著：Antti Ilmanen 译：徐瑞龙
Consider the situation of an investor —— such as a central bank, a commercial bank, an insurance company, or a pension fund sponsor —— that has to choose the neutral benchmark duration for its US dollar portfolio. This choice depends on the long-run reward-risk trade-off offered in the US bond market (as well as on the investor's investment horizon and risk tolerance) and not on any tactical interest rate views. Three directors of the investing institution meet to discuss their combined knowledge about the long-run bond risk premium. One director argues that the typical upward slope of the yield curve is evidence of a positive risk premium. Another director points out that the curve shape might reflect expectations of rising rates instead of a risk premium. It is better to look directly at historical return data, he argues, and presents the others some data that show how average returns over the past decade increased strongly with duration. The third director recalls that over a very long period (1926-94) long-term bonds earned only somewhat higher average returns than one-month bills and lower average returns than intermediate-term bonds. These findings are hard to reconcile until the directors realize that the recent sample reflects findings from a disinflationary period that was exceptionally favorable for long-term bonds. In contrast, the poor returns of long-term bonds in the longer sample partly reflect the yield rise over the decades. What should the directors conclude?
The goal of this paper is to help investors assess whether duration extension is rewarded in the long run. We present extensive empirical evidence mainly from the US Treasury bond market over the past 25 years. All findings about historical returns depend on the interest rate trend in the sample period, but we alleviate concerns about sample-specific results by studying a period without a strong trend. Further, by examining the historical returns over many subperiods, across markets and from several perspectives, we can give as conclusive answers about long-run expected returns as possible.