Browsing by Author "Hua, Xuanying"

When stocks crash together, diversification fails. The options market prices this risk through the Dispersion Spread: the gap between the implied volatilities of an index's individual constituents and the implied volatility of the index itself. A wide spread means the market expects stocks to move independently; a narrow spread means it expects them to move in lockstep.

This thesis asks whether sudden, coordinated jumps in stock prices, detected in real time from high-frequency equity price data, predict how the options market reprices correlation risk by the end of the same day. We apply the Caporin-Kolokolov-Renò (CKR) multi-asset jump detection framework to one-minute mid-prices for the 50 largest U.S. equities from July 2021 to August 2025, fitting Student-t copulas to 30-minute windows to extract co-jump severity and tail dependence. We then test whether these intraday signals, derived entirely from the equity market, predict same-day changes in the dispersion spread, constructed from the options market.

The Maximum Severity Ratio, which captures how far the most extreme intraday co-jump exceeded its locally calibrated threshold, is the only signal that predicts the end-of-day Dispersion Spread, compressing it on high-severity days (p = 0.022). On the next-day horizon, this severity signal predicts directional reversal: when the spread widens on a high-severity day, the probability of continued widening drops by 64% (p = 0.006). An expanding-window logistic model achieves a 63.4% out-of-sample hit rate, rising to 69.1% on high-severity days, with a frozen first-half model matching performance over two years of unseen data. Much of the strategy's profitability reflects mean reversion, but the CKR signal provides incremental value by identifying high-conviction reversal days.