The Perils of Past Performance

The role of a fund investor is to identify active managers (individuals or teams) with some form of edge or skill that can deliver excess returns relative to a specific index or hurdle rate.  We seek to uncover a process (application of skill) that delivers a positive outcome (outperformance).  Whilst this may appear to be a simple concept, isolating skill in the field of investment is a complex task, particularly given the undue focus on how well a fund has performed historically – which as a standalone measure provides no indication of skill and can materially skew our (seemingly) objective judgements.

To evidence skill there must be a direct and deliberate link between process and outcomes. If I am playing golf, take aim at the flag and proceed to shank the shot, but the ball ends up in the hole after ricocheting off a tree, that is not skilful – even though the ultimate outcome is both the one I intended and positive.  Furthermore, there must be a large sample to validate skill – a single successful golf shot could be mere fortune, after one hundred attempts the influence of luck is significantly reduced.

The ability to discern skill also depends on the type of activity undertaken – some activities are dominated by randomness, others can be influenced by skill.  Michael Mauboussin (2012) suggested an intuitive test for judging whether something is more subject to luck or skill – is it possible to fail deliberately? Take a lottery – it is impossible to purposely fail when playing; provided the ticket is correctly completed, the probability of success cannot be impacted by the combination of numbers selected, and the result is entirely arbitrary. Contrastingly, chess is a game dominated by skill, with limited influence from chance or randomness; it is easy to intentionally lose a game by recklessly sacrificing key pieces.

We can apply such a framework to active fund management; however, we rapidly arrive at definitional problems.  Even if we assume that outperformance of a benchmark is a robust indicator of a successful outcome – over what time horizon should this apply? For example, an active equity manager could attempt to deliberately fail (or underperform their benchmark) by selecting a range of companies with characteristics contrary to those in which they typically invest; yet if the efficacy of this strategy was assessed over a short time period (one day, for example), the outcomes would be driven by luck – did the market rise or fall on the day? What was the key macro-economic news? If, however, we extend the time period then the impact of random short-term market vacillations should wane and fundamentals exert more influence.

Lengthening time horizons in our assessment of active management certainly improves our ability to distinguish between luck and skill – but it is by no means failsafe. Even over the long-term, a randomly selected portfolio can (and often will) improve upon a market cap allocation (absent fees) and active managers frequently outperform for reasons that are not directly related to their process. Thus, whilst long-term past performance information for a fund is preferred, as a measure of skill it is severely limited.

The problem with historic fund performance is that it is devoid of context and impacted by a blizzard of variables over which the manager has no control. Returning to the golfing analogy – viewing headline fund performance in isolation is akin to judging that I am a skilful golfer by observing on my scorecard that I recorded an eagle, without knowing that I dented a tree.  It is all outcomes and no process.

This presents a challenge for investors; we seek coherent and consistent narratives and want to believe that strong active fund performance is a direct result of skill as this is the simple, coherent explanation.  The world becomes incredibly complicated when a good process delivers a bad outcome, or vice-versa.

The issue of outcome bias – our propensity to judge the quality of a decision or process by its outcome – has been consistently evidenced in behavioural research (See Baron & Hershey, 1998), yet our awareness of the phenomenon seemingly does little to shake its influence. Survivorship bias, return driven ratings and performance screens still suggest an undue focus on historic performance. Whilst the common behaviour of investors selling funds that have struggled after three years and investing in those that have outperformed over the same period is well-documented.

Professional fund investors are aware of this issue, of course, and most have thorough due diligence processes for analysing active investors, which downplay the influence of the historic excess returns delivered by a fund. The problem, however, is that as soon as we are aware of the headline performance / outcomes of a strategy then it will impact our perception of all elements of the approach.  Outcome bias would suggest that our assessment of the quality of two identical strategies would differ markedly if one reported strong historic headline performance and the other weak. Therefore, it is not just that past performance can take undue precedence in our assessment of an active fund manager, but that knowledge of it can influence all other considerations.

This is not to suggest that past performance analysis contains no relevant information, or that it should be ignored entirely; however, where considered it must be framed by the appropriate context, rather than employed as a binary good / bad marker. The primary consideration should be – is the behaviour of the strategy consistent with expectations, given my knowledge of the philosophy and process adopted?  Even then, whilst providing some useful evidence about the historic characteristics of the strategy, assessment of past performance in this more nuanced manner provides minimal guidance on whether a manager possesses skill because it remains difficult to link the outcomes to any underlying process.

Identifying skill in an active fund manager is incredibly difficult – long-time horizons are a pre-requisite and past performance alone does more to mislead us than guide us – but it is not impossible. Crucial to the endeavour is thinking about different types of outcomes; focusing on specific decisions through time, rather than top-level fund performance. This involves understanding the rationale for holdings and positioning within a portfolio and observing their impact – this provides us with a far greater sample from which to judge skill and offers a more direct link between process and outcome (though admittedly not free from noise)

Such a method also allows us to pinpoint areas of skill and locate potential weakness.  There is no skill of ‘active fund management’, rather individuals or teams may possess expertise in certain facets (this could be areas of the market, types of security or particular disciplines, such as portfolio construction); disaggregating decision making and outcomes is a crucial means of understanding where competitive advantages or particular abilities may exist.   Blunt past performance evaluations provide no information on this aspect.

That isolating skill requires us to find evidence linking process to outcomes also highlights the importance of maintaining due diligence once invested in a particular active strategy.  When initially researching a manager much of the questioning will relate to historic decisions, and the manager will be vulnerable to post-hoc rationalisation and hindsight bias, where (consciously or unconsciously) the rationale for historic decisions is prone to revision.  Regular monitoring meetings with active managers offer reams of ‘live’ decision making evidence, which is far less susceptible to the aforementioned problems. This is particularly crucial when invested in a fledgling managers / approaches where the level of confidence that skill exists and will offer persistent advantage must necessarily be limited.

Given that outperformance is the ultimate objective of active fund selection, it is unsurprising that so much attention is lavished upon historic results; yet in a chaotic and unpredictable system focusing on such a context-free number is highly problematic.   Not only does it leave us exposed to mean reversion and conflate skill with luck; it serves to influence our perspective on all other relevant evidence. Although it is impossible to expect investors to research funds without an awareness of its past performance, we should continue to strive to reduce its relevance in any due diligence process and move to a point where investing in a fund with a strong recent track record might be seen as much a problem as it is a validation.

Key reading:

Baron, J., & Hershey, J. C. (1988). Outcome bias in decision evaluation. Journal of personality and social psychology54(4), 569.

Mauboussin, M. J. (2012). The success equation: Untangling skill and luck in business, sports, and investing. Harvard Business Press.