By Megan Czasonis, Huili Song, and David Turkington
We use a statistical method that combines a stock’s attributes in a nonlinear and conditional way in order to predict its relative vulnerability or resilience to market drawdowns.
By definition, a stock market crash corresponds to a severe, market-wide drawdown. However, below the surface, there is often considerable dispersion in the performance of individual stocks during these events. In a recent paper, we explore whether a stock’s unique attributes can help predict its relative performance in future market drawdowns. We use a rigorous, statistical method that compares a stock’s unique circumstances—reflected as popular factor attributes—to stocks that performed relatively poorly during past market drawdowns, and those that performed relatively well. The result is a precise score indicating the relative probability that a stock will be vulnerable to future drawdowns, and another indicating the probability it will be resilient. We find that these scores are powerful predictors of relative stock performance during market crashes, more so than any individual stock attribute or their linear combination. Moreover, the least vulnerable stocks also outperform the most vulnerable stocks during non-crash periods, suggesting that investors may not be compensated for bearing the risk of high vulnerability
