Over the past few months, numerous articles have appeared in the financial press declaring that Treasuries are no longer an effective portfolio diversifier with stocks. In the simplest terms, the rationale is that yields are now so low that there’s just not much room for bonds to rise in price when stocks fall.

There’s some truth in this claim. But while Treasuries may not have quite the near-mirror-image anticorrelation they have sometimes shown in the past, the rationale for simply giving up on them is far from complete.

First of all, since their recent high water mark in early August, prices have retreated around 8%, as measured by the Vanguard fund VGLT, which covers the 10-30 year Treasury maturity range. This means that to just return to highs already established, long UST have over 8% upside headroom.

Second, the correlation argument can only be established with more than a few month’s data. And it is not necessary to have a full -1 correlation for assets to be useful cross-diversifiers. Anything less than a perfect correlation of 1 is an improvement over nothing.

Third, correlation is not even precisely defined absent specification of sampling frequency. A statement that two sets of data have a correlation X is incomplete. The same data can have a low correlation on a monthly basis and a high correlation on an annual basis, etc.

Fourth, correlation can only be defined with respect to a specified unit of measure.  As we have shown in prior posts, two uncorrelated assets may appear to be correlated merely by using a common unit of account whose own fluctuations impart an illusion of correlation to them.  

Often large drawdowns in stock prices are due in part to the currency they’re being quoted in rising in value.  This was a major factor in the 2008 episode, as the rush to dollars to satisfy debt caused a surge in value, as revealed by the Financology Dollar Index.  In such a case, even if the dollar price of Treasuries didn’t change, they would rise in value by virtue of the rising value of dollars. 

But more significantly, the assumption that diversification power can only be established by statistical arguments is in itself invalid. We only know past performance with certainty. So if we were to try and predict future correlation using only the statistical past we could certainly find a basis for projecting just about any correlation we wished.  Past performance is no guarantee of future results.

The better argument isn’t merely statistical, but fundamental. Treasuries and stocks are the securities of different issuers. Things which impair the prospects of businesses and especially their common stock – the lowest and most contingent of claims on corporate assets – aren’t the same as those which affect the value of securities backed by the full faith and credit of the US government. The risks are fundamentally different. And consequently the fundamental rationale for diversification power stands.

None of this is of course to say that Treasuries will provide the kind of strong returns they have in the past. Or even that returns will be positive.  It’s not even to say that they’re adequate on their own to diversify an allocation to stocks. Other asset classes, such as commodities (precious metals, industrial commodities) may be more important than ever as a third pillar of diversification.  But the case for US investors to completely omit Treasuries isn’t there.