Equilibrium High Frequency Trading

Abstract

Algorithms enable investors to locate trading opportunities, which raises gains from trade. Algorithmic traders can also process information on stock values before slow traders, which generates adverse selection. We model trading in this context and show that, for a given level of algorithmic trading, multiple equilibria can arise, some of which generate market exclusion for slow traders and sharp increases in the price impact of trades. We offer a theoretical interpretation for the flash-crash of may 2010. Next, we analyze the equilibrium level of investment in algorithmic trading. Because when others become fast it increases adverse selection costs for slow investors, algo-trading generates negative externalities. Therefore the equilibrium level of algo-trading exceeds its utilitarian welfare maximizing counterpart. Furthermore, since it involves fixed costs, investment in algorithmic trading is more pro table for large institutions than for small ones. This generates equilibrium informational asymmetries between large fast traders and small slow traders.