Research

Job Market Paper

Lendable Inventory Concentration, Recall Risk, and Equity Pricing Anomalies (sole-authored)

Presentations: AFA Annual Meeting, PhD Student Poster Session (Washington, D.C., January 2027, scheduled); EUROFIDAI–ESSEC Paris December Finance Meeting (December 2026, scheduled); EFMA Annual Meeting (Kristiansand, 2026); AWG Workshop (Graz, 2026); VGSF Conference (2025, 2026); VGSF PhD Research Seminar (2025).

Award: EFMA 2026 GARP Risk Management Award

I find that the returns of equity pricing anomalies remain large and significant net of fees in stocks whose lendable shares are concentrated among a few lenders. I argue that recall risk explains this pattern. Lenders can recall shares at any time, and when lendable inventory concentration (IC) is high, one withdrawal can recall more shares than borrowers can replace in time. Within firms, a one-standard-deviation higher IC predicts a 2.1 percentage point higher next-month recall event probability (mean 10.9%), controlling for the fee and lendable supply. This recall risk deters short selling even when unlent shares are available. At the highest IC, the same overpricing signal attracts about one quarter as much short interest as at the lowest IC. Less of the overpricing is therefore corrected. Average anomaly returns are 0.43% per month higher in the top IC quintile than in the bottom quintile, mainly through the short leg, and the gap is still 0.33% per month net of fees. Russell 3000 inclusion, which broadens the lender base, lowers recall risk and raises short interest most in overpriced stocks with high pre-inclusion IC.

Key result

Anomaly returns are concentrated in stocks whose lendable inventory is held by a few lenders (high IC). For each stock, NetLeg is the number of anomalies that place it in the long leg minus the number that place it in the short leg. Within each IC group, I sort stocks into NetLeg quintiles and go long Q5 (most underpriced) and short Q1 (most overpriced).

Cumulative log returns of the Q5 minus Q1 NetLeg portfolio in high- and low-IC stocks
Cumulative log returns of the Q5 − Q1 NetLeg portfolio in the top and bottom IC quintiles, before and net of fees. US stocks, 2006 to 2023.
Q5 − Q1 NetLeg return (bp a month, t-statistics in parentheses)
Bottom IC quintile Top IC quintile
Before fees 31 (0.8) 122 (2.7)
Net of fees 27 (0.7) 100 (2.2)
DGTW-adjusted, net of fees 28 (1.2) 102 (2.9)

Working Papers

Short-Sale Constraints and the Return Predictability of Idiosyncratic Variance: Evidence from Crypto Perpetuals (with Jia Zhai)

Presentations: WU Brown Bag Seminar (2026)

High idiosyncratic variance predicts low future returns in cryptocurrency markets, and the relation comes mainly from large, recent upside price shocks in coins that cannot be shorted. To test whether short-sale constraints drive this relation, we use the staggered expansion of access to short selling across coins and compare the same coin's returns after shocks before and after it can be shorted. After upside shocks, prices reverse over the following week, and access to short selling reduces this reversal by 2.55 percentage points, about three quarters of its size. After downside shocks, by contrast, access matters mainly when the decline follows an upside shock. Consistent with these results, access also weakens the return predictability of recent upside shocks, by an amount similar to the reduction in the reversal. Idiosyncratic variance thus predicts low returns in part because, when pessimistic investors cannot sell short, prices stay above the level implied by the average belief after a shock and fall back as disagreement fades.

Prediction Targets for Cross-Sectional Bond Selection (work in progress)