I think the biggest issue is that too many investors (and investment advisors/consultants) took the output from a mean-variance optimization as gospel. Just about everyone realizes (now) that this is not a good idea. One solution is to use more intuition (i.e. a little more art and a little less science). Most academics aren’t going to concede this as the solution, so it would be interesting to hear from the man himself what he thinks the solution is to the shortcomings of MPT.

There are a few obvious adjustments that can be made such as accounting for higher moments (skewness and kurtosis) and adjusting for stale pricing of private investments (lagged pricing regressions, ARCH, GARCH, etc.). Unfortunately most investors/advisors do not have the software at their fingertips that make such adjustments. Supposedly JP Morgan has developed this and I am trying to line up a meeting with them, but haven’t scheduled anything yet.

A few other questions that I think would be interesting are;

Can MPT, or some other model account for changing correlations?

at what level does Markowitz think MPT should be used: for example, is the best use to set broad asset allocation targets (stocks, bonds and real assets), or should it be used to model finely categorized investments (large cap vs. small cap stocks, investment grade vs. non-investment grade bonds, etc.).

thoughts on other concepts developed to building portoflio’s such as Black-Litterman and Bill Sharpe’s expected utility.

As the main inputs to any mean/variance model are mean and variance, what are your thoughts on the recent past given alternative investments, and that these historical estimates have shown to be wildly wrong? An optimizer will obviously allocate to alternatives, but without understanding what drives these returns, is this not totally misleading? For example, hedge funds use leverage and tricks like collecting option premium to looks smart and fool the Sharpe ratio, and occasionally even tell their investors about it. Private equity instead chooses to tell investors basically nothing, and when they do it is 90 days too late and the valuations are, um, suspect. The wonderfully uncorrelated pvt equity returns everybody brags about at cocktail parties are very correlated if adjusted for leverage and time lags. The distributions of these returns are far from normal as well. So how do we adjust our expectations of an optimizer, or the inputs to the model, to incorporate these problems looking forward?

Why do so HFoF allocate so few assets to global macro, especially compared to the early ’90s and even less to managed futures?

Would it be fair to suggest that in the taxonomy of investing HFoFs consider themselves “better” than CTAs somehow due to the difference in regulation or background of the different “classes” of investors?