Monday, April 22, 2013

Fischer Black: We should be paid for teaching alone. Those who really want to do research will do it nonetheless. Also, how to test models...

In Perry Mehrling's Fischer Black,

Fischer Black, "I see our university system as similar to the former Soviet empire, and as having similar problems . . . teaching and research are too uniform. They do not respond quickly to shifts in tastes and technology. . . . And, most important, teaching and research cost too much. . . .  The basic problem is that we have too much research, and the wrong kind of research, because governments, firms, foundations, and generous alumni support it." Namely, pay professors for their teaching only, since those not interested in research will stop producing it, and those who want to do research will do it anyway. (pp. 300-301)

p. 112. Black would stress his theories of portfolio management using past stock behavior. If the theories did not work well, then you might learn how to fix them. Empirical work is not so much to test theories, as to guide us to likely theoretical enhancements. Hence, the usual sophisticated econometric procedures and statistical tests are less useful than is running the theory "practically" on a given data set and seeing if there is a substantial effect. (This reminds me of Tukey's Exploratory Data Analysis, or the way particle physicists work with their theories. In each case, what you are looking for is what is missing, and it is remarkable when things work reasonably well {as well as being dispiriting since there is nothing more to think about, hence the problems with the Standard Model of particle physics--it works much too well}.)  Black was skeptical of econometrics, with its use of (linear) regression models: misspecification, identification problems, collinearity, and lack of independence of the independent variables--so the meaning of the estimated coefficients is not at all apparent. (pp. 117-118).

Note: Fischer Black was an innovator in finances (Black-Scholes equation), and worked at Arthur D Little, U of Chicago, MIT and Goldman Sachs. His earliest work was in artificial intelligence.