Skewness Dispersion and Stock Market Returns

Battle Count: 224

Highly relevant for quantitative trading as it provides a novel, robust predictor of aggregate equity market returns with strong out-of-sample performance. The predictor is simple to construct from return data alone, computable at high frequencies, and delivers substantial economic gains (CER gains of 709-875 bps, Sharpe ratios of 0.82-0.91). The FOMC concentration finding is particularly actionable for event-driven strategies. However, it predicts aggregate market returns rather than individual stock returns, making it most suitable for market timing and tactical asset allocation rather than stock selection.

The core methodology is relatively straightforward: compute daily realized skewness from 5-minute intraday returns for each stock, then calculate the inter-percentile range (e.g., 75th minus 25th percentile) across stocks to get skewness dispersion. Monthly aggregation uses the last 5 trading days. Predictive regressions are standard OLS with robust inference. The main complexity lies in data acquisition (high-frequency prices for 6,770 stocks) and the comprehensive comparison with 50 alternative predictors. The portfolio allocation framework is standard mean-variance optimization.

The paper uses publicly available data (Kibot for high-frequency prices, CRSP for stock identification, and multiple publicly available predictor datasets from authors' websites). The methodology is clearly described with explicit formulas for realized skewness and skewness dispersion construction. All 50 alternative predictors are documented with sources. However, the specific Kibot data subscription and some predictor datasets may require access. The sample period (2000-2022) and construction methodology are fully specified.