Welcome guest author Mark Thurston, founder and partner at consultancy firm Alpha Signal. Mark was previously head of equity manager research at Russell Investments, where he worked for 25 years.
Understanding the true nature of our biases is essential when it comes to making better decisions.
The idea that the future is unpredictable is undermined every day by the ease with which the past is explained. – Daniel Kahneman
Most manager research analysts understand that performance frequently reverts to the mean and therefore selecting managers based primarily on past performance is counterproductive. Even so, it remains a common practice.
But while a portfolio’s past performance may be no indication of future returns, a more nuanced study of a fund’s history can tell us a lot about what its future might look like.
Holdings-based analysis can improve forecasts by explaining how returns were generated, and behavioral analytics can further enhance allocators’ understanding of why portfolio managers made the decisions they did.
For the latter, there are quantitative tools that can identify signs of behavioral bias, further helping analysts to separate skill from luck and then to select managers who are likely to do better in the future.
The disposition effect
One example of behavioral analytics was carried out by Inalytics, a leading analytics provider. The firm studied managers’ sell discipline and found that potential winners were frequently sold too early. Meanwhile, poor performers were retained because they were showing a loss, particularly if they involve feelings of regret.
The disposition effect was first identified in 1985 and relates to the propensity of investors to lose more money when selling than they would be expected to lose by chance alone. The extensive Inalytics database of transaction-level data provides significant evidence of a lack of skill among professional money managers.
Behavioral finance literature cites ‘prospect theory’ and ‘mental accounting’ to describe the psychological and behavioral challenges that investors face. Prospect theory highlights that investors are risk-averse when looking at profits but tend to be risk takers when confronted with losses. Meanwhile, mental accounting points out that investors view each position within a portfolio as an entirely separate item and treat them in an inconsistent manner. In particular, investors tend to bucket ‘winners’ and ‘losers’ separately, and the chances of something being sold increases if a profit has been made on the investment.
"Inalytics found that a clear majority of the stocks that had been sold (57%) had outperformed over the previous 12 months. This result certainly supports the disposition effect – namely that fund managers tend to sell their winners."
Inalytics used 45,000 individual trades, representing a broad spread in terms of industries, regions and benchmarks. First, the firm investigated whether the majority of stocks being sold had in fact been winners or, contrary to the literature, whether investors were being disciplined in running the winners and systematically weeding out the losers. Inalytics found that a clear majority of the stocks that had been sold (57%) had outperformed over the previous 12 months. This result certainly supports the disposition effect – namely that fund managers tend to sell their winners.
Having found evidence supporting this thesis, Inalytics then examined what impact it has on performance. The firm calculated the impact of every purchase and sale on the performance of the portfolio. The stocks sold impacted negatively on performance by a highly significant 94 basis points per year.
Inalytics also investigated the shorter-term performance of the stocks to see whether there was an added dimension to this phenomenon. It found that performance typically turned negative in the month prior to the sale. This could suggest that short-term momentum is being used as a proxy for research, meaning that managers are oversimplifying the sell decision-making process.
The data on selling supports what academics have frequently noted: Far less time is spent on considering whether to sell a strategy than is spent on assessing securities to be purchased. While this detailed selling analysis provides significant evidence of shortcomings in many managers’ sell discipline, other behavioral analytics have shed light on some relative skill in research and stock selection.
The place of performance
A greater focus on a manager’s past behaviors – rather than on their fund’s performance – may be more telling about the future likelihood of returns. Of course, performance should not be dismissed from analysts’ work, but it is the peer-relative comparisons that are the key to identifying those managers who are the most skilled at stock selection, positioning and trading.
Effective portfolio and performance analysis should help to identify behavioral bias too, but they require sufficient evaluation horizons. Analysis should encompass a long time horizon for the given strategy. Finally, it is imperative to validate the softer, supporting elements, such as the lead portfolio manager’s motivation and the level of ongoing organizational support.
Why do researchers focus on performance?
Investors’ focus on past performance above other, possibly more helpful parts of a fund’s history is partly down to what data is easily available to analyze:
Understanding benchmark- and peer-relative performance is a necessary starting point in the evaluation process. For years, most quantitative manager evaluation was simply based on historical performance. Even today, performance remains the single most important factor in manager selection. Since returns are widely available, inexpensive to collect and easily comparable, it makes sense that analysts and allocators would rely heavily on them.
Historically, holdings data was expensive to collect and store. Now, with the success of Morningstar on the retail side and eVestment on the institutional side, reviewing holdings and holdings-based analytics is common and relatively inexpensive. Factor-based analyses, both fundamental and risk-based, are helpful in understanding what has driven historical performance relative to the benchmark.
Transaction-based analysis, which is comparatively expensive, remains less common. Collecting transactions with holdings is beneficial when trying to assess implementation skill and the impact of behavioral bias. It is now more possible than ever to understand the direct impact of major and common behavioral biases such as overconfidence and loss aversion.
As with traditional performance analysis, it is important to consciously promote objectivity. Manager research is best approached holistically, and misinterpreting individual elements can be as problematic as overly simplistic analysis.