You're overfitting. This is exactly the problem I'm describing. Case in point, you've omitted 2000-2010. Real (CPI as inflation) averaged annualized return on

You're overfitting. This is exactly the problem I'm describing. Case in point, you've omitted 2000-2010. Real (CPI as inflation) averaged annualized return on the DJIA 2000-2010: -2.219% This is with relatively mild inflation. Of course, 1960-2000 looks better when you literally put in the apex of the software bubble when P/E ratios shot skywards...part of which corrected over the decade following. Expecting a 8% nominal or real return is simply nuts in my opinion because you are staking yourself on a really small sample size with crazy inflation to boot. Nominal or real is a nuanced distinction to what is plainly just bad overfit thinking. The introduction of just one decade with low inflation and bad returns can wreck you -- case in point: Dallas Firefighters/Puerto Rico ($43bn + of unfunded pension obligations)/Detroit etc. etc. etc.