Trade size, high-frequency trading, and colocation around the world

We examine the impact of changes in market microstructure, particularly algorithmic trading (AT) and high-frequency trading (HFT), on trade size across 24 stock exchanges around the world. Using colocation services as a proxy for AT and HFT, we find mixed results on the impact of AT and HFT on the average trade size. Furthermore, we test whether the presence of HFT leads to the introduction of colocation services. The data are consistent with the view that HFT pre-dates colocation by at least eight months on most exchanges, and has strong power in explaining the introduction of colocation services. In effect, our results show that colocation services do not properly measure effective AT and HFT; rather, colocation services are the result of HFT. Exchanges choose to offer colocation services due to the fact HFT requires higher speed transactions. Finally, we show there have been substantial changes in trade size in other countries such as China where there is no HFT and offer explanations for these changes and suggest avenues for future research.