Bulk Volume Classification and Information Detection
44 Pages Posted: 11 Oct 2014 Last revised: 26 Jul 2018
Date Written: July 16, 2018
Using European stock data from two different venues and time periods, for which we can identify each trade’s aggressor, we test the performance of the bulk volume classification (Easley et al. (2016); BVC) algorithm. BVC is data efficient, but may identify trade aggressors less accurately than “bulk” traditional trade-level algorithms. We find, however, that using out-of-sample calibrated BVC mitigates accuracy disadvantages. In addition, BVC-estimated trade flow is related to proxies of informed trading, while bulk tick test is not. Interestingly, we also document that the identification of individual trade aggressors no longer captures information. In the new era of fast trading, sophisticated investors, and smart order execution, BVC appears to be the most versatile algorithm.
Keywords: Classification algorithms; bulk volume; informed trading strategies
JEL Classification: G11, G12, G14, G18
Suggested Citation: Suggested Citation