You are on page 1of 4

Portfolio Diversification:

Where It Goes Wrong


The way most portfolios are constructed violates
the very basic premises of modern portfolio theory.

By John Davenport

18 Financial Planning • Disclosures • July/August


Diversification is one of the fundamental ideas similarly painful losses. As mentioned, the cry has
in investing. However, the idea of diversification as grown louder with 2008 serving as proof that asset
we know it in the financial services industry is only allocation does not work.
a little over half a decade old. I would argue, however, that it was not that asset
The concept of asset allocation, in which one allocation did not work, it’s that the way most port-
builds a portfolio of different investments in order folios are constructed violate the very basic premises
to maximize return and minimize risk, was originally of modern portfolio theory.
proposed by Harry Markowitz in 1952 in his classic In particular, most portfolios are constructed of
paper “Portfolio Selection.” highly correlated asset classes. Remember, one of
Since that initial paper, asset allocation, which the core tenets of modern portfolio theory is that
is formally known as Modern Portfolio Theory the portfolio comprises assets that are not highly
(MPT), has become the very backbone of portfolio correlated to one another. If you look at Table 2,
construction in the financial services industry. The you will see that with respect to correlations, the
belief is that by combining assets that are not highly assets that make up most portfolios provide little in
correlated to one another (meaning that they act the way of real diversification.
differently than one another in response to market As you will note, every one of the asset classes
forces), one can help control risk. represented in Table 2 are positively correlated to
However, following years like 2008, public the S&P 500. All domestic equity asset classes are
outcry arises, pronouncing that asset allocation and more than 80 percent correlated to the S&P 500.
buy-and-hold portfolio strategies are dead. International equities (as represented by the MSCI
It’s not hard to concede these arguments if you EAFE) are more than 60 percent correlated and
bonds are still positively correlated at 24 percent.
This means that all the elements of the portfolio
are highly dependent on the S&P 500 experiencing
positive returns to also perform well.
Without becoming too academic, there are three
basic flaws in the way that traditional portfolios are
constructed that result in portfolios behaving as they
did in 2008.
The first flaw is the process of selecting and ex-
cluding various asset classes. The second element
is the artificial constraints that are then placed on
the included asset classes. The third element con-
cerns the constraint to long-only investing in most
portfolios.

Asset class selection


In Modern Portfolio Theory, portfolios should
be constructed of a variety of non-correlating asset
classes. However, when looking at the distribution
of investments across asset classes in most retail port-
folios, the vast majority of assets are concentrated
in large cap, mid cap, small cap, international and
aggregate bonds (ICI Factbook, 2007).
According to the Investment Company Institute,
at the close of 2007, approximately 40 percent of
U.S. mutual fund assets were in domestic stock
funds. Twenty-six percent of assets were in money
look at how most portfolios are constructed. Look- market funds, while international equities and bond
ing at the major asset classes included in the typical funds each represented 14 percent of U.S. mutual
portfolio, we see that nearly every one of the asset fund assets.
classes posted negative returns in 2008 (see Table Many in the industry would call those distribu-
1). A similar phenomenon happened in 2001 and tions a well-diversified portfolio. However, given
2002, in which most asset classes posted negative their high correlations to one another, the resulting
returns. portfolios constructed using these basic asset alloca-
Given the negative returns in most of the major tions are then by default directionally dependent
asset classes, client portfolios as a whole suffered on the S&P for its returns. Essentially, these w

Financial Planning • Disclosures • July/August 19


portfolios are not as diversified as they may the system. It is not constructing an efficient First, only a few asset classes are used
appear. frontier based on all available investment for consideration in the optimizer. Then the
The underlying problem with this gen- choices. It is, rather, optimizing a portfolio optimizer is told to put limits on how much
eralized approach to asset allocation is the of preferences. can be invested in any one of those select few
exclusion of numerous other viable asset asset classes. Again, this is more in line with
classes (commodities, currencies, managed creating portfolios of preference rather than
futures, structured notes, long/short equi- Artificial constraints process.
ties, inflation-linked assets, etc.), which can The second issue that distorts the results As stated above, most asset allocation
increase the diversification of the portfolio. of optimizing portfolios is the practice of ap- portfolios for U.S. investors will be largely
Many of these “non-traditional” asset classes plying artificial constraints to the asset classes biased to the domestic asset classes (which
offer much lower correlations to both the S&P that have been included in the optimizer. If also biases the portfolio to the U.S. dollar
500 and the other traditional asset classes. you think through how portfolios are being — but that’s another discussion), with in-
Eliminating available asset classes forces constructed, you can begin to see the true ternational investing relegated to a smaller,
the optimization process to create an “opti- limitations that are being placed on the ability “satellite” allocation. However, researchers
mal” solution of only those things put into to optimize the portfolio. have demonstrated that when an uncon-
strained optimization is performed, even the
most conservative of portfolios will allocate
Table 1: A typical portfolio
portions of the portfolio to the most risky asset
2000 2001 2002 2003 2004 2005 2006 2007 2008
classes (small cap and international equities).
S&P 500 -9.1 -11.89 -22.1 28.68 10.88 4.91 15.79 5.49 -37
They also note that in an unconstrained opti-
Russell 2000 -3.02 2.49 -20.48 47.25 18.33 4.55 18.37 -1.57 -33.79
mization, large cap U.S. stocks are relegated
Russell Mid Cap 8.25 -5.62 -16.18 40.06 20.22 12.65 15.26 5.6 -41.46
to a supporting role.
MSCI EAFE -14.17 -21.44 -15.94 38.59 20.25 13.54 26.34 11.17 -43.38
This is often the exact opposite of how
BarCap US Agg Bond 11.63 8.44 10.26 4.1 4.34 2.43 4.33 6.97 5.24
client assets are deployed. The artificial con-
Russell 2000 Growth -22.43 -9.23 -30.26 48.54 14.31 4.15 13.35 7.05 -38.54
straints that are applied in the optimization
Russell 2000 Value 22.83 14.03 -11.43 46.03 22.25 4.71 23.48 -9.78 -28.92
process will essentially dictate the outcomes
Russell 1000 Value TR 7.01 -5.59 -15.52 30.03 16.49 7.05 22.25 -0.17 -36.85
in line with the constraints, not along a true
Russell 1000 Growth -22.42 -20.42 -27.89 29.75 6.3 5.26 9.07 11.81 -38.44
ML US HY Master II -5.12 4.48 -1.89 28.15 10.87 2.74 11.72 2.24 -26.39
efficient frontier. This process essentially
constructs portfolios along the lines of the
Table 2: Where’s the diversification? investor’s or advisor’s preferences, not in
1/1979-7/2009 S&P 500 Russell Russell Russell Russell Russell MSCI US Agg Wilshire accordance with a true mean-variance opti-
1000 1000 Mid Cap 2000 2000 EAFE Bond US REIT mization.
Growth Value Growth Value As an example, an unconstrained portfolio
S&P 500 1 built to mimic the standard deviation (risk)
Russell 1000 Growth 0.96 1.00 of the S&P 500 would be constructed of
Russell 1000 Value 0.95 0.83 1.00
small cap equities and bonds. Using historical
Russell Mid Cap 0.94 0.91 0.91 1.00 returns and standard deviations, neither the
Russell 2000 Growth 0.81 0.86 0.72 0.91 1.00 S&P 500 nor the MSCI World Index lies on
Russell 2000 Value 0.80 0.74 0.84 0.91 0.87 1.00 the efficient frontier so they are completely
MSCI EAFE 0.63 0.60 0.61 0.62 0.56 0.56 1.00 excluded from the unconstrained optimized
BarCap US Agg Bond 0.24 0.21 0.26 0.24 0.11 0.19 0.18 1.00 portfolio (see Table 3).
Wilshire US REIT 0.55 0.48 0.63 0.65 0.56 0.75 0.44 0.19 1.00 Most investors would simply be uncom-
fortable with a portfolio made up of only
Table 3: Unconstrained optimized portfolio bonds and small caps. “Common sense” dic-
tates that they further diversify the portfolio.
However, most of the time investors will not
include additional asset classes to solve for the
efficient frontier; they will simply apply con-
straints to the existing asset classes, restricting
allocations to those perceived as riskier.
Following this logic, if the efficient fron-
tier is constrained with a limit of 10 percent
to small caps, the resulting portfolio is then
forced to include large cap and international
stocks — asset classes that do not reside on
our original efficient frontier.
Further, this process of forcing large cap
and international stocks into the portfolio

20 Practice
Financial
Management
Planning • Disclosures • July/August
moves the efficient frontier down and to the Table 4: Investing comparisons
right on the risk return spectrum, assuming
Common Period 2/1993-1/2009 Geometric Mean Standard Deviation Alpha
more standard deviation (risk) for less return. (%) (%) (%)
However, in the eyes of many investors and
Hennessee HF Market Neutral TR USD 5.21 3.96 4.7948
advisors, this portfolio is more “palatable”
Hennessee HF Long/Short Equity TR USD 9.11 8.68 6.6373
because it seems more “diversified.”
S&P 500 TR 6.04 15.81 0.0006
This process of applying constraints may
result in portfolios that are more intuitively
appealing, but they are by no means an op-
timal solution. Unfortunately, this process Table 5: A superior risk/return profile to the S&P 500
of portfolio construction is to force into the
portfolio the existing set of asset classes,
whether or not they offer a compelling risk/
return trade-off.

Long-only investing
Most portfolios are constrained to long-
only investments, which by default excludes
the use of short sales within the portfolio.
Despite the evidence that short sales can im-
prove the risk/return profile of the optimal
investment mix, strategies that employ short
sales are allocated very little space within the
vast majority of portfolios. This can impede
the ultimate efficacy of a portfolio; limiting
portfolios to long-only investments does
not allow portfolios to take advantage of all annualized return over the same period with gies increase portfolio risk in spite of the
available market information. a standard deviation of 3.96. The long/short evidence to the contrary. This is clearly seen
In a traditional long-only investing ap- index delivered a return of 9.11 percent with in the portfolio construction behaviors that
proach, the investor only has two options: a standard deviation of 8.68. Both indices tend to rely more on ‘intuition’ than sci-
hold a security or not hold a security. Even in delivered similar to better returns with a ence. In investing, we attempt to maximize
the face of compelling negative evidence (bad fraction of the risk (see Table 4). return while minimizing risk. However,
news), the only thing the long-only investor What happens to portfolios when these the processes by which most retail investor
can do to take advantage of adverse news is three constraints are removed from the opti- portfolios are constructed are counter to this
to not hold the security (sell it). However, mization process? In Table 5, the available as- basic premise.
by removing the long-only constraint and set classes include shorting strategies and asset By highlighting these behavioral flaws in
allowing short sales, the investor can now classes that present a better correlation profile the portfolio construction process, perhaps
also short a security in the face of negative to the broad market and bond asset classes. investors can return to the true essence of
news and has the opportunity to realize an The new array of asset classes in addition to modern portfolio theory and attempt to
absolute positive return on the trade even the domestic asset classes includes long/short build truly diversified portfolios. 
though the stock or security may decline in equity strategies, commodities, managed
price. Again, the ability to employ shorting futures and market neutral strategies.
strategies allows investors to take advantage The resulting efficient frontier allows us
of all available information. to construct a portfolio of long/short, man-
The historical returns from the Ibbotson aged futures and the aggregate bond index
database also suggest that the inclusion of that delivers a far superior return to the S&P John Davenport is
shorting strategies within a portfolio could 500 with a standard deviation more in line director of research with
improve the risk/return profile of an overall with the aggregate bond index. As Table 5 the Actuarial Consulting
portfolio. Comparing the S&P 500 Index, indicates, the domestic equity asset classes Group in Midlothian.
which is a long-only index, to indices that that have garnered the bulk of U.S. mutual He is also an adjunct professor,
do allow shorting of stocks in the large cap fund assets in traditional portfolios do not teaching finance, investing and
universe, we can see the ability of short sales reside on the efficient frontier and thus are other business courses for several
to shift the risk profile of a portfolio. excluded from the portfolio. local colleges. Additionally, John
From February 1993 through January is a doctoral student. Contact him
2009, the S&P 500 Index posted an aver- at jdavenport@acgworldwide.com
age annualized return of 6.04 percent with Returning to diversification or www.linkedin.com/pub/john-
a standard deviation of 15.81. The market Many advisors have the perception that davenport/5/744/593.
neutral index posted a 5.21 percent average alternative investments and shorting strate-

Practice Management
Financial Planning • Disclosures
• Disclosures
• July/August
• July/August 21

You might also like