Red-Blooded Risk: The Secret History of Wall Street

Back to the Dissertation

I went back to my dissertation. In graduate school at the University of Chicago, I had worked for Professor Craig Ainsley. Craig is a Bayesian and was working on the problem of analyzing time series data—that is, data measured at time intervals like daily stock price changes or monthly rainfall—when the volatility of the series was both unknown and changing. Stock prices, for example, can spend a long period rarely moving more than half a percent in a day, and then suddenly start moving 2 percent or more on most days. Most common statistical methods assume homoskedasticity, that volatility is constant.

One day, a professor from California came to Chicago to give a seminar on his way to New York University for a job interview. He was working on the same problem as Craig and me, but from a frequentist point of view. Chicago seminars are notoriously rough, but this one still stands out in my memory. Craig went after the guy like it was a religious war, questioning—no, attacking—every assumption, every equation, every conclusion. Afterward he pulled me aside and said, “That guy is way ahead of us; we have to drop everything so we can publish first.” I realized it was not academic ambition animating Craig; it was the battle between Bayesians and frequentists, a battle that Bayesians take much more seriously than frequentists do. This was going to be a powerful new method, and whichever camp published first would win points for a major advance. The second publication, ...