Siegel States Equity Premium 3.1%

Jeremy Siegel is touting stocks, with his latest at last week's TD Ameritrade Institutional conference:

U.S. stocks have had an average annual real (after inflation) return of 6.7% between 1802 and 2010, compared with 3.6% for bonds. While there are good decades and bad for all asset classes, Mr. Siegel thinks that the long-term trends will hold true for equities.

He didn't say it, at least in the newspaper quote, but the implication is that is the expectation, as well (even greater for today, because of current undervaluation). Now, Jeremy does not correct his number for adverse flows, the fact that people, in aggregate, put more money into stocks before bad times than before big increases. He also does not adjust for taxes, or transaction costs. The latter must have been terrible, because before computers, the brokers surely could have given clients prices near their daily lows and no one would now (currently there's a little kerfluffle over this in FX bank pricing). The survivorship bias of always using the US and other markets that were extremely fortunate (unlike Poland, Japan, or Russia), is also ignored. Those are each major adjustments, taking out a couple percent of the return.

Further, looking at the data back to 1802, one must remember our grandparents didn't appreciate survivorship bias. As Bill Schwert of the University of Rochester observed about one portion of the data:

From 1825 through 1845, [Smith and Cole (1935)] created an equal-weighted portfolio of six New York banks and one insurance stock. For both of these portfolios, Smith and Cole omitted most of the stocks for which they had collected price data. They chose stocks in hindsight to repreesnt typical movements in the period. Further, to think these data had only tens or less of stocks for any one time highlights how precarious these estimates are. The sample selection bias caused by only including stocks that survived and were actively quoted for the whole period is obvious.

Basically, in the bad old days, every couple of generations a couple professors would look at an industry that survived, on the one exchange that flourished (Charleston, S.C., New Orleans, and Norfolk, Va. had vibrant exchanges at one time), and then take within that those stocks with complete data over a period.

Obviously, this make the sample cleaner with regard to annual fluctuations, which were the major concern, but the averages of this clearly inflate sample returns. One cannot reject the hypothesis that the rate of return on stocks is no greater than that for bonds.