3  Sector Composition Inside the Lottery-Stock Population

Why the Aggregate Payoff-Asymmetry Result Does Not Survive Excluding Healthcare

Author

Jean-Marc Choufani

Published

August 9, 2026

SSRN: abstract_id=7024778 Status: Corrected historical analysis; prospective replication pending


WarningCorrection to the posted working paper

The earlier claim that payoff asymmetry independently identifies a profitable subset of lottery-like stocks is withdrawn. The aggregate result is dependent on Healthcare composition, and the terminal-return cap previously described as a stop loss is not an implementable trading test.

3.1 Abstract

In a survivorship-conditioned historical panel and one exact 60-session outcome horizon, the top-IVOL payoff-asymmetry spread is +6.08% per monthly formation period with Newey-West \(t=2.65\). After excluding Healthcare, it is -0.27% with \(t=-0.09\). Skewness, tail index, and MAX also fail the same exclusion. The naive mean long-short is unprofitable. These results establish sector-composition dependence in this sample, not an independent payoff-asymmetry premium or a tradable strategy.

3.2 1. What the historical panel shows

Signal All sectors spread / NW t Ex-Healthcare spread / NW t
Payoff asymmetry +6.08% / +2.65 -0.27% / -0.09
Skewness +5.94% / +4.36 +2.07% / +1.58
Tail index -2.96% / -1.79 -0.59% / -0.41
MAX +4.59% / +3.18 -0.19% / -0.11

In 200 seeded equal-size random ticker exclusions, the MAX spread has a median of +3.89% and 63% retain \(t>2\). Excluding Healthcare instead produces -0.19% with \(t=-0.11\). This makes sector composition a specific threat to validity, although it does not establish Healthcare as a causal mechanism.

3.3 2. What the panel cannot establish

The sample contains only names that survived into the present universe, so delistings and point-in-time membership are not represented correctly. The exact-60 horizon is useful for an auditable comparison but does not establish robustness across horizons. The aggregate MAX spread is statistically clear only in 2014–2018; the ex-Healthcare result is null in every examined subperiod. Multiple testing, dependence across company and formation dates, and time-varying sector membership remain unresolved.

3.4 3. Why terminal capping is not a stop loss

The naive equal-weight mean long-short returns -1.1% per period (\(t=-0.4\)). Capping each name’s terminal 60-session return changes an accounting payoff after the path has occurred. It does not model an executable stop, gap-through loss, borrow availability and recall, spreads, impact, locates, option prices, or portfolio rebalancing. No tradability or risk-managed implementation claim follows from that sensitivity calculation.

3.5 4. Prospective test

The hypothesis remains research-only. Within point-in-time high-volatility cohorts, TrailMap will compare payoff asymmetry within sector—especially within Healthcare subindustries—at 20, 60, and 120 sessions. It will count canonical episodes, preserve delisted and unresolved outcomes, cluster company and sector-week dependence, and publish null results. Historical results do not alter a score or approval weight.

3.6 Conclusion

The useful finding is narrower than the posted paper claimed: a striking aggregate relation can dissolve under one economically important sector exclusion. That is a warning about composition, not evidence of an independent premium.