We can depend on human emotions -- greed and fear. No stats needed.
There are a few other choices here, too:
We can:
-assume the null hypothesis (stock market returns aren't possible to predict)
-depend on human beings finding patterns that don't really exist, then bet against them.
-Explain 19x earnings as natural in a low-rates environment.
-Be contrarians, bet against other peoples' prior views, and make money off of it.
I really think the smartest answer here is "I don't know; maybe."
I think that betting your life savings on eight independent data points (only 2-3 of which are similar to yours') that someone else calls stats is... suboptimal.
I think it's irresponsible to say we can pick out anything statistically significant yet on stock price returns over the 10 year period without at the very least trying to justify it using 3 year or 5 year returns.
I'm dubious about the stats these guys are using. If you count cars for 10 minutes on I-95 at 3AM and get 15 cars, and then come out and say "I-95 averages 1.5 cars per minute", that's not really accurate. To be fair, these guys have taken eight samples of I-95, but they may not be at completely different random times of the day, and if one of those samples is an outlier, now you need several dozen more samples to see how thick the tail is.
Here's another way of putting it. An insurance company studies 8 houses to see how much risk there is in insuring them. Last year, one house filed $200 a claim for a broken window when the neighborhood kids were playing baseball. The insurance company concluded that next year, they should expect about $25/house in claims, and billed everyone $30 for their insurance premium the next year.
The moral of the story is that when you're dealing with these heavy-tailed distributions, you need more samples. Normal distributions need about 20-30 data points; the market has fatter tails; these guys are trying to make some sort of statement with eight points, which is a tad dangerous.