Diversification is not

*automatically*an improvement. And even if it is an improvement it is not automatically a

*meaningful*improvement.

A good example of an asset class that has been a diversifier, yet did not improve traditional portfolios, is "commodities" (collateralized commodities futures) funds, over the time period when they have been available in the form of mutual funds and ETFs. Our specific example will be PCRIX, once widely recommended as a diversifier, suitable for inclusion in otherwise-traditional retirement savings portfolios.

As I'll show below, PCRIX (PIMCO Commodity Real Return Strategy) has behaved very

*differently*from VTSAX (Vanguard Total Stock Market Index Fund). Since inception, it has had a good claim on being a diversifier, a truly distinct asset class, independent from stocks.

Over the lifetime of PCRIX their correlation has been 0.40. (That's in the range that people call "low" when they are advocating for diversifiers; for example, over that same time period, the Vanguard Real Estate fund had a correlation of 0.68 with stocks; Vanguard Total International, 0.88; and DFA US Small-cap Value Portfolio, 0.91). (Source)

There are other ways to measure diversification, but I doubt that anyone can find a measure by which PCRIX hasn't qualified as "a diversifier."

Yet, if we take VSMGX, Vanguard LifeStrategy Moderate as a good example of a "traditional" portfolio (Portfolio 1, blue) and compare it to the effect of adding PCRIX (Portfolio 2, red, 80% VSMGX, 20% PCRIX), over the life of PCRIX, the effect of adding PCRIX would have been to make the portfolio worse by almost every measure:

Source

Lower return (CAGR). Higher volatility (as measured by standard deviation). True, better "best year." But worse "worst year," worse drawdown, and lower risk-adjusted return as measured by Sharpe and Sortino ratios.

Now, some will say "you deliberately chose stinker of a diversifier to make your point," and so I did. The point is, it

*was*a diversifier, yet it did

*not*improve the portfolio as a whole? Why not? The answer is obvious: PCRIX, since inception, has been

*such a lousy investment in itself*.

Improvement through diversification is a

*balancing act*with

*two*pans.

On the one side, low correlation means that the volatility of the whole will be less than the weighted average of the volatility of the parts. That's the "free lunch." You can get your free lunch in the form of reduced volatility. Alternatively, you can adjust the portfolio to hold more of the now-less-risky mix, and get it in the form of increased return.

But on the other hand, the annual

*return*of a portfolio is

*always*just the weighted average of the returns of its parts. If it is 2014 and if fund X goes up 32.00% and and fund Y goes down -14.00%, then a 50/50 mix of the two will go up by 16.00%-7.00% = 9.00%,

*exactly*, and nothing about the correlations or standard deviations of the two funds can change it.

Dragging down the return by a small amount may be worth it if it also cuts volatility by a large amount. The Sharpe ratio measures that relationship, and as noted if we get an improvement in Sharpe ratio, we can turn that into improved return rather than reduced volatility if we want to.

To return to my main point, though,

*diversification is a balancing act*with

*two*sides of the scale.

The "magic" of MPT is that

*if things are right*, an asset that is not so great in itself can improve a portfolio as a whole. This is true because it is just math, but "if things are right" can be a strong requirement. The idealized example is two assets that have

*about the same return*and low correlation. If they have about the same return, then the low correlation is roughly cost-free way to reduce the volatility. If they don't have about the same return, then the effect of diversification is to reduce volatility, but the lower-return asset also reduces return.

It's a balancing act, the balancing scale is the Sharpe ratio (or other measure of risk-adjusted return). A pretty good asset with very low correlation makes a big improvement. An almost-equal-return asset with not low but not 1.00 correlation makes a small improvement. In theory, a low return asset can make an improvement if the correlation is low enough. In theory a

*zero*-return asset

*can*make an improvement, but the correlation must be not just low, but actually negative. In theory, an asset

*could*lose money yet improve a portfolio if the correlation were sufficiently negative.

But it is always a balancing act, and if the diversifier has low returns, it will not automatically improve a portfolio--it has to be an

*extremely powerful*diversifier.

Now, the problem is that none of the things that govern whether a diversifier improves a portfolio are

**any more predictable**than return itself is. "Past performance is no sure guide to future results" is just as true of volatility and correlation as it is for return.

Many diversifiers lie in the "grey zone of endless debate" and the usual reason is that the balancing pans--between "diversifier" and "drag" have, in the past, been so close level--either way depending on choice of time period--that nobody can produce a truly compelling case to believe that in the future, the good effects of diversification will outweigh the bad effects of drag.

If someone point out some fund or asset class as being "a strong diversifier, with low correlation with stocks," don't let that lull you into thinking "so it must be good." You have to ask hard questions. It needs to be at least a

*fairly good*investment in itself, and you need to believe that both "decent return" and "low correlation" will continue into the future.