Working Paper: CEPR ID: DP2762
Authors: Serhuang Poon; Michael Rockinger; Jonathan Tawn
Abstract: In the finance literature, cross-sectional dependence in extreme returns of risky assets is often modelled implicitly assuming an asymptotically dependent structure. If the true dependence structure is asymptotically independent then existing finance models will lead to over-estimation of the risk of simultaneous extreme events. We provide simple techniques for deciding between these dependence classes and for quantifying the degree of dependence in each class. Examples based on daily stock market returns show that there is strong evidence in favour of asymptotically independent models for dependence in extremal stock market returns, and that most of theextremal dependence is due to heteroskedasticity in stock returns processes.
Keywords: Asymptotic Independence; Extreme Value Theory; Hills Estimator; Tail Index
JEL Codes: C13; C22; G11; G15
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| true dependence structure among stock returns is asymptotically independent (C58) | overestimation of risk of simultaneous extreme events (C46) |
| heteroskedasticity in stock return processes (C58) | extremal dependence in stock returns (G17) |
| assumption of asymptotic dependence (C29) | bias in previous studies (C90) |
| left-tail dependence stronger than right-tail dependence (C46) | implications for risk assessment (D81) |