Research

We derive generalized bounds on conditional expected excess returns that can be computed from option prices. The generalized lower bound (GLB) may serve as an expected excess return proxy for individual and basket-type assets, is conditionally tight, accounts for the entire risk-neutral distribution of returns, and outperforms existing variance-based models in out-of-sample predictions. Bounds calibrated to realized returns correspond to reasonable risk aversion and prudence. On average, expected stock returns given by the bounds decrease on even weeks of the FOMC cycle. Cross-sectional tests deliver a reasonable market risk premium. 

Download GLB Data 

In statistics, samples are drawn from a population in a data-generating process (DGP). Standard errors measure the uncertainty in estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence-generating process (EGP). We claim that EGP variation across researchers adds uncertainty: Non-standard errors (NSEs). We study NSEs by letting 164 teams test the same hypotheses on the same data. NSEs turn out to be sizable, but smaller for better reproducible or higher rated research. Adding peer-review stages reduces NSEs. We further find that this type of uncertainty is underestimated by participants. 

This paper documents that 56% of nonprofessional social media investment analysts (SMAs) are skilled and declare beliefs that generate positive abnormal returns, while 44% produce negative abnormal returns. 13% of all SMAs are high-skill type and produce a one-week three-factor alpha of 61 bps, while the remaining 87% generate only 6 bps. The distinctive features of high-skill SMAs are primarily firm and industry specializations. Although SMAs tend to extrapolate and herd, their expectations are not systematically wrong. For higher-skilled SMAs compared to the less-skilled ones, extrapolation fades more quickly, and herding is lower, consistent with theory.

Forecasters who are optimistic about an asset react to negative news by shifting their optimistic expectations to a longer forecast horizon. To document this novel pattern of optimism shifting in belief updating, we rely on CAPS, a social-finance platform offering the unique opportunity to observe individuals' beliefs about stocks alongside their chosen forecast horizon. Additional analysis indicates that optimism shifting leads to large underperformance, and it is consistent with forecasters’ motivation to retain optimistic beliefs about their skill (confidence channel), the value of their financial assets (financial-stakes channel), and the value of their accrued knowledge about an asset (intangible-stakes channel).

We study the recent explosion in trading of same-day expiry (0DTE) options on the S&P500 index. 0DTE positions can destabilize the underlying market when delta-hedging requires trading in the same direction as realized returns. We address this concern by investigating whether measures of trading activity propagate volatility. We find no evidence that aggregate open interest and trading volume increase volatility. On the contrary, market makers' inventory gamma is significantly and negatively associated with future intraday volatility. This evidence is consistent with delta-hedging by market makers because, in our sample, they hold a predominantly positive inventory in 0DTEs.

We propose a novel high-frequency measure of sovereign default risk that can be used when traditional metrics like CDS spreads are unavailable. The measure exploits the information in news text, can be computed in real-time for any country, and is highly informative about sovereign default risk. It predicts sovereign CDS spreads, rating downgrades, and realized defaults over long horizons. Consistent with theories on sovereign risk spillovers, an increase in the index is associated with higher firm default probability, default protection cost, and lower equity valuation. The measure is valuable for equity market-timing, and its informativeness is driven by macroeconomic concerns.

Using data from The Motley Fool's social prediction platform, CAPS, we document substantial differences in stock predictions across investment horizons. Short- and long-horizon investors respond differently to macroeconomic events and firm news announcements. At the onset of the Covid-19 pandemic, the sentiment of short-horizon predictions became sharply more negative while long-horizon predictions remained optimistic. Short-horizon investors also react more than twice as strongly as long-horizon investors to earnings surprises and technical view events. Around acquisition rumors, short- and long-horizon investors update in opposite directions about the target: short-term investors become more optimistic, while long-term investors become more pessimistic. Motivated by these findings, we develop a firm-day measure of horizon disagreement, spanning from 2006 to 2022, and find it relates significantly to abnormal trading. Additionally, the disagreement-trading relation strengthens on earnings announcement days, providing new evidence on the role of model disagreement.

We use empirical Bayes (EB) to mine data on 140,000 long-short strategies constructed from accounting ratios, past returns, and ticker symbols. This "high-throughput asset pricing" produces out-of-sample performance comparable to strategies in top finance journals. But unlike the published strategies, the data-mined strategies are free of look-ahead bias. EB predicts that high returns are concentrated in accounting strategies, small stocks, and pre-2004 samples, consistent with limited attention theories. The intuition is seen in the cross-sectional distribution of t-stats, which is far from the null for equal-weighted accounting strategies. High-throughput methods provide a rigorous, unbiased method for documenting asset pricing facts.

We investigate how investors react differently to the information provided by male and female non-professional analysts on investment social media and the financial market consequences. Although male and female contributors exhibit similar informativeness about future abnormal returns and publish content with comparable novelty, female-authored perspectives receive significantly lower engagement, lower trust, and higher disagreement from platform users. The less favorable engagement and lower attention result in lower retail trading following female-authored views than males’. Consequently, the incorporation of information into asset prices can be slower and information aggregation less efficient, given that female-authored views are informative and non-redundant.

We show that an increase in stock return exposure to media attention to narratives, measured with standard methods for extracting topic attention from news text, leads to a lower stock price informativeness about future fundamentals. Empirically, narrative exposure explains over 86% of idiosyncratic variance in the cross-section, and both narrative exposure and non-systematic information channels—idiosyncratic variance and variance related to public information—decrease stock price informativeness. Moreover, stocks with high narrative exposure demonstrate elevated trading volume. To rationalize these results, we develop a theoretical model based on investor disagreement stemming from differential interpretations of media narratives.

We study learning and uncertainty under the factor investing paradigm using an endogenous information model with correlated assets. As investors shift attention from firms towards systematic risk factors, stock prices become less informative, increasing systematic uncertainty and incentivizing learning about the systematic risk. This learning complementarity leads to multiple regimes in systematic uncertainty and attention allocation. We specify and estimate a model-based, forward-looking measure of attention to systematic versus firm-level information. Consistent with the model, the measure follows a regime-switching process. The high-level regime is linked to lower stock price sensitivity to firm-specific information and a higher systematic risk concentration.