Documente Academic
Documente Profesional
Documente Cultură
Harlow
C ENTRAL
TOMODERNportfoliotheory overall magnitude of a loss, should one occur.
is the premise that investment decisions These seemingly disparate notions of risk, as
are made to achieve an optimal risk/ well as other possible definitions, serve as a
return tradeofffrom the availableopportunities. reminder that simple return variance (or stan-
In order to meet this objective, the portfolio dard deviation)-the traditional measure of
manager must first evaluate capital market in- risk-is sometimes deficient for dealing with the
formationand quantify ex antemeasures of both rich set of portfolio objectives and constraints
risk and expected return for the appropriateset that investment managers often formulate.
of assets. The next task is to isolate those This articlediscusses and demonstratesa gen-
combinations of assets that are the most "effi- eral approach to asset allocationbased on defi-
cient," in the sense of providing the lowest level nitions of risk that are attractivealternatives to
of risk for a desired level of expected return, and variance.1These alternativesall capture the ap-
then to select one combinationthat is consistent pealing notion of "downside risk" and provide
with the risk tolerance of the investor. a more robust approach to portfolio optimiza-
While the principle of identifying portfolios tion. Using an asymmetric measure of risk that
with the requiredrisk and return characteristics focuses on the returns below a specified target
is certainly clear, the appropriate definition of or benchmark return level, this framework in-
risk is more ambiguous. One manager might cludes as special cases such well known mea-
view risk as the probability of shortfall below sures as the probability of loss, expected loss
some benchmark level of return, for example,
while another may be more sensitive to the 1. Footnotes appear at end of article.
0% 6% 15%
Target Mean
As researchersin finance, economics and psy- where pp is the probability that return, Rp,
chology have noted over the past three decades, occurs.6The type of "moment," n, specified in
individuals view return dispersion in an asym- Equation (1) captures an investor's preferences
metric manner;that is, losses weigh more heav- by determining the manner in which the return
ily than gains.4 The examples of risk mentioned dispersion below the target is characterized.7
earlier-the probabilityof shortfall below some Figure A depicts a normal distributionwith a
benchmark level of return and the expected mean of 15 per cent and a standarddeviation of
magnitude of a loss-typify the concerns of 10 per cent. For a target rate of 6 per cent, the
portfolio managers. While these measures differ applicabledownside portion of the distribution
significantly,they both capturesome element of is representedby the shaded area. Forn = 0, the
downside risk. We focus our attentionon down- risk measure becomes a Oth-ordermoment (de-
side-risk measures in order to alleviate some of noted LPMO),with the term in brackets being
the shortcomings of variance and establish a
raised to the 0th power (i.e., equal to 1). Hence
more general approach to asset allocation.
the measure is simply the probabilityof falling
below the target rate. For this particularcase,
Downside-Risk Measures
Several classes of downside-risk measures are LPMoequals 0.184. In other words, there is an
18.4 per cent chance that the return perfor-
of particularinterest in finance. All involve the
tail of the relevant distributionof returns below mance will fall short of the desired minimum
some specific threshold level or target rate. level signified by the target rate.
These risk measures are referred to as "lower Forhigher-ordermoments, the shaded areain
partialmoments" (LPMs),because only the left- Figure A remains pertinent for the risk calcula-
hand tail of the returndistributionis used in the tion. However, for n = 1, LPM1becomes the
calculation.5 expected deviation of returns below the target,
Computationally, the LPM for an empirical or the target shortfall.8 LPM1for this example
(discrete) distribution of portfolio returns, Rp, equals 1.01 per cent. For n = 2, LPM2is analo-
with a target rate, i, is described by: goits to variance, in that it is a probability
weighting of squared deviations. Rather than
T computed around the mean of the distribution,
LPMn = > pp(t - Rp (1) however, the deviations are determined with
Ro = - xc respect to the target rate. LPM2 can thus be
1991El 30
FINANCIALANALYSTSJOURNAL/ SEPTEMBER-OCTOBER
referred to as a target semivariance. In our of portfolio optimization. The objective within
example, this measure equals 9.38 (or a 3.06 per this frameworkis essentially the same as in any
cent target semideviation, using the square root approach-i.e., select a portfolio of assets in
of the LPM2measure). some combinationso as to minimize risk subject
Many popular notions of risk are special cases to a specified level of expected return. In this
of the generalized LPMnmeasure. For example, case, however, LPMnis used as the appropriate
with n = 0 and a target rate equal to 0 per cent, characterizationof risk.
LPMois simply the probabilityof a loss. For n = Figure B illustrates this decision problem
2 and a target rate equal to the mean of the graphically.If Rp*is the investor's desired level
distribution, LPM2 becomes the traditional of return, then P* represents the portfolio that
semivariance measure. Furthermore, for nor- provides the lowest risk. For alternativevalues
mal, or symmetric, distributions, LPM2is ex- of Rp*,the resulting set of solutions trace out a
actly proportional to variance [LPM2(r = R) is convex mean-LPMnefficient frontier (denoted
equal to one-half of variance]. Using it as a risk MLPMn), reflecting the optimal risk/return
measure would be equivalent to using variance tradeoffs in exactly the same manner as the
and result in the same ordering of risky assets. traditionalmean-varianceapproach.
While the LPMnmeasure of risk has obvious Stated more formally, an investor who is
intuitive appeal, it is important to consider the averse to downside risk and who has a target
economic justificationfor its use and the general rate of return, r, must determine the allocation
conditions under which it is appropriate. The weight, Xj, for each relevant asset, j, to achieve
distinction between this framework and the an efficient point within the investment oppor-
traditional mean-variance approach lies in the tunity set. The nonlinear MLPMn optimization
assumptions regarding the distributionalprop- problem is represented by:
erties of returns and investor preferences. As
Select X to minimize: LPMn(T;X)
mentioned earlier, the use of variance as a
measure of risk requires a somewhat restrictive T
E(R)
MLPMn Efficient Frontier
Rp* /
portfolio construction. The primary reason for target, thus establishing a level of aversion to
this is that the LPMomeasure does not differ- below-target exposure. This constraint inter-
entially weight the returns in the lower tail of sects or is tangent to the efficient frontier, indi-
the distributionbased on their "distance"from cating which portfolios are attractive to the
the target. In other words, the shortfall proba- investor. The probability of target shortfall,
bility is an incomplete measure of risk, because therefore, can be thought of as a risk-tolerance
it fails to provide any indication of how severe assessment tool when coupled with another
the shortfallwill be, should it occur. As such, it measure of risk. This is true whether the mea-
is inappropriateto use this measure directly to sure is variance or a downside-risk criterion
define an efficient frontier.11 such as LPM1or LPM2.
Note that, while the probability of target The appeal of the downside-risk frameworkis
shortfallis not a complete measure of risk, it can certainlynot diminished because target shortfall
be incorporated into the asset allocation deci- probabilitiescannot be used directly as a risk
sion. Specifically, LPMo provides a means of measure in a portfolio optimization problem.
isolating the segment of the investment oppor- Indeed, beyond LPMO,there exists a very large
tunity set most relevant to the investor. As class of theoreticallyattractiveand flexible risk
Leibowitz, Henriksson and Kogelman have measures. LPM1 (target shortfall) and LPM2
shown, the shortfallprobabilitycan be specified (targetsemivariance)provide an intuitive set of
as a constraint in a portfolio optimization prob- risk definitions that are more useful than tradi-
lem that utilizes some other measure of risk- tional approachesand are valid under a broader
for example, variance-to delineate the efficient set of economic conditions.
frontier.12This shortfall restriction essentially
provides information regarding the risk toler- Downside-Risk Optimization
ance of the investor. To appreciate the downside-risk framework
For the mean-variance formulation of Equa- more fully, it is useful to examine its implemen-
tion (2), the shortfall constraintis equivalent to tation under alternative formulations. As with
an upward-sloping line in Figure B, with an any asset allocation decision, the analysis can
intercept on the return axis representing the proceed in a variety of ways, based on the
investor's target rate and a slope tied to the manner in which informational inputs are de-
probability of falling below the target. The in- veloped. In particular,the optimizationproblem
vestor specifies both the probability and the stated in Equation (2) can be solved using ex-
FINANCIALANALYSTSJOURNAL/ SEPTEMBER-OCTOBER
1991 O 32
plicit return forecasts with probabilitiesof real- loss of initialcapitaland thus has a target rate of
ization derived through some independent return, r, equal to 0 per cent. These frontiersare
modeling process. Alternatively, the analysis computed using Equation (3) and 60 months of
can be accomplished using historical data as recent historical return data (i.e., T = 60) to
proxies for ex ante asset behavior. derive estimates of the LPMnrisk measure and
Equation (2) can be modified to utilize histor- expected asset returns. Panel A depicts the
ical returns, as follows: MLPM1(target shortfall)frontier, while Panel B
illustrates the MLPM2frontier using the same
Select X to minimize: LPMn(r;X) target rate of 0 per cent. In the latter case, the
T1 square root of the LPM2risk value is plotted to
provide a measure comparable to percentage
=E (r-R )n (3) return (i.e., target semideviation). Both panels
Rp T-
also indicate the risk and return of a global
Subject to: n=1 or 2 capitalization-weighted benchmark composed
of 60 per cent equity and 40 per cent fixed
{fjXjE(Rj)= Rp*}and income securities, based on the Salomon-
Russell Global Equity Index and the Salomon
11AX= 1, Xj> 0}, BrothersWorld Government Bond Index.13
where T is the number of return observations. In Panel A, the 60/40 benchmark has an
In other words, within the optimization pro- expected return of 13.5 per cent and a target
cess, LPMncan be computed by constructingan shortfall of 9.07 per cent. In other words, be-
empirical portfolio distribution given a set of cause the investor's target rate has been speci-
asset weights, X, and T previously observed fied as 0 per cent, the magnitude of the expected
returns. This value is then minimized subject to probability-weightedloss is 9.07 per cent. By
an expected returnconstraint,Rp*.The expected comparison, an MLPM1efficient portfolio deliv-
return for each asset, E(Rj),is determined using ering the same benchmark expected return of
its historicalmean risk premium in conjunction 13.5 per cent has only a 4.53 per cent expected
with the current risk-free rate of interest, Rfo. target shortfall. In Panel B, which uses target
That is, the asset's excess return at time t, Rjt- semideviation to define the frontier, the risk of
Rft, is first averaged over T observations. This the benchmark is estimated at 23.20 per cent,
value is then added to Rfoto obtain the expected while the risk of the comparable MLPM2effi-
return, so that E(Rj)= [NT (Rjt - Rft)/T] + Rfo. cient portfolio is projected at 12.40 per cent.
Keep in mind that, for purposes of this dis- Figure D indicates the effect that altering the
cussion, we are interested primarilyin evaluat- target rate of return, T, has on the magnitude of
ing the usefulness of an alternative portfolio the estimated risk. For ease of comparison, we
constructiontechnique, as opposed to a method have included the previous frontiers, with tar-
of producing return forecasts. As most asset get rates equal to 0 per cent as well as those
allocationprocedures rely to some extent on the constructed using target rates of 8 and 16 per
information contained in historical returns, the cent. Forthe lattertwo cases, the specificationof
problem formulated in Equation (3) is repre- risk is based on measures of dispersion that
sentative of typical implementationsand is thus capture underperformance relative to higher
the focus of our attention. benchmark target levels of return. While the
We consider a global asset allocationproblem investor with a target return equal to 0 per cent
relevant to many portfolio managers. We in- is concerned with the loss of any principal,
clude fully currency-hedged equity and fixed these higher targets represent the risk of return-
income markets in 11 countries-the United ing less than the risk-freeasset (r = 8%)or less
States, the United Kingdom, Japan, West Ger- than the market portfolio (r = 16%).
many, Switzerland, France, the Netherlands, Inspection of the figures indicates that, as the
Sweden, Australia, Canada and Hong Kong. target rate increases, both the MLPM1 and
The returns used in this analysis span the 11- MLPM2frontiers shift to the right. This occurs
year period from January 1980 to December because more of the distributionof returns falls
1990. below the specified target rate. The downside
Figure C presents actual MLPMn efficient component of returns thereforebecomes larger,
frontiers for an investor who defines risk as any increasing the numerical value of the risk mea-
MLPM,Frontier
15
8 - (Target=0%)
Benchmark(60/40)
x; 10
5
0 5 10 15
15 MLPM2Frontier
(Target=0%) /
Benchmark(60/40)
a)
a.)
X 10
5 11 11 1 I 1 11 1 11 11il11llll
0 5 10 15 20 25 30
sure. Accompanying this shift is an increase in for example, the LPM1stock allocation is ap-
the allocation to stocks. This is certainly not an proximately10 per cent higher than that for the
intuitive result;rather,it represents the complex LPM2frontier (42.87 versus 32.83 per cent). In
interactionbetween the portfolio distributionof fact, the relative stock allocations for the LPM1
returns and the target rate of return. optimization are higher for all values of ex-
Another interesting result is that, for a given pected return, with the differences becoming
target rate and expected return, the LPM1and smaller toward each end of the frontiers.
LPM2allocationsdiffersubstantially.For T-equal This disparity highlights the importance of
to 0 per cent and a 15 per cent expected return, risk definition to the asset allocation decision.
Stocks:42.87%
Bonds:57.13%
154
av / / Stocks:43.72% S
x 10 Bonds:56.28% Bonds:53.59%
5-
0 5 10 15
LPM1 Target Shortfall (%)
Q) ~~~~~~~~~~~~~~~~~~~~~~~~Stocks:
36.73%
Q) 10 ; / / / \ |Bonds: 63.27%
X 10
5-
5 10 15 20 25 30
LPM2 Target Semideviation (%)
Simply changing the focus of attention from variance setting must be determined empiri-
target shortfall to target semideviation dramati- cally. If return distributions are approximately
cally alters the composition of the investor's normally distributed, then variance is a suffi-
portfolio. Indeed, this choice has a much larger cient measure of risk, and differences between
effect on the allocation decision than the choice the two approaches will be small. But if returns
of a target rate for a specific type of risk mea- are not symmetrically distributed about the
sure. mean, but instead possess some degree of skew-
The actual benefits of the downside-risk ness, then the asset allocation decision reached
framework relative to the traditional mean- within the downside-risk framework can be
PanelA: TargetShortfall(MLPM1)EfficientFrontier
20 -
- MLPM2Frontier(Target= 0%) .....*.
_ [ Stocks:32.83%
_ | Bonds:67.17%
Stocks:40.53%
Bonds:59.47%
;x J 10 Mean-Variance
Frontier
5-
5 10 15 20 25 30
LPM2TargetSemideviation(%)
significantly different from that obtained using normally distributed. The fact that they do not
mean-variancerelationships. It is to this poten- indicates the presence of skewness in the his-
tial distinction that we now turn our attention. torical return distribution. The variance risk
measure cannot capture the asymmetry in re-
Downside-Risk vs. Mean-Variance turns relative to the target rate.
Efficient Frontiers Figure E also indicates the stock/bond mix at
To make a fair comparison between the two the 15 per cent expected return level. For this
asset allocation frameworks, we use only the particularset of return data, the asset allocation
LPM2risk measure, because it, like variance, is under the MLPM2decision rule leads to a much
a second-order measure in which deviations higher bond allocation than that using the
from some return level are being squared. Fig- mean-varianceapproach(67.17 versus 59.47 per
ure E compares the MLPM2(r = 0%) efficient cent). The allocationdifferencesdiminish as the
frontier previously discussed with that com- frontiers converge toward each end. But at
puted using variance as the risk measure. (In intermediate return values, the two portfolio
order to display both frontiers on the same construction methodologies can lead to very
graph, we made LPM2 calculations for the differentasset allocation decisions.
mean-variance portfolios.) For both frontiers,
the square roots of the risk measures (i.e., target Downside-Risk vs. Mean-Variance
semideviation and standard deviation) are used Strategies
in order to reflect return-comparableunits. The differences between MLPM2and mean-
It is immediately apparent that the mean- variance portfolios, given a particular set of
variance frontier lies inside the MLPM2oppor- historicaldata, do not necessarily represent dif-
tunity set. Hence the MLPM2 portfolios are ferences in realized portfolio performance.It is
more efficient in that, for the same level of important to investigate the characteristicsof
expected return, they provide more downside these two alternative allocation strategies in
protectionthan those determined using a mean- terms of their ex post returns and risk.
variance decision rule. This holds over the en- To accomplish this task, we constructed port-
tire range of risk/return possibilities, with the folios using both optimization techniques. Be-
largest differences at intermediatevalues. ginning in January 1985, we used return data
Note that these frontierswould coincide if the over the preceding 60 months to develop the
historical returns used to construct them were required inputs. The return objective for both
umns (3) and (8), are also largerfor the MLPM2 S?SSSeE , =<
strategies. The higher ex post results for the \5 -
MLPM2approachappear to be an artifactof the n~~~~~~~~~c
U) m
t._ ,$CN
lower downside exposures of the MLPM2strat-
F
Q~~~4'
_ < s
1.75 -
1.50 A
tJ~~~ A
> 1.25 -A
1.00-
A
0.75 -
0.50 - A
0.25 AA A
0.00
-0.25
1.4 -0.50
-0.75 -
-1.00 -
-1.25 -
-1.50
-1.75
-5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0
RiskDifference(%)MLPM2vs MV
egies when compounded over multiple holding fixed income allocation, while column (9) pres-
periods. 4 ents a t-test of this value's significance.
As a separate evaluation of the differences In all cases, the average allocationto bonds is
between the MLPM2and mean-variance deci- higher for the MLPM2approach than for the
sion rules, the last two columns of Table I mean-variancemethodology and is significantly
summarize the below-targetreturns of both sets differentfrom zero. The differences range from
of portfolios. Specifically, column (12) indicates a low of 1.53 per cent (0/100 benchmark) to a
the percentage of times the MLPM2portfolios high of 2.68 per cent (90/10 benchmark). The
outperformed the mean-variance portfolios month-to-month differencesare as large as 9.49
when the latterportfolios returnedless than the per cent (90/10benchmarkin column (11)).
target. For all benchmarks,the MLPM2portfolio The MLPM2 approach's tendency to favor
returns exceed the mean-variancereturns in a bonds is also evident in the large percentage of
majorityof the 72 months. A t-test of this ratio, times in which the MLPM2fixed income alloca-
in column (13), shows that it is significantin six tion exceeds the mean-variance allocation (re-
of the 11 cases. ported in column (12) of the table). This per-
Figure F plots the annualized return differ- centage is in excess of 79 per cent for all
ence for each set of strategiesnoted in TableI, as benchmarks, reaching a high of 90.3 per cent
well as the reduction in target semideviation (20/80 benchmark). A t-test of this ratio, re-
obtained by using the downside-risk optimiza- ported in column (13), shows that it is signifi-
tion. Note that all observations fall in the north- cant for all sets of portfolios.
west quadrantof the figure, indicating the ben- These results, and those in Table I, suggest
efits of risk reduction and return enhancement that asset allocation in a downside-risk frame-
documented previously. work provides an attractive and powerful al-
Table II compares details of the asset mixes ternative to the traditional mean-variance
that generated the results reportedin TableI. In approach. The differences between the down-
particular,columns (2) through (7) indicate the side-risk and mean-variance approaches arise
mean fixed income allocation over the 72 because returns are not strictlynormallydistrib-
months, as well as the minimum and maximum uted. More important, the departuresfrom nor-
bond allocations for both strategies. Column (8) mality in past asset returns appear to contain
shows the average monthly difference in the informationuseful in ex anteportfolio decisions.
Conclusions
Downside risk is an intuitive measure of risk,
focusing on the asymmetry of returns about
< Q _
ocr
ec N rs N
oo a\cr o
et "D
\Cs N o oa, c\
ct CD \ N
L
some target level of return. The LPMn, as a
tn d I formal definition of downside risk, further en-
hances the appeal of this approach because of
u
the very simple theoretical assumptions re-
u u u Q
Q v
. U
l
U
q
U
f , U U
00 n 0
U
D
U U
:
.
quired to justify its use. Of equal import, a
broad set of well known risk measures, includ-
ing the traditionalvariance(standarddeviation)
> r ~~~~~~~~~~>
measure, are special cases of the LPMnframe-
work. Asset allocationin a downside-riskframe-
m x a,\ O H N 0 m 'IC N S
work, therefore, determines an investment op-
portunityset for downside-averse investors that
is at least as efficient as that derived using
conventional techniques.
o .E D ~~~~~N ri q \o \ q~o Ln o q More generally, the downside-risk approach
offers the potential for portfolios that are more
attractive than mean-variance portfolios. That
~~~~~~~aN ON ON aN ON (7 C?N x x x 00 o,
is, a downside-risk approach can lower risk
while maintaining or improving upon the level
of expected return offered by mean-variance
approaches.
u
4~~~~~~F t u o .
crtoN C: ooo o t, nH C
Beyond these appealing features, the down-
0 0 00~ ci t- lZ 1D 1D a? Cn tIn 0 b
>
side-risk approachallows the portfoliomanager
to define risk in a way appropriateto the objec-
W: S ~~~~
gC N 6 06 n ei 6 tf f- 0o UQ 14 -
tives and constraints of the portfolio. By being
able to specify a target rate of return and the
manner in which return deviations below this
target are considered, the manager can more
o~~~~~~~~~o
to - Ln 00 cn o, \0a fully appreciate the risk/returntradeoffs avail-
able from a given set of assets.'5 U
X~~~~~~~
C< C4 u6t C;L C;
Q e Wn ~~~N (OHNCN x x x x O O ;s
Footnotes
1. Extensionsof this frameworkto asset pricing are
discussed in W. V. Harlow and R. K. S. Rao,
"Asset Pricingin a GeneralizedMean-LowerPar-
X~~~~~~~~~~6L aL
, r
, 8L >t
tial Moment Framework,"Journalof Financialand
QuantitativeAnalysis, September 1989, and V.
co I t|t% Bawa and E. Lindenberg,"CapitalMarketEqui-
D
libriumin a Mean, Lower PartialMoment Frame-
U