Clustering Effect in Higher-Order Moments Across Various Timescales in the Cryptocurrency Market

This paper examines the risk cascade in high-order moments between investors and traders operating at different investment horizons in the cryptocurrency market. After constructing realized skewness and kurtosis on five major cryptocurrencies (BTC, ETH, XRP, LTC, and XLM) based on high-frequency (5-minute) price data, this paper estimates the transition probability of high or low skewness/kurtosis from long timescale to short timescale using the wavelet hidden Markov tree model. The results show that the negative skewness (low kurtosis) in the long timescale is likely to be followed by the negative skewness (low kurtosis) in the short timescale. However, high kurtosis in the long timescale does not necessarily lead to high kurtosis in the short timescale. These results are robust across various sub-periods and market statess. In addition, it is found that in a market with higher market efficiency, high-frequency traders can absorb risk faster and thus show a lower probability of maintaining a high-risk state when a low-frequency trader is exposed to high risk. At the beginning of the COVID-19, the extended analysis indicates, for the case of skewness, the average transition probability of low-to-low and high-to-high across different timescales has increased. These findings indicate the presence of a higher-order moments risk transmission from long-period (low-frequency) investors to short-period (high-frequency) investors, which have implications regarding market efficiency and predictability of higher-order moments across timescales in the debatable cryptocurrency market