estimates of other parameters-like individual investor risk aversions-that subjects these results to much debate. However, observable market data do reveal important information about the range of equity return expectations. In particular, we observe yields on corporate bonds for the same companies for which we desire the expected return on equity. Corporate bond yields-along with an estimate of the long-run expected loss on these bonds due to default-deliver reasonably accurate estimates of the expected return premium on corporate bonds. Since equity is a subordinated claim on the same assets of the firm, in equilibrium equity holders demand a premium above corporate bonds. Using data from June 2002, we estimate the equity market capitalization weighted U.S. corporate bond yield above Treasuries is approximately 2.25 percent.5 Using a rough estimate of historical default losses on U.S. corporate bonds of 0.75 percent, we arrive at an expected U.S. corporate bond risk premium of 1.5 percent. This provides, at a minimum, a lower bound on the current ERP. Considering that the volatility on equities is two to three times that on corporate bonds, we cautiously suggest that investors are currently demanding an ERP in the neighborhood of 3 percent or more. You might call this a casual empirical estimate! THE EQUITY PREMIUM AND INVESTMENT POLICY_____________ Why are investors so concerned about the level of the equity premium? The principal reason is very straightforward: Practically every important decision that an investor makes is driven by the equity premium assumption. Decisions like the split between equity and bond holdings, the allocation to alternative investments, and the level and structure of active risk taking all depend on the equity premium assumption. Given the importance of this assumption, it is not terribly surprising that so much time is spent in analyzing the historical record. Unfortunately, however much time we spend analyzing the historical record, it will not be enough to estimate the equity premium with any level of certainty. For example, with 130 years of data from 1872 to 2001 and stock market volatility of 20 percent per year, the standard error in Fama and French's average return estimate is 1.75 percent. Therefore, an estimate of 3.5 percent is only two standard errors from zero. This permits us to reject the null hypothesis that the ERP is zero with a confidence level of 5 percent. Let's turn the problem around, however, and test, at the same level, how different the equity premium is from 3.0 percent. For this test, we would need another 6,270 years of data! Thus, from a practical perspective, a significant level of uncertainty is bound to accompany any estimate of the long-run equity premium. The equity premium clearly plays an important role in setting investment policy. The goal of this chapter has been to provide some guidance that investors can use to set their own equity premium assumptions. As the discussion has indicated, 5Using option-adjusted spreads over the U.S. Treasury curve on a broad portfolio of corporate bonds, including high-yield bonds.