Sunday, April 29, 2018

The battle with ambiguity - It is constant within investment management


...Choices in situations of extreme uncertainty, which I termed "ambiguity": sparse information, unprecedented or unfamiliar circumstances, lack of reliable frameworks for understanding processes, conflicting evidence of testimony, or contradictory opinions of experts....I felt that existing theories of appropriate behavior ("rational choice") in these circumstances were inadequate, in fact misleading...

- Daniel Ellsberg, The Doomsday Machine: Confessions of a Nuclear War Planner 

Daniel Ellsberg was a leading thinker on decision theory when he worked with the RAND corporation. The Ellsberg paradox, which he developed, is not given as much attention as it should for those focused on decision-making. The paradox is simple. People will take a specific bet with known odds than an alternative where the odds are unknown. If give a choice between picking from an urn with 50% winners and 50% losers or an urn with an unknown number of winners and losers, most will choose the urn with the known probabilities even though the unknown urn may have a higher number of winners and chance of success. You could say investors are always willing to take the deal with the devil they know against the devil they don't know.

Ellsberg in his recent new book, which is a sobering discussion of his study of nuclear war in the 50's and 60's, provides his broad definition of ambiguity. It struck me as a good description of the problem that investment managers constantly face. We don't always have all the information we desire. We are faced with unfamiliar choices. The explanatory power of our models is mixed, and there are contradictory opinions of experts. I can think of at least a half a dozen decisions that could be characterized by ambiguity in the first quarter of this year. 

The frameworks that are needed for decision-making have to be reviewed and improved. This improvement could be through the better use of scenarios building, forming ensembles of models, discounting experts, or adding information, but any current process should be reviewed to determine if there is a better way of making disciplined decisions.

No comments: