Unravelling the Trading Invariance Hypothesis
Michael Benzaquen,Jonathan Donier,Jean-Philippe Bouchaud
DOI: https://doi.org/10.1142/s238262661650009x
2016-12-01
Market Microstructure and Liquidity
Abstract:We confirm and substantially extend the recent empirical result of Andersen et al. (Andersen, T. G., O. Bondarenko, A. S. Kyle and A. A. Obizhaeva, 2015, Unpublished), where it is shown that the amount of risk [Formula: see text] exchanged in the E-mini S&P futures market (i.e., price times volume times volatility) scales like the 3/2 power of the number of trades [Formula: see text]. We show that this 3/2-law holds very precisely across 12 futures contracts and 300 single US stocks, and across a wide range of time scales. However, we find that the “trading invariant” [Formula: see text] proposed by Kyle and Obizhaeva is in fact quite different for different contracts, in particular, between futures and single stocks. Our analysis suggests [Formula: see text] as a more natural candidate, where [Formula: see text] is the average spread cost of a trade, defined as the average of the trade size times the bid–ask spread. We also establish two more complex scaling laws for the volatility [Formula: see text] and the traded volume [Formula: see text] as a function of [Formula: see text], that reveal the existence of a characteristic number of trades [Formula: see text] above which the expected behavior [Formula: see text] and [Formula: see text] hold, but below which strong deviations appear, induced by the size of the tick.