I've critiqued economists who believe in purely rational actors and markets, but this is not to critique number crunching in general. Indeed, a recent NYT Magazine article on an NYU political scientist is quite tantalizing. An article by Clive Thompson titled "Can Game Theory Predict When Iran Will Get the Bomb?" focuses on the work of Bruce Bueno de Mesquita, who trained at Michigan and developed his ideas based on the work of William Striker at Rochester University. Bueno de Mesquita claims quite a track record forecasting events using an acute analysis of the players in the decision-making and an analysis of their levels of interest and commitment and power. Also, he uses a theory of coalitions. It seems that this works very well. I must say that I'm happy to learn that Bueno de Mesquita doesn't make forecasts by percentages: it either happens (his particular forecast) or it doesn't.
I was very interested to learn that he interviews players and learns as much about them as possible before making any predictions (i.e., feeding data into his proprietary program). Thus, a very human element remains. It appears that a computer can perform calculations of coalitions and run scenarios that no single human mind could manage. So be it. But how does one filter out "the passions"? I can understand that reason and interests sort themselves out over the long run, but the passions can hit like a sudden summer thunderstorm: brief but forceful. Perhaps those are the forecasts that don't work out--or that suffer from the random event. Query: how would his prediction about approval of a health care reform bill be affected by Ted Kennedy's death? (Apparently forecasting Clinton's effort was a miss for him.) So, in the end, I admire the work but I still believe that we can go wrong placing too much emphasis on the predictable based on induction. Too many Black Swans swim around in reality ready to appear out of nowhere.