A modern-day investment professional looking back at accepted doctrine in the early 1900s would rightfully characterize much of it as misguided. Do you remember when it was not considered appropriate, or even legal, to hold risky securities in a portfolio, even if the portfolio was not risky?
Will investment professionals fifty years from now view our performance measurement regulatory structure, guidelines, theory, and practice as misguided? Let me try to make the case that they will. If you agree with part of what I say, then I challenge you to think of how our profession can be changed for the better.
The orientation of today's performance measurement, at least in principle, is to rank portfolio managers from best to worst. Does this make sense?Suppose it is possible to rank portfolio managers from best to worst. Imagine that you are a pension fund officer, with hordes of investment management firms that want to manage your money knocking at your door. You measure each firm's performance. According to the best to worst paradigm, it makes sense to hire the best one and forget about the others.
Isn't this thinking behind what our consultants do? Do they ever bring you managers that look mediocre, or bad? Do they bring you a large group of managers? No.Consultants might respond by pointing out that they almost never offer you a choice of only one manager. This may be because consultants can't measure a portfolio manager's true performance or talent. All they can measure is his past performance. Performance measurement's goal of providing a useful ranking requires being able to estimate reasonably accurately a portfolio manager's future performance. Of course, consultants may deny this, since it not clear that they can ascertain future performance.
Worse, there is a theoretical reason for thinking that forecasting a portfolio manager's performance reasonably accurately is impossible. To see why, posit that a portfolio manager has been identified who is virtually certain to provide good future performance. The fact that he has been identified changes his ability to deliver performance. Anyone who trades with him at the prices he wants to trade at will be buying overpriced stocks and selling underpriced ones. Consequently, his ability to trade will be impaired. There is a theorem lurking hereabouts that goes something like this: To the extent that it is possible to ascertain that a portfolio manager's past performance is due to talent, then to that extent he will not be able to deliver future performance.The problem is not bad consultants but the impossibility of the task. However, things are even worse. The notion of providing a ranking misses the point.
Even if you could identify the portfolio manager with the best future performance, you wouldn't give all your funds to him any more than you would construct a one-stock portfolio. Allocations to managers are like allocations to stocks in a single portfolio. As long as there is risk, it pays to balance expected reward against risk with managers' portfolios just as with stocks. The method for doing this is portfolio selection, not performance measurement. Moreover, performance measurement as it currently is done is not relevant for portfolio selection.Allocating funds among managers' portfolios requires the same kind of information that is needed to allocate a portfolio among stocks. For example, a mean-variance context requires expected returns, variances, and covariances for all securities. Performance measurement does not provide this information for managers' portfolios. Nor do consultants.
Can this information be obtained? The outlook is not good. Success at active management requires beating the competition, an inherently dynamic and unstable process. It would be surprising if actively managed portfolios had usefully stable relationships of the sort needed to do a good job of allocating a pension fund's money to managers as an ordinary portfolio is allocated to stocks.Is there a reasonable expectation that such information ever will be available? Suppose it was. Then one implication is that managers would be identifiable as reasonably talented or not which, as we have already seen, is self-destructing.