Conventional studies of long-run returns on capital market assets, because of changes in valuation between the starting date and the ending date, obscure the basic return each asset earns. Consequently, both absolute returns and measured risk premiums are distorted. The basic return can be extracted by selecting widely separated dates with identical valuation levels. Over nearly 200 years, the analysis for equities produced 63 episodes averaging 35 years with a mean nominal basic return of 9.6 percent and standard deviation of 1.6 percent; 63 bond episodes averaging 43 years produced a mean nominal basic return of 4.9 percent and standard deviation of 2.3 percent. Equities revealed a tendency to regress to the mean over time, but no such tendency was apparent in the bond data. Thus, long-run equity returns were more predictable than long-run bond returns. This conclusion applies with even greater force to real returns.