As cryptocurrency becomes more mainstream, understanding its behavior is a new take on weighing up risk. KAUST’s analysis of price movements for the top five cryptocurrencies has used statistics designed to account for rare events to provide new insights into their market behavior. The research will help investors assess levels of diversification, and the risk of simultaneous price falls and gains.
Understanding risk is critical to investing. One of the ways investors manage their risk exposure and mitigate against potential losses is to diversify their holdings across assets that are not correlated. Ideally, this means that the market prices of diversified assets are completely independent. Across disparate asset classes like healthcare versus gold, for example, this independence is quite apparent. Among different assets in the same class, however, independence can be more difficult to judge.
Ph.D. student Yan Gong and Professor Raphael Huser from KAUST’s Extreme Statistics Research Group have now applied their expertise in flexible statistical tools used to characterize rare events in large data sets to tease out the potential relationships in price movements among the top five cryptocurrencies.
“For investors, it is crucial to uncover the dependence relationships between cryptocurrencies for more resilient portfolio diversification,” says Gong. “In order to assess risk, we developed a flexible copula model that is able to distinctively capture dependence or independence in the lower and upper tails simultaneously.”
These “tails” are the rapid price rises (upper tail) and falls (lower tail) that occur occasionally as part of market trading. While such wild price movements make headlines, they are exceedingly rare in the sequence of trading data that is dominated by more pedestrian incremental price movements. Gong and Huser’s model can capture these extreme movements separately for increases and decreases at the same time, while allowing the trends to be compared and correlated among multiple pairs of cryptocurrencies.
“Our analysis suggests that the upper tail dependence strength has remained relatively stable at a moderate level for most pairs of cryptocurrencies,” says Gong, “whereas the lower tail representing the big joint losses has become more and more dependent in recent years, transitioning from a weak independence regime to a strong dependence regime in some cases, such as between Bitcoin and Ethereum.”
Interestingly, the researchers found that this regime switch coincided with the 2017 boom followed by the 2018 cryptocurrency crash.
“The asymmetric tail dependence phenomenon is common in financial data modeling, and so our model has the potential to help with generic financial risk management and mitigation,” Gong says.
Gong, Y. & Huser, R. Asymmetric tail dependence modeling, with application to cryptocurrency market data. Annals of Applied Statistics (2022).| article