Risk Parity Portfolio vs.

Other
Asset Allocation Heuristic
Portfolios
DENIS CHAVES, JASON HSU, FEIFEI LI, AND OMID SHAKERNIA
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

T
DENIS CHAVES raditional strategic asset alloca- with alternative asset classes, at the margin,
is a senior researcher at tion theory is deeply rooted in the garnering only modest weights. It is unlikely
Research Affiliates, LLC,
mean–variance portfolio optimi- that this portfolio posture falls out of an exer-
in Newport Beach, CA.
[email protected] zation framework developed by cise in constrained portfolio mean–variance
Markowitz [1952] for constructing equity optimization;3 rather it is a hybrid child of
JASON HSU portfolios. However, the mean–variance legacy portfolio practice and return tar-
is CIO at Research optimization methodology is difficult to geting. Using historical realized risk premia
Affiliates, LLC, in New- implement due to the challenges associated to guide our capital market return expecta-
port Beach, CA, and pro-
fessor of finance at UCLA
with estimating the expected returns and tions, assuming a 9.0% equity return and a
Anderson Business School covariances for asset classes with accuracy. 6.5% bond returns, the 60/40 portfolio con-
in Los Angeles, CA. Subjective estimates on forward returns and veniently achieves the 8% portfolio return
[email protected] risks can often be inf luenced by behavioral target that is common to most pension funds.
biases of the investor, such as over-estimating As more asset classes, such as real estate, com-
FEIFEI LI expected returns due to the recent strong per- modities, and emerging market securities, are
is a director of research at
Research Affiliates, LLC, formance of an asset class or under-estimating added to the investment universe, weights are
in Newport Beach, CA. risk due to personal familiarity with an asset reallocated from stocks and bonds modestly to
[email protected] class. Empirical estimates based on historical these alternative assets. Most pension funds
data are often far too noisy to be useful, espe- hold a 60/40 equity/bond variant portfolio
OMID SHAKERNIA cially if risk premia and correlations for asset despite the significantly larger universe of
is a senior researcher at
Research Affiliates, LLC,
classes are time varying.1 Additionally, the investable asset classes. Undoubtedly, these
in Newport Beach, CA. possibility of “paradigm shift” in the capital incremental allocations improve portfolio
[email protected] market makes historical data far less relevant mean–variance eff iciency by improving
for forecasting the future evolution of asset diversification; however, it is also likely that
returns. This last concern is especially rel- more-optimal asset allocation methods or
evant today given the hypothesis on a “new heuristics can be created.
normal” for the global economy postulated
by Gross [2009]. RISK PARITY ARGUMENT
The challenges in the implementation of
Markowitz’s portfolio optimization have led Empirically, the risk (variance) of the
to a wide gap between the theory of the prac- traditional 60/40 equity/bond portfolio
tice and the practice of the theory.2 In prac- variants is dominated by the equity market
tice, institutional pension portfolios largely risk, since stock market volatility is signifi-
take on a 60/40 equity/bond allocation, cantly greater than bond market volatility.

108 R ISK PARITY PORTFOLIO VS. OTHER A SSET A LLOCATION H EURISTIC PORTFOLIOS SPRING 2011

JOI-CHAVES 108 2/16/11 12:27:07 AM

 Additionally, at the margin, the allocations to alternative In Exhibit 1, we show the historical return of
asset classes are too small to contribute meaningfully to the 60/40 S&P 500 Index/Barclays Capital Aggregate
the portfolio risk. In this sense, a 60/40 portfolio variant Bond Index (BarCap Agg) portfolio versus a risk parity
earns much of its return from exposure to equity risk and portfolio constructed from the same two assets. From
little from other sources of risk, making this portfolio a Sharpe ratio perspective, the risk parity construction
approach under-diversified in its risk exposure. does appear to be superior. While the unlevered risk
Proponents of the risk parity approach argue that a parity portfolio has a lower return, it can be levered up
more efficient approach to asset allocation is to equally to the same volatility as the 60/40 portfolio to provide
weigh the asset class by its risk (volatility) contribution to a better return than 60/40.7
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

the portfolio. This essentially allocates the same volatility A major benefit of risk parity weighting over mean–
risk budget to each asset class; that is, under the risk parity variance optimization is that investors do not need to
weighting scheme, each asset class contributes approxi- formulate expected return assumptions to form port-
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

mately the same expected f luctuation in the dollar value folios. The only input that needs to be supplied is asset
of the portfolio. Theoretically, if all asset classes have class covariances, which usually can be estimated more
roughly the same Sharpe ratios and same correlations, accurately than expected returns using historical data
risk parity weighting could be interpreted as optimal (Merton [1980]). Certainly, the covariance estimates
under the Markowitz framework.4 There is no official can have an impact on portfolio allocation; however,
definition for the risk parity methodology; product pro- it is unclear whether poor quality covariance estimates
viders use varying definitions of “risk contribution” and would bias the resulting portfolio returns downward.
different assumptions on the joint distributions for asset When compared with asset allocation products
classes; many even model the joint distributions as time (whether tactical or strategic, qualitative or quantita-
varying. In the two-asset case, all interpretation would tive), which are heavily focused on forecasting capital
roughly lead to the same portfolio, which is one that is market returns, the risk parity portfolio heuristic may
simply weighted by the inverse of the portfolio volatility. be considered more transparent and mechanical, which
In the multi-asset case, the portfolio constructions can mitigates the risk of behavioral biases inf luencing asset
differ very significantly and (time-varying) correlation allocation decisions. However, we do note that the com-
assumptions between assets can play a critical role. A sim- mercial products generally can and do involve some (if
plified risk parity approach that has anchored the practice not significant) manager discretion and that the exact
of some of the biggest players in this space is weighting method for measuring risk contribution and allocating
by inverse asset class volatility.5 Regardless of the exact the risk budget may not be fully disclosed. A recent
approach, the risk parity portfolio generally is fixed-in- report by Hammond Associates concludes, with regard
come heavy, which results in lower portfolio volatility to the managed commercial products, that “… there
and returns. Investors can then target the desired port- appears to be a lot of art involved.”
folio expected return by levering up the portfolio.
The strategy, of course, has its critics. Inker
[2010] questions whether asset classes like commodi-
ties and government bonds provide a positive risk
EXHIBIT 1
premium over cash in the long run; in the absence 60/40 vs. Risk Parity Portfolio Heuristic for Stock
and Bond, January 1980–June 2010
of a risk premium for a number of the asset classes
included for investment, the risk parity approach
would result in very a suboptimal portfolio. Lovell
[2010] and Foresti and Rush [2010] point out that
leveraging introduces new risks into the investor
portfolio, such as variability in financing costs and
availability of financing; it also amplifies the impact
Notes: The risk-free rate is the three-month Treasury bill from St. Louis FED
of tail events (like a liquidity crisis) on the investor (http://research.stlouisfed.org/fred2/series/TB3MS). S&P 500 Total Returns
portfolio.6 are from Global Financial Data (http://www.globalfinancialdata.com). BarCap
Agg Total Returns are from Barclays Capital Live (http://live.barcap.com).

SPRING 2011 THE JOURNAL OF INVESTING 109

JOI-CHAVES 109 2/16/11 12:27:07 AM

 OTHER COMPELLING PORTFOLIO U.S. investment-grade bonds, global bonds, U.S. high-
HEURISTICS yield bonds, U.S. equities, international equities,
emerging market equities, commodities, and listed real
Risk parity weighting is, of course, not the only estates. These asset classes are represented by the fol-
alternative asset allocation heuristic to the 60/40 equity/ lowing investable indexes, respectively: Barclays Capital
bond portfolio. In this article, we also consider two addi- U.S Long Treasury Index, Barclays Capital U.S. Invest-
tional asset allocation strategies which are more trac- ment Grade Corporate Bond Index, JP Morgan Global
table than the Markowitz mean–variance optimization Government Bond Index, Barclays Capital U.S. High
strategy and offer better risk premium diversification Yield Corporate Bond Index, S&P 500 Index, MSCI
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

than the 60/40 equity/bond strategy.8 EAFE Index, MSCI Emerging Market Index, Dow
Equal weighting. One of the most naive portfolio Jones UBS Commodity Index, and FTSE NAREIT
heuristics is equal weighting. Investors do not need to US Real Estate Index.
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

assume any knowledge regarding the distribution of For the mean–variance optimized strategy, we
the asset class returns. The equal-weighted portfolio is use the average return from the past five years as a
mean–variance optimal only if asset classes have the same forecast for future asset class returns. We also use the
expected returns and covariances. This strategy, empiri- monthly data from the past five years in conjunction
cally, provides superior portfolio returns when applied to with a standard shrinkage technique to estimate the
the U.S. and global equity portfolio construction.9 covariance matrix.11 The same covariance matrix is also
Minimum variance. Another popular approach for used to construct the minimum variance portfolio. We
constructing equity portfolios without using expected also construct a model U.S. pension portfolio with a
stock return information is the minimum variance 60/40 anchor, consisting of 55% stocks (80% U.S. and
approach. The approach utilizes the covariance infor- 20% international), 35% bonds (60% U.S. Long Trea-
mation but ignores expected returns information. Cova- sury, 20% investment-grade corporate, and 20% global
riances can also be estimated with a higher degree of bonds) and 10% alternative investments (2.5% each
accuracy using historical data (Merton [1980]) than commodities, REITs, emerging market equities, and
expected returns; the minimum variance methodology high-yield bonds). All strategies are rebalanced annu-
therefore focuses on extracting information that can be ally and are long-only portfolios.12 The weights in the
extracted with some accuracy from the historical asset mean–variance optimal strategy are constrained to less
return data. Note that the minimum variance portfolio than 33% to avoid extreme allocations.
is mean–variance optimal only if asset classes have the We simulate portfolio returns using asset class
same expected returns. Again, the minimum variance return data from 1980 through June 2010. The con-
strategy has demonstrated success when applied to equity structions are such that there are no look-ahead and sur-
portfolio construction.10 Chopra and Ziemba [1993] vivorship biases. Note that prior to 1989, the high-yield
show that, for stocks, the stark assumption that all stock index does not exist; prior to 1993, the emerging market
returns are equal can actually result in a better portfolio equity index does not exist. We simply omit those asset
than formulating an optimal portfolio based on noisy classes in the portfolio construction prior to their exis-
stock return forecasts. tence. We report the performance of the asset allocation
strategies in Exhibit 2. Admittedly, our choice of annual
A HORSE RACE BETWEEN RISK PARITY rebalancing is an arbitrary one—we would expect the
AND OTHER ASSET ALLOCATION Sharpe ratios to decrease slightly with more frequent
STRATEGIES rebalancing due to asset class momentum effect.13 By
comparing strategies according to their respective
In this section, we compare the risk parity strategy Sharpe ratios, we are implicitly assuming that investors
against other asset allocation strategies. In this horse will use leverage to achieve a required rate of return.14
race, we consider equal weighting, minimum variance, The time series of portfolio weights are reported in the
and a naïve mean–variance optimization, in addition Appendix.
to two variants of the 60/40 portfolio. The universe of
investible asset classes includes long-term U.S. Treasury,

110 R ISK PARITY PORTFOLIO VS. OTHER A SSET A LLOCATION H EURISTIC PORTFOLIOS SPRING 2011

JOI-CHAVES 110 2/16/11 12:27:08 AM

 EXHIBIT 2 lowest portfolio volatilities; only the minimum
Risk Parity vs. Other Portfolio Heuristics (with Nine Asset variance portfolio has a lower volatility. How-
Classes), January 1980–June 2010 ever, unlike our initial example in Exhibit 1
(and what is referenced in most studies on the
risk parity strategy), the Sharpe ratio of the
more diversified and comprehensive risk parity
portfolio is not higher than the 60/40 portfolio
variants, or a simple equal weighting of the
nine asset classes. Additionally, note that when
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

these portfolios are levered up to achieve the

same 5.1% excess return of the 60/40 bench-
mark, it is unclear whether their Sharpe ratios
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

would remain the same after financing costs.

Notes: The risk-free rate is the three-month T-bill from St. Louis FED (http:// More interestingly, the Sharpe ratio for the
research.stlouisfed.org/fred2/series/TB3MS). S&P 500 Total Returns are from Global stock/bond risk parity portfolio in Exhibit 1 is
Financial Data (http://www.globalfinancialdata.com). The BarCap Aggregate, U.S.
Long Term Treasury, U.S. Corporate Investment Grade, and U.S. Corporate High higher than the Sharpe ratio for the, arguably,
Yield Bond Total Returns are from BarCap Live (http://live.barcap.com). Global more diversified nine asset class risk parity port-
Bonds Total Returns through 1985 are from Global Financial Data and since 1986 are folio (0.62 vs. 0.51). This calls into question the
from Bloomberg ( JP Morgan Global Government Bond Index). REITs Total Returns
are from FTSI NAREIT Equity REITS series (http://www.REIT.com). MSCI
robustness of the methodology’s performance
EAFE and MSCI EM Total Returns are from MSCI (http://www.mscibarra.com/ advantage noted in different studies. We also
products/indices/global_equity_indices/performance.html). Commodities returns are the compare our results to a different horse race
Dow Jones-AIG Commodity Index from Global Financial Data (http://www.global- performed by Maillard, Roncalli, and Teiletche
financialdata.com).
[2010], who study portfolios constructed from
different asset classes and over a shorter horizon
DISCUSSION
(1995–2008) than in our study. They report the highest
Similar to previous findings based on U.S. and Sharpe ratio for their risk parity portfolio followed by
global equities, the mean–variance optimal approach minimum variance, with equal weighting coming in
underperforms the non-optimal strategies in out-of- last. This further substantiates one of the key messages
sample horse races, giving support to the claim that with in our article—that the observed risk parity performance
noisy inputs, optimized portfolio strategies are not nec- characteristics relative to other asset allocation alterna-
essarily optimal (Michaud [1989]). The mean–variance tives can be highly dependent on the time period and
optimized portfolio based on five-year historical aver- the asset classes included.
ages has a relatively low Sharpe ratio of 0.43, contrary In Exhibit 3, we take a closer look at the robust-
to the objective of the methodology, which is to have ness of the strategies by computing the subperiod Sharpe
the highest attainable Sharpe ratio. Using recent asset ratios for each decade since 1980. We see that the 60/40
class performance leads the mean–variance optimizer to strategy had a full sample Sharpe ratio of 0.50. How-
allocate aggressively to asset classes with high past-five- ever, the Sharpe ratio during the 1990s was nearly
year returns and/or low past-five-year risk. However, twice that at 0.99 and was only 0.04 during the 2000s;
this approach results in significantly lower risk-adjusted the 60/40 portfolio experience was dominated by the
future returns and seems to suggest mean-reversion in equity market performance, despite the massive bond
15
asset class returns. The second optimization approach, market rally in the 2000s. The Sharpe ratios for the
minimum variance, also produces disappointing results. equal-weighting and the risk parity portfolios have been
Although it achieves its objective of producing a low- comparably more stable over the last three decades than
volatility portfolio, its Sharpe ratio, which is the lowest the other strategies. This suggests that the full-sample
of all, is only 0.24. Sharpe ratio for the risk parity or equal-weighting port-
As expected, the risk parity strategy favors more folios would be good predictor of strategy performance
of the lower-risk asset classes, resulting in one of the for the next 10 years; whereas the full-sample Sharpe
ratio for the 60/40 benchmark, minimum variance, or

SPRING 2011 THE JOURNAL OF INVESTING 111

JOI-CHAVES 111 2/16/11 12:27:08 AM

 EXHIBIT 3 similar in terms of risk allocation. At the
Subsample Analysis of Sharpe Ratios: Risk Parity vs. Other Portfolio other extreme, we see that the minimum
Heuristics (with Nine Asset Classes), January 1980–June 2010 variance portfolio puts the bulk of its risk
allocation in less volatile bonds.

SENSITIVITY TO ASSET CLASS

UNIVERSE

Comparing the performance of the

It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

risk parity portfolios in Exhibits 1 and 2,

we find that the performance of the strategy
can be highly dependent on the universe of
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

asset classes we include. Which asset classes

and how many to include can be an art with
the risk parity strategy (as would be the case
Note: See Exhibit 2 Notes.
with equal weighting). The sensitivity to
asset class inclusion can also bring to ques-
the mean–variance optimal portfolios would not predict tion the validity of the documented superior
future strategy performance with high accuracy. empirical performance. The very act of selecting asset
We now turn our attention to one of the claims classes for the risk parity portfolio construction can add
by risk parity proponents, which is that the strategy elements of data mining and look-ahead bias into the
provides true diversification by allocating risk equally empirical research.
across asset classes. To evaluate whether that is indeed We illustrate the sensitivity to the asset class inclu-
the case, for each strategy we compute the percentage sion decision in Exhibit 5 and Exhibit 6. Specifically,
of the ex-post total portfolio variance attributed to each in Exhibit 5 we reduce the number of asset classes from
asset class. Since the portfolio return can be decomposed nine down to five, keeping only U.S. long Treasury,
to the weighted asset class returns, rp Σ iN=1w i ri , the port- U.S. investment-grade corporate, S&P 500, commodi-
folio’s total variance can be decomposed into sums of ties, and REITs. For the five asset class scenario, the
covariances of the weighted returns. Thus, the ex-post Sharpe ratios for both the risk parity and the equal-
risk allocation for each asset class is weighting strategies drop from 0.51 to 0.45 in the full
sample. In Exhibit 6, we add one new index into the
original nine asset class and the five asset class universes
∑
N
cov(w i ri , w j r j )
A cation to Asset i =
Risk Allo
j =1 of investments—the BarCap Aggregate Bond Index, an
var (rp ) index that is largely invested in intermediate-term U.S.
Treasuries. This is not a special asset, except that it has
Exhibit 4 shows the percentage of ex-post total had one of the best historical Sharpe ratios (0.82), pro-
variance attributed to each asset class for the portfolio ducing 7.3% return with 4% volatility in the last 30 years.
strategies under consideration. Although the risk allo- The BarCap Aggregate is also the driver of the impres-
cation for the risk parity portfolio is not exactly equal sive Sharpe ratio (0.62) for the stock/bond risk parity
across asset classes, ex post, it is indeed much more bal- portfolio reported in Exhibit 1; the S&P 500/BarCap
anced than the other strategies. Notice that the equal- Agg risk parity portfolio, on average, invests 80% of the
weighting portfolio has a higher risk allocation to the portfolio in the BarCap Agg index. The inclusion of this
riskiest asset classes. Since those risky assets typically low-risk bond index results in an improvement in Sharpe
demand a higher risk premium, the mean–variance ratios for both the equal-weighting and risk parity meth-
optimal strategy also tends to have more risk alloca- odology (from 0.51 to 0.54 for the nine asset class case
tion to the riskiest assets; hence the equal-weighting and from 0.45 to 0.50 for the five asset class case). Fur-
and the mean–variance optimal portfolios look quite thermore, this difference is especially pronounced in the
last decade. For shorter-horizon studies, the last decade

112 R ISK PARITY PORTFOLIO VS. OTHER A SSET A LLOCATION H EURISTIC PORTFOLIOS SPRING 2011

JOI-CHAVES 112 2/16/11 12:27:08 AM

 The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

JOI-CHAVES 113
SPRING 2011
EXHIBIT 4
Ex-Post Risk Allocation, January 1980–June 2010

THE JOURNAL OF INVESTING

113

2/16/11 12:27:09 AM
 EXHIBIT 5
Sensitivity of the Risk Parity Portfolio to Asset Class Inclusion, January 1980–June 2010
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

Note: See Exhibit 2 Notes.

EXHIBIT 6
Sensitivity of the Risk Parity Portfolio to Asset Class Inclusion, January 1980–June 2010

Note: See Exhibit 2 Notes.

would have disproportional inf luence on the empirical Exhibit 5 and Exhibit 6 suggest that, perhaps,
result. Investors should apply caution when examining including more asset classes produces better risk parity
the empirical benefit of leveraging up a fixed-income- portfolios. However, this is not generally the case. The
heavy risk parity portfolio. 2 asset class (S&P 500/BarCap Agg) risk parity portfolio

114 R ISK PARITY PORTFOLIO VS. OTHER A SSET A LLOCATION H EURISTIC PORTFOLIOS SPRING 2011

JOI-CHAVES 114 2/16/11 12:27:12 AM

 has a significantly better Sharpe ratio than the 10 asset problematic because there is little in the way of theory
class (9 + BarCap Agg) risk parity portfolio (0.62 vs. to guide the asset inclusion decision. It is not the case
0.54). Also the nine asset class risk parity portfolio has that including more asset classes leads to better portfolio
only an insignificant performance advantage over the results. Empirically, we also know that including low-
six asset class (5 + BarCap Agg) risk parity portfolio volatility fixed-income asset classes, which tend to have
(0.51 vs. 0.50). Further research is required to deduce high Sharpe ratios historically, can lead to better back-
a general relationship between the number of asset tested results. However, this is unlikely to be a sound
classes included and the resulting risk parity portfolio rule for investment; there may be reasons to question
performance. whether the high historical Sharpe ratio for bonds can
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

persist into the future. We believe that more research

CONCLUSION on methods for evaluating asset classes for inclusion into
a risk parity portfolio would provide tremendous value
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

Risk parity is an investment strategy that has to the industry.

attracted significant attention in recent years. We show
that this strategy has a higher Sharpe ratio than well
established approaches like minimum variance or mean–
APPENDIX
variance optimization, but it does not consistently out- PORTFOLIO WEIGHTS
perform a simple equal-weighted portfolio or even a
60/40 equity/bond portfolio. It does have some inter- These charts compare the time series of portfolio
esting characteristics such as a balanced risk allocation and weights for the different strategies. Mean–variance optimi-
less volatile performance characteristics (Sharpe ratios) zation clearly has the highest turnover, followed by minimum
over time. However, we also find that risk parity is very variance. Risk parity and equal weighting have similarly low
sensitive to the inclusion decision for assets. The meth- turnover. Not only do these two strategies have the best ex-
post performance, but the lower turnover also implies lower
odology is mute on how many asset classes and what
rebalancing costs.
asset classes to include. This last point is particularly

EXHIBIT A1
Time Series of Portfolio Weights

SPRING 2011 THE JOURNAL OF INVESTING 115

JOI-CHAVES 115 2/16/11 12:27:13 AM

 The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

JOI-CHAVES 116
116
E X H I B I T A 1 (continued)

R ISK PARITY PORTFOLIO VS. OTHER A SSET A LLOCATION H EURISTIC PORTFOLIOS

SPRING 2011

2/16/11 12:27:15 AM
 ENDNOTES We reference their results in a later section to arrive at a
conclusion regarding the robustness of the risk parity in-
1
See Merton [1980] for a discussion on the impact sample outperformance.
of time-varying volatility on the estimate for expected 9
See DeMiguel, Garlappi, and Uppal [2009] and Chow,
returns. See Cochrane [2005] for a survey discussion on the Hsu, Kalesnik, and Little [2010].
time-varying equity premium and models for forecasting 10
See Chopra and Ziemba [1993], Clarke, de Silva, and
equity returns. See Campbell [1995] for a survey on the time- Thorley [2006], and Chow, Hsu, Kalesnik, and Little [2010].
varying bond premium. See Hansen and Hodrick [1980] and 11
See Clarke, de Silva, and Thorley [2006].
Fama [1984] for evidence on time-varying currency returns. 12
The no-shorting constraint on the minimum vari-
See Bollerslev, Engle, and Wooldridge [1987] and Engle, ance and mean–variance optimal strategies is necessary for an
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

Lilien, and Robins [1987] for evidence on time-varying vola- apples-to-apples comparison, since both equal weighting and
tility in equity and bond markets. risk parity weighting implicitly start with no shorting.
2
See Michaud [1989] and Chopra and Ziemba [1993] 13
With monthly rebalancing, the Sharpe ratios for the
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

for discussions on problems with using the mean–variance 60/40, U.S. pension, risk parity, equal-weighting, minimum
optimization methodology for constructing portfolios. variance, and mean–variance optimal portfolio strategies are
3
Using 9.0% and 6.5% as expected stock and bond returns, 0.52, 0.50, 0.50, 0.47, 0.24, and 0.46, respectively.
respectively, the mean–variance optimal portfolio would 14
We used bootstrap resampling to compute standard
invest 9.3% in stocks and 90.7% in bonds; which would produce errors and compute t-tests of the differences of Sharpe ratios.
a portfolio with a Sharpe ratio of 0.67. The 60/40 equity/bond As one would expect given the similarity of the Sharpe ratios,
portfolio, by comparison, has a Sharpe ratio of 0.41. none of the strategies’ Sharpe ratios were statistically signifi-
4
For an exact mathematical proof for this statement, see cantly different from each other.
Maillard, Roncalli, and Teiletche [2010]. 15
See De Bondt and Thaler [1985] for evidence on
5
See Maillard, Roncalli, and Teiletche [2010] for details equity market mean-reversion and Asness, Moskowitz, and
on one reasonable execution of the risk parity portfolio con- Pedersen [2009] for evidence on mean-reversion for various
cept—an equal-weighted risk contribution portfolio; this asset classes.
methodology includes as a special case the inverse volatility-
weighted risk parity portfolio. Also see the research papers REFERENCES
by Qian [2005, 2009] and Peters [2009], which are product
provider white papers that provide discussions on their respec- Asness, Clifford S., Tobias J. Moskowitz, and Lasse Heje
tive risk parity strategies. Bridgewater promotes a version of Pedersen. “Value and Momentum Everywhere.” Working
risk parity that only focuses on the volatility and ignores the paper, 2009.
correlation information (or assumes a special case of constant
correlation for assets), which produces one of the simplest risk Bollerslev, Tim, Robert F. Engle, and Jeffrey M. Wooldridge.
parity methodologies. In our article, we adopt this simpler “A Capital Asset Model with Time-Varying Covariances.”
portfolio construction. We believe that the qualitative con- Journal of Political Economy, 96 (1987), pp. 116-131.
clusions are robust to the exact specification of the risk parity
methodology. Campbell, John Y. “Some Lessons from the Yield Curve.”
6
In a recent research report, Meketa Investment Group, Journal of Economic Perspectives, 9 (1995), pp. 129-152.
a U.S.-based institutional asset consultant, highlights these
very same risks to its clients. Chopra, V.K., and W.T. Ziemba. “The Effect of Errors in
7
We note that our data sample (1980–2010) coincides Means, Variances and Covariances on Optimal Portfolio
with a period of declining interest rates that is favorable to Choice.” Journal of Portfolio Management, Vol. 19, No. 2
the risk parity portfolio. We’d expect that the performance of (Winter 1993), pp. 6-11.
the risk parity strategy would be somewhat degraded during
rising interest rates. Furthermore, by performing subsample Chow, Tzee-man, Jason C. Hsu, Vitali Kalesnik, and Bryce
analysis, we see that the results can be highly dependent on Little. “A Survey of Alternative Equity Index Strategies.”
sample period. Working paper, 2010.
8
Maillard, Roncalli, and Teiletche [2010] also consider
a horse race between risk parity, equal weighting, and min- Clarke, Roger G., Harindra de Silva, and Steven Thorley.
imum variance. They use a different universe of assets and “Minimum-Variance Portfolios in the U.S. Equity Market.”
a shorter time period (1995–2008) whereas our data cover Journal of Portfolio Management, Vol. 33, No. 1 (Fall 2006),
1980 to 2010 and find different performance order ranking. pp. 10-24.

SPRING 2011 THE JOURNAL OF INVESTING 117

JOI-CHAVES 117 2/16/11 12:27:17 AM

 Cochrane, John H. Asset Pricing. Princeton, New Jersey, Maillard, Sebastien, Thierry Roncalli, and Jerome Teiletche.
2005. “The Properties of Equally Weighted Risk Contribution
Portfolios.” Journal of Portfolio Management, Vol. 36, No. 4
De Bondt, Werner F.M., and Richard Thaler. “Does the (2010), pp. 60-70.
Stock Market Overreact?” Journal of Finance, 40 (1985), pp.
793-805. Markowitz, Harry. “Portfolio Selection.” Journal of Finance,
Vol. 7, No. 1 (1952), pp. 77-91.
DeMiguel, Victor, Lorenzo Garlappi, and Raman Uppal.
“Optimal Versus Naïve Diversification: How Inefficient Is the Merton, Robert C. “On Estimating the Expected Return on
1/N Portfolio Strategy?” Review of Financial Studies, Vol. 22, the Market: An Exploratory Investigation.” Journal of Financial
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.

No. 5 (2009), pp. 1915-1953. Economics, Vol. 8, No. 4 (1980), pp. 323-361.

Engle, Robert F., David M. Lilien, and Russel P. Robins. Michaud, Richard. “The Markowitz Optimization Enigma:
The Journal of Investing 2011.20.1:108-118. Downloaded from www.iijournals.com by Joel Chernoff on 12/13/11.

“Estimating Time Varying Risk Premia in the Term Struc- Is Optimized Optimal?” Financial Analysts Journal, Vol. 45,
ture: The ARCH-M Model,” Econometrica, 55 (1987), pp. No. 1 ( January/February 1989), pp. 31-42.
391-407.
Peters, Ed. “Balancing Betas: Essential Risk Diversification.”
Fama, Eugene F. “Forward and Spot Exchange Rates.” Journal First Quadrant Perspective, Vol. 6, No. 2 (February 2009).
of Monetary Economics, 14 (1984), pp. 319-338.
“Portable Beta.” Presentation, AQR Capital Management.
Foresti, Steven J., and Michael E. Rush. “Risk-Focused
Diversification: Utilizing Leverage within Asset Allocation.” Qian, Edward. “Risk Parity Portfolios: Efficient Portfolios
Wilshire Consulting, February 11, 2010. through True Diversification,” Research Paper, Panagora,
September 2005.
Gross, William H. “On the ‘Course’ to a New Normal.”
PIMCO Investment Outlook, September 2009. ——. “Risk Parity Portfolios: The Next Generation,”
Research Paper, Panagora, November 2009.
Hansen, Lars Peter, and Robert J. Hodrick. “Forward
Exchange Rates as Optimal Predictors of Future Spot Rates: “Risk Parity.” White Paper, Meketa Investment Group, 2010:
An Econometric Analysis.” Journal of Political Economy, 88 http://meketa.com/documents/RiskParityWP_000.pdf.
(1980), pp. 829-853.

Inker, Ben. “The Hidden Risk of Risk Parity Portfolios.” To order reprints of this article, please contact Dewey Palmieri
GMO White Paper, March 2010. at dpalmieri@ iijournals.com or 212-224-3675.

Little, Pat. “Risk Parity 101.” Research Note, Hammond

Associates, July 2010.

Lovell, Christopher A. “Risk Parity: In the Spotlight after

50 Years.” NEPC, March 2010.

118 R ISK PARITY PORTFOLIO VS. OTHER A SSET A LLOCATION H EURISTIC PORTFOLIOS SPRING 2011

JOI-CHAVES 118 2/16/11 12:27:17 AM