تجارت با فراوانی بالا و سازندگان بازار جدید

This paper characterizes the trading strategy of a large high frequency trader (HFT). The HFT incurs a loss on its inventory but earns a profit on the bid–ask spread. Sharpe ratio calculations show that performance is very sensitive to cost of capital assumptions. The HFT employs a cross-market strategy as half of its trades materialize on the incumbent market and the other half on a small, high-growth entrant market. Its trade participation rate in these markets is 8.1% and 64.4%, respectively. In both markets, four out of five of its trades are passive i.e., its price quote was consumed by others.

This paper benefits from proprietary Chi-X and Euronext datasets that contain anonymized broker IDs for trades in Dutch index stocks for a sample period that runs from September 4, 2007 to June 17, 2008. One particular set of broker IDs matched across markets shows the characteristics of an HFT that acts as a market maker in both the entrant market (Chi-X) and the incumbent market (Euronext). In each market, four out of five of its trades are passive i.e., the HFT was the (liquidity-supplying) limit order in the book that got executed. Its lowest inter-message time is at most one millisecond. It trades actively with an average of 1397 trades per stock per day. It makes money on the spread but loses money on its positions. If this positioning loss is decomposed according to duration, one finds that positions that last less than five seconds generate a profit whereas the ones that last longer generally lose money. The HFT is equally active in both markets as roughly half of its trades are on Chi-X and the other half are on Euronext. Overall, it is a significant market participant as it shows up in 14.4% of all trades (aggregated across markets). It is particularly active in Chi-X where it participates in roughly every other trade. The paper also focuses on the capital required for the operation. The fee structure and margin requirement of the each of the two clearing houses associated with the two markets were retrieved to the extent possible. The HFT cannot net its positions across markets and therefore is estimated to have to put up a 100 times more capital than what would have been needed if netting were possible. This might in effect explain why the U.S. equity market is most fragmented as netting is possible in the U.S. The maximum (across my sample period) of the capital tied-up due to margining is then taken as a natural measure for the standby capital needed in the operation. This capital measure along with the daily gross profits and the risk-free rate imply an average annualized Sharpe ratio of 7.62. Round-the-clock price changes are modeled to analyze whether the HFT is visible in the security's market prices. The market making literature suggests that it might be in two ways: permanent price changes should correlate negatively with HFT position change (the HFT is adversely selected on its quotes) and transitory pricing errors should also correlate negatively with the HFT position (the HFT skews quotes to get out of its position). The evidence is supportive. In the trading day, the HFT position generates significant (transitory) price pressure. It is an economically meaningful amount as it is, for example, larger than half the average bid–ask spread. Also, the (surprise) HFT position change correlates negatively with permanent price changes throughout the trading day, but not in the overnight period for which the sign is reversed. Finally the results show that fees are a substantial part of a high frequency trader's profit and loss account. It is therefore not surprising that new, low-fee venues have entered the exchange market as they are attractive to these ‘modern’ market makers. This evidence adds to the regulatory debate on high frequency traders and highlights that a subset is closely linked to the rapidly evolving market structure that is characterized by the entry of many new and successful trading venues.