neuralgap.io
Can we solve Cognitive Biases in Investment Research with AI? (Part I)
Cognitive biases are mental shortcuts that can lead us astray in various aspects of life, including the high-stakes world of finance and equity analysis. As investors and analysts navigate complex markets and make crucial decisions, they often fall prey to these hidden biases. In this article series, we will attempt to locate these biases, the environments they appear in, and how artificial intelligence (AI) can help us recognize and overcome them, ultimately leading to writing more objective investment research reports, and hence allowing clients to make more objective investment decisions.
In the first part of the series, we’ll delve deeper into specific biases identified in recent studies, how and why they emerge and their consequences.
The Primary Cognitive Biases
Cognitive biases are mental shortcuts that have been engineered into us as survival mechanisms. These biases evolved to help our ancestors make quick decisions in dangerous or complex environments, conserve mental energy, and process vast amounts of information efficiently. However, in our modern world, particularly in complex domains like financial decision-making, these biases can lead to systematic errors in judgment.
Let’s examine a few that will directly affect the scope of this article:
- Confirmation Bias: The tendency to seek out information that supports pre-existing beliefs while ignoring contradictory evidence. This bias evolved as a mental shortcut to quickly process information and make decisions in a complex world.
- Overconfidence Bias: Overestimating one’s own abilities or the accuracy of one’s predictions a trait that emerged to promote action and risk-taking in uncertain environments.
- Anchoring Bias: Relying too heavily on the first piece of information encountered when making decisions, developed as a way to simplify decision-making processes
- Availability Bias: Overestimating the probability of events based on how easily they come to mind a shortcut evolved to help humans quickly assess risks and opportunities in their immediate environment.
- Loss Aversion: The tendency to prefer avoiding losses over acquiring equivalent gains which has roots in our evolutionary history, where losses had more severe consequences for survival than missed gains.
While potentially beneficial in certain historical contexts, these biases can lead to suboptimal decision-making in modern financial markets. They stem from our brain’s attempt to process complex information efficiently, often at the expense of accuracy. Understanding these biases is crucial for investors and analysts seeking to make more objective, data-driven decisions in today’s fast-paced financial landscape.
The Consequences of Biases
Understanding cognitive biases is crucial in making objective observations and subsequent decisions in low-validity (i.e. events tend to be more unique and have less intrinsic patterns), high-noise environments such as capital markets. Biases lead to systematic errors in judgment, decision-making processes, and development of investment strategies.
Bias in Portfolio Composition and Trading
Let’s start with how cognitive biases affect our ability to compose and trade a portfolio. We refer to a seminal study by Barber and Odean (2001), published in the Quarterly Journal of Economics, which analyzed the trading patterns of 35,000 households from 1991 to 1997. The study reveals several key finding including;
- Overconfidence bias is evident primarily in males who traded 45% more frequently than females, leading to reduced returns due to transaction costs. Moreover, despite suggestions that optimal portfolios require over 100 stocks, many investors over-confidently held only 3-4 stocks
- Loss aversion shows that investors experience 2-2.5x more pain from losses than pleasure from equivalent gains. This coupled with a disposition effect meant investors tend to sell winners prematurely and hold losers too long.
It’s clear that hidden biases affect our trading and composition performance. However, these biases are more easily recognisable (and thus correctable) because the feedback loop (i.e. gain/ loss) tends to be more short term. When it comes to observational analysis however, the feedback loops are longer and even more riddles with noise.
Bias in Research Reports
In order to understand how biases affect investment analysts ability to write objective reports, let’s refer to a comprehensive study conducted by Vesa Pursiainen, published in January 2020 that analyzed 1.2 million analyst-firm-month observations from 15 European countries over the period 1996-2018. The study observed a couple of biases including.
Cultural Trust Bias:
- Analysts were more likely to issue positive stock recommendations for companies based in countries toward which they have a more positive cultural trust bias. From 1996 to 2018, a one standard deviation increase in trust bias was associated with a 2.25% to 3.5% increase in the recommendation score (on a 1-5 scale).
- This bias is even stronger for “eponymous” firms whose names mention their home country. For these firms, the effect is about 7.25% to 7.75% larger than for non-eponymous firms.
- During the European debt crisis (Q4 2011 – Q1 2013), Northern European analysts were 10-23% less likely to assign buy recommendations to Southern European firms
Overconfidence and Experience:
- Analysts with more years of experience show a stronger cultural bias effect. The study found that both overall analyst experience and time covering a specific firm were associated with increased bias, suggesting that overconfidence may increase with experience.
- This phenomenon contradicts the expectation that biases would diminish with greater expertise, highlighting the persistent nature of cognitive biases.
Market Reactions to Biased Recommendations:
- Buy recommendations from analysts with higher trust bias are associated with 1.8-2.0% lower announcement returns, indicating that the market may discount recommendations perceived as overly optimistic. This effect is more pronounced for upgrades to buy recommendations, suggesting that the market is particularly sensitive to changes in analyst sentiment.
These are just a couple of high-level examples to demonstrate the pervasive impact of psychological bias, and by no means are exhaustive. Moreover, On top of these cognitive biases, we also have incentive-based biases that affect analysts’ ability to write objective reports.
Conflicts of Interest
Unlike cognitive biases, which are inherent mental shortcuts, incentive-based biases arise from external factors such as career concerns or conflicts of interest. While both types of biases can impact financial analysis, they require different mitigation strategies (since we are focusing primarily on hidden cognitive biases, we won’t delve too deeply into incentive-based biases here).
In a comprehensive study published in the Journal of Business Finance & Accounting analyzing ~14,000 stock recommendations from top US brokerage firms over 1997-2003, researchers found evidence of cognitive biases affecting analyst judgments. Only 31% of new buy recommendations outperformed their benchmark by 10% or more after 12 months, indicating frequent overoptimism. In contrast, 59% of new sell recommendations correctly predicted underperformance, suggesting analysts were more accurate with negative predictions.
The study revealed several biases in buy recommendations: analysts showed greater linguistic optimism, favored stocks with recent strong performance (momentum bias), preferred growth stocks over value stocks (representativeness bias), and were more positive about companies with investment banking relationships (potential conflict of interest). These findings highlight how cognitive biases can lead analysts to make overly optimistic predictions, especially for buy recommendations.
In conclusion, cognitive biases significantly impact equity analysis and investment decision-making, leading to suboptimal outcomes. The persistence of these biases, even among experienced professionals, underscores the need for innovative solutions. In our next part, we will explore how artificial intelligence can help identify and counter these biases, enabling more rational, data-driven decision-making in finance. Stay tuned!
Looking to integrate AI into your Research?
Our flagship product Sentinel is designed to be and end-to-end platform that helps Investment Research companies 2x their productivity.
©2023. Neuralgap.io