Mean Aversion in Fixed Income Returns: Evidence and Implications for Long Horizon Investors

Abstract

This study documents strong mean aversion in U.S. fixed income returns (but not stock returns) at 5-20 year horizons. These results are only slightly weaker for nominal returns than for real returns and prevail regardless of the period examined (1926-2011, 1951-2011, 1857-1925 or 1857-2011). I show that the presence of mean aversion, along with a historically sizable equity premium, dramatically increases the risk of holding fixed income securities for long horizon investors, to the point where they are often actually riskier than stocks in a downside-risk framework for investment horizons as short as ten years. These results are fairly robust to the use of nominal vs. real returns, different sample periods from which simulated returns are drawn and even to sizable reductions in the historical equity premium.