Download this Working Paper (PDF)
Abstract
Many participants, regulators, and observers of commodity and security markets have a sense that something in recent years has gone awry: that the explosive growth of high-frequency digital trading is somehow excessive, costly, unfair, and/or destabilizing. Several ideas for changing the rules have been discussed. Without a coherent explanation of exactly what is wrong, however, it can be very difficult to develop a promising remedy.
The object of this paper is to offer one such explanation: that the digitization of the trading infrastructure, in combination with ubiquitous but fleeting information asymmetries, has stimulated a dramatic expansion of racing. By racing I mean the wasteful expenditure of resources in a contest to trade ahead of other market participants; that is, racing – like its cousin, queuing – is an example of a directly unproductive profit-seeking (DUP) activity whose costs erode the gains from trade that otherwise would be available to participants in the market. The paper also offers a specific remedy: the optional use of randomizing temporal buffers in the order flow. By slightly slowing the pace of trading, such buffers will allow market-data dissemination processes to saturate (i.e., will allow information asymmetries to dissipate) a little bit faster than order execution processes, so that price discovery and trading can operate more efficiently in an environment with more symmetrical information. By decoupling order flow from market-data flow, this remedy should also help reduce the likelihood of chaotic feedback instabilities in automated trading markets.