By applying artificial intelligence and Chat GPT to statements made by active fund managers, researchers have found that their underperformance can be partly explained by overconfidence that led to, among other things, excessive risk taking.
Behavioral finance is the study of how psychological factors affect financial decision-making, examining how emotions, biases, and cognitive limitations can lead to making irrational decisions in markets. Behavioral finance research has shown that investors are often susceptible to a variety of biases, such as:
- Loss aversion: tending to experience more pain from losses than pleasure from equivalent gains;
- Herding: tending to follow the crowd, even when doing so is irrational;
- Overconfidence: often overestimating their own knowledge and abilities; and
- Recency bias: giving too much weight to recent information and events.
Another well-documented bias is self-attribution: a cognitive bias where individuals tend to attribute their successes to internal, personal factors (skill) and their failures to external, situational factors (bad luck). Self-attribution bias is viewed as a learning bias that hinders investors from objectively updating beliefs about their own ability based on their past investment performance. It results in overweighting their successes and underweighting their failures – leading to overconfidence, excessive risk taking, excessive trading, and underperformance.
Meng Wang contributes to the behavioral finance literature with his September 2023 study, “Heads I Win, Tails It’s Chance: Mutual Fund Performance Self-Attribution,” in which he investigated the presence of self-attribution bias among mutual fund managers and evaluated its impact on performance. He sought to determine if sophisticated institutional investors were able to avoid biased attribution and evaluate their performance objectively or perceived themselves as knowledgeable (making them perhaps overconfident of their skills) and exhibit stronger attribution bias.
To find the answer, Wang examined the narrative attribution of performance by mutual fund managers in their N-CSR filings. Under Rule 30e-1 of the Investment Company Act of 1940, mutual funds that are registered with the SEC are required to disclose performance information, including the management’s narrative discussion of fund performance in shareholder reports at least semiannually. In that disclosure, managers typically highlight what significantly contributed to, and detracted from, fund performance as well as their views on the attributing factors behind those contributions and/or detractions. Those attributions might include internal factors, such as stock selection, sector weighting, and deviation from the benchmark as well as external factors like the economic environment, conditions in specific sectors, and common exposure with the benchmark. For instance, Wang explained, “the statement ‘The fund experienced a positive contribution from its overweight exposure in industrials, which we attribute to the effects of individual stock selection’ implies an internal factor, suggesting the fund’s stock selection was a key contributor to its performance.” On the other hand, “the statement ‘During the last six months, this was an impediment to the performance of the funds, as value stock returns have continued to outpace growth returns’ suggests an external detractor.”
To accurately extract attribution information from the textual content of such disclosure, Wang developed a two-layer GPT-based natural language processing (NLP) model that could read a sentence and 1) identify performance-attribution information (i.e., perception of causality), and 2) classify that information as a performance contributor versus a performance detractor and an internal factor versus an external factor. His model was able to achieve an overall out-of-sample accuracy of 90%.
To understand the dynamics of self-attribution bias among mutual fund managers, Wang constructed a filing-level self-attribution bias measure, self-attribution score (SAS), computed by taking the difference between the percentages of internal and external contributors in performance attribution-related sentences (IC - EC) and subtracting the difference between the percentages of internal and external detractors (ID - ED). The SAS ranged from -1 to 1 and was meant to capture the discrepancy in a manager's perception of causality between what contributes to versus what detracts from performance over the reporting period.
Wang’s data set included mutual fund shareholder reports from 2006 to 2018 from the SEC’s EDGAR database and included 15,434 shareholder reports associated with 1,969 unique funds. Here is a summary of his key findings:
- On average, 41% of the factors attributed to performance contributors were external, while 59% were internal. Conversely, 83% of the factors attributed to performance detractors were external, with 17% being internal. This result supports the hypothesis that mutual fund managers tend to internalize successes (i.e., performance contributors) – attributing them to skill – and externalize failures (i.e., performance detractors) – attributing them to bad luck.
- 74% of observations had an SAS value greater than 0, with a mean SAS of 0.23 and a standard deviation of 0.37.
- On average, mutual fund managers exhibited a significant self-attribution bias, as they were 40.6% more likely to attribute performance contributors versus performance detractors to internal factors.
- Funds displaying stronger self-attribution bias engaged in increased risk-taking and tended to engage in excessive trading in the subsequent reporting period, negatively affecting their performance and increasing the fund’s volatility – a fund’s performance in the subsequent reporting period decreased in the level of self-attribution bias.
- Using the Carhart four-factor (beta, size, value, and momentum) model as a benchmark, a one standard-deviation increase (0.37) in SAS resulted in a 0.8% decrease in cumulative alphas over the subsequent reporting period. The results held across various factor models.
- Funds exhibited a higher self-attribution bias following higher performance even though biased attribution only influenced fund flows when funds performed poorly.
Wang noted that the last finding was interesting because, while psychology literature suggests that the perception of self-attribution bias in others can indeed elicit negative reactions such as feelings of frustration and dissatisfaction, investors only reacted to high SAS scores after periods of poor performance. Perhaps unsurprisingly, investors ignored the self-attribution bias when the fund performed well – exhibiting self-attribution bias themselves in their choice of funds.
Investor takeaways
Behavioral finance insights can be used to improve financial decision-making by helping investors become more aware of their own biases and develop strategies to mitigate them. For example, investors can use behavioral finance concepts to:
- Avoid making investment decisions based on emotions;
- Be more skeptical of consensus opinion;
- Be more mindful of their own knowledge and limitations; and
- Give more weight to long-term trends and less weight to recent events.
While behavioral finance is a relatively new field, it has made a significant impact on the way financial markets are understood and analyzed. And most importantly, knowledge of their own biases can help investors make better financial decisions.
Another takeaway is that sophisticated institutional investors are subject to some of the same biases as individual investors, providing yet another explanation for the lack of persistence in performance of actively managed funds and for their overall underperformance – providing further support for the use of index funds and systematic quant-based funds whose managers are not subject to behavioral biases.
Larry Swedroe is head of financial and economic research for Buckingham Wealth Partners, collectively Buckingham Strategic Wealth, LLC and Buckingham Strategic Partners, LLC.
For informational and educational purposes only and should not be construed as specific investment, accounting, legal, or tax advice. Neither the Securities and Exchange Commission (SEC) nor any other federal or state agency have approved, determined the accuracy, or confirmed adequacy of this article. Certain information may be based on third party data and may become outdated or otherwise superseded without notice. Third party information is deemed to be reliable, but its accuracy and completeness cannot be guaranteed. By clicking on any of the links above, you acknowledge that they are solely for your convenience, and do not necessarily imply any affiliations, sponsorships, endorsements or representations whatsoever by us regarding third-party websites. Buckingham is not responsible for the content, availability or privacy policies of these sites, and shall not be responsible or liable for any information, opinions, advice, products or services available on or through them. The opinions expressed here are their own and may not accurately reflect those of Buckingham Wealth Partners. LSR-23-557
A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out our most recent white papers.
Read more articles by Larry Swedroe
By applying artificial intelligence and Chat GPT to statements made by active fund managers, researchers have found that their underperformance can be partly explained by overconfidence that led to, among other things, excessive risk taking.
