Investment Decisions under Almost Complete Causal Ignorance

Abstract

This article investigates investment decision making under conditions of almost complete causal ignorance. Using two basic notions of causal dependence, probabilistic and counterfactual dependence, as building blocks, a formal notion of causal distance is presented that gives decision makers the ability to quantitatively assess the proximity that different causal graphs have to each other. The latter are directed acyclic graphs that can also be used to represent causal relations among economic events. Once the causal distance of each graph in a set of causal graphs is determined, it is possible to select the graph with the shortest total distance to the other graphs. This in turn allows decision makers to select a course of action that will be beneficial regardless of the particular set of causal relations that is actually driving observed economic events. The article describes how causal distance values can be used formally within an optimization to facilitate portfolio construction.