Practical analysis for investment professionals
07 February 2019

Diversification: High Dispersion Beats Low Correlation

When advisers talk about diversification, their go-to variable is correlation. Finding an asset with low correlation with equities and bonds is a key consideration of every asset allocator and chief investment officer (CIO) of an institution. But correlation is not everything.

What really matters for diversification is dispersion.

A big concern of my friends who advise US clients these days is that their clients are losing faith in international diversification. After more than a decade of outperformance by US stocks, US investors are wondering why they should even bother with European or emerging market equities. European investors, on the other hand, are tempted to shift more and more of their equity allocation toward the United States. After all, the US market seems to be the only one doing well.

So when advisers or consultants show up and tell investors that international diversification is beneficial in the long run, they either quote John Maynard Keynes or suggest that the correlation between US stocks and those of other developed markets is so high that there is little benefit to diversifying.

As an example, over the last 100 years, the correlation between US and UK stocks was a whopping 0.7 — not perfectly aligned but close enough to warrant an arbitrage position from an arb hedge fund. Not much diversification benefit to harvest from UK stocks if you are a US investor, is there?

Subscribe Button

The chart below shows the difference in rolling 10-year total returns for US and UK stocks since 1920. Whenever the line is above zero, US stocks outperformed their UK counterparts over the past 10 years. When the line is below the zero, the opposite occurred.

US vs. UK Stocks: Annual 10-Year Total Return Differentials

US vs. UK Stocks: 10-Year Total Return Differentials

Source: Fidante Capital

While the average performance difference between the two markets over the last 100 years is zero, the 10-year return differentials can be extremely large. From June 1942 to June 1952, for example, US equities eclipsed UK stocks by 11% per year. For 10 years! That adds up to 185% outperformance over a decade. This was clearly driven by the destruction World War II inflicted on Europe and the United Kingdom and the stimulative impact it had on the United States. But even if we ignore World War II, there are 10-year periods when US equities well outpaced those of the United Kingdom. The 5.6% outperformance the US market enjoyed in the last 10 years is hardly exceptional.

On the other hand, there are plenty of intervals when UK equities outperformed by a wide margin. During the high-inflation era from 1975 to 1985, the UK market beat the US market by 17% per year (!) thanks to its tilt toward energy and mining companies.

This range of outcomes is what we think of when we talk about dispersion. While there is no difference in the two markets’ average performance, the one standard deviation return difference between them is 4.4% per year — or 54% after 10 years.

That’s why the focus on correlation over dispersion is shortsighted. It ignores the material diversification benefits that can be achieved with assets that seem to move in tandem even though the steepness of the trend might vary considerably, giving rise to significant performance differences.

Dispersion is also forgotten by those who dismiss return forecasts anticipating wide variations between US and non-US equities in the future. Their typical argument is that the United States dominates global equity markets, so if the US stock market has low returns, then European and emerging markets cannot have high returns. The comparison between US and UK equities over the last century shows how faulty this logic is and why we can reasonably expect big return differentials between US and European or emerging market stocks in the years ahead.

For more from Joachim Klement, CFA, don’t miss Risk Profiling and Tolerance: Insights for the Private Wealth Manager, from the CFA Institute Research Foundation, and sign up for his regular commentary at Klement on Investing.

If you liked this post, don’t forget to subscribe to the Enterprising Investor.

All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer

Image credit: © Getty Images/chaluk

About the Author(s)
Joachim Klement, CFA

Joachim Klement, CFA, is Head of Investment Research at Fidante Capital and a trustee of the CFA Institute Research Foundation. Previously, he was CIO at Wellershoff & Partners Ltd., and before that, head of the UBS Wealth Management Strategic Research team and head of equity strategy for UBS Wealth Management. Klement studied mathematics and physics at the Swiss Federal Institute of Technology (ETH), Zurich, Switzerland and Madrid, Spain, and graduated with a master’s degree in mathematics. In addition, he holds a master’s degree in economics and finance.

8 thoughts on “Diversification: High Dispersion Beats Low Correlation”

  1. Alec Ferguson says:

    Wonderful article, thank you for the insight. I suppose p is a popular metric in portfolio management because of its simplicity in usage. It’s a single metric that wraps up historical pair-wise return relationships and while imperfect, does guide investors decisions meaningfully based on past performance of assets.

    My question would be how do you account for dispersion in a single metric? Dispersion in fixed income cash flows is accounted for in the convexity figure, but I assume these dispersions are not created equal(?). Any thoughts on ratios/metrics/statistics one could use to demonstrate equity portfolio dispersion?

    Thank you

  2. Arka Bose says:

    The post is good, but how do you go about it?
    Moreover, I believe looking at last 3-5 years of correlation will accommodate the dispersion you are talking about?

    1. Norbert says:

      Why should the importance of high dispersion for diversification be pointed out in contrast to low correlation, if longer correlation intervals would serve the same purpose? Two assets of similar risk and long-term return levels, which mostly move into the same direction but with currently different rates, have a high correlation over time periods of all lengths but a high dispersion of returns in the medium-term at the same time. This can be observed with US and international equities since the last crisis. Thus, it is beneficial to allocate both assets for better diversification, and, thus, for stabilizing the portfolio and for generating rebalancing alpha from their divergence in the medium-term. However, this only holds true if mean reversion can be expected with a high probability for the considered assets.

      “How do you go about it?” Assets of different world regions and of other different equity market sectors, such as Large and Small Cap or Growth and Value, as well as of alternatives, such as REITs and Hedge Funds, can fulfill these requirements for diversification with high dispersion. Try to combine anything that is sure to have similar risk and return levels in the long-term as well as high dispersion of returns in the medium-term, taking turns due to mean reversion like the assets I mentioned. I would avoid individual industries and countries maybe with the exception of the USA (I am German ;-). Because mean reversion is less certain for them.

      1. Arka Bose says:

        Correlation shift does occur and hence looking at returns of smaller period like 3-5 years should take care of this problem? After all its all about asset allocation and you allocate the asset based upon your model of risk vs return.

        1. Norbert says:

          No, I am afraid that return differences of smaller periods like 3-5 years is just statistical dispersion. RTM effects may already take much longer than that to set in, usually around 10 up to 20 years just for a half cycle.

          Also statistically, small deviations need many decades or an ordinary investor‘s life time in order to become sufficiently significant and to support a sound prognosis of the further fate for assets below world regions without betting or even guessing.

          Thus, sound asset allocation may well include world regions in order to harvest rebalancing alpha exploiting RTM. But at least for me it definitely ends there without much better insights than the best active managers taking similar bets with long successful track records.

  3. James Rich says:

    My impression is that most published correlations are for one-month periods. Perhaps useful for short-term planning horizons but not so useful for intermediate and longer term horizons.

    I have calculated very different correlations in the past based upon various horizons – and much investment planning should probably be based upon longer term horizons.

  4. Micah says:

    Excellent article. I have purposely taken this route to higher dispersion vs low correlation but never knew how to put it into words. We run an all equity portfolio. Therefore, some of the asset classes we invest in are highly correlated, yet, we are able to frequently perform tactical asset reallocation (arbitrage) due to our high dispersion rate. Thank you for putting into words a concept that should be readily evident to many investors.

  5. Norbert says:

    This excellent explanation of dispersion and it’s little understood function and high importance in portfolio construction is to be commended. However, in order to provide for some protection of the portfolio from severe crisis declines, low correlation of assets in the short-term is at least as important for diversification and for overall performance in the long-term as high dispersion is in the medium-term. This can be well achieved by alternatives with low correlation, particularly well by pure alpha investments such as market neutral and trend-following Managed Futures strategies from CTAs, which can both go long and short with leverage. The short-term divergence to beta of these uncorrelated pure alpha instruments during severe crises can stabilize the portfolio in the short-term well. In addition, they can also contribute at least as much rebalancing alpha in the long-term due to their mostly high dispersion as beta instruments with high dispersion but high correlation can do.

    Therefore, I would consider low correlation at least as important for diversification as high dispersion. Only if both are utilized in a balanced way, the portfolio can be sufficiently stabilized over all relevant time periods. And a higher possible risk-adjusted excess return, compared to the proper benchmark index, can be achieved in the long-term at the same time.

    It makes no sense to me to mainly manage rather benign medium-term risks by focussing on assets with high dispersion, but paying less attention to actually much more severe tail risks, where a similar allocation to assets with appropriately low correlations would be very beneficial for both short-term risk and long-term return. Particularly, the highest short-term risk with devastating long-term consequences for clients, leaving the supposedly sinking ship near the bottom of an unexpectedly severe crisis due to panic, if their portfolio values decline more than tolerated due to insufficient crisis protection and/or due to usually overestimated risk tolerances, would then be neglected!

    These severe short-term risks can and should be appropriately reduced and pure alpha returns harvested with at least the same effort as other longer term risks and returns by sufficient diversification through a balanced allocation of assets with low correlation as well as with high dispersion, respectively. Similarly, excellent enterprises pay at least equal attention to surviving in the short-term as well as to growing more then their industry benchmarks in the long-term by balancing all their risks from short- to long-term with best practices in order to assure a sustainable development.

Leave a Reply

Your email address will not be published. Required fields are marked *