Welcome. I am a fifth-year PhD Candidate in Finance at Brandeis
University.
My research is in empirical asset pricing. I use large-scale factor replication, out-of-sample testing, and portfolio analysis to identify return predictors that are statistically reliable, economically meaningful, and implementable. My job-market paper develops a multidimensional approach to evaluating factor quality by combining t-statistics with return monotonicity.
Two Heads Are Better Than One: t-Statistics and Monotonicity in the Factor Zoo
Job Market Paper
Sole-authored
Abstract
This paper shows that two in-sample diagnostics used to judge factor quality the conventional t-statistic and monotonicity capture distinct information and enhance predictive power when combined. The two measures are nearly un-correlated in-sample (ρ = −0.08). Factors with the highest in-sample summed rank across both dimensions (≥70th percentile) deliver 33.4/34.2 bps/month in out-of-sample return/alpha, while factors with the lowest summed rank (≤30th percentile) deliver only 10.2/10.9 bps/month. High-summed-rank factors also perform better in real-time trading: a strategy built on the top 30% by summed rank dominates on a certainty-equivalent basis at every risk-aversion level considered and continues to hold up once trading costs and market impact are priced in. Finally, the combined strategy results in higher diversification in factor rotation.
The literature finds that U.S. equity returns are higher overnight, on Fridays, in January, around the turn-of-the-month, and during macro announcements (e.g., FOMC, GDP, NFP, and ISM releases). We replicate 153 long–short factors and find that their premia are lower in all of these periods. On average, a factor loses 2.01 bps overnight but earns 5.45 bps intraday (t = 8.54); earns 5.77 bps on Mondays, 3.44 bps on midweek days, and 1.24 bps on Fridays (midweek−Monday difference: t = −7.06; Friday−midweek difference: t = −5.34); earns 0.22 bps in January but 3.71 bps on non-January days (t = 3.66); earns 1.89 bps during turn of-the-month days versus 3.61 bps on other days (t = 6.01); earns 2.99 bps during three-day windows surrounding major macroeconomic announcements, compared with 3.71 bps on other days (t = 4.62). The results are robust in-sample and out-of-sample. Our results are consistent with sentiment-driven mispricing, as we find that the calendar effects of factors are stronger when news sentiment is high.
Presented at:
Chinese Economists Society (CES) China Annual Conference 2026
Stock return anomalies tend to concentrate (1) in the first month following information releases, (2) on earnings announcement days, and (3) on corporate news days. These patterns suggest that biased expectations contribute to asset mispricing, which is gradually corrected in the weeks after new information becomes public. In this paper, we replicate 94 stock return anomalies using intentionally lagged accounting data—delayed by 12 months—to eliminate short-term behavioral biases such as overreaction, underreaction, and limited attention. With stale data, information is presumed to be fully diffused and incorporated into prices. We find that the median monthly alpha of these anomalies remains significantly positive at 28 basis points. This alpha is robust out-of-sample and aligns with risk-based explanations. Moreover, we document that the alpha gap between stale and fresh data narrows over time, consistent with faster information diffusion, technological advancements, and behavioral interpretations of anomaly returns.
