Behavioural portfolio theory
The paper6 that introduced financial economists to BPT is a strange one. After some considerable reflection, the source of the strangeness becomes apparent. At first, when you pick up the paper, you are expecting a method or a tool comparable to MPT but incorporating behavioural insights. You are expecting something that you can use to compute a behavioural portfolio efficient frontier analogous to Markowitz’s bullet-shaped set. The ample discussion and formal analysis presented by Shefrin & Statman (2000) never actually gets to the point of providing such a tool and one could be forgiven for setting BPT aside for now to await its further development. This would be a mistake because while the paper does not provide a BPT replacement for MPT, it provides ample insights into how we can use advances in decision theory to understand more fully the decision-making process involving combinations of risky prospects and how that decision-making process may lead to systematic departures from the optimal benchmark provided by MPT. Shefrin & Statman (2000) have provided a pathway for adding deeper behavioural narratives to the MPT framework that we presented earlier.
One thing that usually bothers people about MPT is the idea that decisionmakers develop some understanding of the degree to which the payoffs to different risky prospects move up and down together over time (Levy & Markowitz 1979; Tversky & Kahneman 1986; Kroll et al. 1988). That is, that they are aware of and use the correlation structure. If we doubt this, a natural conclusion might be that the correlation structure can be ignored in analysis of decision-making because decision-makers are unaware of it. BPT highlights the folly of this position. The correlation structure is critically important to the risk-reward trade-off and the outcomes that decision-makers will achieve whenever combinations of risky prospects are possible, notwithstanding the fact that they might neglect it. The key point is that even if we think that terrorist groups which form combinations of attack methods neglect the correlation structure, it does not follow that the correlation structure is irrelevant to the outcomes of those combinations of attack methods and can be ignored.7