This article revisits some work I did for a quant conference in 2007 which looked at the challenges of variation in trading volume, particularly for VWAP strategies. This remains topical: a few years later I had to ensure an existing range of algorithms for a bank could execute futures contracts which often only traded once a day. The expectation is that all listings can be traded electronically.

The chart below shows data for 1,365 main board listings at the HKEx during August 2014, excluding ETFs, depositary receipts, warrants and other derivatives.

The chart plots the coefficient of variation for trade counts and traded volumes: in both cases the coefficient is the standard deviation divided by the mean, so the chart can show relative trading consistency. The traded volume excludes the morning auction, as the focus of this analysis is on the five and a half hours of continuous trading.

The HSI listings are clustered at the lower left of the chart, where daily variations in trade count and trading volume are up to 50% of the average for each listing. Outside of the index most listings cluster in group which has variations in daily volume of 100% or more.

Note that the chart shows the relationship between trade and volume variation is broadly linear, which implies there is a typical trade size for each listing.