Shows how ordinary profitability mandates can lead LLMs to discount ambiguous safety signals even when risk remains visible in their reasoning.
Identifies context anxiety as a source of reasoning failure when capable models misjudge the effort required to complete long-horizon tasks.
Examines how people trade off expertise against fear of judgment when deciding whether to disclose information to human or AI advisors.
Tests when purposefully designed LLM conversations can durably correct financial misconceptions rather than reinforce them.
Projects how AI-driven changes in productivity, growth, and public spending could reshape the long-run U.S. fiscal outlook.
Finds that heavier financial-media coverage predicts deteriorating fundamentals and adverse events; a negative editorial tilt attracts attention but causes bad news to enter prices only gradually.
Finds that retail investors concentrate option purchases before high-volatility earnings announcements, overpay for volatility, trade at wide spreads, and close too late—losing 5–9% on average and 10–14% around the highest-volatility announcements.
Using Shenzhen disclosure data, finds that abnormally frequent corporate visits predict 70–100 basis points of monthly outperformance, especially for neglected firms, as visiting institutions add holdings and fundamentals improve.
Finds that when arbitrage capital is scarce, levered investors rely more on earnings as a lower-risk signal, strengthening announcement reactions and reducing post-earnings-announcement drift.
Shows that material non-core earnings are dispersed throughout 10-Ks; removing them improves forecasts of future performance, while slow investor response—especially to footnote items—supports an 8% annual abnormal-return strategy.
Uses quasi-exogenous calendar rotations to show that moving earnings announcements earlier increases media, analyst, and investor attention while reducing pre-announcement information leakage.
Shows that implied-cost-of-capital proxies perform best over time, characteristic-based proxies perform best across firms, and factor-based proxies perform poorly—choices that can change estimated treatment effects.
Finds that firms with stronger incentives to manage expectations earn lower pre-announcement and higher announcement returns across fiscal quarters, consistent with managers lowering expectations to manufacture positive surprises.
Introduces a simple measure based on abnormal option-versus-stock volume that tracks spreads, price impact, and order imbalance, predicts volatility, and detects shocks to information asymmetry.
Finds that later-than-expected earnings dates foreshadow worse news; option markets efficiently price the timing-related volatility, but equity prices fail to absorb the earnings signal until announcement.
Documents a sixfold increase in short-term return reversals around earnings announcements, consistent with liquidity providers demanding greater compensation for bearing inventory risk ahead of anticipated news.
Develops characteristic-based earnings forecasts that predict analyst errors, revisions, recommendations, and abnormal returns, showing that investors systematically overweight analyst forecasts.
Shows that value-versus-glamour returns and later expectation revisions concentrate where valuation-implied expectations conflict with firm fundamentals—and disappear where the two are aligned.
Shows that the unsigned option-to-stock volume ratio contains private information: low-ratio firms outperform high-ratio firms by 0.34% weekly, especially when short-sale costs are high or option leverage is low.
Using mobile-device GPS data, shows that earnings announcements increase foot traffic to firms' locations—especially when the news captures attention—and that this consumer response raises revenue and advertising effectiveness.
Finds that Egan-Jones ratings are more optimistic, slower to downgrade, and less accurate for bonds held by more subscribers, showing that subscriber-paid ratings replace issuer conflicts with client-revenue conflicts.
Introduces an option-based measure of how disclosures change investors' risk expectations; it predicts firms' future risk, costs of capital, and investment while exposing pitfalls in raw implied-volatility changes.
Finds that voluntary guidance and mandatory 8-K filings act as substitutes, with the negative relation more than doubling after the 2004 expansion made 8-Ks broader, timelier, and less concentrated in bad news.
Shows that trading on negative news becomes relatively costlier before earnings announcements, creating a predictable pre-announcement upward price bias that later reverses and can distort measured announcement returns.
Finds that investors pay through option prices to hedge earnings announcements expected to raise market-wide volatility; the premium concentrates among bellwether firms and predicts announcement-period straddle returns.
Finds that firms with more central boards earn 4.68% higher annual risk-adjusted returns and later improve profitability, especially when network information and resources are most valuable.
Shows that a firm's first analyst adds mainly industry and market information, increasing return synchronicity, whereas later analysts distinguish themselves with firm-specific information that reduces synchronicity.
Uses 15 years of student surveys across 51 treatment and 50 control departments to identify how the Mellon Graduate Education Initiative changed PhD programs and which program features affected attrition and graduation.
Offers practical guidance for incorporating AI into economics teaching while preserving active learning and judgment.
Reviews how incentives shape analyst forecasts and finds that forecasts contain cash-flow information but predictable biases that markets neither fully absorb nor filter out; it also assesses their use in estimating expected returns.
Builds a practical framework for market efficiency through noise trading, investor sentiment, fundamental analysis, arbitrage constraints, and methods for distinguishing mispricing from risk.