[navanit]

There's an implicit assumption in his argument that the mean is stable and finite. This need not be the case, and is often not the case with tail events under power-law distributions: http://navanitarakeri.com/Diversification.pdf

Also see the following: Didier Sornette and Daniel Zajdenweber, The economic return of research: the Pareto law and its implications, European Physical Journal B, 8 (4), 653-664 (1999)

here: http://xxx.lanl.gov/abs/cond-mat/9809366

There's also an excellent (and old) study on whether restaurants with specialized menus did better than restaurants with diverified menus. Turns out the specialized ones did well in the short term, but were vulnerable to sharp changes in taste by patrons. The more diversified ones had much higher longevity but did relatively poorly in the short term. I can't find the reference right now.

Finally, this "Fox vs. Hedgehog" problem is widely studied, and the research shows that the decision is very dependent on the market and domain - i.e it is difficult to generalize.