10. The empirical bond risk premiums are based on the compound average returns of various maturity-subsector portfolios of Treasury bills, notes, and bonds between 1952 and 2009. This period does not have an obvious bullish or bearish bias because long-term yields were at broadly similar levels at the beginning and end of the sample. Moreover, the sample period begins just after the “Treasury-Fed accord” ended a decade of regulated (capped) bond yields. Arithmetic average returns would be somewhat higher than these compound average returns (at the longest maturities 0.2-0.4% higher). This evidence is from Chapter 9 in Antti Ilmanen, Expected Return (Hoboken, NJ: John Wiley & Sons, 2011), which discusses bond risk premiums but focuses mainly on forward-looking measures of such premiums.