stocks have performed better than bonds over every 30-year period since the beginning of the 20th century
Non-overlapping sample size = 3
This is not just a nitpic. I do not think your central claim, that riskiness declines with time horizon, is true outside of some special cases. If you have other arguments for this I would be happy to discuss them.
I do not think your central claim, that riskiness declines with time horizon, is true outside of some special cases. If you have other arguments for this I would be happy to discuss them.
The obvious argument is the Central Limit Theorem. Unless you want to argue that the log-variance of stock movements is secretly infinite, which seems doubtful.
Yes your average return will mean revert but I do not think there is any reason to think total wealth will revert to trend. e.g. if SPY falls by 10% this year, there is little to no catch up growth in future, though obviously future positive returns will dilute the 10% drop.
My recommendation to use the maximally aggressive allocation may not be reasonable for everyone, but my impression is that it’s the best for most EAs. Do you disagree about this?
You may be right here—I haven’t researched this claim much. It’s also true that stocks performed unusually well during the 20th century and that this may not continue.
Non-overlapping sample size = 3
This is not just a nitpic. I do not think your central claim, that riskiness declines with time horizon, is true outside of some special cases. If you have other arguments for this I would be happy to discuss them.
Otherwise, I recommend The long-term risks of global stock markets
The obvious argument is the Central Limit Theorem. Unless you want to argue that the log-variance of stock movements is secretly infinite, which seems doubtful.
Yes your average return will mean revert but I do not think there is any reason to think total wealth will revert to trend. e.g. if SPY falls by 10% this year, there is little to no catch up growth in future, though obviously future positive returns will dilute the 10% drop.
Oh, huh, you’re right—the relative log-variance of a product of iid lognormals shrinks, but the relative variance doesn’t. TIL, thanks!
My recommendation to use the maximally aggressive allocation may not be reasonable for everyone, but my impression is that it’s the best for most EAs. Do you disagree about this?
The maximally aggressive strategy is to lever up 10x and bet on momentum!
I think there are credible arguments for this view, but these are mainly about EAs having low risk aversion, not the strategy being low risk.
Really the sample size is closer to 4 ;)
You may be right here—I haven’t researched this claim much. It’s also true that stocks performed unusually well during the 20th century and that this may not continue.