Asset Allocation: Is It Really Different This Time?

    by Mark Hulbert

    This decade’s frustrating investment environment has led many to re-examine their core beliefs about how they should construct a portfolio.

    And while it is always healthy to engage in such a re-examination, the process tempts investors into proclaiming what, in effect, are the four most dangerous words in the investment world: “This time is different.”

    A case in point is the notion that there has been a fundamental shift in the relationships between the various asset classes. Because of the increased inter-connectedness of modern global markets, so the argument goes, assets that were previously uncorrelated have now become significantly correlated. As a result, traditional asset allocation strategies do not immunize investors from risks the way that they used to.

    The investment implication of this argument is that investors need to diversify more broadly than before, to go beyond stocks and bonds and add other asset classes, including commodities, currencies, hedge funds, and other alternative assets.

    For this column, I will remain agnostic on whether investing in these alternative asset classes is a good idea. Instead, I want to focus on what I believe are the faulty premises underlying the arguments being presented to support investing in them.

    And it turns out that I can find little evidence that the traditional asset classes are becoming more correlated.

    Correlations: The Last 20 Years

    I started by calculating the correlations between five different asset classes since mid-1980, which is when I began tracking the performance of investment newsletters. The benchmarks I used were as follows:

    • Domestic Stocks: The Dow Jones Wilshire 5000 Total Return index;
    • International Stocks: Morgan Stanley Capital International’s Europe Australia Far East (EAFE) Total Return index;
    • Domestic Bonds: The Shearson Lehman Treasury index (a total return index representing a composite of all U.S. Treasury securities having at least one year of maturity);
    • International Bonds: J.P. Morgan’s Global Government Bond index (ex-U.S.), which also is a total return index; and
    • Gold: The London P.M. Gold Fixing price for an ounce of gold bullion.

    When calculating the correlation between these various asset classes, one can focus on the last year, last five years, last decade, or even longer. And once the time period has been chosen, one has to choose how many sub-periods the overall period is to be broken into—monthly, quarterly, or yearly, for example. For this column, I looked at monthly observations over the trailing five-year period. (By the way, the conclusions I reach in this column would have been identical if I had focused on monthly observations over the trailing 10-year period.)

    Figure 1.
    Correlation Between U.S
    Stocks and Other
    Asset Classes

    Figure 1 reports the five-year correlation of monthly returns between stocks and the other four asset classes. Correlation figures, reported along the vertical axis, range from +1 to -1 and indicate how one asset class performed relative to the other. A figure of +1 would mean that the two classes zig and zag in perfect lockstep with each other, each going up and down at the same time and by the same magnitude; a figure of –1 would mean that the two asset classes are perfect mirror images of each other, with one zigging at precisely the same time and extent that the other is zagging. A 0 indicates there is no correlation between the two.

    The chart begins in mid-1985, the point at which my database has monthly returns for the trailing 60-month period. Each new month’s point on the chart represents the rolling forward of that trailing 60-month window. (Note that the plot for the correlation between U.S. stocks and international bonds does not begin until the early 1990s, since data for the international bond benchmark do not extend back to 1980.)

    A quick review of the chart does not show an across-the-board increase in correlation. With the exception of the correlation between domestic and international stocks—about which I will have more to say in a moment—the figures have either stayed more or less constant over the last six years, or actually declined.

    U.S. vs. Foreign Stocks

    That leaves just the increased correlation between U.S. and international stocks to support the view that something fundamental has changed in the relationships between various asset classes. On its face, this increase certainly appears to be dramatic. Over the 13-year period between 1985 and 1998, the correlation ranged from 0.25 to just below 0.6. Since 1998, in contrast, it has never gotten below 0.6, and has risen has high as 0.77.

    But it turns out that there is less here than meets the eye, for two reasons.

    First, as has been noted by researchers before, correlations tend to grow in the equity arena during bear markets. This is nothing new, but a feature that has appeared in the historical record for many decades. And, not surprisingly, as you can see from the chart, the jump in the correlation between U.S. and international stocks occurs in August 1998, which is when the Long Term Capital Management meltdown spooked equity markets worldwide, giving rise, in fact, to the term “Asian contagion.”

    The impact of the Long Term Capital Management saga on the correlation between U.S. and international stocks would have otherwise disappeared from the chart in 2003, since that chart looks at correlations over the trailing five years. But by 2003 the bear market in U.S. stocks between 2000 and 2002 was part of the trailing five-year period. These two factors together conspire to make it look as though there was a permanent shift in the relationship of U.S. and international stocks. In fact, however, this change can be explained in terms of patterns that have long been part of the financial landscape.

    And there is another factor to keep in mind when interpreting the big increase in the correlation between U.S. and international stocks: Much of the increase may be a function of focusing on monthly observations. A different picture would have emerged if I had focused on decade-long holding periods. The reason I didn’t do that is that, with 26 years of data, I would have at most about three independent observations. But a couple of years ago, several researchers did study the impact of internationally diversified stock portfolios for an investor with a 10-year holding period.

    The study, which is unpublished, is “International Diversification: Have We Missed the Forest Through the Trees?” Its authors are Clifford Asness, Robert Krail, and John Liew, all of whom are principals at AQR Capital Management in Greenwich, Connecticut. They found that for an investor focused on one-month holding periods, international diversification has little benefit. But an investor who is only worried about losses over 10-year horizons, such diversification does a wonderful job reducing risk—just as financial planners have been saying for years.

    What About Newsletters?

    When looking at the relationships between various market benchmarks, I can’t find the basis for concluding that the world has changed—that “this time is different.”

    As an additional reality check on this conclusion, I calculated the extent to which the U.S. stock market is correlated with investment newsletter portfolios. Might there be something different about the markets that show up only in the portfolios that advisers actually are recommending, as opposed to overall market benchmarks? In other words, are newsletters having a more difficult time diversifying? If that were true, one might expect to see an increase in the correlation between newsletter portfolios and the U.S. stock market.

    To find out, I relied on the Hulbert Financial Digest’s tracking of more than 500 model portfolios recommended by nearly 200 newsletters. For each month since 1980, I calculated the average performance of all portfolios the Hulbert Financial Digest tracked in that month. I then calculated the correlation between that series and the U.S. stock market, focusing on monthly observations over the trailing five-year period.

    Figure 2.
    Correlation Between U.S
    Stocks and Average
    Newsletter Portfolio

    The results appear in Figure 2. The correlation between newsletters’ model portfolios and the U.S. stock market has varied in a relatively small range over the last 20 years, between 0.80 on the low side to 0.95 on the high side. I can detect no significant increase in this correlation in recent years.

    To be sure, the chart shows that the average newsletter is not particularly diversified out of U.S. stocks. But it has always been this way. The high correlation between the average newsletter’s portfolio and the U.S. stock market is a function of the newsletter editors’ desire to be heavily invested in that asset class, not any increased difficulty in recent years of diversifying away the risks of that asset class.


    The bottom line? I don’t see any justification for concluding that “things are different this time.” The asset classes appear to be inter-related to each other no differently in recent years than in prior decades. My hunch is that the argument that “things are different” is coming, at least in part, from the purveyors of the alternative asset classes. Needless to say, they have a vested interest in our believing that things are different, since if that is true investors should be investing a lot more in those alternatives.

→ Mark Hulbert