As promised, we are posting the results of our recent time-series analysis of the correlations among global asset categories.

Each exhibit shows the average weekly correlations of two asset classes over the most recent calendar year (as in the previous post, the data is current through mid-February). The time series analyzed represent a global selection of assets in each category (e.g., global stocks, global bonds, a broad basket of world currencies and commodities, and even worldwide property assets). Click on a thumbnail below to view the full image.

Stocks with bonds

Stocks with commodities

Stocks with real estate

Stocks with currencies

Bonds with commodities

Bonds with real estate

Bonds with currencies

Commodities with real estate

Commodities with currencies

Real estate with currencies

World Enough and Time

These are “three-dimensional” charts, in the sense that they show the relationships among three variables at once (and thus have three axes). One axis is correlation, which we graph across the left-hand vertical.

There are two time axes. The one along the bottom (the horizontal) speaks to holding period — here corresponding to how long one typically maintains a position. Each chart supports comparisons for holding periods ranging in duration from one day to a full calendar year. A swing-trading strategy, for example, may call for one (one investor, portfolio, managed account, etc.) to exit a position completely after hours, days, or weeks. Portfolios et cetera with lower turnover hold positions for quarters or years. We use the horizontal axis to determine what holding periods of different length mean to asset-category correlations.

The vertical time axis (on the right-hand side) speaks to our comparison or “look-back” period. That is, are we measuring correlation over the previous week? Six months? It is fine for me to tell you that “the average temperature in New York is 55°,” but you may wonder to what period I am referring — the average for March? Annually? Since the Pleistocene? This is what the z-axis tells you.

In combination, the use of two time axes allows us to make statements like, “global bonds and currencies demonstrate a three-month correlation of about 50% over a six-month holding period,” or “commodities and real estate have shown a correlation of -43% over the last year.” (Note that the previous statement assumes our holding period is the span over which we want to measure correlation, though we could have used the self-same chart to state average weekly correlations over a one-year, or any, holding period.)

Of course, we are measuring directional relationships among asset categories, not positions. So while it is difficult to imagine any investor switching among these categories with the deftness of a position trader, all positions fall into one or another such category, and might be expected to demonstrate some (typically high) degree of correlation with like assets, eo ipso.

We have tried to select the most revealing angles of rotation and y-axis demarcations for each study.

Some Buy the Low Tide

Notice how the waves or “blankets” of data tend to ripple out from most of the exhibit’s left-hand sides, settling down as they extend rightward or back in time. This indicates that those asset categories tend to move in tandem over short periods, and that this tendency declines to the point where different classes may actually move in opposite directions over sufficiently long holding and/or look-back periods. Low or even negative correlations among investable assets is how one begins to diversify — to distribute factor risk among a set of determinants (asset class, sector, geography, time period, and so on) which are themselves, like family, not of one’s choosing but come with the territory.

Now Therefore…

We entertained some conclusions based on these studies in our earlier post. We have a long drive ahead of us tomorrow; it is getting late; and as Homer Simpson says of the Blue Man Group, “don’t get me started!” (That’s also what we say about Blue Man Group — it’s not about the colors, you see, it’s a caricature of groupthink and modern structures that…)

You can draw your own conclusions about the benefits of a broad, global allocation strategy maintained over long periods with low portfolio turnover. You can, we hope, draw all sorts of conclusions from these studies, the graphical depiction of which is so much more valuable than tinkering around in the excruciatingly vast data sets on which they are based.