AM Non-Normality of Market Returns

Non-normality of Market Returns
A framework for asset allocation decision-making

About J.P. Morgan Asset Management—Strategic Investment Advisory Group
The Strategic Investment Advisory Group (SIAG) partners with clients to develop objective, thoughtful solutions
to the broad investment policy issues faced by corporate and public defined benefit pension plans, insurance
companies, endowments and foundations. Our global team is one of J.P. Morgan’s primary centers for thought
leadership and advisory services for institutional clients in the areas of asset allocation, pension finance and
risk management.
The team’s expertise is supported by powerful analytical capabilities for conducting asset/liability, risk
budgeting and optimal asset allocation analysis, in line with client-specific investment guidelines, risk tolerance
and return requirements. In response to the changing needs of CFOs, treasurers and CIOs, our suite of tools
has been expanded to include corporate finance-based risk management analytics for assessing and
proactively managing the impact of the pension plan on the corporation as a whole.
SIAG brings a deep knowledge and understanding of capital markets behavior to all its advisory services,
ensuring that results and recommendations have real-world consistency and can be tested under a variety
of market scenarios. Strategic Investment Advisory Group services are offered as part of an overall asset
management relationship to complement the many ways in which J.P. Morgan Asset Management provides
clients with value-added insights.
About J.P. Morgan Asset Management

For more than a century, institutional investors have turned to J.P. Morgan Asset Management to skillfully
manage their investment assets. This legacy of trusted partnership has been built on a promise to put client
interests ahead of our own, to generate original insight, and to translate that insight into results.
Today, our advice, insight and intellectual capital drive a growing array of innovative strategies that span
U.S., international and global opportunities in equity, fixed income, real estate, private equity, hedge funds,
infrastructure and asset allocation.
Foreword
When market stress and disruptions hit portfolios period2 and normally distributed. In reality, we can empirically
observe that in many cases returns are not independent, and
especially hard, it is natural for investors to ask
in all cases they are not normally distributed.
whether anything could have been done differ-
ently. We revisit our tools and strategies looking • Inadequate risk measures: If one adopts an asset allocation
framework that incorporates non-normality, then standard
for ways to improve them. And often, it is in such deviation becomes ineffective as the primary quantifier of
times of reflection that we arrive at important portfolio risk.
insights and begin to craft effective solutions.
Using the latest statistical methods, however, we believe that
The turmoil of 2007/2008 is just one among many such these shortcomings can be addressed. And based on the results
financial crises over the past 30 years—rare and unpredictable presented here, we argue that a modified risk framework may
events that can, and should, challenge conventional ideas about help investors improve portfolio efficiency and resiliency.
portfolio construction. Of course, no single statistical framework—no matter how
Of course, we can never know in advance when such events will robust—can render portfolios immune to such extreme events.
occur and how bad they will be. We do know, however, that we Even the most important concepts and discoveries in finance
empirically observe such events with much greater frequency continue to evolve through a process of dialogue and
than current models allow for. So, the proper line of questioning refinement.
for investors becomes: Can current risk management frame- But our hope is that the work presented here will be an ef-
works be modified to better capture the long-term, downside fective step toward confronting and clarifying some critically
risk associated with these rare but dangerous anomalies—i.e. unresolved issues in asset allocation—specifically, the habit of
frameworks that take into account more than just one-standard- sacrificing accuracy in favor of ease-of-use when addressing the
deviation event, and better reflect their observed frequency? issue of non-normality of returns. This trade-off is potentially
The goal of this paper is to present such a revised framework. very harmful as it understates portfolio risk, and given advance-
ments in statistical techniques, it is also no longer necessary.
In broad scope, we believe that two specific weaknesses in con-
ventional risk assessment may be contributing to a quantifiable The challenge of asset allocation is one of both art and science,
underestimation of portfolio risk: and to this end we hope that this work will prompt both discus-
sion and further research, all geared toward helping investors
• Frameworks assuming “normality”1: Conventional asset better address the most pressing issues of the day. We appreci-
allocation frameworks typically make a range of assumptions ate all comments, questions, and contributions that this paper
about the “normality” of asset returns, the most problematic might prompt from clients and colleagues alike.
of which are that returns are independent from period to
1
The normal distribution—recognizable by its bell shaped probability density function—
is a statistical distribution commonly used to model asset class returns in traditional
Mean-Variance Optimization frameworks.
Abdullah Z. Sheikh, FIA, FSA
2
Though “independence” is not in strict statistical terms a form of “normality”, we include it
here because the assumption of “independence” is one of the central tenets of conventional Director of Research
asset allocation frameworks built around the concept of “normality” of asset returns.
Strategic Investment Advisory Group
J.P. Morgan Asset Management | 1

Table of Contents
1 Foreword
3 Overview
7 Empirical evidence of non-normality

in asset returns
14 Incorporating non-normality into an asset

allocation framework
24 Estimating downside portfolio risk in a

non-normal framework
27 Implications of non-normality on optimal

portfolio solutions
31 Conclusion: Incorporating non-normality

can lead to more efficient portfolios
32 Appendix
2 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

Overview
Non-Normality and Its Impact on Portfolio Risk which they can be implemented. To implement an asset alloca-
As sudden market disruptions go, the events of the past tion framework based on normal asset return distributions,
year—the sub-prime mortgage crisis and ensuing global financial practitioners need only make two assumptions for each asset
meltdown—are not without precedent. Just in the last three class (mean and standard deviation) and one assumption for
decades, investors have been faced with a number of financial each pair (co-variance). The latter applies because one implicitly
crises: assumes the relationship between each pair of asset classes is
linear (another problematic issue discussed later in the paper).
• Latin American debt crisis in the early 1980s
• Stock market crash of 1987 But what if asset returns are not normally distributed?
• U.S. Savings & Loan crisis in 1989–1991 In fact, in the real world we can observe empirically that returns
• Western European exchange rate mechanism crisis in 1992 are not normally distributed. This leads us to ask: How would
• East Asian financial crisis of 1997 incorporating non-normality affect the strategic asset allocation
• Russian default crisis and the LTCM Hedge Fund crisis in 1998 process?
• Bursting of U.S. Technology bubble in 2000-2001
While these anomalous events are rare, we observe such Empirically Identifying Non-Normality
extreme “non-normality” in real-world markets more frequently
Each of the market events mentioned above was due to an
than current risk management approaches allow for. Said
incidence of non-normality of one sort or another3. And the
another way, we believe that conventionally derived portfolios
practical implication is that incorporating non-normality into
carry a higher level of downside risk than many investors
the asset allocation process would force recognition of greater
believe, or current portfolio modeling techniques can identify.
downside risk to the portfolio, precisely from such extreme,
The primary reason for this underestimation of risk lies in the unexpected negative events.
conventional approach to applying mean-variance theory, which
Our focus here is on capturing the impact of non-normality
was pioneered by Harry Markowitz in 1952. Traditional mean-
on downside portfolio risk—as well as the asset allocation/
variance frameworks have become the bedrock of top-down
optimization process—rather than identifying the specific sources
asset allocation decision making. As suggested above, a standard
of non-normality itself (although we do discuss economic or
assumption in the mean-variance framework, and indeed many
behavioral factors where relevant).
other holistic asset allocation frameworks, is that future asset
class returns will be independent from period to period and Specifically, we test seven asset classes4 and empirically confirm
normally distributed. three primary categories of non-normality:
Despite being widely recognized as overly simplistic, such broad • Serial Correlation: A critical pillar of many traditional asset
assumptions of “normality” have appeal due to the ease with allocation frameworks (i.e. frameworks built on a premise
of “normality”), is the assumption that asset returns from
3
In this context, the converse is also true, i.e. empirical observations of non-normality are period to period are independent and identically distributed.
the result of extreme market scenarios. However, if one month’s return is ‘influenced’5 by the previous
4
Asset classes include U.S. Aggregate Bonds, U.S. Large Cap Equity, International Equity,
Emerging Markets Equity, Real Estate Investment Trusts, Hedge Fund of Funds and Private month’s return, then there may be a need to account for this
Equity. effect in future asset projections. Typically, traditional asset
5
“Influenced” in this context refers to a statistically significant coefficient when one month’s
return is regressed against the previous month’s return.
allocation frameworks do not allow for serial correlation,

but we identify serial correlation in four of the seven asset An asset allocation framework based on the normal distribu-
classes we model. tion would understate both the frequency and magnitude of
extreme negative events, as well as their potential effects on
Serial correlation, if not adjusted for in the underlying data, portfolio returns and efficiency.
masks true asset class volatility and biases risk estimates down-
wards, leading to underestimation of overall portfolio risk. • Correlation Breakdown: The simple correlations often used
in traditional asset allocation models assume a linear rela-
• “Fat” Left Tails (Negative Skewness and Leptokurtosis): Our tionship between asset classes—i.e. they assume that the
second form of non-normality relates to observing negative relationship between the variables at the extremes is similar
returns in greater magnitude and with a higher probability to their relationship at less extreme values. Using simple
than implied by the normal distribution. This phenomenon linear correlations is the equivalent of assuming that the
is commonly referred to as “fat” left tails. ‘joint’ distribution of asset returns is (multivariate) normal.
Exhibit 1 illustrates this phenomenon for monthly dollar-hedged Joint distributions capture how asset classes behave together
International Equity returns over the ten years to October 2008. rather than individually.
The chart plots the empirical (i.e. observed) probability density However, we find that in many cases correlations under extreme
function of the data relative to a normal distribution. conditions are quite different than under normal conditions. In
One can see that the observed return series (blue line) is more other words, the expected linear correlations breakdown and
peaked, has a higher density at the extreme left, and leans asset classes exhibit quite different joint behavior. The relation-
further to the right than the normal distribution (orange line). ships, in fact, are not linear and the assumption of linearity
(by using linear correlation matrices) underestimates the
The rightward lean is its “negative skewness”. The consequence probability of joint negative returns under extreme conditions.
of this skewness is that the left slope of the blue line is longer
than the left slope of the orange line—i.e. it has a longer tail, Relying on linear correlation matrices tends to overestimate
which indicates a greater magnitude of extreme negative events. the benefits of portfolio diversification during periods of high
market volatility. This leads to a systematic underestimation of
In addition, the blue line is taller at its apex and shows a higher downside portfolio risk.
density at the extreme left end (i.e. leptokurtosis). In particular,
the higher density at the left tail indicates a higher probability
of extreme negative events. Statistical Approaches for Incorporating Non-Normality
Fortunately, we have at our disposal sophisticated statistical
tools that allow us to correct for these types of non-normality.
• Unsmoothing Serial Correlation: Serial correlations can be

Exhibit 1: International Equities—“Fat” Left Tails in historical returns
“unsmoothed”. That is, we can correct for the influence of
12
Empirical
prior-period returns and restore independence to single-
10 period returns. To arrive at our new adjusted return series
Normal
we apply a variation of Fisher-Geltner-Webb’s well-established
8
“unsmoothing” methodology6. Our ‘new’ adjusted return
Density
6 series is better reflective of the risk characteristics of the

asset class. Notably, the new unsmoothed return series has
4
“Fat” left tails the same mean as our original return series, but shows a
2 higher volatility, and thus higher downside risk.
0
-20% -15% -10% -5% 0% 5% 10% 15% 20% 6
For a more detailed treatment, please refer to Fisher, J., D. Geltner, and B. Webb. 1994.
Monthly return Value Indices of Commercial Real Estate: A Comparison of Index Construction Methods.
Also, Fisher, J. and D. Geltner. 2000. De-Lagging the NCREIF Index: Transaction Prices and
Source: J.P. Morgan Asset Management. For illustrative purposes only. Reverse-Engineering.

• Modeling “Fat” Left Tails (Negative Skewness and Lepto- are mathematical functions that allow us to model the joint
kurtosis): “Fat” left tails can be addressed using Extreme distribution of asset returns separately from the marginal
Value theory—a body of work specifically designed to look at (i.e. individual asset class) distributions.
the probability of high-risk, but low-probability, events such
as floods, earthquakes and large insurance losses. In other • By considering joint distributions, we turn our focus to how
words, its focus is estimating tail risk. asset classes behave together rather than individually. In
particular, copulas allow us to model an increased incidence
By applying Extreme Value theory we can create asset return of joint negative returns (i.e. the “fatter” joint left tails) in
distribution models that are a closer “fit” to the return series we our simulation results, just as we observe empirically in
observe empirically, much more similar than normal distributions. real-world market data. And again, using a more accurate
Exhibit 2 shows a probability density function for dollar-hedged proxy for observed behavior results in recognizing higher
International Equities, calculated using Extreme Value Theory downside portfolio risk, specifically due to increased depen-
(orange line). dence of asset returns during periods of market stress.
Exhibit 2: International Equities—Applying Extreme Value theory

for a better fit
A Better Risk Quantifier: Conditional Value at Risk
12
Empirical
We believe that empirical evidence suggests an imperative to
10 Extreme value
incorporate various types of “non-normality” into the asset
8
allocation and portfolio modeling process, specifically to better
understand and model downside portfolio risk. Yet if we take
Density
6 this step, we have to ask whether or not our conventional risk

Much better fit
measure (i.e. standard deviation) is up to the new task.
4
for left tail
2
We would argue that, in a framework based on non-normality,
standard deviation may not be investors’ most appropriate
0 measure of portfolio risk because it rewards the desirable
-20% -15% -10% -5% 0% 5% 10% 15%
Monthly return upside movements as hard as it punishes the undesirable
Source: J.P. Morgan Asset Management. For illustrative purposes only.
downside movements. This is generally inconsistent with
investor risk preferences—primarily as observed in the field
of behavioral finance8.
It is a much closer approximation to empirically observed
Conditional Value at Risk (CVaR95) overcomes many of the draw-
performance in terms of negative skewness (rightward tilt) and
backs of standard deviation as a risk measure. Primarily, as it
leptokurtosis (“fat” left tail). This process can be applied to all
only measures risk on the downside, it captures both the asym-
assets in the portfolio, the overall result being an increase in
metric risk preferences of investors and the incidence of “fat”
the portfolio’s downside risk profile.
left tails induced by skewed and leptokurtic return distributions.
• Simulating Correlation Breakdown: Finally, we can address Further, given the widespread use by major institutional inves-
the issue of correlation breakdown using ‘copula’ theory7—a tors and regulators of its first cousin—Value at Risk—we judge it
body of work that explicitly looks at the impact of contagion to be the most appropriate risk measure to incorporate it into
or converging correlations at the total portfolio level. Copulas our framework.
We define Conditional Value at Risk (CVaR95) as the average real9

7
Copulas have been applied extensively to the pricing of Collateralized Debt Obligations—in
addition to other areas in finance. For a detailed treatment, please refer to “An Introduction portfolio loss (or gain) relative to the starting portfolio value in
to Copulas” by Nelson R. R. 1999. the worst five percent of scenarios, based on our 10,000 Monte
8
A key tenet of behavioral finance is the idea of loss aversion, i.e. a tendency of investors to
prefer avoiding losses than making gains. This translates to risk preferences that are
Carlo simulations. It is simply the average real loss (or gain) in
asymmetric in nature. the worst 500 (5% of 10,000) scenarios i.e. the left tail of the
9
Our model calculates real portfolio value by discounting the nominal portfolio value using
projected inflation. Inflation itself is projected stochastically in our framework.
portfolio distribution.

Incorporating Non-Normality: A Potentially Better Way
to Assess and Manage Downside Risk
Through rigorous analysis, we examine the impact of incor-
porating non-normality into the asset allocation process and
determine that it reveals increased downside risk, compared
with portfolios optimized using conventional mean-variance
frameworks.
These results are compelling: ignoring empiric observations of

non-normality in equity (and equity type) return distributions
understates portfolio downside risk. Likewise, using standard
deviation, rather than more behaviorally attuned conditional
value-at-risk measures, may in fact inadvertently increase
rather than decrease downside risk. Such an underestimation
of downside risk can have severe consequences for investors,
even in extreme cases presenting a solvency risk.
The remainder of this paper is focused on statistical proofs of

non-normality and the steps we apply to allow for them in the
asset allocation and optimization process. The applied method-
ologies can be, statistically speaking, quite intricate; but they
are very logical in their flow, and following the sequence of
steps is very straightforward.
But again, what we would emphasize here is the goal and over-
all impact of this work: to better capture downside portfolio risk
that is observed in real-world markets, but missed by tradition-
al asset allocation frameworks. The intended result is a more
robust portfolio modeling approach that may help investors
improve portfolio efficiency and resiliency, in light of a clearer
understanding of portfolio risk.

Empirical evidence of non-normality in
asset returns
Addressing the problem of non-normality We address three categories of non-normality10 in the
following order:
entails employing quantitative techniques that
are rather more sophisticated compared to A. Serial correlation in asset returns—this occurs when one
traditional asset allocation techniques. However, period’s return is correlated to the previous period’s return,
inducing dependence over time.
our end goal is conceptually very simple: to
produce an asset allocation framework that B. “Fat” left tails (negative skewness and leptokurtosis)—this
occurs when extreme negative returns are observed, with a
balances future portfolio return expectations
magnitude and frequency greater than implied by the ‘normal’
with the potential for more robust downside distribution.
risk management.
C. Correlation breakdown in joint asset class returns—this
First, however, we must test to see whether and what kinds of occurs during periods of high market volatility and is typically
non-normality may be present in empirically observed asset not captured by linear correlation matrices.
returns. In this section, we present the results of such tests,
which indicate the presence of several types of non-normality Throughout this section, we base our tests on ten years of
that are typically not allowed for in traditional asset allocation monthly return data to October 31, 200811. In later sections we
frameworks—to the detriment of risk estimation. then incorporate these empirical results into a revised asset
allocation framework.
Exhibit 3: Testing for serial correlation in monthly asset class returns

For those familiar with this
Reject null
hypothesis of no Statistically statistical procedure, we note that
Benchmark index Test statistic P-value serial correlation significant lag the complete official moniker for
the “Q-statistic” is the Ljung-Box
Aggregate Bonds Barclays Capital U.S. 4.46 0.62 No None
Q-statistic.
Aggregate Index
U.S. Equity S&P500 Total Return Index 3.59 0.73 No None It is computed as follows:
International Equity MSCI EAFE (hedged) Index 4.48 0.03 Yes One
Emerging Markets MSCI Emerging Markets 5.73 0.02 Yes One
Equity Index
Real Estate Investment FTSE / S&P NAREIT Index 1.48 0.96 No None
Trusts (REITs)
Hedge Fund of Funds HFRI Fund of Funds 16.18 0.00 Yes One Where Tj is the j-th serial
Diversified Index correlation and T is the number
of observations.
Private Equity Dow Jones Wilshire Microcap 4.50 0.03 Yes One
Total Return Index
10
There are other forms of non-normality one can observe in asset returns. For example, 11
Our sources for these return streams are Datastream, Bloomberg, Barclays Capital, MSCI,
heteroskedasticity or serial correlation in volatilities (also called volatility clustering) is HFRI, FTSE, and Dow Jones.
another such form.

A. Serial Correlation of Returns In the case of International Equity and Emerging Markets Equity,
For each asset class, we formally test for serial correlation by the presence of serial correlation is a very recent phenomenon.
calculating what is called the “Q-statistic.” In this test, we are In fact, it is a direct consequence of the financial crises experi-
looking to either affirm or reject the “null hypothesis,” which enced globally by investors over 2008 and into 2009.
asserts that serial correlation does not exist in the data. Public equity markets around the world have fallen precipi-
If the Q-statistic for a given asset class has significance at a 5% tously over 2008 and into 2009, displaying month-upon-month
confidence level (i.e. a p-value of less than 0.05), we can con- of negative returns. In Developed International Equity, nine of
clude that there is sufficient evidence to reject the null hypoth- the last twelve monthly returns to October 2008 were negative.
esis of no serial correlation. In this case, we must allow for the In the case of Emerging Markets Equity, eight of the last twelve
effect of serial correlation on future asset class returns. If the monthly returns were negative. This has had the effect of induc-
p-value is higher than 0.05, the null hypothesis is not rejected ing statistically significant serial correlation in monthly public
and we conclude that there is insufficient evidence to reject the equity market returns for International and Emerging Markets
null hypothesis of no serial correlation. Equity.
In Exhibit 3 on the previous page, we show the results of testing Unfortunately, the presence of serial correlation distorts the
for serial correlation in the return series of our universe of seven true risk characteristics of an asset class. In particular, it is
asset classes. incorrect to model future returns assuming independence, if
in fact returns are serially correlated (as our statistical tests
We find statistical evidence of first-order serial correlation12 in indicate). If it is not corrected, serial correlation can reduce risk
the returns of several asset classes: International Equity, Emerg- estimates from a time series by inappropriately smoothing
ing Markets Equity, Hedge Fund of Funds, and Private Equity. asset class volatility
We hypothesize that serial correlation is common in alternative

investment strategies—such as Hedge Fund of Funds and Private B. Evidence of “Fat” Left Tails
Equity—because they often hold illiquid and hard-to-price as- Our second form of non-normality relates to observing nega-
sets. The difficulty in valuing the underlying assets at regular in- tive returns in greater magnitude and with a higher probability
tervals requires managers or administrators to estimate prices
(for example, with reference to the closest marketable security Exhibit 4: histogram of U.S. equity returns—ten years ending
or based on certain economic indicators). october 2008
25 Key Statistics
If current asset prices are derived—for example—by updating Mean 0.1% Std Dev. 4.4%
last month’s asset prices (after allowing for changes in the Median 0.7% Skewness -0.67
Maximum 9.8% Kurtosis 4.20
economic environment since the last valuation), then serial 20
Minimum -16.8% Jarque-Bera 16.0
correlation reflects a gradual recognition of the true underlying
value of the asset. Such a valuation methodology would explain
15
our empirical observation of first order serial correlation in the
Frequency
asset returns of our alternative asset classes. And a conse-

quence of gradual (rather than instantaneous) recognition of 10
the true underlying value of an asset class is that conventional
risk estimates derived from a serially correlated return stream
will be underestimated. 5
0
-18% -15% -12% -9% -6% -3% 0% 3% 6% 9% 12%
12
First order serial correlation implies that the return from one month is correlated with the Monthly return
return from the previous month. Source: J.P. Morgan Asset Management. For illustrative purposes only.

than implied by the normal distribution. As an example, let us Exhibit 5: U.S. Bonds—“Fat” left tails in historical returns
consider U.S. Equities. 50
Exhibit 4 shows a simple histogram of monthly returns for U.S. 40

Equities over the last ten years. It illustrates that the distribu-
tion is clearly not symmetric. In fact it has a very noticeable 30
Density
negative skew (i.e. leans to the right) as well as excess kurtosis
20 Empirical
(i.e. is more peaked than a normal distribution; it has a kurtosis
Normal
value of 4.15, which is above the “normal” value of 3.0). We can
10
confirm this intuitive observation with a simple statistical test:
the Jarque Bera (J-B) test13 is a simple test of non-normality 0
and conclusively rejects the null hypothesis of normality in the -6% -4% -2% 0% 2% 4% 6%
Monthly return
return distribution.
When we repeat the J-B test using our other asset classes, we
find that those results, too, reject the null hypothesis. Their
returns also exhibit non-normality. We can illustrate the under-
statement of downside risk by comparing two types of density Exhibit 6: U.S. Equities—“Fat” left tails in historical returns
functions: 12
1. The density function of the normal distribution14 10
Empirical
2. The kernel (i.e. empirical) density function ‘implied’ by actual 8
Normal
Density
historical returns 6
In Exhibits 5 through 11, we show density function results for all 4

seven asset classes. Our focus, in particular, is on the left tails
2
of the density function and on whether the normal distribution
underestimates the probability of outcomes at these extremes. 0
-30% -20% -10% 0% 10% 20%
These kernel (or empirical) density functions clearly indicate “fat” Monthly return
left tails relative to those implied by the normal distribution. The

table in Exhibit 12 summarizes the results of our tests, which
identify non-normality in all seven asset classes and “fat” left tails
across all asset classes. Exhibit 7: International Equities—“Fat” left tails in historical
returns
The “fat” left tail phenomenon implies that using a normal 12
distribution to model returns underestimates the frequency

10
and magnitude of downside events.
Empirical
8
Normal
Density
13
The Jarque-Bera (J-B) test measures the departure from normality of a sample, based on its
0
skewness and kurtosis. The J-B test statistic is defined as N/6 * (S2 + (K-3)2 / 4) where N is the
sample size, S is the skewness, and K is the kurtosis. -20% -15% -10% -5% 0% 5% 10% 15% 20%
Monthly return
14
The normal distribution density function is parameterized using the Method of Maximum
Likelihood based on our sample of underlying historical data. Source: J.P. Morgan Asset Management. For illustrative purposes only.

Exhibit 8: Emerging Markets Equities—“Fat” left tails in historical Exhibit 9: U.S. REITS—“Fat” left tails in historical returns
returns
7 9
8
6
7
5
6
4
Density
5
Density
Empirical Empirical
3 4
Normal Normal
3
2
2
1
1
0 0
-40% -30% -20% -10% 0% 10% 20% 30% 40% -40% -30% -20% -10% 0% 10% 20% 30% 40%
Monthly return Monthly return
Exhibit 10: Hedge Fund of funds—“Fat” left tails in historical returns Exhibit 11: Private Equity—“Fat” left tails in historical returns
32 8
28 7
24 6
20 5
Empirical Empirical
Density
Density
16 4
Normal Normal
12 3
8 2
4 1
0 0
-100% -75% -50% -25% 0% 25% 50% 75% 100% -30% -20% -10% 0% 10% 20% 30%
Exhibit 12: Empirical evidence of “fat” left tails in asset returns
Skewness Kurtosis J-B Test Statistic Rejects Normality “Fat” Left Tail Compared to Normal
Aggregate Bonds -0.70 4.03 15.15 Yes Yes
U.S. Equity -0.67 4.20 15.99 Yes Yes
International Equity -0.91 4.01 21.66 Yes Yes
Emerging Markets Equity -0.82 4.44 23.93 Yes Yes
Real Estate Investment Trusts (REITs) -2.30 14.18 731.05 Yes Yes
Hedge Fund of Funds -0.36 7.26 93.44 Yes Yes
Private Equity 0.0 4.17 6.81 Yes Yes

C. Correlation Breakdown: Converging Correlations If this proves to be the case, it means that the benefits of port-
During Periods of Market Stress folio diversification may be overestimated by traditional frame-
It is a common observation in financial literature that correlations works that rely on linear correlation matrices. In other words,
tend to be quite unstable over time. Depending on the historical diversification may not materialize precisely when an investor
period, different asset classes can provide varying degrees of needs it most—i.e. during periods of high market volatility.
diversification. Market events such as the 1987 stock market To test our hypothesis15, we analyze simple “Pearson” correla-
crash and the 1997 Asian Financial crisis illustrate how conta- tions16 over two distinct historical periods—a full ten-year period
gion—the risk of one market negatively impacting and destabi- versus a shorter, high-volatility period of market stress. Our
lizing another—can lead to cases of converging correlations. chosen volatility measure is the the Chicago Board Options
In this section, we test the hypothesis that correlations and vola- Exchange Volatility Index (CBOE VIX). Exhibit 13 plots CBOE VIX
tility are positively related. In particular, we investigate whether values since its inception.
correlations between asset classes tend to increase during The CBOE VIX tracks expected equity volatility over a 30-day
periods of high market volatility and stress, compared to periods period17. It is widely regarded as one of the most reliable mea-
of relative calm. sures of market uncertainty available, and it is traded in real
time on a daily basis. We identify “high-volatility” periods in
Exhibit 13: CBOE VIX since inception
relation to the long-term VIX average.
70
60 More specifically, our two test periods are defined as follows:

Expected volatility %
50 1. The control period is a full ten-year history, from October

40 1997 to September 2007, during which the VIX averaged 20.9%.
30
2. The period of market stress used for comparison is from
20 August 1998 to September 1999, during which time the VIX
10 averaged 30.9%—considerably higher than its ten-year average.
This high-volatility period captured both the Asian and the Rus-
0
Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- sian financial crises, which led to significant market uncertainty.
90 92 94 96 98 00 02 04 06 08
Exhibit 14: SIMPLE CORRELATIONS OVER TEN YEARS TO SEPTEMBER 2007
International Emerging Real Estate Hedge Fund of

U.S. Bonds U.S. Equity Equity Markets Equity Investment Trusts Funds Private Equity
U.S. Bonds 1.00
U.S. Equity -0.21 1.00
International Equity -0.35 0.81 1.00
Emerging Markets Equity -0.23 0.71 0.74 1.00
Real Estate Investment Trusts (REITs) 0.05 0.31 0.23 0.33 1.00
Hedge Fund of Funds -0.14 0.46 0.61 0.70 0.19 1.00
Private Equity -0.15 0.61 0.62 0.70 0.33 0.73 1.00
15
Note that although we do not test whether the changes in correlations are statistically signifi-
cant, we do believe the changes illustrate our broad point of correlation convergence during
periods of market stress. 17
More precisely, the CBOE VIX represents the implied volatility of S&P 500 index options.
16
Pearson correlations are used to indicate the strength and direction of a linear correlation The VIX is quoted in terms of percentage points and translates, roughly, to the expected
between two random variables. movement in the S&P 500 index over the next 30-day period, on an annualized basis.

“The results are striking. Nearly all the
correlation coefficients increased to some
degree (during this period of high market
volatility), with only four showing a decrease.
Our main diversifiers did not provide the
diversification benefits anticipated.”

Exhibit 14 shows correlation coefficients (ρ) over the full anticipated. The correlation coefficients with U.S. Equities were
(monthly) ten-year period to September 2007 for all seven as follows:
asset classes.
Long-run Coefficient During Market Stress
Our correlation matrix indicates that investing in certain pairs
U.S. Bonds -0.21 -0.02
of assets should provide diversification benefits—i.e. they
REITs 0.31 0.58
should either be negatively or lowly correlated. For example,
Hedge Fund of Funds 0.46 0.60
let’s consider correlations of asset classes with U.S. Equities.
Private Equity 0.61 0.86
Based on historical correlations over the entire ten years, we
expect the following asset classes to provide some diversification
benefits relative to a U.S. Equity-dominated portfolio: Our analysis reaffirms our intuition, suggesting ample evidence
of correlation convergence during periods of market stress.
• U.S. Bonds (ρ= -0.21)
• REITs (ρ = 0.31) Again, conventional frameworks which assume normality—in
• Hedge Funds of Funds (ρ = 0.46) this case normal distribution of joint returns—prove inadequate
• Private Equity18 (ρ = 0.61) to account for what we observe in real-world markets. And
inappropriately assuming linearity of correlations can lead to
In times of market stress, however, these correlations break a significant underestimation of joint negative returns during a
down. Exhibit 15 shows a simple correlation matrix for the market downturn.
high-volatility period from August 1998 to September 1999.
We have highlighted in orange where correlations have
converged (i.e. reducing diversification) and in green where
correlations have diverged (i.e. increasing diversification).
The results are striking. Nearly all the correlation coefficients

increased to some degree, with only four showing a decrease.
Our main diversifiers did not provide the diversification benefits
Exhibit 15: Correlations during high volatility period (AugUST 1998—SeptEMBER 1999)

U.S. Bonds 1.00
U.S. Equity -0.02 1.00
Hedge Fund of Funds -0.42 0.60 0.73 0.69 0.48 1.0
Private Equity -0.08 0.86 0.70 0.81 0.76 0.72 1.00
18
Private Equity may be seen as a return enhancer rather than a diversification play. We include
it in our diversification analysis—despite a relatively high correlation—for illustrative purposes.

Incorporating non-normality into an asset
allocation framework
Based on empirical testing, we have demon- characteristics of the underlying data generating process. Nota-
bly, our new series should have a higher volatility, demonstrate
strated that conventional asset allocation frame-
no serial correlation and have a cross correlation structure with
works understate downside risk because they do other asset classes similar to our original data.
not allow for non-normality. However, we believe
Our procedure for “unsmoothing” returns (or correcting for
that achieving optimal portfolio efficiency—based serial correlation in the underlying data series) is a two-step
on a more precise estimation of portfolio risk— process:
requires investors to incorporate non-normal
STEP 1:
return distributions into the asset allocation and We determine the correlation coefficient at lag one20
portfolio optimization process. (i.e. previous month’s return) for each return series, by
running the following regression:
In this section, we offer a statistical methodology for incorporat-
ing the three categories of non-normality already discussed. We Rt = a + bR(t-1)
address them in the same order as they were presented in the
previous section: where Rt represents the return at time t
A. Serial correlation The regressions indicate the following estimates (Exhibit 16) for
the “‘serial” correlation coefficients, based on monthly returns.
B. “Fat” left tails (negative skewness and leptokurtosis)
C. Correlation breakdown (converging correlations)
A. Incorporating the Impact of Serial Correlation

Our empirical work in the last section shows that certain asset
Exhibit 16: Correlation coefficients at lag one
classes—International Equity, Emerging Markets Equity, Hedge
^b Statistically significant*
Fund of Funds and Private Equity—test positive for serial cor-
relation. This result implies that, over time, single period returns International Equity 0.21 Yes
are influenced by, and not independent from, previous periods. Emerging Markets Equity 0.25 Yes
Hedge Fund of Funds -0.42 Yes
First-order serial correlation, if not corrected for in the un-
Private Equity 0.21 Yes
derlying data, will mask true asset class volatility and bias risk
Source: J.P. Morgan Asset Management. Note the coefficient b reflects the strength of the auto-
estimates downwards, leading to underestimation of asset correlation. For illustrative purposes only.
* Statistically significant at the 5% level
class risk and, hence, portfolio risk.
In order to ‘adjust’ for the impact of serial correlation on

asset returns, we apply a variation of Fisher-Geltner-Webb’s 19
For a more detailed treatment, please refer to Fisher, J., D. Geltner, and B. Webb. 1994.
“unsmoothing” methodology19, whereby we re-calculate an Value Indices of Commercial Real Estate: A Comparison of Index Construction Methods.
Also, Fisher, J. and D. Geltner. 2000. De-Lagging the NCREIF Index: Transaction Prices and
“unsmoothed” return stream from our original data. The new Reverse-Engineering.
adjusted return series should be more reflective of the risk 20

Note International Equity, Emerging Markets Equity, Hedge Fund of Funds and Private Equity
returns test positive for serial correlation at lag one.

STEP 2: Finally, we evaluate the impact on asset correlations to assess
We then produce our “unsmoothed” return series as follows, the viability of our unsmoothed data series—as our unsmooth-
based on the “serial” correlation coefficient already derived: ing procedure should not affect correlations. We compare our
^
^ ) / (1—b) original correlation matrix using index data (Exhibit 19) with
Rt (corrected) = (Rt—bRt-1 our new correlation matrix derived from unsmoothed data
After applying the methodology outlined above, we obtain (Exhibit 20).
a new data series Rt (corrected) that should display no serial We see that, as anticipated, the impact of unsmoothing on asset
correlation. Once again we use the Q statistic to test the correlations is negligible.
unsmoothed data for serial correlation. The results are shown
below (Exhibit 17).
B. Incorporating the Impact of “Fat” Left Tails into our
Exhibit 17: Tests for serial correlation on corrected data for up to Framework
six lags
All of our asset classes show “fat” left tails, compared to the
Evidence of serial normal distribution. Our approach to dealing with such left tail
Test statistic p-value correlation
risk involves what is called a “semi-parametric approach”: that
International Equity 0.03 0.86 No
is, instead of modeling returns using a single distribution (such
Emerging Markets Equity 0.06 0.80 No
as is traditionally done with the normal distribution), we fit the
Hedge Fund of Funds 0.04 0.85 No
left and right tails separately from the interior of the distribution.
Private Equity 0.05 0.82 No

Hence, we segment each return distribution into three parts—
the left tail, right tail and interior (See Exhibit 21). We then
“fit” each segment of the distribution separately. This approach
The test indicates that our revised data is no longer serially cor- enables us to more accurately capture the potential distribu-
related. More important, our transformations of the underlying tions associated with the empirical data in each segment of
data series do not alter the mean return of the series over the the distribution.
time period considered. However, removing the effects of serial
correlation does increase the standard deviation of returns, as To fit the left and right tails of the distribution we turn to
shown in Exhibit 18 below. As discussed, the higher volatility of Extreme Value theory—a body of work specifically designed to
our “unsmoothed” return series is in line with our expectations, look at the probability of low-probability but high-risk events.
as serial correlation masks volatility and understates risk.
The central premise behind Extreme Value theory is that the
probability of observing extreme values (such as negative
Exhibit 18: Volatilities before and after “unsmoothing” for serial
correlation
financial market returns), can be modeled using any one of
Annualized volatility Annalized volatility
several Extreme Value distributions.21
before after
“unsmoothing” “unsmoothing” Our choice of Extreme Value distribution is the Generalized
International Equity 15.25% 18.93% Pareto distribution (GPD). We choose the GPD distribution,
Emerging Markets Equity 24.21% 31.42% rather than other Extreme Value distributions, due to the ease
Hedge Fund of Funds 6.51% 10.30% with which it can be adapted to modeling financial market
Private Equity 23.51% 29.17% returns.
21
An Extreme Value distribution is defined as the limiting distribution of extreme values sampled
from any independent randomly distributed process.

Exhibit 19: Correlations based on smoothed (or raw) data over ten years to October 2008

U.S. Bonds 1.00
U.S. Equity -0.07 1.00
Hedge Fund of Funds 0.08 0.57 0.69 0.78 0.30 1.00
Private Equity -0.03 0.65 0.68 0.71 0.42 0.76 1.00
Exhibit 20: Correlations based on unsmoothed data over ten years to October 2008

U.S. Bonds 1.00
U.S. Equity -0.07 1.00
Hedge Fund of Funds 0.10 0.57 0.69 0.77 0.25 1.00
Private Equity -0.02 0.64 0.67 0.69 0.40 0.74 1.00
Exhibit 21: Hypothetical dissection of return distribution
Left tail Interior Right tail

The “Fitting” Process is Done in Two Steps: • What we are looking for is a better fit to the actual distribu-
• First: Fitting the left (and right) tail of the Generalized Pareto tion of empirically observed data (blue line)—as shown in the
distribution (GPD) to our historical data. This is done by cali- kernel density function.
brating the tail of the GPD to extreme values in the sample • What we see is that in each case (Exhibits 22 through 28), the
using the Method of Maximum Likelihood. Extreme values are GPD distribution (orange line) is a much more accurate fit to
defined by a threshold. This method is commonly referred to real-world data than the “normal” distribution.
as the Peaks-Over-Threshold method within Extreme Value
theory. • The improved fit is particularly noticeable in the left tail,
where the normal distribution understates risk.
• Second: Fitting the interior of a distribution is accomplished
using a non-parametric (kernel or empirical) approach.
Hence, in aggregate, we derive a semi-parametric probability

density function to describe the data generating process. We re-
fer to this broad approach as a semi-parametric GPD approach.
We can test the results of the “fitting” process against both the
normal and kernel (or empirical) distributions used to establish
“fat” left tails in the previous section:
Generalized Pareto Distribution

For those interested in the Generalized Pareto distribution,
we provide supporting equations here.
The probability density function for the generalized Pareto
distribution with shape parameter k ≠ 0, scale parameter σ,
and threshold parameter θ, is given by:
Exhibit 22: u.s. bonds—fitting the semi-parametric gpd distribution

to historical data
50
Empirical
Semi-parametric GPD for θ < x, when k > 0, or for θ < x <–σ/k when k < 0.
40
In the limit for k = 0, the density is
30 for θ < x.
Density
1
-1
k
20 f(xI k,σ,θ) 1+k
10 If k = 0 and θ = 0, the Generalized Pareto distribution is

equivalent to the exponential distribution. If k > 0 and θ =
0 σ, the generalized Pareto distribution is equivalent to the
-.60% -40% -20% 0% 20% 40% Pareto distribution.
Monthly return

Exhibit 23: U.S. Equities—Fitting the semi-parametric GPD Exhibit 24: International Equities—Fitting semi-parametric GPD
distribution to historical data distribution to historical data
12 12
Empirical Empirical
10 Semi-parametric GPD 10 Semi-parametric GPD
8 8
Density
Density
6 6
4 4
2 2
0 0
-40% -30% -20% -10% 0% 10% 20% -20% -15% -10% -5% 0% 5% 10% 15%
Exhibit 25: Emerging Markets Equities—Fitting the semi-parametric GPD Exhibit 26: U.S. REITs—Fitting the semi-parametric GPD distribution
distribution to historical data to historical data
7 9 Empirical
Empirical
8
6 Semi-parametric GPD Semi-parametric GPD
7
5
6
4
Density
Density
3 4
3
2
2
1
1
0 0
-100% -80% -60% -40% -20% 0% 20% -100% -80% -60% -40% -20% 0% 20%
Exhibit 27: Hedge fund of Funds—Fitting the semi-parametric Exhibit 28: Private Equity—Fitting the semi-parametric GPD
GPD distribution to historical data distribution to historical data
32 8
Empirical Empirical
28 Semi-parametric GPD 7 Semi-parametric GPD
24 6
20 5
Density
Density
16 4
12 3
8 2
4 1
0 0
-20% -10% 0% 10% 20% 30% -40% -30% -20% -10% 0% 10% 20% 30% 40%

To enhance our conviction in these results, we need to statisti- Given our focus is on ‘left’ tail risk, however, we can also gauge
cally test this “goodness of fit” both for the whole distribu- fit by considering the Quantile-Quantile (QQ) plots of the
tion as well as specifically for the tails. Below we apply the empirical data against those implied by the semi-parametric
Kolmogorov-Smirnov (KS)22 empirical distribution test under two GPD distribution23.
hypotheses: first, the null hypothesis that returns are normally
distributed, and second, under the hypothesis that they are We see that in each case, the lowest quintiles are a much bet-
sampled from a semi-parametric GPD. Exhibit 29 below sets out ter fit with the semi-parametric GPD distribution than with the
our results. We also show the probability that our null hypoth- normal distribution, indicating that the semi-parametric GPD
esis stands statistical significance. distribution more accurately captures the frequency of extreme
negative events.
Our empirical tests confirm our view that the semi-parametric
GPD distribution provides a better fit for the historical data—
when compared to the normal distribution—for all our asset
classes.
Exhibit 29: Testing goodness of fit of Normal and semi-parametric GPD distribution
Normal Semi-parametric GPD

Reject Semi-
KS statistic p Value Reject Normality KS Statistic p Value parametric GPD
U.S. Aggregate Bonds 0.50 0.00 Yes 0.05 0.89 No
U.S. Large Cap Equity 0.49 0.00 Yes 0.06 0.75 No
EAFE Unhedged Equity 0.46 0.00 Yes 0.12 0.08 No
Emerging Markets 0.46 0.00 Yes 0.11 0.13 No
Real Estate Investment Trusts (REITs) 0.45 0.00 Yes 0.05 0.89 No
Hedge Fund of Funds 0.47 0.00 Yes 0.05 0.92 No
Private Equity 0.47 0.00 Yes 0.10 0.18 No
Exhibit 30: U.S. Bonds—QQ Plot (Normal Distribution) Exhibit 31: U.S. Bonds—QQ Plot (Semi-parametric GPD distribution)
.04 .03
Quantiles of semi-parametric GPD
.03 .02
.02 .01
Quantiles of Normal
.01 .00
.00 -.01
-.01 -.02
-.02 -.03
-.03 -.04
-.04 -.02 .00 .02 .04 -.04 -.02 .00 .02 .04
Quantiles of data Quantiles of data
QQ plots illustrate how different quantiles of a sample compare to those implied by the theo-
23
22
The Kolmogorov-Smirnov test is a goodness of fit test to test whether a sample comes from a retical density function. All data points being on a straight 45 degree line indicate a perfect fit
particular distribution. Our null hypothesis is tested at 5% significance. of the data with the distribution.

Exhibit 32: U.S. Equities—QQ Plot (Normal Distribution) Exhibit 33: U.S. Equities—QQ Plot (semi-parametric GPD distribution )
.12 .15

.08 .10
Quantiles of Normal
.04 .05
.00 .00
-.04 -.05
-.08 -.10
-.12 -.15
-.20 -.15 -.10 -.05 .00 .05 .10 -.20 -.15 -.10 -.05 .00 .05 .10
Exhibit 34: International Equities—QQ Plot (Normal Distribution) Exhibit 35: International Equities—QQ Plot
(semi-parametric GPD distribution )
.15 .15
.10 .10
Quantiles of Normal
.05 .05
.00 .00
-.05 -.05
-.10 -.10
-.15 -.15
-.15 -.10 -.05 .00 .05 .10 -.15 -.10 -.05 .00 .05 .10
Exhibit 36: Emerging Markets Equities—QQ Plot (Normal Distribution) Exhibit 37: Emerging Markets Equities—QQ Plot
(semi-parametric GPD distribution )
.20 .2
.15
.1
.10
Quantiles of Normal
.05 .0
.00
-.05 -.1
-.10
-.2
-.15
-.20 -.3
-.3 -.2 -.1 .0 .1 .2 -.3 -.2 -.1 .0 .1 .2

Exhibit 38: REITs—QQ Plot (Normal Distribution) Exhibit 39: REITs—QQ Plot (semi-parametric GPD distribution )
.15 .1

.10
.0
Quantiles of Normal
.05
.00 -.1
-.05
-.2
-.10
-.15 -.3
-.4 -.3 -.2 -.1 .0 .1 -.4 -.3 -.2 -.1 .0 .1
Exhibit 40: Hedge fund of Funds—QQ Plot (Normal Distribution) Exhibit 41: Hedge fund of Funds—QQ Plot
(semi-parametric GPD distribution)
.06 .12

.04
.08
Quantiles of Normal
.02
.04
.00
.00
-.02
-.04
-.04
-.06 -.08
-.08 -.04 .00 .04 .08 -.08 -.04 .00 .04 .08
Exhibit 42: Private Equity—QQ Plot (Normal Distribution) Exhibit 43: Private Equity—QQ Plot (semi-parametric GPD distribution )
.20 .3
.15
.2
.10
Quantiles of Normal
.05 .1
.00
-.05 .0
-.10
-.1
-.15
-.20 -.2
-.3 -.2 -.1 .0 .1 .2 .3 -.3 -.2 -.1 .0 .1 .2 .3
Note that the results shown are for illustrative purposes and are based on applying Extreme Value theory to the raw ‘smoothed’ data. In our framework—because we require independence of asset
returns—we remove the effect of serial correlation before fitting a semi-parametric GPD.

C. Incorporating the impact of converging correlations This approach better reflects the empirical observations of con-
into our framework verging correlations during periods of high volatility, which tend
One of our empirical findings in the previous section was that to be accompanied by negative returns. Specifically, we model
correlations converge during periods of high market volatility joint “fat” left tails by adjusting the copula’s “degree-of-freedom”
and stress. Our approach to dealing with this phenomenon is parameter: 30 degrees-of-freedom is virtually identical to a
to apply ‘copula’ theory—a body of work that explicitly looks at normal distribution. By lowering the parameter, we can increase
the impact of contagion or converging correlations at the total asset class dependence to model observed asset class behavior
portfolio level. in times of market stress.
Copulas are mathematical functions that allow us to model the In Exhibit 44, we illustrate the impact of modifying the normal
joint distribution of asset returns separately from the marginal marginal distributions by adding a t-copula with a low degree-
(or individual asset class) distributions. By considering joint of-freedom parameter.
distributions, we turn our focus to how asset classes behave
together (rather than individually, as we did in the last section).
In a traditional framework, joint distributions are captured by

simple (Pearson) correlations. Unfortunately, simple correla-
tions assume a linear relationship between individual asset
Exhibit 44: Modeling joint dependence in asset returns using copulas
classes, and we have already shown empirically that linearity
breaks down at the extremes. Copulas, on the other hand, Illustration of traditional linear correlations
10%
Monthly simulated Hedge Fund of Funds returns
allow us to differentiate the relationship between asset

8%
classes in times of market stress, from more normal times.
6%
In particular, we observe empirically that not only do individual 4%
asset classes have “fatter” left tails than allowed in a normal 2%
distribution, combinations of asset classes exhibit “fat” joint left 0%
tails—i.e. a higher incidence of joint negative returns in times of -2%
market stress. Copulas allow us to model both the “fat” marginal -4%
left tails for individual asset classes (using the semi-parametric -6%
-8%
GPD approach explained earlier), while at the same time modeling
-10%
more accurately the “fat” joint left tails (joint negative returns) -20% -15% -10% -5% 0% 5% 10% 15% 20%
as empirically observed. Monthly simulated U.S. Large Cap Equity returns
Just as the application of Extreme Value Theory involves choos- 10%

Illustration of increased joint dependence through copulas
Monthly simulated Hedge Fund of Funds returns
ing from among several types of Extreme Value Distributions, 8%

applying copula theory requires us to choose a copula type. 6%
Here we use the t-copula (based on the Student t distribution)24 4%
because it enables us to better capture the effects of converg- 2%
ing correlations—i.e. allows fatter tails than the normal 0%
distribution. -2%
-4%
-6%
-8%
-10%
24
The Student t distribution comes from the same family of distributions as the Normal. A Stu- -20% -15% -10% -5% 0% 5% 10% 15% 20%
dent t distribution is calibrated on only one parameter i.e. the degrees of freedom. The higher
the degrees of freedom, the closer it is to a normal. At 30 degrees of freedom, the Student t Monthly simulated U.S. Large Cap Equity returns
distribution is virtually identical to the normal distribution. Hence, assuming a lower degree of
freedom parameter imposes significant increased dependence between asset classes returns. Source: J.P. Morgan Asset Management. For illustrative purposes only.

In order to show the effect of modeling increased joint depen-
Exhibit 45: Incorporating non-normality—Summary of model
dence, we produce a scatter plot of 2,500 simulations of U.S.
Large Cap Equity returns and Hedge Fund of Funds returns
Step 1: Input historical data
first, assuming normality with simple correlations; and second,
assuming normality with a Student t-copula added (with two
degree of freedom25).
Step 2: Remove serial
correlation from returns
Compared to the traditional multivariate normal framework,
our simulations show the impact of increased dependence,
through a higher joint incidence of negative returns between
U.S. Large Cap Equities and Hedge Fund of Funds (highlighted). Unsmoothed data
This illustrates the phenomenon of “fat” joint left tails we
alluded to earlier.
Our simulations also illustrate increased dependence of other

extreme joint occurrences. In other words, we can see an
Step 3: Fit marginal distributions
increased incidence of negative U.S. Large Cap Equity returns (using Extreme Value theory) to allow for
with positive Hedge Fund of Funds returns (the top left quadrant) “fat” left tails
or positive U.S. Equity returns with positive Hedge Fund of
Funds returns (the top right quadrant). This is visible as the
“star” pattern in the scatter plot.
This occurs because our Student t distribution is symmetric

around the mean (like the Normal distribution) and, hence,
Step 4: Calibrate Student t-copula
incorporating “fat” joint tails increases the probability of to allow for joint “fat” tails—i.e. increased
observing all extreme joint outcomes (positive and negative)— dependence during periods of
not just extreme joint negative outcomes. This can be seen as market stress
a drawback of our choice of the Student t distribution, as—in
reality—correlations do exhibit asymmetric behavior, i.e. the Source: J.P. Morgan Asset Management. For illustrative purposes only.
convergence properties are different in extreme positive
markets compared to extreme negative markets.
However, our choice of t-copula is a vast improvement when

compared to linear correlation matrices used in traditional
models, as the latter completely ignore the effect of correlation
convergence. Incorporating “fat” joint left tails using copula
theory allows us to more robustly model the empirical
phenomenon of correlation breakdown during periods of
market stress.
25
Note our model calibrates the degrees of freedom parameter for the Student t distribution
assuming the marginal distributions are semi-parametric GPD. However, to illustrate the effect
of adding the copula, we choose a low degree of freedom parameter of two and assume the
marginal distributions are normal (rather than semi-parametric GPD). This allows us to isolate
and illustrate the effect of fat ‘joint’ tails separately from fat ‘marginal’ tails.

Estimating downside portfolio risk in a
non-normal framework
Now that we have a statistically solid framework We define Conditional Value at Risk (CVaR95) as the average
(real) portfolio loss (relative to the starting value) in the worst
for incorporating non-normality into the asset
five percent of scenarios, based on our 10,000 Monte Carlo
allocation process, we can apply this framework simulations. It is simply the average real loss in the worst 500
to a hypothetical U.S. domiciled investor. We will (5% of 10,000) scenarios (i.e. the left tail of the portfolio
assume a well-diversified portfolio with an initial distribution).
value of $1 billion and allocations across our
seven major asset classes. Detailed allocations to A Better Risk Quantifier: Conditional Value at Risk
each asset class are shown in Exhibit 46 below. We believe that empirical evidence suggests an imperative to
incorporate various types of “non-normality” into the asset
We can now generate forward looking projections of our hypo- allocation and portfolio modeling process, specifically to better
thetical portfolio’s real value26 in our newly revised Monte Carlo understand and model downside portfolio risk. Yet if we take
simulation framework. For each of our ten forward years, we this step, we have to ask whether or not our conventional risk
generate 10,000 simulations27. We then measure downside port- measure (i.e. standard deviation) is up to the new task.
folio risk as the Conditional Value at Risk at the 5th percentile of
the portfolio. We would argue that, in a framework based on non-normality,
standard deviation may not be investors’ most appropriate
measure of portfolio risk because it punishes the desirable
Exhibit 46: Hypothetical portfolio allocation upside movements as hard as it punishes the undesirable down-
side movements. This is generally inconsistent with investor risk
Asset class Current Allocation
preferences—primarily as observed in the field of behavioral
Total bonds 30%
finance28.
U.S. Aggregate Bonds 30%
Total equity 55% Conditional Value at Risk (CVaR95) overcomes many of the
U.S. Large Cap Equity 40% drawbacks of standard deviation as a risk measure. Primarily,
International Equity (hedged) 10% as it only measures risk on the downside, it captures both the
Emerging Markets Equity 5% asymmetric risk preferences of investors and the incidence of
“fat” left tails induced by skewed and leptokurtic return distri-
Total alternatives 15%
butions. Further, given the widespread use by major institutional
REITs 5%
investors and regulators of its first cousin—Value at Risk—we
Hedge Fund of Funds 5%
judge it to be the most appropriate risk measure to incorporate
Private Equity 5%
it into our framework.
Key statistics
Expected arithmetic return 9.1%
Expected volatility 10.0% 26
Our model calculates real portfolio value by discounting the nominal portfolio value using
Expected compound return 8.7% projected inflation. Inflation itself is projected stochastically in our framework.
Sharpe ratio 0.51

27
We believe 10,000 simulations should reduce simulation error sufficiently to allow us to
draw robust inferences from the results.
Source: J.P. Morgan Asset Management. For illustrative purposes only. 28
A key tenet of behavioral finance is the idea of loss aversion i.e. a tendency of investors to
Sharpe ratio calculated based on expected risk free return of 4.0% per year as per prefer avoiding losses than making gains. This translates to risk preferences that are
J.P. Morgan Asset Management Long Term Capital Market Assumptions (please see asymmetric in nature.
Appendix for details).

The calculation for CVaR95 is illustrated graphically for the current We should note that while the increased risk associated with
allocation below. Exhibit 47 shows the projected frequency of modeling non-normality is striking, incremental risk in itself is
the portfolio’s real gains (or losses) at the end of ten years. We not necessarily a reason for changing a plan’s asset allocation
calculate CVaR95 by taking the average real cumulative loss in (unless an absolute threshold value has been breached). More
the worst 5% of simulations. specifically, if one assumes an arbitrarily (but equally) higher
downside risk for all asset classes, this in itself would not
Exhibit 47: Histogram of projected cumulative portfolio gain (loss) impact the efficiency of the portfolio—i.e. its efficiency would
at the end of ten years assuming non-normality
800
not change at all, albeit the CVaR95 would now be higher.
700 The reason non-normality can impact asset allocation is that

the downside risk associated with different asset classes is very
600
different. Most obviously, equity and equity type asset classes
entail greater degrees of downside risk than, for example, fixed
Frequency
500
income type investments. Hence, because the downside risk
400
characteristics of different asset classes are different—and
300 cannot be accounted for using traditional modeling techniques
or risk measures such as standard deviation—the efficient
200
allocations in our CVaR95 motivated non-normal framework
100 must also be different from a traditional framework.
0 For this reason—not merely due to higher CVaR95 figures—we
(489)
(289)
(90)
110
309
509
709
908
1,108
1,307
1,507
1,706
1,906
2,106
2,305
2,505
2,704
2,904
3,104
3,303
3,503
3,702
3,902
4,101
4,301
4,501
4,700
4,900
5,099
5,299
5,498
5,698
5,898
believe investors need to quantitatively incorporate the impact

of non-normality into the asset allocation process.
Expected portfolio gain (loss) at the end of ten years ($ mm)
In the next section, we consider the impact of non-normality
CVaR95 is defined as the average real cumulative loss in the worst 5% or 500 simulations.
This is equal to $168 million for the current portfolio. on optimized portfolio solutions for investors with a long-term
investment horizon.
The CVaR95 of the current allocation, based on our new method-

ology, is $168 million. In other words, the portfolio can expect
to lose (on average) $168 million in the worst five percent of
cases (based on our simulation results). This risk is significant. It
indicates a real return (i.e. after allowing for inflation) of -16.8%
on the portfolio over an extended time horizon—a result our
investor is unlikely to be very happy with.
For the same portfolio, however, risk calculations that assume

normality29 would result in a CVaR95 figure of $74 million. Incor-
porating non-normality more than doubles our prior estimate
of CVaR95. In absolute terms, the risk underestimation is $94
million or 9.4% of the portfolio’s initial value.
29
Our estimates of CVaR95 under a traditional framework were derived with an identical asset
allocation, except that asset returns were assumed to be individually and jointly normally
distributed. Risk and correlations were derived based on the same ten year historical period
from November 1998 to October 2008.

“The reason non-normality can impact asset
allocation is that the downside risk associated
with different asset classes is very different. Most
obviously, equity and equity type asset classes
entail greater degrees of downside risk than, for
example, fixed income type investments.”

Implications of non-normality on optimal
portfolio solutions
In this section we assess the impact of Exhibit 48: Efficient frontier based on traditional and non-normal
framework
non-normality on portfolio efficiency, using Expected compound return (% per year)
CVaR95 as the risk measure of choice30. 12%
Our optimization process involves minimizing CVaR95 for a given 11%
expected return (or equivalently, maximizing return for a given 10%

CVaR95). In any case, the portfolios we derive here are “efficient”
in that they offer investors the optimal tradeoff between 9%
(downside) risk and reward. The optimization process is based 8%

on linear programming techniques31 applied to 10,000 Monte
Carlo simulations (generated in the previous section). 7%
Normal framework
6%
We construct an efficient frontier comprising ten “efficient” Non-normal Mean CVaR framework
portfolios: 5%
(200) 0 200 400 600 800 1,000
• Our first portfolio comprises our minimum CVaR95 portfolio— Conditional Value at Risk 95 ($ millions)
such that there is no other asset allocation (derived from
our simulations) that minimizes the portfolio’s CVaR95 further
over ten years. It is our minimum risk portfolio. In aggregate, a plot of the expected compound return (in % per
year) versus the CVaR95 (in $ millions) constitutes our efficient
• Our tenth portfolio comprises our highest return portfolio— frontier (Exhibit 48).
such that there is no other asset allocation (derived from
our simulations) that maximizes the return of the portfolio The efficient frontier created using a non-normal framework
further over ten years. differs markedly from the equivalent frontier produced from a
traditional model32 built on assumptions of normality. Notably all
• The intermediate eight portfolios are equally distributed points along the “new” efficient frontier demonstrate—for a given
between the first and tenth in terms of expected return, but expected return—significantly higher risk, as captured by CVaR95.
minimizing CVaR95 over ten years at each point along the Detailed allocations underlying each efficient frontier are shown
efficient frontier. in the Appendix.
These results confirm our intuition that capturing the non-

normality associated with equity and alternative return distribu-
tions should increase the magnitude of CVaR95 of any portfolio
30
It is worth noting that when optimizing under the assumption of normality, the choice of with an allocation to these asset classes. Crucially, as traditional
risk measure (whether standard deviation or CVaR95) is irrelevant. This is because a normal
distribution is symmetric about its mean—so choosing a different percentile (e.g. two standard
frameworks fail to capture non-normality, they underestimate
deviations)—does not impact the derived solution. Under non-normality, the choice of risk the downside risk associated with such portfolios. Hence, the
measure is critical to the derived solution.
31
Linear programming is required because as we move away from traditional “normal”
efficient frontier derived from a traditional model is likely to
assumptions, there are no shortcuts to computing total portfolio risk. Individual downside show a misleading, lower risk figure for portfolios with a similar
risk measures—such as standard deviation—can no longer be aggregated to the total
portfolio level using analytical formulae. expected return.
32
Our traditional efficient frontier was derived under a Monte Carlo simulation framework
except that asset returns were assumed to be individually and jointly normally distributed.

Optimal Portfolio Solution for our Hypothetical Investor
Optimization Constraints In this section, we ask ourselves how incorporating non-normality
Optimizations in general—whether based on into our return distributions affects our optimal portfolio solution
modern or, as in our case, post-modern portfolio for our hypothetical investor.
theory—suffer from certain inherent draw-
backs. A particular drawback associated with Exhibit 49 compares our current hypothetical investor with a
traditional frameworks is the sheer number of portfolio allocation using an identical target arithmetic return of
constraints applied to the optimization process. 9.1%, but optimized using our revised non-normal framework33.
Constraints reduce the credibility of the optimi- For illustrative purposes, we also show the optimized allocation
zation process as, by applying constraints, the derived from a traditional-mean variance framework. We do not
practitioner is guiding the model towards a impose any constraints on our optimization process.
particular solution—rather than allowing the
model to calculate one itself independently.
33
All optimizations are based on J.P. Morgan Long Term Capital Market Return Assumptions.
In deriving our optimal portfolios, we do not
apply any constraints. Hence, we let the model
dictate the direction and magnitude of results.
Exhibit 49: Optimization results
Optimized, Unconstrained Optimized, Unconstrained

Current allocation
Normal allocation Non-normal allocation
Total bonds 30.0% 0.0% 34.5%
U.S. Aggregate Bonds 30.0% 0.0% 34.5%
Total equity 55.0% 18.3% 36.0%
U.S. Large Cap Equity 40.0% 8.9% 21.7%
International Equity 10.0% 7.6% 5.8%
Emerging Markets Equity 5.0% 1.9% 8.5%
Total alternatives 15.0% 81.7% 29.5%
REITs 5.0% 11.9% 11.1%
Hedge Fund of Funds 5.0% 64.1% 7.0%
Private Equity 5.0% 5.6% 11.5%
Total 100.0% 100.0% 100.0%
Key statistics
Target expected arithmetic return 9.1% 9.1% 9.1%
Expected volatility 10.0% 8.6% 9.5%
Expected compound return 8.7% 8.6% 8.7%
Sharpe ratio 0.51 0.59 0.54
CVaR95 ($ million) allowing for non-normality $168 mm $206 mm $148 mm
CVaR95 vs. current allocation - 23% 12%
Return per unit of CVaR95 0.30 0.25 0.35
Sharpe ratio calculated assuming risk free return of 4.0%.

Our results indicate that the ‘optimal’ portfolio using a tradi-
tional mean-variance framework actually increases (rather than
decreases) risk by 22.6% relative to the current allocation—
as defined by CVaR95. As a traditional framework minimizes
standard deviation34—which we argue is an inadequate risk
measure—it inadvertently exposes our investor to even worse
scenarios on the downside than the current allocation.
However, the more significant issue by far—and the biggest

drawback of the traditional approach—is that it produces a highly
concentrated and impractical asset allocation. This is because
in the absence of formal constraints, it over allocates to asset
classes based on small differences in assumptions. This limits
our ability to draw useful insights into the portfolio construction
process, using such a framework.
On the other hand, our non-normal CVaR95 based framework

produces a diversified solution with allocations across the asset
class spectrum. No single asset class significantly dominates the
portfolio. Despite the fact that our investor already holds quite
a diversified portfolio, our framework suggests that there is still
scope for our investor to improve portfolio efficiency further.
The “optimal” portfolio identified by the non-normal frame-

work improves both the expected Sharpe ratio and reduces the
CVaR95 relative to the current allocation. This signifies that our
optimal portfolio is more efficient (in Sharpe ratio and CVaR95
space) than the current hypothetical portfolio. Based on our
Long Term Capital Market Return assumptions, the allocations
to fixed income and alternatives increase, while the allocation to
equities decreases.
34
Minimizing standard deviation for a given expected return target is the equivalent of
maximizing Sharpe ratio.

Comparing a Traditional Model with Serial correlation
a Non-normal Framework The impact of serial correlation is negative for International
Equity, Emerging Markets Equity, Hedge Fund of Funds and
So, how does each form of non-normality affect our optimal
Private Equity as these asset classes test positive for first order
portfolio solution relative to a traditional mean-variance
serial correlation. Recall that the presence of serial correlation
optimization framework?
increases the risk associated with the particular asset class—
Exhibit 50 shows the marginal and then combined impact of hence its negative marginal impact. As U.S. Bonds, U.S.
each form of non-normality on the optimal portfolio solution, Equity and REITs do not test positive for serial correlation,
when compared to a traditional mean-variance framework35. the marginal impact for these asset classes is positive
(relative to a traditional mean-variance framework).
(Note the analysis below is based on J.P. Morgan’s Long Term Expected Return
Assumptions. It is worth noting that the results below will vary depending on the
expected return assumptions i.e. relative attractiveness of the asset classes.) “Fat” left tails
Each arrow in the table indicates the broad directional impact Our second form of non-normality—“fat” left tails—has a
of the particular form of non-normality on the allocation to the negative marginal impact on optimal allocations to U.S. Equi-
asset class—relative to the allocation implied by a mean-variance ties, REITs and Hedge Fund of Funds (relative to a traditional
approach. For example, an up arrow indicates the allocation to mean-variance framework). This is driven by the asset classes’
the asset class increases in a non-normal framework, relative to distributional properties—its negative skewness, excess kurtosis
the allocation implied by a traditional framework. Each subsequent and extreme negative values (all of which are bad for investors).
column allows for the prior form (or forms) of non-normality—
hence, the attribution in each case is truly marginal. In particular, our model indicates higher excess kurtosis for
U.S. Equity, REITs and Hedge Fund of Funds compared to U.S.
Bonds, International Equity, Emerging Markets Equity and
Exhibit 50: Attribution analysis of non-normality Private Equity (after allowing for serial correlation). This
Phenomenon 3: implies that these asset classes have “fatter” left tails than the
Phenomenon 1: Phenomenon 2: Converging other asset classes, when compared to a normal distribution.
Serial correlation “Fat” left tails correlations
Hence, our model reduces allocations to U.S. Equity, REITs
U.S. Bonds and Hedge Fund of Funds (relative to a traditional framework)
U.S. Equity due to the prevalence of this phenomenon.
International
Equity
Emerging Converging correlations
Markets Equity
REITs Our calibration results for the copula and optimization results
suggest correlation convergence during periods of market stress
Hedge Fund
of Funds is much more pronounced for International Equity and Hedge
Private Equity Fund of Funds (after allowing for serial correlation and “fat”
left tails). Hence, the model reduces allocations to these asset
Source: J.P. Morgan Asset Management. For illustrative purposes only. classes relative to traditional mean-variance frameworks.
Note the impact of the form of non-normality on the optimal portfolio solution varies
35 In other words, the degree of non-linearity in correlations is
along the efficient frontier. The analysis in this section is broadly representative of a much more severe for International Equity and Hedge Fund
portfolio with an expected (arithmetic) return of 9.0%—based on our Long Term
Capital Market Return Assumptions. of Funds, compared to the other asset classes. This implies
that diversification benefits from these asset classes do
not materialize to the extent implied by linear correlation
matrices (during periods of market stress). This detracts from
relative allocations.

Conclusion
Incorporating non-normality can lead to more efficient portfolios
Until recently, investors have been constrained

in their ability to incorporate non-normality into
the asset allocation process. But now, with the
availability of sophisticated statistical tools, we
can meet this challenge. Why should we change?
The most straightforward answer is that this is how the world
really works—i.e. we empirically observe non-normality with
much greater frequency that current mean-variance frame-
works allow for. The more important answer is that ignoring
non-normality in equity (and equity type) return distributions
significantly understates downside portfolio risk—in the worst
of the worst-cases, potentially posing a solvency risk for the
investor.
We also believe that investors need to allow for downside risk in

a more robust fashion than standard deviation measures have
traditionally assumed. For this reason we recommend CVaR95.
This measure is a better fit for investors’ asymmetric risk pref-
erences, as well as the “fat” left tails recognized by non-normal
asset allocation frameworks.
Finally, an asset allocation incorporating non-normality has the

benefit of reducing the need for external constraints. Investors
often impose such constraints in an effort to get normal frame-
works to provide non-normal solutions—i.e. to better reflect the
non-normality we see in the real world. A framework that builds
in non-normality up front, however, provides much a more
direct, efficient, and elegant way of addressing the problem.
Ultimately, we believe the quantitative results illustrate the

point best, and speak for themselves: incorporating non-normality
may reduce the portfolio’s volatility, improve its efficiency
(Sharpe ratio), and reduce its risk relative to unpredictable, Limitations of reliance
extreme negative events.
It should be noted that a quantitative framework is only one input into the asset al-
location process and cannot replace the professional skill and judgment necessary to
So, we argue for a new asset allocation framework because
arrive at an appropriate strategy. The importance of allowing for subjective—and often
beyond its pure statistical merit, there lies a significant, qualitative—factors in decision making remains. Further, there is always an explicit
practical benefit for investors: the potential to improve portfolio need to account for the investor’s specific circumstances, including liabilities, when
efficiency and resilience, in light of a clearer understanding of arriving at an appropriate portfolio allocation.
portfolio risk.

Appendix
Exhibit 51: J.P. Morgan Asset Management Asset Allocation Framework—Graphical User Interface

Detailed Asset Allocations and Assumptions Underlying Efficient Frontiers
Exhibit 52: Efficient frontier portfolio allocations (Traditional Mean-Variance based on J.P. Morgan Long Term Expected Returns)
Mean CVaR Allocations Portfolio 1 Portfolio 2 Portfolio 3 Portfolio 4 Portfolio 5 Portfolio 6 Portfolio 7 Portfolio 8 Portfolio 9 Portfolio 10
U.S. Aggregate Bonds 80.2 52.0 24.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
U.S. Large Cap Equity 0.0 3.0 7.0 10.3 7.9 5.3 2.7 0.2 0.0 0.0
International Equity 2.5 2.4 0.1 2.2 8.8 11.8 15.0 18.1 4.4 0.0
Emerging Markets Equity 0.0 0.0 0.0 0.0 4.8 12.7 20.5 28.4 43.3 100.0
Real Estate Investment Trusts 0.0 1.8 6.0 10.4 12.6 14.4 16.2 18.1 15.2 0.0
Hedge Fund of Funds 17.3 40.7 62.9 77.1 57.7 40.7 23.6 6.5 0.0 0.0
Private Equity 0.0 0.0 0.0 0.0 8.2 15.1 22.0 28.8 37.1 0.0
Exhibit 53: Efficient frontier portfolio allocations (CVaR95 motivated non-normal framework based on J.P. Morgan Long Term Expected Returns)
Mean CVaR Allocations Portfolio 1 Portfolio 2 Portfolio 3 Portfolio 4 Portfolio 5 Portfolio 6 Portfolio 7 Portfolio 8 Portfolio 9 Portfolio 10
U.S. Aggregate Bonds 76.3 57.3 43.4 36.8 29.5 21.5 10.7 0.0 0.0 0.0
U.S. Large Cap Equity 7.5 15.8 20.9 21.5 22.6 23.7 25.1 24.7 6.8 0.0
International Equity 2.5 4.4 6.8 6.1 5.2 5.0 2.7 2.8 0.0 0.0
Emerging Market Equity 0.0 0.0 0.0 6.5 12.5 17.7 22.8 28.2 48.1 100.0
Real Estate Investment Trusts 3.6 7.1 9.7 10.8 11.7 12.8 15.0 16.7 12.2 0.0
Hedge Fund of Funds 10.1 15.4 15.1 9.2 4.3 0.0 0.0 0.0 0.0 0.0
Private Equity 0.0 0.0 4.1 9.1 14.1 19.3 23.7 27.6 32.9 0.0
j.p. morgan asset management Long-term capital market return assumptions
Historic
Expected annualized Historic
Assets arithmetic return volatility correlations
U.S. Inflation 2.76% 1.30% 1.00
U.S. Aggregate Bonds 5.59% 3.58% -0.03 1.00
U.S. Large Cap Equity 10.36% 15.15% 0.00 -0.06 1.00
International Equity 10.61% 15.27% 0.01 -0.16 0.87 1.00
Emerging Market Equity 13.50% 24.04% 0.04 -0.03 0.76 0.82 1.00
Real Estate Investment Trusts 9.95% 18.17% 0.10 0.17 0.46 0.39 0.44 1.00
Hedge Fund of Funds 8.12% 6.49% 0.10 0.08 0.57 0.68 0.78 0.30 1.00
Private Equity 13.09% 23.36% 0.00 -0.03 0.65 0.68 0.72 0.42 0.75 1.00

Based on ten years of monthly returns from November 1998–October 2008.

Bibliography
Bacmann, F., Gawron, G. 2004 Fat tail risk in portfolios of hedge funds Nystrom, K., Skoglund, J. (2002), “Univariate Extreme Value Theory,
and traditional investments. GARCH and Measures of Risk”, Preprint, Swedbank.
Bouye, E., Durrleman, V., Nikeghbali, A., Riboulet, G., Roncalli, T. Patton, A. (2006) Copula-Based Models for Financial Time Series
(2000), “Copulas for Finance: A Reading Guide and Some Applications”.
Rockafellar, R. and Uryasev, S. (1999) Optimization of Conditional
Christoffersen, P., Diebold, F.X., and Schuermann, T. (1998), Horizon Value at-Risk.
Problems and Extreme Events in Financial Risk Management,” Economic
Polcy Review, Federal Reserve Bank of New York, October, 109-118. Roncalli, T., Durrleman, A., Nikeghbali, A., (2000), “Which Copula Is
the Right One?”
Coleman, M., Mansour, A. Real Estate in the Real World: Dealing with
Non-Normality and Risk in an Asset Allocation Model. Zeevi, A., Mashal, R., (2002), “Beyond Correlation: Extreme Co-
movements between Financial Assets”, Preprint, Columbia University.
David Geltner, 1999, Using the NCREIF Index to Shed Light on What
Really Happened to Asset Market Values in 1998: An Unsmoother’s
View of the Statistics.
Dorey, M., Joubert, P. Modelling Copulas: An Overview (The Staple Inn

Actuarial Society).
Embrechts, P., McNeil, A., Straumann, D. (1999), “Correlation and

Dependence in Risk Management: Properties and Pitfalls”.
Fernandez, V. (2008) Copula-based measures of dependence structure

in assets returns.
Fisher, J., D. Geltner, and B. Webb. 1994. Value Indices of Commercial

Real Estate: A Comparison of Index Construction Methods. Also, Fisher,
J. and D. Geltner. 2000. De-Lagging the NCREIF Index: Transaction
Prices and Reverse-Engineering.
Georges Gallais-Hamonno, Huyen Nguyen-Thi-Thanh, 2007, the neces-

sity to correct hedge fund returns: empirical evidence and correction
method.
Giliberto, M. (2004) Assessing Real Estate Volatility, Journal of Portfolio

Management–Real Estate 2004.
Hu, W. 2007. Portfolio optimization for t and skewed t returns.
Idzorek, T.M. 2006. Developing Robust Asset Allocations.
Longin, F. 2004 The choice of the distribution of asset returns: How

extreme value theory can help?
Markowitz, H., Portfolio Selection, New York, NY: John Wiley & Sons, 1959.
Miller, M., Muthuswamy, J., Whaley, R. (1994) Mean Reversion of Standard &
Poor’s 500 Index Basis Changes: Arbitrage-Induced or Statistical Illusion?
Nelsen R.R. 1999. An Introduction to Copulas. Springer, New-York.

Authors
Abdullah Z. Sheikh, FIA, FSA Hongtao Qiao, FRM

Director of Research Strategic Advisor
Strategic Investment Advisory Group Strategic Investment Advisory Group
abdullah.z.sheikh@jpmorgan.com hongtao.j.qiao@jpmorgan.com

IMPORTANT DISCLAIMER
This document is intended solely to report on various investment views held by J.P. Morgan Asset Management. All charts and graphs are shown for illustrative purposes
only. Opinions, estimates, forecasts, and statements of financial market trends that are based on current market conditions constitute our judgment and are subject to
change without notice. We believe the information provided here is reliable but should not be assumed to be accurate or complete. The views and strategies described
may not be suitable for all investors. References to specific securities, asset classes and financial markets are for illustrative purposes only and are not intended to be,
and should not be interpreted as, recommendations. Indices do not include fees or operating expenses and are not available for actual investment. The information con-
tained herein employs proprietary projections of expected returns as well as estimates of their future volatility. The relative relationships and forecasts contained herein
are based upon proprietary research and are developed through analysis of historical data and capital markets theory. These estimates have certain inherent limitations,
and unlike an actual performance record, they do not reflect actual trading, liquidity constraints, fees or other costs. References to future net returns are not promises or
even estimates of actual returns a client portfolio may achieve. The forecasts contained herein are for illustrative purposes only and are not to be relied upon as advice
or interpreted as a recommendation. The value of investments and the income from them may fluctuate and your investment is not guaranteed. Past performance is no
guarantee of future results. Please note current performance may be higher or lower than the performance data shown. Please note that investments in foreign markets
are subject to special currency, political, and economic risks. Exchange rates may cause the value of underlying overseas investments to go down or up. Investments in
emerging markets may be more volatile than other markets and the risk to your capital is therefore greater. Also, the economic and political situations may be more
volatile than in established economies and these may adversely influence the value of investments made.
J.P. Morgan Asset Management is the marketing name for the asset management businesses of JPMorgan Chase & Co. Those businesses include, but are not limited to,
J.P. Morgan Investment Management Inc., JPMorgan Investment Advisors Inc., Security Capital Research & Management Incorporated and J.P. Morgan Alternative Asset
Management, Inc.
245 Park Avenue, New York, NY 10167
© 2009 JPMorgan Chase & Co.
J.P. Morgan Asset Management
245 Park Avenue I New York, NY 10167
www.jpmorgan.com/insight

AM Non-Normality of Market Returns

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

AM Non-Normality of Market Returns

Transféré par

Droits d'auteur :

Formats disponibles

Non-normality of Market Returns

A framework for asset allocation decision-making

About J.P. Morgan Asset Management

J.P. Morgan Asset Management | 1

7 Empirical evidence of non-normality

14 Incorporating non-normality into an asset

24 Estimating downside portfolio risk in a

27 Implications of non-normality on optimal

31 Conclusion: Incorporating non-normality

2 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

J.P. Morgan Asset Management | 3

• Unsmoothing Serial Correlation: Serial correlations can be

6 series is better reflective of the risk characteristics of the

4 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

Exhibit 2: International Equities—Applying Extreme Value theory

6 this step, we have to ask whether or not our conventional risk

We define Conditional Value at Risk (CVaR95) as the average real9

J.P. Morgan Asset Management | 5

These results are compelling: ignoring empiric observations of

The remainder of this paper is focused on statistical proofs of

6 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

Exhibit 3: Testing for serial correlation in monthly asset class returns

J.P. Morgan Asset Management | 7

We hypothesize that serial correlation is common in alternative

asset returns of our alternative asset classes. And a conse-

8 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

consider U.S. Equities. 50

Exhibit 4 shows a simple histogram of monthly returns for U.S. 40

1. The density function of the normal distribution14 10

In Exhibits 5 through 11, we show density function results for all 4

left tails relative to those implied by the normal distribution. The

distribution to model returns underestimates the frequency

J.P. Morgan Asset Management | 9

Source: J.P. Morgan Asset Management. For illustrative purposes only.

Exhibit 12: Empirical evidence of “fat” left tails in asset returns

10 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

60 More specifically, our two test periods are defined as follows:

50 1. The control period is a full ten-year history, from October

Source: J.P. Morgan Asset Management. For illustrative purposes only.

Exhibit 14: SIMPLE CORRELATIONS OVER TEN YEARS TO SEPTEMBER 2007

International Emerging Real Estate Hedge Fund of

Source: J.P. Morgan Asset Management. For illustrative purposes only.

J.P. Morgan Asset Management | 11

12 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

The results are striking. Nearly all the correlation coefficients

International Emerging Real Estate Hedge Fund of

Source: J.P. Morgan Asset Management. For illustrative purposes only.

J.P. Morgan Asset Management | 13

C. Correlation breakdown (converging correlations)

A. Incorporating the Impact of Serial Correlation

In order to ‘adjust’ for the impact of serial correlation on

adjusted return series should be more reflective of the risk 20

14 | Non-normality of Market Returns—A Framework for Asset Allocation Decision-Making

Source: J.P. Morgan Asset Management. For illustrative purposes only.

Source: J.P. Morgan Asset Management. For illustrative purposes only.

J.P. Morgan Asset Management | 15

International Emerging Real Estate Hedge Fund of

Source: J.P. Morgan Asset Management. For illustrative purposes only.

International Emerging Real Estate Hedge Fund of

Source: J.P. Morgan Asset Management. For illustrative purposes only.

Exhibit 21: Hypothetical dissection of return distribution

Left tail Interior Right tail

Source: J.P. Morgan Asset Management. For illustrative purposes only.