A1298
Title: Duration analysis of Bitcoin trade with high-frequency transaction data
Authors: Makoto Nakakita - RIKEN (Japan) [presenting]
Teruo Nakatsuma - Keio University (Japan)
Abstract: The aim is to understand the time series structure of duration between consecutive transactions of Bitcoin and identify its similarities and differences with other conventional financial assets such as stocks and commodities. For this purpose, a stochastic conditional duration (SCD) model is estimated using Bitcoin's high-frequency transaction data. To capture the effects of trade prices and volumes on the duration between transactions, those are incorporated into the SCD model as explanatory variables. Furthermore, the intraday seasonality of the duration is modeled with a Bernstein polynomial and simultaneously estimated with other parameters in the SCD model. The model estimation was performed using a Bayesian Markov chain Monte Carlo (MCMC) method. The estimation results suggested a positive relationship between the duration and the changes in trade prices but a negative relationship between the duration and the absolute values of the price changes, which are also known for the volatility in the stock market. The duration process of Bitcoin is also found to be strongly persistent, which is also found in other financial assets. In contrast, no clear pattern of intraday seasonality could be found in the Bitcoin market.
