Quantile regression analysis is a statistical technique that examines the relationship between variables across different portions of the return distribution — not just the average relationship captured by ordinary least squares (OLS) regression. When applied to the asymmetric return-volatility relation, quantile regression reveals how the relationship between equity returns and volatility behaves differently in the tails of the distribution than in the middle — specifically, how the asymmetric effect (volatility rising more sharply with falling prices than with rising prices) varies across different quantiles of the return distribution.

Why Ordinary Regression Is Insufficient

The classic leverage effect — the empirical finding that equity volatility tends to rise more when prices fall than when they rise by an equivalent amount — is typically documented using average (conditional mean) relationships. But investors care most about behavior at the extremes: what happens to volatility during the worst 5% or 10% of return outcomes? Quantile regression addresses this by modeling the relationship at different quantiles — the 5th, 25th, 50th (median), 75th, and 95th percentiles — revealing how the asymmetric return-volatility relation varies across the full distribution of market conditions.

Key Findings

Research applying quantile regression to the asymmetric return-volatility relation consistently finds that the asymmetry is most pronounced in the lower quantiles of the return distribution: during the worst market periods, the inverse relationship between returns and volatility is strongest. When returns are in the bottom quantile (worst outcomes), volatility tends to be extremely elevated. In the upper quantiles (best market periods), the relationship is weaker and more symmetric. This finding reinforces the importance of asymmetric risk management: the markets where protection is most valuable are precisely those where the standard (average-based) estimates of risk are most severely underestimated.

---