Trading in fixed income assets is a profitable business in global investment banks. Besides providing market liquidity through market-making activities, investment banks also devote significant amounts of proprietary capital to trade a wide variety of fixed income instruments, such as Treasury bills to 30-year government bonds, corporate bonds and mortgage-backed securities, etc. Besides investment banks, hedge funds and dedicated bond funds also actively pursue trading opportunities in fixed income assets.
The strategies deployed range from simple arbitrage trading, to complex trades based on technical or market views on the term structures of interest rates and credit risks. These yield curve trading strategies are essentially bets on changes in the term structure. These trading strategies can be broadly classified as directional and relative-value plays. Directional trading, as the name implies, are bets on changes in the interest rates in specific directions.
Relative-value trading, by contrast, focuses on the market view that the unconditional yield curve is upward sloping, and that the current yield curve would mean-revert to an unconditional yield curve. A wide variety of trading techniques are used to construct relative-value trades based on this market view. However, there have been few efforts to examine the performance of these trading strategies or to compare them with equity investment strategies. Litterman and Scheinkman (1991), Mann and Ramanlal (1997) and Drakos (2001) are recent studies on the subject.
In this paper, we analyze the performance a specific class of such relative-value trading techniques that are directly implied by the notion that mean-reversion of the yield curve occurs. We consciously avoid “data-snooping” by not searching through a large number of possible strategies to find a few that are profitable. Instead, we start from the market view that the yield curve mean-reverts and derive trading strategies that follows most naturally from such a view—if the level, spread or curvature is higher (lower) than the historical average, bet that the level, spread or curvature, respectively, will decrease (increase) towards the historical average. We shall refer to this class of technical trading strategies as “mean reverting” trading strategies.
Following Litterman and Scheinkman (1991), we consider the three aspects of the yield curve namely, the interest rate level, the slope (i.e. yield spread) and the curvature and construct a portfolio of yield-curve trading strategies centering on each aspect. To facilitate a consistent comparison of their performance, we impose cash neutrality and consider one-month holding period for each category of strategies, and adjust the payoff for risk, as measured by the standard deviation of the payoffs. Our study abstract from credit risk in particular, default risk and chose as our dataset the U.S. Treasury interest rates, from the period 1964 to 2000 for our study. For each aspect of the yield curve, we consider strategies that trade on the whole yield curve, as well as strategies that trade on individual portions of the yield curve.