Automation versus Intermediation: Evidence from Treasuries Going Off the Run

This paper examines the choice of trading venue by dealers in U.S. Treasury securities to determine when services provided by human intermediaries are difficult to replicate in fully automated trading systems. When Treasury securities go “off the run” their trading volume drops by more than 90%. This decline in trading volume allows us to test whether intermediaries’ knowledge of the market and its participants can uncover hidden liquidity and facilitate better matching of customer orders in less active markets. Consistent with this hypothesis, the market share of electronic intermediaries falls from 81% to 12% when securities go off the run. ELECTRONIC TRADING SYSTEMS HAVE BEEN steadily increasing their share of securities trading in almost all financial markets. As the amount of automation in financial markets increases, it is natural to ask what services human intermediaries provide that are difficult or impossible to replicate in a fully automated trading system. Existing theory suggests that human intermediaries play at least two important functions. First, an intermediary’s knowledge of the market and its participants may uncover hidden liquidity that facilitates quicker and more efficient matching of customer orders (Grossman (1992)). The value of this matching function is greater when trading volume is low and matches are difficult to find. Second, when information asymmetry is high, the repeated interaction between an intermediary and its customers allows the intermediary to protect itself against informed trades and offer better prices to its customers (Seppi (1990)). These two factors explain, for example, why electronic communications networks (ECNs) have a greater market share for the largest and most actively traded Nasdaq stocks while Nasdaq market makers have a larger share in smaller, less liquid stocks characterized by greater information asymmetry (Barclay, Hendershott, and McCormick (2003)), and why NYSE specialists have higher participation rates in smaller, less liquid stocks (Madhavan and Sofianos (1998)). ∗Simon School of Business, University of Rochester, Haas School of Business, University of California, Berkeley, and Forensic Economics, Inc., respectively. We thank Bruno Biais, Ken Garbade, Charles Jones, Rich Lyons, Haim Mendelson, Christine Parlour, Tunay Tunca, and seminar participants at the University of Rochester, the University of Washington, Stanford University, the 2004 NBER Market Microstructure meetings, the 2004 Workshop on Information Systems and Economics, and the 2004 MTS Conference on Financial Markets for helpful comments and suggestions. We are especially grateful to Michael Fleming for sharing numerous insights on the Treasury market. Hendershott gratefully acknowledges support from the National Science Foundation.