1984issue C021-9
Evaluating managed account portfolios on the risk-return frontier
This editorial walk-through treats allocator evaluation as a three-layer audit: whether a selection rule only ranks average reward, whether a reward-to-variability shortcut is treated as generally optimal, and whether an explicit market frontier is matched to a stated preference map so the tradeoff can be checked.
- A manager with a 10 percent average quarterly result can still have posted no quarter at 10 percent, so the average alone need not describe any outcome an allocator would have received.
- As periodic results spread out, that spread is treated as risk, and the standard deviation of returns is one formal measure of how widely those results scatter around their average.
- A reward-to-variability ranking coincides with full mean-variance selection only when utility of money is quadratic, and even that construction answers only a variability-of-profit question.
- One frontier book is chosen only after taste is stated as a utility curve or an indifference map on the same axes, with the preferred point at the tangency of the market frontier and the highest reachable curve.
A three-layer audit
This is an editorial reading of a historical evaluation workflow. Allocator evaluation is taught here as three tests of the selection rule itself, not as three separate contests for the same managed-account book.
The first test asks whether the rule only ranks average reward. The second asks whether a single reward-to-variability ranking is being treated as generally optimal. The third asks whether an explicit market frontier is matched to a stated preference map, so reward, dispersion, and the willingness to exchange one for the other can be checked as one procedure.
Whether the rule only ranks average reward
A manager whose average quarterly result is 10 percent can still have posted no actual quarter at 10 percent. The average alone need not describe any outcome an allocator would have received.
The less informative the average becomes as periodic results spread out, the more that spread is treated as risk. The standard deviation of returns is one formal measure of that spread. It is used here as the risk axis: a numerical summary of how widely periodic results scatter around their average.
Editorial reading: if the selection rule stops at a ranking of averages, the first layer of the audit fails. The received path has not been tested.
Whether a reward-to-variability shortcut is treated as generally optimal
A reward-to-variability ranking is presented as giving the same answer as full mean-variance selection only when the decision-maker’s utility of money is quadratic. That ranking is a shortcut through the same return-versus-dispersion plane, not a separate theory of risk.
Even the broader construction still answers only a well-specified variability-of-profit question. It does not accommodate other risk types or utility schedules that kink or reverse.
Editorial reading: if a book is chosen solely because it ranks first on the ratio, the second layer asks whether quadratic utility was actually assumed. Treating the shortcut as generally optimal is a claim the rule has to make explicit.
Whether the market frontier is matched to a stated preference map
Mean-variance optimization, as used here, is the construction of the set of attainable expected-return and return-dispersion pairs, then selection of the single frontier book that touches the highest reachable preference curve.
In expected-return versus standard-deviation space, only combinations on or to the right of a market-frontier line are treated as attainable. The market frontier is the left-hand boundary of those combinations. A frontier point dominates any interior point that has the same return with more dispersion, the same dispersion with less return, or both.
When several managers are profitable on average and their results are not perfectly positively correlated, combining them is presented as producing smoother, faster account growth than any one of them. The same combination is presented as shifting the frontier up and to the left until the attainable set is closed.
How a preference map picks one book
Choosing one book from the many frontier combinations requires stating taste for risk and reward unambiguously as either a utility curve or a risk-return indifference map drawn on the same axes as the frontier.
A utility curve is a schedule of satisfaction from successive units of money. Its shape decides which frontier point is preferred. Marginal utility is the added satisfaction from the next unit of money. Most people are described as having decreasing marginal utility of money. Constant marginal utility is assigned to a narrow theoretical type, and increasing marginal utility to high-stakes gamblers. The same frontier can therefore map to different preferred points.
An indifference map is a family of curves in the same return-versus-dispersion plane. Each curve holds satisfaction fixed, and higher curves rank as more desirable. Points on one indifference curve are defined as equally satisfying. The single preferred portfolio is the tangency of the market frontier with the highest reachable curve.