The paper extends firm-specific evidence to the aggregate level by examining whether well-diversified portfolios respond to sentiment extracted from posts on the social media platform - X. Firm-specific X sentiment could be related to the covariance structure of returns, or it could be a priced characteristic. To distinguish between these two hypotheses, we construct two aggregate X-sentiment measures: an X-sentiment factor mimicking portfolio and an aggregate excess X-sentiment index of all stocks included in the S&P 500 index. Our evidence indicates that the factor mimicking portfolio is related to the covariance structure of returns through its ability to mimic a mispricing factor.
We study how sustainability preferences and information about climate impacts shape investment choices. In a sequential discrete-choice experiment, participants repeatedly chose between a sustainable and an unsustainable asset, with the sustainable option offering equal, lower, or higher returns. In a representative US sample (N=1,003), respondents were assigned to a control group or to four treatments providing climate-consequence information in unspecific text, specific text, graphics, or an external webpage. Text and webpage information lowered the probability of choosing the unsustainable asset by 0.109-0.431 percentage points in an unincentivized round. Effects attenuate when choices affect payoffs. We discuss implications for ESG disclosure.
We examine executive responsiveness during earnings calls and its impact on stock returns. Using large language model embeddings, we measure semantic similarity between analyst questions and executive responses, capturing direct answers versus deflection. Executives who provide semantically aligned responses generate 3.9% annual alpha (t = 3.41), robust to sentiment, firm characteristics, and market factors. Human validation on 1,642 Q&A pairs shows low-similarity responses are rated evasive 67% of the time versus 22% for high-similarity responses (r = 0.360, p < 0.001; Cohen's d = 1.01).
This study examines how the alignment between objective financial literacy and subjective financial knowledge shapes household investment behavior. Using Survey of Consumer Finances, we classify households as overcalibrated, undercalibrated, or well-calibrated. We analyze differences in stock market participation, risk tolerance, trading behavior, and portfolio diversification. Contrary to expectations, overcalibrated investors participate less frequently in equity markets and trade less than well-calibrated investors. Conditional on participation, however, they allocate a larger share of wealth to equities and hold more diversified portfolios. Undercalibrated investors exhibit lower participation rates and lower risk tolerance. These findings indicate that misalignment between perceived and actual financial knowledge produces investment patterns not explained by financial literacy or confidence alone. Incorporating calibration into behavioral finance provides a more complete understanding of household portfolio decisions.
This study draws on bandwagon perception theory and social presence theory to examine how fund managers' personal profile information (PPI)-specifically profile photos and online reputation-on third-party internet fund platforms influences investors' purchase intention via trust. An online experiment involving 404 participants reveals that online reputation has a significant impact on both affective and cognitive trust, whereas profile photos primarily affect affective trust. Trust, in turn, positively influences purchase intention. The findings underscore the importance of establishing investor confidence before any face-to-face contact and clarify the distinct roles of different PPI elements in subsequent investment decision-making, offering actionable insights for online fund management practice.
Corporate disclosures are not monolithic: managers communicate differently across earnings calls and regulatory filings, and markets fail to immediately reconcile these differences. We examine cross-channel divergence in managerial tone between earnings calls and Management's Discussion and Analysis (MD&A) sections of 10-K filings, and show that this divergence predicts post-disclosure stock returns. Using Loughran-McDonald (2011) dictionaries applied to 22,366 matched disclosure pairs from 1,762 firms over 2006-2025, we find that sentiment divergence generates temporary return predictability consistent with interpretive conflict, while complexity divergence, capturing readability gaps between conversational calls and dense regulatory filings, produces more persistent effects consistent with processing frictions. Both effects strengthen under firm fixed effects and vary systematically with analyst participation, supporting a limited attention interpretation. The findings indicate that informational inefficiencies arise not only from tone within individual disclosures, but from the difficulty of integrating signals across institutional communication formats.
We investigate the link between ticker memorability and the exchange-traded fund (ETF) investment from 1993 through 2023. Issuers often favor salient trading symbols for marketing reasons. Ticker symbol selection is nonrandom consistent with the strategic adoption of salient tickers to capture retail interest and justify higher fees. Using our familiarity proxies, we find ticker memorability is strongly associated with higher liquidity, idiosyncratic risk, expense ratio, and retail ownership. Upgrades from non-memorable to memorable constitute the most common symbol change in our sample, exhibiting the largest post-change increases in liquidity and idiosyncratic risk. Results are robust to logit, fixed effects, correlated random effects, and difference-in-difference estimations.
We introduce the Fear Generating Function (FGF), a behavioral model of volatility that captures asymmetric fear decay following market shocks. The model incorporates a momentum-sensitive persistence parameter, predicting longer fear duration under negative market momentum. Using VIX and S&P 500 data from 2015-2025, we identify 38 systematic volatility shock events and estimate decay durations under a rule-based framework. Regression analysis with standard errors provides statistically significant support for the model's central prediction that negative momentum regimes are associated with longer volatility persistence. Robustness checks across alternative momentum windows and shock thresholds confirm the stability of this relationship. A benchmark comparison with a purely exponential decay specification shows that incorporating behavioral persistence improves duration prediction. We also include a time-stamped illustration using a March 2025 volatility spike to demonstrate operational use. The FGF offers a structured behavioral framework for modeling the temporal dynamics of market fear.
Conventional rule-based advisory systems typically overlook investor psychology and real-time sentiment dynamics, limiting their effectiveness in mitigating emotionally driven trading behaviors. This paper presents a proof-of-concept emotion-aware decision support system that integrates market sentiment detection (FinBERT), behavioral classification (DistilBERT), and personalized risk advisory (DeepSeek R1) to reduce psychologically reactive trading. Results show significant risk-mitigation benefits in simulated trading environments across three distinct market regimes: high-volatility crises, speculative phases, and low-volatility conditions. Findings highlight the potential of explainable, real-time AI to improve investor discipline, with promising avenues for future research on user validation, adherence, and regulation. Simulation code available at: t.ly/SQl82.
This study provides evidence that gambling investors embed a lottery premium into stock valuations. I use the fraction of opposing votes on 431 U.S. reverse stock split proposals as a proxy for the proportion of gambling investors in each firm (the gambling investor concentration, or GIC). An event study shows that firms with above-median GIC experience cumulative abnormal returns approximately ten percentage points lower than those with below-median GIC, with stronger effects in contexts of elevated gambling sentiment. These findings help explain the valuation losses around reverse splits.
We explore how CEOs' pro-nature values-imprinted through early-life experiences-affect green technology innovation (GTI) in heavily polluting industries. These environmental preferences serve as behavioral priors influencing strategic choices, even in the face of economic trade-offs. Using Chinese firm-level data from 2013-2023, we find that CEOs with a strong nature orientation promote greater green technology innovation (GTI). This relationship is amplified by higher CEO educational level and stronger corporate governance but attenuated by firms' degree of internationalization and financial constraints. Our findings support the view that executive personality traits and environmental beliefs shape firm-level innovation through behavioral pathways.
This study examines the dynamics of US stock-specific news sentiment relative to aggregate market sentiment, leveraging a dataset of 2.7 million news articles. Focusing on negative tail events among S&P 100 constituents from 2003 to 2021, we identify a significant association between firm-specific sentiment changes and subsequent one-month stock performance. Specifically, a positive news sentiment score during adverse events suggests the absence of fundamental issues and correlates with superior post-event returns. These findings support the development of a profitable reversal trading strategy and indicate a potential regime shift in market behavior.
We investigate the decision-making behaviour of fund managers within the framework of Cumulative Prospect Theory (CPT). Employing monthly data from nearly 200,000 funds across all asset classes, investment styles and geographical regions between 1990 and 2022, we estimate the parameters of the CPT value and probability-weighting functions using hierarchical Bayesian estimation. Our findings reveal that fund managers consistently exhibit many of the behavioural traits that have been widely documented in the experimental psychology literature, although often with parameter values that are significantly different those that have been reported in these studies and which have been subsequently used in empirical research. We show that there are statistically significant differences in fund managers' behavioural characteristics between different asset classes, and between different fund categories within each asset class. We also show that fund size plays an important role in determining fund managers' behavioural traits, while manager tenure has only limited impact.
Does a higher level of ignorance lead to higher subjective confidence in making stock price predictions? 150 subjects make stock price forecasts for three listed companies. A Likert scale is then used to measure how confident the subjects are that their predictions will actually come true. Additionally, the level of knowledge and experience relevant to the stock market was measured by knowledge questions. The results show that individuals with limited specialist knowledge and experience are particularly confident in their forecasts and vice versa. This finding is evident for men and is statistically highly significant, but it is not for women.
Financial surveys often include planning horizon questions to understand people's financial decisions. Response options vary, seemingly assuming no effects on reported planning horizons. However, our U.S.-wide survey (N = 5,175) revealed shorter reported planning horizons when response options were short-range (from less than a week to longer than next year) rather than mid-range (from next month or less to longer than 10 years), or long-range (from next year or less to longer than 20 years). The mid-range condition elicited planning horizons that were most similar to an open-ended condition thought to capture natural thinking, had better predictive validity, and low respondent burden.
This paper explores how behavioral priors from entrepreneurial experience influence CEOs' attitudes toward corporate green innovation. Drawing on imprinting theory and upper echelons theory and using a sample of Chinese manufacturing firms, we find that CEOs with entrepreneurial experience tend to underinvest in green innovation, favoring familiar, non-environmental initiatives. Further analysis shows that external pressures, such as media tone and regulatory stringency, moderate this effect, highlighting the role of external framing in correcting managerial heuristics. Our findings contribute to understanding how experience-induced cognitive biases shape strategic innovation decisions.
This study investigates herding behavior among retail investors by analyzing sentiment in YouTube Shorts comments on investment-related content through web scrapping in Python. Using an ensemble sentiment analysis model Lexicon based Valence Aware Dictionary and Sentiment Reasoning (VADER) + BERT and unsupervised clustering (Hierarchical Density Based Spatial Clustering of Applications with Noise or HDBSCAN), the study reveals that 56% of comments exhibit latent convergence indicative of herding, far exceeding the 3% detected through rule-based sentiment classification alone. Grounded in Social Learning Theory, the research highlights how digital platforms facilitate observational learning and group mimicry. The findings offer methodological advancements for behavioral modeling and practical implications for fintech advisors, regulators, and platforms aiming to monitor sentiment-driven investor behavior in real time.
Using Morningstar's fund-level sustainability (ESG) rating data for a large sample of mutual funds domiciled in Australasia, we present strong evidence of herding in mutual funds with high sustainability ratings (high-ESG-rated) relative to unrated or low-ESG-rated funds. Herding in funds with high sustainability ratings is intentional, not driven by fundamental information, and independent of fund age, size, past performance, and investment focus. More importantly, we show that herding in funds with high ESG ratings dampens the correlation between the returns on these funds and their conventional counterparts, thus creating diversification potential via a sustainable investment approach. Finally, fund herding behavior is associated with lower subsequent fund performance, implying that herding toward sustainability does not necessarily reflect informed trading by fund managers while no effect is found on subsequent fund flows. Our findings provide novel insight into investor behavior in the growing market for sustainable investments.
This study examines how the sentiment embedded in AI risk disclosures influences stock market reactions to the DeepSeek_R1 event. Using FinBERT, a state-of-the-art language model tailored for financial texts, we construct a novel AI risk sentiment measure from firms' 10-K disclosures. Empirical results show that firms with more optimistic AI risk disclosures experience significantly more negative abnormal returns in response to the DeepSeek event. Our findings underscore the behavioral bias inherent in investor reactions to managerial communication and highlight critical challenges in transparent disclosure of emerging technological risks. The study offers practical insights for investors on how disclosure sentiment can shape market dynamics amid disruptive innovation.
The study investigates how the Korean stock market's underreaction to earnings announcements reflects differences in investor sophistication and attention, controlling for firm characteristics and estimating hidden private information. The empirical findings reveal that foreign investors are net sellers, whereas institutional non-members are net buyers in the short term. Conversely, institutional members are net sellers in the medium term, and individuals are net sellers in both. Foreign, institutional non-members, and individual investors exhibit distinct attitudes toward extreme earnings surprises. The delayed responses to Friday announcements are attributed to sample selection bias, underscoring the importance of information choices in trading quarterly earnings announcements.
