Publication: Pricing the Storm: Hurricane Landfall Effects on Industry-level Returns and Volatility in the U.S. Equity Market
Files
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Access Restrictions
Abstract
This thesis applies an ARMAX–EGARCH(1,1) framework to 17 Fama-French industry portfolios from 1980 to 2024, using 59 continental U.S. hurricane events to estimate short-window abnormal returns and conditional volatility responses around landfall, and to test whether these effects change after 2008. The results show a clear asymmetry in the pricing of hurricane risk. Across industries, abnormal returns are generally small and rarely statistically significant, suggesting that average price effects are limited and difficult to detect in nationally aggregated portfolios. In contrast, volatility responses are large and persistent, and most industries exhibit significant increases in conditional variance after landfall. This divergence indicates that hurricanes are not primarily reflected in equity markets through directional price movements, but through increased uncertainty and risk. Sectors tied to reconstruction and capital replacement, such as construction and machinery, tend to experience positive responses, while consumption-oriented industries, including food and consumer non-durables, show negative effects consistent with demand disruption and short-run economic strain. These patterns strengthen with storm severity and differ by geographic exposure. Oil and Utilities produce large sign reversals between Gulf Coast and non-Gulf landfalls, suggesting that pooling across geographic zones produces misleading aggregates for both sectors. After 2008, long-window mean return effects that were significant across six industries in the earlier period mainly disappear, while volatility responses remain strong and persistent. The overall pattern suggests that hurricane effects on equity markets are conditional on landfall geography, storm magnitude, and the prevailing market regime, and are more reliably detected through changes in conditional variance than through changes in mean returns.