create certain kinds of attitudes under existing conditions.

All such statements tend to be tautological, since you are trying to infer the causes of action (the independent variable) from the actions themselves (the dependent variable). Peter Cave's comment that rational choice tends to be tautolgical is quite correct - in effect, all too often they "just so" stories that provide post-hoc rationalist explanations of how the elephant got his trunk, or won the election.
Unfortunately, in the past Culturalist explanations were singled out with some justification by Brian Barry and others as inherently post-hoc, tautological form of analysis that reflected the analyst's biases or plain laziness ("they just can't do the math"). Rationalist theories that rely on supposedly easily measurable variables - such as votes or profits - were supposed to remedy the problem. Unfortunately, as many in this thread have already noted, they just don't stand up all that well empirically.

Frustrating, I know, but the best that I have been able to come up with. There are, of course, people who persistently claim they can do more, but their track record on prediction or post-diction is just not very good.

Thomas U. Berger
Boston University