The perils of diversification

[bigbadwolf]

In the NYT:

The attractiveness of spreading investment eggs in many baskets is fading fast in the short term as strengthening correlations of different asset classes are aggravating losses on a diversified portfolio.

During the credit boom of 2002-early 2007, investors were encouraged to diversify their traditional equity-bond portfolio, spread risks and seek extra "alpha" returns by buying commodities, hedge funds, real estate which were seen as having almost zero correlation with traditional stocks and bonds.

However, since the credit crisis began in August 2007, these alternatives fell in lockstep with, or sometimes faster than, equities, driving volatility higher and amplifying losses of a risky portfolio.


"A diversified approach worked like a charm until October last year... But diversification failed in 2008," said Terence Moll, head of multi-asset strategy at Investec Asset Management.


"Some of alternative assets went through bubbles and precisely these bubbles got punished. Assets that are overpriced do not give diversification."


His analysis shows returns to eight-asset class portfolio -- comprising of stocks, bonds, emerging market stocks and bonds, real estate, commodities, hedge funds and managed futures -- lost 29.1 percent since October 2007.


This compares with a loss of 26.3 percent in a simple equity-bond portfolio.

 

[Sanket Patel]

Reminds me of the adage "in a crisis, correlations go to 1."

 

[Bastian Gross]

That sounds suspiciously like "herd behaviour" to me.

Asymmetric aggregation of animals under panic conditions has been observed in many species, including humans and portfolio-manager! It's a byproduct of communication skill of modern mankind and neighbour copying.

 

[PatM]

It also shows a lack of recognition between ex ante expectation, and ex post results. Diversification doesn't mean that in every single time interval, you are guaranteed to win, relative to other more concentrated choices. There are too easy counterexamples that show that in other time periods, diversified portfolios did in fact beat concentrated ones. So what are we to conclude from all that?

If I knew which one of the eight asset classes would go up the most, I wouldn't need diversification. Arguably, diversification did work in this case, because it beat many of the concentrated portfolios available: you didn't need to just pick a concentrated portfolio of stocks, or bonds. You could have picked a concentrated portfolio of commodities or private equity. Some people did, and really, really got pounded. They would have benefitted from a little diversification.

Trader's adage about expectations and results:
there are good trades and bad trades, and there are winning trades and losing trades,
but the good trades aren't always the winning trades, and the bad trades aren't always the losing trades.

It's not the math of diversification that is flawed - that is indisputable, pace journalists everywhere - but garbage in, garbage out for the expectations on individual items. If you can accurately project expectations for the future, portfolio theory can show optimal ways to order your portfolio (ex ante). But it's not a cure-all for errors in the individual expectations. It can't defend itself against things it never claimed to be.