Average U.S. buyout fund performance has exceeded that of public markets for most vintages for a long period of time. The outperformance versus the S&P 500 averages 20% to 27% over the life of the fund and more than 3% per year. Average U.S. venture capital funds, on the other hand, outperformed public equities in the 1990s, but have underperformed public equities in the 2000s.
These strategies should have higher 'operational risk,' and thus higher risk from tail events, Knightian/Keynesian uncertainty. It highlights that merely noting an asset class is opaque and illiquid does not mean it generates an above-average return. This is especially important because many pension funds are stretching to alternative assets (read: hedge funds) under the illusion that these areas have less clear data, so are riskier, therefore generate higher returns.
Most interesting are the data issues, which are very complex. If you read the paper, you see it is all about what data to include, which to exclude, etc. It has very little to do with asymptotic standard errors, or overidentifying restrictions that are presented as the really important tool for empirical financial researchers in graduate school.
I mean, consider this graph, and you can see there's something obvious going on, and that data prior to 2000 was probably not the same as that afterward. That isn't an abstruse statistical issue.
Click to enlarge