Use much larger international samples, including data from markets that shut down or had long interruptions. Doing this brings the return down several percent from the USA estimates; for example, Jorion 2003.

(It stands to reason that to maximally avoid survivorship bias and diversify, you would want to invest equally in every country, at which point your portfolio would grow something like the growth rate of the global economy which is ~2% annually over the last century or two IIRC; ironically, this apparently has happened to the Norwegian sovereign wealth fund - it's too big to invest significantly in any particular market, so they are ultra-diversified and have what looks like low annual returns.)