Академический Документы
Профессиональный Документы
Культура Документы
T
A
M
ROGER CLARKE, HARINDRA DE SILVA, AND STEVEN THORLEY
R
FO
Y
N
A
IN
P
ROGER CLARKE ortfolio construction techniques or other portfolio constraints, which leads to
LE
is chairman of Analytic based on predicted risk, without easily replicable results.
Investors, LLC, in
C
expected returns, have become pop- The concept of risk parity has evolved
Los Angeles, CA.
TI
rclarke@aninvestor.com ular in the last decade. In terms of over time from the original concept that
individual asset selection, minimum-variance
R
Bridgewater embedded in research in the
H ARINDRA DE SILVA
is president of Analytic
and (more recently) maximum diversification
objective functions have been explored, moti-
A 1990s. Initially, an asset allocation portfolio
was said to be in parity when weights are pro-
IS
Investors, LLC, in vated in part by the cross-sectional equity risk portional to asset-class inverse volatility. For
TH
STEVEN T HORLEY objective functions to large (e.g., 1,000 stock) income subportfolio has a volatility of just 5
C
is the H. Taylor Peery investable sets requires sophisticated estimation percent, then a combined portfolio of 75 per-
U
Professor of Finance at techniques for the risk model. cent fixed income and 25 percent equity (i.e.,
D
in Provo, UT.
principal of risk parity, traditionally applied to be in parity. This early definition of risk parity
R
steven.thorley@byu.edu
small-set (e.g., 2 to 10) asset allocation deci- ignored correlations, even as the concept was
EP
sions, has been proposed for large-set security applied to more than two asset classes.
R
structures is based on standard unconstrained couching the property in terms of a risk budget
portfolio theory, matched with long-only sim- where weights are adjusted so that each asset
L
ulations. The empirical results in such studies has the same contribution to portfolio risk.
A
are specific to the investable set, time period, Maillard, Roncalli, and Teiletche [2010] call
G
maximum weight constraints, and other port- this an equal risk contribution portfolio,
LE
folio limitations, as well as the risk model. and analyzed properties of an unconstrained
IL
This article compares and contrasts analytic solution. Lees [2011] equivalent
risk-based portfolio construction techniques portfolio beta interpretation says that risk
IS
using long-only analytic solutions. We also parity is achieved when weights are propor-
provide a simulation of risk-based portfolios
IT
40 R ISK PARITY, M AXIMUM DIVERSIFICATION, AND M INIMUM VARIANCE : A N A NALYTIC P ERSPECTIVE SPRING 2013
42 R isk Parity, M aximum Diversification, and Minimum Variance : An Analytic Perspective Spring 2013
typically had betas of 0.3 to 0.5, but the minimum pre- As is common practice, the realized returns of each
dicted beta went into negative territory for several years asset class are shown as average annual (multiplication
in the mid-1990s. For the minimum-variance portfolio, by 12) monthly excess returns, although investors with
the long-only threshold beta varied between 0.6 and 0.8 longer holding periods may be more interested in com-
over time, meaning that only stocks with lower predicted pound annual returns. Thus, Exhibit 2 also provides an
beta values were admitted into solution each month. x for the position of the compound annual return to
The result is that the long-only minimum-variance each portfolio, calculated over the entire period of 1968
portfolio averaged only about 62 stocks in solution, with to 2012.3 The risk-based portfolios realized average-
a low of 30 and a high of 119. Similarly, the long-only return performance is interesting, but not part of the
maximum-diversification portfolio averaged about 82 portfolio construction process. Conclusions regarding
stocks in solution, with a low of 50 and a high of 137. the relative performance of the various portfolio con-
Although these counts may indicate higher portfolio struction techniques depend on the cross-sectional rela-
concentration than some managers are comfortable tionship between risk and return among the investable
with, they are largely a function of the aggressiveness of securities during the simulation period.
the shrinkage process used to translate historical security In order to make this point about historical perfor-
risk values into predicted values. mance in empirical simulations clear, Exhibit 3 shows
Exhibit 1 tabulates and Exhibit 2 plots the average the performance of risk-sorted quintile portfolios for
performance from 1968 to 2012 of the three risk-based the same 1,000 U.S. stocks over the same 1968 to 2012
portfolios, as well as the market (value-weighted) port- period. Each quintile portfolio contains two hundred
folio. Based on DeMiguel, Garlappi, and Uppal [2009], equally weighted stocks, assigned to portfolios each
we also include an equal-weighted 1,000 stock portfolio month based on their prior 60-month standard devia-
as a form of naive diversification. The returns throughout tion of excess returns. Exhibit 3 is consistent with the
this study are reported in excess of the contemporaneous low-risk anomaly first documented by Ang, Hodrick,
risk-free rate, measured by one-month Treasury bill Xing, and Zhang [2006] within the cross-section of U.S.
returns from Ibbotson Associates. stocks.4 The lowest and second-to-lowest risk portfolios
Although high average returns are not the explicit (i.e., quint 1 and quint 2) have the traditional risk return
goal of risk-based portfolio construction, all three risk- pattern, but the other three portfolios have an inverse
based portfolios outperform the excess market return relationship between risk and return.
of 5.3 percent. The risk-parity portfolio had an excess This potentially perverse relationship between
return of 7.4 percent, closely matching the return on the risk and reward is even worse when measured by com-
market-wide equal-weighted portfolio. The maximum pound returns, as shown with an x for each portfolio
diversification and minimum-variance portfolios had an in Exhibit 3. The high ex ante risk quintile does in fact
excess return of 5.7 percent, similar to the capitalization- have the highest realized risk, at about 27.1 percent,
weighted market portfolio. but with a compound annual excess return of only 2.7
EXHIBIT 1
Performance of Risk-Based Portfolios from 1968 to 2012
percent over the last 45 years. Although inconsistent and Sharenow [2000]. The market betas (single time-
with long-standing academic views of risk and return, series regression of realized portfolio returns) for the
Baker, Bradley, and Wurgler [2011] provided a reason- various portfolios are about 1.00, but with a notably
able explanation for the low-risk anomaly based on lower beta of 0.51 for the minimum-variance portfolio,
individual investor preferences for high-risk stocks, a key source of its low realized risk. Though the market
along with constraints associated with benchmarking portfolio includes all 1,000 securities each month, mar-
that prohibit larger institutions from fully exploiting the ket-capitalization weighting leads to an average Effective
anomaly. Frazzini and Pedersen [2013] also provided an N of only 138.5.
explanation based on lack of leverage access for indi- Alternatively, the average effective N of the risk-
vidual investors, who bid up the price (i.e., lower the parity portfolio is 934.1, close to an equally weighted port-
subsequent return) of high-risk stocks. folio. The effective N of the maximum diversification
For the three risk-based portfolios in Exhibit 1, and minimum-variance portfolios is even lower than the
the explicit objective of low risk is best achieved by the market portfolio, with average values of 46.3 and 35.7,
minimum-variance portfolio, with a realized risk of 12.4 respectively.
percent compared to the market portfolio risk of 15.5 Some investors traditionally view more securities
percent. As a result, the Sharpe ratio for 1968 to 2012 is in solution as equivalent to better diversification and
highest for the minimum-variance portfolio, followed lower risk. However, the relatively low number of secu-
by the risk parity and then the maximum diversification rities in the maximum diversification and minimum-
portfolios.5 variance portfolios illustrate that risk reduction is best
Exhibit 1 also reports each portfolios market achieved by selecting fewer, less correlated and less risky
beta, the average number of positions over time, and securities, rather than just adding more securities.
the average effective N, as defined by Strongin, Petsch,
44 R ISK PARITY, M AXIMUM DIVERSIFICATION, AND M INIMUM VARIANCE : A N A NALYTIC P ERSPECTIVE SPRING 2013
Of course, realized risk minimization depends on The security-risk ranges for 2013 are high by historical
the accuracy of security risk forecasts, specifically the standards, although not as high as the around the turn
spread in systematic and idiosyncratic risk. Alternatively, of the 21st century.
if a manager does not want to infer much from histor- As is typical over time, the scatter-plot in Exhibit 4
ical differences in security risks, the predicted security shows that the two measures of asset risk are somewhat
risk parameters would be shrunk towards a common correlated and that common factor beta, as well as
value, and the risk-based portfolios described in Equa- idiosyncratic risk, is positively skewed. The predicted
tions (3), (4), and (5) would converge to equal-weighted common factor risk (i.e., market portfolio) in January
portfolios.6 2013 is 19.5 percent, relatively high by historical stan-
dards. As we shall see, the impact of higher systematic
WEIGHT DISTRIBUTIONS AS risk leads to even more concentration in the minimum-
A FUNCTION OF RISK variance and maximum-diversification portfolios than
is typical over time.
We now turn to Equations (3), (4), and (5) for a Exhibit 5 shows asset weights plotted against beta
closer examination of the weight structure of the three for the three risk-based portfolios; minimum variance
risk-based portfolios at a specific point in time: Jan- ( markers), maximum diversification ( markers), and
uary 2013. We chose January 2013 because the weights risk parity (+ markers). Only 42 of the 1,000 invest-
employ the most recent five years of historical data: Jan- able stocks are in solution for the long-only minimum-
uary 2008 to December 2012. To support the subsequent variance portfolio in January 2013, ranging from a
weight charts, Exhibit 4 plots the predicted beta and weight of about 8.9 percent to zero.
idiosyncratic risk of all 1,000 stocks as of January 2013. The stocks with positive minimum-variance weights
Beta ranges from about 0.5 to 2.9, and idiosyncratic risk all have betas below the long-only threshold beta of 0.7
ranges from about 15 percent to more than 81 percent. for this period, in accordance with Equation (3). The
minimum-variance weights tend to fall on a negatively shown) instead of factor beta, the positive weights are
sloped line, also in accordance with Equation (3), although all associated with stocks that have correlations below
the correspondence is not perfect because of the impact of the long-only threshold correlation of about 0.4 for this
idiosyncratic risk. For example, the stock with the lowest period, in accordance with Equation (4). For example,
beta of 0.4 in Exhibit 4 has a minimum-variance weight the stock with a beta of 1.4 mentioned above has the
in Exhibit 5 of about 1.3 percent, much lower than the highest idiosyncratic risk in this set leading to a relatively
largest weight of 8.9 percent. low correlation to the common risk factor and thus a
Notice that the impact of idiosyncratic risk on this positive maximum diversification weight in Exhibit 5.
minimum-variance portfolio weight is greater than for As shown using the right-hand scale in Exhibit 5,
the same 0.4 beta stock in the maximum diversification the risk parity portfolio weights are positive for all 1,000
portfolio, because stock-specific idiosyncratic risk squared securities, and the largest weight of about 0.22 percent is
in the denominator of the first term in Equation (3) is not given to the stock with the lowest beta. The risk parity
offset by total stock risk, as in the numerator of the second weights in Exhibit 5 are remarkably aligned with factor
term in Equation (4). beta along a curve that has the hyperbolic functional
The long-only maximum diversification portfolio form indicated by Equation (6).
has 52 of the 1,000 investable stocks in January 2013, Because the risk parity portfolio includes all 1,000
with weights ranging from about 6.0 percent to zero. stocks, no single stock has an exceptionally large weight,
The stocks with positive weights tend to have low betas, and the idiosyncratic risks impact on the weights is
but several have high betas, including one maximum negligible. The lower concentration of the risk parity
diversification weight of about 1.5 percent for a stock portfolio does not mean, however, that it has less risk
with a beta of 1.4. When the maximum diversifica- ex ante than the other two portfolios. By design, the
tion weights are plotted against factor correlation (not minimum-variance portfolio has the lowest ex ante risk
46 R ISK PARITY, M AXIMUM DIVERSIFICATION, AND M INIMUM VARIANCE : A N A NALYTIC P ERSPECTIVE SPRING 2013
of all three (or any other long-only portfolio) and will the other hand, a key determinant of the concentration
generally have the lowest risk ex post depending on the in the maximum diversification and minimum vari-
risk models accuracy. ance portfolios is the value of the long-only threshold
Finally, as discussed in Lee [2011], the risk parity parameters, which are in turn a function of the level of
portfolio weights are perfectly aligned with the inverse systematic risk, in addition to the range of betas.
of another asset beta: the beta of each stock to the final For example, under the simplifying assumption of
risk parity portfolio (not shown). homogenous idiosyncratic risk (i.e., ,i is the same for
Charts of risk-based portfolio weights similar to all stocks), Equation (A-11) in the technical appendix
Exhibit 5 for dates other than January 2013 illustrate the for the minimum-variance portfolio long-only threshold
dynamic nature of market risk, and thus risk-based port- beta is
folio construction over time. In the mid-1990s, some of
2
the largest 1,000 U.S. common stocks exhibited negative + i2
market factor betas, and are consequently given relatively F i
2
L = L (6)
large weights in each of the three risk-based portfolios,
but particularly the risk parity portfolio. For example,
i
i
L
the Best Buy Corporation, a potentially countercyclical
stock, had a predicted beta of about -0.2 in 1995 and a Holding the cross-sectional distribution of betas
weight in the risk parity portfolio of more than 1%. fixed, the long-only threshold beta increases with the
As explained in the technical appendix, risk parity ratio of idiosyncratic to systematic risk in Equation (6),
portfolios can only accommodate assets with a limited increasing the number of securities in solution. Intui-
magnitude of negative common factor beta before the tively, higher idiosyncratic risk relative to systematic
goal of parity across all assets becomes untenable. On risk in the marketplace allows for more diversification
48 R ISK PARITY, M AXIMUM DIVERSIFICATION, AND M INIMUM VARIANCE : A N A NALYTIC P ERSPECTIVE SPRING 2013
the 1990s with broad asset allocation in mind, but now P2 = w i w j i , j (A-5)
i =11 j 1
applied to large-set security portfolios, may also be less
susceptible to an ex post bias.
where i,j are the elements in the asset covariance matrix
. A portfolio is said to be in parity if the total (rather than
APPENDIX marginal) risk contribution is the same for all assets
(A-9) k = K
2MV
(A-13)
+ (/ ) ' MV ,l Vk,l
1 2
2
F l =1
and can be substituted into the optimization solutions in In Equation (A-13), MV,l is the portfolio beta with
Equations (A-2) and (A-4). respect to the l th risk factor, and V k,l is an element of the
Specifically, the individual optimal weights in the K-by-K factor covariance matrix. The K-factor solution
long-only minimum-variance solution are given in Equa- in Equation (A-12) also has a long-only constrained ver-
tion (3), where L is a long-only threshold beta that cannot be sion, although the criterion for inclusion in the portfolio
exceeded in order for an asset to be in solution. The long-only involves multiple sorts and is therefore more complicated to
threshold beta is a function of final portfolio risk estimates calculate.
Substituting the inverse covariance matrix from Equa-
tion (A-9) into Equation (A-4) gives a relatively simple
2MV
L = (A-10) solution for individual weights in the maximum diversifi-
MV F2 cation portfolio, as shown in Equation (4), which includes
the term L as a long-only threshold correlation that cannot
where MV is the risk-factor beta of the long-only minimum- be exceeded for an asset to be in solution. The long-only
variance portfolio. As a more practical matter, the threshold threshold correlation can be described as a function of final
beta can be calculated from summations of individual asset portfolio risk estimates
risk statistics that come into solution,
2MD
L = (A-14)
1 2 A MD F
+ i2
F i L ,i
2
L = (A-11) where MD is the factor beta and A is the average asset risk,
i
i
2 ,i
respectively. As a more practical matter, the long-only
L threshold correlation can be calculated from summations of
individual asset correlations that come into solution,
Equation (A-11) is more practical than Equation (A-10),
in the sense that assets can be sorted in order of ascending i2
factor beta, and then compared to the summations until an
1+ 1
i <L
2
50 R ISK PARITY, M AXIMUM DIVERSIFICATION, AND M INIMUM VARIANCE : A N A NALYTIC P ERSPECTIVE SPRING 2013
solution to the maximum-diversification portfolio is
w RP , i = RP (A-19)
2 , i k =1 k N 2RP k =1 k
2MD i K
i
w MD , i =
,i A
2
1
k =1 k
(A-16)
2 2RP
where
r k = K
(A-20)
where k is a portfolio (i.e., not asset-specific) parameter
l =1
RP ,l
Vk ,l
2MD
k = K
(A-17) The general K-factor risk parity portfolio is long-only
A MD ,l Vk,l
l =1
by definition and can be used to inform a quick calcula-
tion routine similar to the one described for the single-factor
model.
In Equation (A-17), MD,l is the portfolio beta with Under the single-factor model, the implications of dif-
respect to the l th risk factor, and V k,l is an element of the ferences in beta and idiosyncratic risk on the relative magni-
K-by-K factor covariance matrix. The K-factor solution in tude of the weights in the minimum-variance and maximum
Equation (A-16) has a long-only constrained version, although diversification portfolios is fairly evident from the algebraic
criterion for inclusion in the portfolio involves multiple sorts form of Equations (3) and (4). The more complex form for
and is therefore more complicated to calculate. risk parity portfolio weights in Equation (5) makes the impact
Risk-parity portfolio weights under a single-factor risk of the different asset risk parameters less apparent, motivating
model can be derived by substituting the individual covari- formal calculus. The partial derivative of the risk parity asset
ance matrix terms of Equation (A-8) into Equation (A-6), weight in Equation (5) with respect to asset beta is
and then applying the quadratic formula. Using the positive
root in that formula gives Equation (5) in the body of the w RP , i w RP , i
article, where = (A-21)
i
1/2
/
2 1 2
, i
i2 +
N 2RP
2 2RP
= (A-18)
RP F2
and thus always negative, because weights in the long-only
solution are positive, as is the square-root term in the numer-
is a constant term across asset weights that includes RP, the ator of Equation (A-21).
risk parity portfolios beta with respect to the risk factor. The derivative is partial in the sense that Equation (A-21)
Unlike the equations for individual weights in the applies to the magnitude of asset weights compared to one
minimum-variance and maximum diversification portfolios, another in a risk-balanced set. A full derivative would be more
Equation (5) cannot be used to perform a simple asset sort complex, because a change in any single assets risk parameter
to calculate optimal weights, because final portfolio terms will change the risk parity weights for the entire set.
are embedded in the equation, not simply part of the scaling The functional form of Equation (5) (i.e., squared
factor. terms and square-roots) for risk parity portfolios indicates
However, Equation (5) does suggest a quick numerical that weights asymptotically approach zero with high beta
routine for even large investable sets. The numerical process and increase with low beta, including negative betas (if any).
starts with an equally weighted portfolio, and then iteratively Indeed, under the assumption of homogenous idiosyncratic
calculates the portfolio parameters (RP and RP ) and asset risk, risk parity weights form the positive side of a non-rect-
weights until the weights converge to one over N times their angular hyperbola, centered at the origin, with linear asymp-
beta with respect to the final portfolio, i,RP, in accordance totes of the X-axis for larger beta stocks, and a line given by
with Equation (A-7). We also note that the quadratic formula
allows for a negative root that is real. However, the positive
RP F2
root specified in Equation (5) is the only root that maintains w ( aasymptote ) = (A-22)
the budget constraint that the asset weights sum to one. 2
For a general K-factor risk model, the solution to the
risk parity portfolio is
52 R ISK PARITY, M AXIMUM DIVERSIFICATION, AND M INIMUM VARIANCE : A N A NALYTIC P ERSPECTIVE SPRING 2013