Decision attribution: Portfolio manager skill vs past performance

A PORTFOLIO manager’s job is to make decisions — all day, every day. Some of those decisions result in trades, but many more do not. So, an important question for a portfolio manager is which of their decisions are helping and which are hurting performance? Which types of decisions are they skilled at making, and which would be better made by someone, or something, else? And could they be using their own energy more efficiently by making fewer, better decisions? Enter decision attribution analysis – the largest and, for investors, most consequential area of behavioural analytics.

Until recently, these questions were nearly impossible to answer. The best performance attribution analysis — the primary evaluative tool for many investors and fund managers — starts with the outcome and works backwards to explain it by comparing it to the performance of an index alternative. But that doesn’t really help the manager: While it is useful for explaining why the portfolio performed the way it did during a certain period, this analysis cannot identify what the fund manager could do differently to achieve a better result.

Decision attribution analysis has been greatly refined in recent years with the exponential growth in machine learning capabilities. Decision attribution is a bottom-up approach, compared to the top-down approach provided by performance attribution analysis. It looks at the actual, individual decisions a manager made in the period being analysed, along with the context surrounding those decisions. It assesses the value those decisions generated or destroyed and identifies the evidence of skill or bias within them.