Asymmetric Extremal Dependence Modeling, with Application to Cryptocurrency Market Data

• Activity Number: 512 - Predicting and Evaluating Risk Models Within Distributions and Across Time
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Risk Analysis
• Author(s): Yan Gong* and Raphaël Huser
• Companies: KAUST and King Abdullah University of Science and Technology
• Keywords: Asymmetric tail dependence; Asymptotic dependence and independence; Copula model; Censored likelihood inference; Extreme event; Lower and upper tails

Abstract

In order to jointly assess the asymmetric risks related to both low and high extremal events, we develop a flexible copula model that is able to distinctively capture asymptotic dependence or independence in its lower and upper tails. Our proposed model is parsimonious and smoothly bridges both extremal dependence classes in the interior of the parameter space. The inference can be performed using the full likelihood or various types of censored likelihood approaches. Specifically, we implement three different censoring schemes that are designed to provide a good fit in the lower and upper joint tail regions and we investigate their relative efficiency in an extensive simulation study. We illustrate our methodology by studying the dependence strength that governs extreme log-returns in cryptocurrency market data. Our analysis shows that our model provides a better fit than alternative copula models, and reveals that the joint lower tail of Bitcoin and Ethereum cryptocurrencies has become increasingly dependent in the past two years, hence suggesting that big simultaneous losses are now more frequent than before.