By Alan Dunne, CEO and founder of Archive Capital.
Five strategies for overcoming behavioral biases in manager selection
“The investment world worships success, deifying the market seer du jour. Instead of wondering whether the manager at the top of the performance charts made a series of lucky draws from the probability distribution of security returns, the world presumes that good results stem from skill. Conversely poor results follow from a lack of ability”. David Swensen, Pioneering Portfolio Management
“Findings from a multitude of research literature converges on a single point. People are credulous creatures who find it very easy to believe and very difficult to doubt”. Daniel Gilbert, How Mental Systems Believe
“Our bets are only as good as our beliefs” Annie Duke, Thinking in Bets.
Success is seductive. In all walks of life, the winners are celebrated and the losers are quickly forgotten.
The investment industry is no different. Everything revolves around performance. Who, or what, is delivering returns today, this month, this quarter, this year. Yet in contrast to many games where the results are largely determined by skill, the outcomes in trading and investing, particularly over shorter periods of time, are influenced by chance and random influences.
Markets experience booms, bubbles and busts. In a bull market everybody can look like an investing genius. An investor with a naïve strategy can outperform a savvy, sophisticated investor simply due to good luck. As Buffett says, it’s only when the tide goes out do you discover who has been swimming naked.
Randomness and narratives
The challenging influence of randomness and luck when making investment decisions is arguably most apparent in the area of manager selection. Yet the mindset of the allocator often doesn’t reflect the random influence on outcomes.
Selecting investment managers is a bet on their ability to deliver future performance. As former poker player and decision-making expert Annie Duke says, “our bets are only as good as our beliefs”. It is in the process of formulating beliefs about investment managers that we are susceptible to bias.
Instead of approaching the selection decision with a statistical mindset, asking is there is sufficient evidence to reject the null hypothesis that a manager has zero skill, we can easily revert to thinking with a narrative mindset.
It is well documented that stories rather than statistics tend to resonate more strongly with humans. Narratives help us process and makes sense of the world. We observe an occurrence and search for an explanation. The narrative provides the color.
Convincing narratives have compelling characters fitting stereotypes we identify with and cause and effect relationships that fit with our understanding of the world. But by placing everything into a neat narrative we may be oversimplifying reality. That is the essence of the narrative fallacy.
In the world of investing, when we observe that portfolio manager A has generated strong returns, while manager B has struggled, the temptation is to conclude that manager A is a more skillful investor or has more of an edge rather than examining alternative explanations. The role of chance or luck is rarely entertained.
When a portfolio manager then shows improved performance there is also a narrative – there was an improvement in the investment process, a new system was developed or a new portfolio manager hired. The improvement is rarely chalked down to reversion to the mean.
These narratives could be particularly relevant in more complex, less transparent strategies such as hedge fund investing, where the manager will promote his preferred narrative and it may be difficult for an investor to disprove the story.
Stereotypes and overweighting the recent past
When investors assess investment managers, the narrative fallacy tends to intertwine with two other biases: representativeness and recency bias.
The tendency to select based on stereotype, known as the representativeness bias, makes us more vulnerable to compelling narratives. It occurs when people use a judgmental short cut when making a difficult decision amid uncertainty.
The classic example from Daniel Kahneman (1) is:
“Bob is an opera fan who enjoys touring art museums when on holiday. Growing up, he enjoyed playing chess with family members and friends." Which is more likely?
A. Bob plays trumpet for a major symphony orchestra
B. Bob is a farmer”
In this scenario, a large proportion of people will select A as this fits with the stereotype, even though there are statistically more male farmers than trumpet players in major orchestras.
How is this relevant to investment manager selection? Selecting investment managers is difficult. There is a huge amount of potentially relevant information available: performance data, background of manager, investment strategy, research process, number of PhDs in the organization etc.
When faced with the difficult question of whether there is evidence that the portfolio manager has an edge versus others, and the right characteristics to manage risk through different market environments, an allocator could take a mental short cut and ask does this manager fit the stereotype of a successful portfolio manager?
The key ingredient that often prompts us to act on these biases is performance, particularly recent performance. Media stories on investment performance typically focus on who is doing well this month, this quarter or at most this year. It is human to weigh recent events more heavily – they tend to be fresher in our minds.
However, the random element in performance means that recent performance is an imperfect guide to long-term returns. Managers with an edge can underperform for a period of time just by chance. Too often we extrapolate recent performance without examining the full data set.
But the recency bias is more than that. Even if we like a manager’s story or we can see a potential edge we like to see it “proven” in the form of positive recent performance before we are comfortable making the allocation.
Data on mutual fund flows show that investors tend to move in and out of asset classes following periods of positive and negative performance, as they seemingly extrapolate recent positive performance. The net effect is that the returns they extract from the market are less than the long term returns of the asset class (2).
Similarly, chasing performance when selecting managers can be detrimental to the returns achieved allocating to a given strategy.
Motivated reasoning and overconfidence
Even after the decision to allocate is made the behavioral challenges remain. Narratives and recency bias will always influence the decision as to whether to retain or remove a manager.
But other biases complicate this decision. Due to motivated reasoning or confirmation bias we may be more inclined to seek out, or weigh more heavily, data that supports our initial decision.
We can also be prone to overconfidence in our abilities. For example, when surveyed, more than 50% of people see themselves as above average drivers (4). More information tends to increase people’s confidence in forecasts without necessarily improving forecast accuracy (5).
This is a sobering thought when it comes to manager evaluation given the large volume of data now available to allocators (performance data, due diligence questions, on site meetings etc) and the highly sophisticated performance evaluation tools available.
The challenge of overconfidence is particularly relevant when investing in hedge funds due to the existence of performance fees. As performance fees are typically charged on performance above a high-water mark, managers in drawdown do not earn these fees until they exit their drawdown.
Therefore, when removing a manager in drawdown, an investor has to be confident of finding a manager who will not only deliver better performance than the existing manager, but performance must be sufficiently greater to outweigh the higher performance fees paid.
Five Strategies to Overcome Behavioral Biases
How can investors overcome these biases if they are hard-wired into our brains?
The reality is that even leading behavioral scientists acknowledge that overcoming biases is very challenging. It is one thing accepting at an intellectual level that we are all vulnerable to biases, it is a much bigger challenge being cognizant of those biases in the moment of decision making.
It is not just a matter of finding the smartest people to make the decisions. Michael Maboussin (3) argues that smarter people can in fact be more prone to biases as they can be more creative in generating narratives to support their predetermined belief.
The most effective tool is to try and embed behaviors, into the manager selection process, that actively seek to overcome our natural biases.
First, although seemingly obvious, it makes sense to look at as much relevant performance data as possible rather than extrapolating from recent performance. The longer the track record, all else being equal, the larger the sample size and the greater the confidence the allocator can have that that the track record is less influenced by luck or a particular environment.
Two to three years is the typical time period for performance evaluation, but that period may have been characterized by a particular market environment that is unusually favorable or unfavorable for the strategy.
This will be obvious in some cases (e.g. a tail risk strategy) but may be at play in other scenarios. For example, a trend follower who allocates a large percentage of risk to commodity markets should be expected to outperform other trend followers in a period when commodity markets experience strong trends.
Of course, in some instances the fund manager, or some characteristic of the investment program, may have changed over time and this has to be factored in. But the general point is shorter track records are more heavily influenced by chance.
Second, develop a skepticism towards narratives. If one starts with the idea of trying to reject the null hypothesis that the manager has zero skill, it is easier to see that observed outperformance over a period of time may be just due to chance.
In contrast if one starts from the perspective that observed outperformance reflects skill, one naturally searches for a narrative to explain the skill.
As much as possible, it is important to analyze the performance data to derive an independent assessment of the drivers of a manager’s performance. What has driven the gains? Was performance skewed to one period or say from one particularly large position?
Third, objectively evaluate the process as much as the outcome. In her book, Thinking in Bets, former poker player Annie Duke emphasize the importance of avoiding “Resulting” in decision making. In essence, we can’t assess the quality of a decision by just looking at the result.
Try and look beyond the result to assess the quality of the edge and of the investment process. According to David Swenson “active managers worth hiring possess an edge that creates reasonable expectations of superior performance. This edge supplies an advantage stemming from a manager’s personal attributes and organizational characteristics.”
Are there are ex ante reasons to believe that there is an edge and the investment process should deliver returns? Then, did the manager make losses and gains for reasons that are consistent with their approach? If so, each of the portfolio decisions may have been sound and a manager’s relative underperformance may just reflect normal statistical fluctuation rather than a degradation in ability.
Fourth, encourage probabilistic thinking rather than deterministic thinking, by mapping out the probability of different return scenarios at the time of allocation. This should not be just a point estimate of the expected Sharpe ratio but expectations for likely drawdown and an expectation of what type of environment the manager is likely to deliver positive and negative performance.
Just as a pre-mortem can be helpful in project management to anticipate what might go wrong, it can be beneficial for an allocator to consider scenarios a manager may not deliver performance and what would be a reasonable level of drawdown. The very act of mapping of the probabilities of different outcomes can reinforce the idea that that there is the possibility of a manger underperforming just by chance.
Having detailed appraisal reports for the rationale and risks in selecting a manager may reduce the risk of a knee-jerk reaction when a manager suffers the inevitable drawdown.
Fifth, incorporate the independent views of a number of people who are skilled and experienced in manager assessment. For example, getting members of the investment team to conduct separate onsite visits may help the development of independent views.
The reason for doing this is to try and average out the impact of biases and individual blind spots. Not every allocator will be subject to representative bias and selecting on the basis of stereotype in the same way. Equally, different allocators may have had different experiences with a manager in the past and it can be beneficial to capture these insights independently without bias.
Just as it is good practice when brainstorming to get people to put their ideas down on paper independently before having a conversation (to avoid the views of the first or most vocal person dominating), manager selectors should complete and report their evaluation independently to avoid being influenced by each other’s opinions and biases.
Portfolio manager selection is difficult. Large amounts of information and more sophisticated tools may give a false sense of comfort to allocators. The big challenge is not one of insufficient information but instead it is the challenge of overcoming our natural biases in decision making.
Acknowledging the potential for bias upfront, and actively working to embed processes in the decision-making process that can mitigate the possibility of bias, are the best strategies investment teams have in making less biased decisions in manger selection.
1. See Daniel Kahneman (2011) Thinking, Fast and Slow
2. See for example the Morningstar Mind the Gap study.
3. Michael Maboussin (2012) The Success Equation.
4. See McCormick, Walkey and Green (1986) Comparative perceptions of driver ability - a confirmation and expansion.
5. See Philip Tetlock (2015) Superforecasting.
About the Author:
Alan Dunne has worked in the financial markets and investment management industry for over 25 years at hedge funds and large investment banks as a CIO, hedge fund allocator, macro strategist, and technical analyst. He is the Founder and CEO of Archive Capital a boutique multi-asset and investment research firm which focuses on liquid alternative investments. Prior to founding Archive Capital, he was Managing Director and a member of the investment committee at Abbey Capital.
Alan started his career as a foreign exchange analyst and trader, working for Bank of America in London, Hong Kong, and Singapore and for BNP Paribas in emerging markets before returning to Dublin to join Allied Irish Capital Management, a global macro commodity trading advisor. He was subsequently Investment Director of Royal Bank of Scotland's wealth management business in Ireland where he headed the investment team and was responsible for asset allocation.
Alan is a CFA Charter holder and holds an MSc Investment Management from Hong Kong University of Science and Technology, an MBA from Smurfit Business School, and a BA (Mod) in Economics from Trinity College Dublin.
For more on Archive Capital visit: Archive Capital | Dublin | Asset Allocation