Empirical properties of volume dynamics in the limit order book

The study of order volumes in financial markets has shown that these display several non-trivial statistical properties. The majority of studies have focused on the sizes of incoming orders or of realized transactions, the present work is a study of dynamical aspects of volume available at the best ask and best bid. The interest in these volumes stems from their capacity to limit or otherwise affect possible trades in the near future Using limit order book data from the Bitcoin/USDT market we study, among other things, the behavior of the distribution of volume changes as a function of the time scale at which the changes are measured, the autocorrelations of volume changes at each side of the book and the autocorrelations of volume imbalances between asks and bids. We find that several of these properties can be well approximated by power laws. •We investigate the dynamics of volume available at the best ask and best bid in a Bitcoin/USDT and Bitcoin/USDT order book.•A variety of stylized facts found in price returns are also present in the changes of volume. Other properties of volume series, such as a negative autocorrelation function, turn out to be very idiosyncratic.•Some dynamical properties appear to be present only at very short intraday time scales and vanish at daily frequency.