You are on page 1of 58

Accepted Manuscript

Do Commodities add Value in Multi-Asset Portfolios? An Out-of-Sample Anal-


ysis for different Investment Strategies

Wolfgang Bessler, Dominik Wolff

PII: S0378-4266(15)00189-2
DOI: http://dx.doi.org/10.1016/j.jbankfin.2015.06.021
Reference: JBF 4770

To appear in: Journal of Banking & Finance

Received Date: 13 August 2014


Accepted Date: 27 June 2015

Please cite this article as: Bessler, W., Wolff, D., Do Commodities add Value in Multi-Asset Portfolios? An Out-
of-Sample Analysis for different Investment Strategies, Journal of Banking & Finance (2015), doi: http://dx.doi.org/
10.1016/j.jbankfin.2015.06.021

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Do Commodities add Value in Multi-Asset Portfolios?

An Out-of-Sample Analysis for different Investment Strategies

Wolfgang Bessler

Center for Finance and Banking


Justus-Liebig University Giessen, Germany

Dominik Wolff

Center for Finance and Banking


Justus-Liebig University Giessen, Germany

Current Version: June 21, 2015

*Corresponding Author: Wolfgang Bessler, Center for Finance and Banking, Justus-Liebig University
Giessen, Licher Strasse 74, Giessen, Germany, Email: Wolfgang.Bessler@wirtschaft.uni-giessen.de
Do Commodities add Value in Multi-Asset Portfolios?

An Out-of-Sample Analysis for different Investment Strategies

Abstract.

An essential motive for investing in commodities is to enhance the performance of portfolios


traditionally including only stocks and bonds. We analyze the in-sample and out-of-sample
portfolio effects resulting from adding commodities to a stock-bond portfolio for commonly
implemented asset-allocation strategies such as equally and strategically weighted portfolios,
risk-parity, minimum-variance as well as reward-to-risk timing, mean-variance and Black-
Litterman. We analyze different commodity groups such as agricultural and livestock com-
modities that currently are critically discussed. The out-of-sample portfolio analysis indicates
that the attainable benefits of commodities are much smaller than suggested by previous in-
sample studies. Hence, in-sample analyses, such as spanning tests, might exaggerate the ad-
vantages of commodities. Moreover, the portfolio gains greatly vary between different types
of commodities and sub-periods. While aggregate commodity indices, industrial and precious
metals as well as energy improve the performance of a stock-bond portfolio for most asset-
allocation strategies, we hardly find positive portfolio effects for agriculture and livestock.
Consequently, investments in food commodities are not essential for efficient asset allocation.

Current Version: June 21, 2015

Keywords: Commodities, Asset allocation models, Out-of-sample portfolio optimi-


zation, Diversification, Performance evaluation

JEL classification: C61, G10, G11


1. INTRODUCTION

During the last decade commodities have emerged as an attractive asset class for insti-

tutional asset managers and private investors. Consequently, commodity investments more

than doubled from roughly $170bn to $410bn from 2007 to 2013 (Croft and Norrish, 2013).

One explanation for this extraordinary growth is the potential diversification gain from com-

modities. The apparently low correlations of commodities with stocks and bonds is based on

the belief that commodity price changes are related to different risk factors such as weather,

geopolitical events, and supply conditions (Geman, 2005; Daskalaki et al., 2014). Moreover,

commodities are often seen as a natural hedge against inflation and thus seem attractive dur-

ing inflationary periods (Bodie and Rosanky, 1980; Erb and Harvey, 2006; Gorton and

Rouwenhorst, 2006). This study analyzes the in-sample and out-of-sample portfolio benefits

of commodities when added to a stock-bond portfolio for widely implemented asset allocation

strategies. We investigate individual commodities and commodity groups separately as well

as alternative commodity indices. The results suggest that aggregate commodity indices, in-

dustrial and precious metals as well as energy improve the performance of a stock-bond port-

folio for most asset-allocation strategies. Interestingly, we hardly find positive portfolio ef-

fects for agricultural and livestock commodities.

There exists a large body of literature documenting the diversification benefits from

adding commodities to a stock- or bond-portfolio. For different evaluation periods some stud-

ies report that switching from a stock-only portfolio to a portfolio that contains both stocks

and commodities improves the risk-return profile (Bodie and Rosanky, 1980; Greer, 1994;

Conover et al., 2010). Other studies find that the ex-post efficient frontier shifts upwards

when commodity futures complete the investment universe (Fortenbery and Hauser, 1990;

Satyanarayan and Varangis, 1996; Abanomey and Marthur, 1999; Jensen et al., 2000). More

-1-
rigorous approaches such as spanning tests analyze whether the shift of the efficient frontier is

statistically significant.1 Both mean variance (MV) (Scherer and He, 2008; Galvani and

Plourde, 2010) and non-MV spanning tests (DeRoon, Nijman and Werker, 1996; Daskalaki

and Skiadopolous, 2011) mostly reject the hypothesis that traditional asset classes span com-

modity futures returns. Hence, commodities significantly shift the efficient frontier upwards.

In contrast, Cao et al. (2010) report for the 2003 to 2010 period that including commodities in

the investment universe does not significantly shift the efficient frontier. Belousova and

Dorfleitner (2012) conduct spanning tests for 25 individual commodities, concluding that the

diversification benefit of commodities greatly varies between different types of commodities.

Thus, there exists no clear empirical evidence so far whether adding commodities to a stock-

bond portfolio is valuable for investors.

Another reason for the diverse results might be the different research setups. A major

shortcoming of studies analyzing shifts of the efficient frontier, either visually or with span-

ning tests, is that they analyze the contribution of commodities within an in-sample setting.

Therefore, these studies are limited to demonstrating only that commodities would have im-

proved the efficient frontier during a specific time-period [t, t*], if the asset returns for this

period were known in advance. Thus, in-sample analyses implicitly assume perfect forecasts

of expected asset returns, volatilities, and correlations. In reality, however, forecasts usually

contain large levels of estimation errors (Goyal and Welch, 2008). These are the reasons for

the lower performance of the out-of-sample optimized portfolios relative to the ex post effi-

cient frontier (Broadie, 1993; You and Daigler, 2012). As a result, in-sample spanning tests

are limited to investigating the maximum potential benefits of commodities in the absence of

1
Spanning tests investigate whether the return of a test asset is ‘spanned’ by a set of benchmark assets. If the
null hypothesis of spanning is rejected the augmented investment opportunity set of test assets and benchmark
assets leads to a significant enhancement of the initial efficient frontier of the set of benchmark assets
(Huberman and Kandel, 1987; Daskalaki and Skiadopolous 2011).

-2-
estimation errors. Consequently, they tend to overstate the achievable portfolio gains. There-

fore, a more realistic assessment of the attainable portfolio gains from commodities rests on

out-of-sample analyses. Here, the investor has to determine the portfolio weights at time (t)

for the subsequent period [t, t+1] using only data available at time (t).

So far, only a few out-of-sample studies investigate the portfolio effects of commodi-

ties, however, with contradicting results. Some studies report that the benefits of commodities

found in in-sample spanning tests are not observable in out-of-sample analyses (Daskalaki

and Skiadopoulos, 2011). In contrast, others conclude that commodities do improve the out-

of-sample performance of MV optimized portfolios (You and Daigler, 2012). Given these

opposite conclusions, the out-of-sample contribution of commodities remains ambiguous.

Moreover, these two studies are restricted to the sample-based MV and non-MV investment

strategies. These are subject to well known shortcomings including corner solutions (Broadie,

1993), high transaction costs resulting from extreme portfolio reallocations (Best and Grauer,

1991), and estimation error maximization (Michaud, 1989). These weaknesses may not only

distort the performance of out-of-sample portfolios but may also bias the portfolio benefits of

commodities.

In addition, several studies suggest that the growth and financialization of commodity

markets (Domanski and Heath, 2007) have resulted in higher correlations of commodity with

stock and bond returns (Silvennoinen and Thorp, 2013; Büyüksahin and Robe, 2014) and

within different commodity groups (Tang and Xiong, 2012; Gruber and Vigfusson, 2012;

Basak and Pavlova, 2013). Also monetary policy may affect commodity prices, offering a

different explanation for the higher correlations in recent years (Beckman et al., 2014). Higher

correlations, however, diminish the portfolio effects of commodities. In addition, Cheung and

Miu (2010) report that the diversification benefits of commodities are regime-dependent.

-3-
Beyond that, commodity investments particularly into agricultural commodities are

subject to an ongoing ethical debate. Critics claim that financial investors are to blame for the

increase in agricultural prices and in price volatility. The literature so far offers contrary views

with partly contradicting empirical findings (Cheng and Xiong, 2013; Pies et al., 2013) so that

the effects of commodity investments on prices and volatilities remain unsolved. Neverthe-

less, the public currently views investments in commodities very critically. Consequently,

institutional asset managers have to assess the tradeoff between potential benefits and reputa-

tional costs very carefully before offering commodity investments, especially those in agricul-

ture and livestock. Hence, it is important to analyze whether commodities in general and indi-

vidual commodities in particular provided the expected diversification benefits over longer

horizons and during the most recent financial crisis.

This study contributes to the literature in several dimensions. First, we analyze the in-

sample and out-of-sample portfolio benefits resulting from adding commodities to a stock-

bond portfolio for widely implemented asset allocation strategies including 1/N, strategically

weighted portfolios, risk-parity (RP), minimum-variance (MinVar), reward-to-risk-timing

(RRT), mean-variance (MV) and Black-Litterman (BL). By relying on several different asset-

allocation approaches, we overcome the shortcomings of using only MV. Because these asset-

allocation strategies are intensively discussed in the literature and employed by asset manag-

ers, the portfolio effects of commodities for these strategies should be of interest to academics

and practitioners alike. Most importantly, by analyzing the contribution of commodities for

different asset allocation strategies, we offer additional insights whether or not the portfolio

benefits of commodities depend on the implemented asset allocation approach. Second, we

investigate individual commodities and commodity groups separately, thereby evaluating

their potential portfolio effects. Third, we propose alternative commodity indices, excluding

-4-
controversially discussed livestock and agricultural investments. Fourth, we evaluate com-

modities for conservative and aggressive investment strategies as their portfolio effects might

depend on the investment style. Finally, we analyze the portfolio benefits of commodities in

different market environments (sub-periods) and over time using rolling Sharpe ratios.

Our empirical results confirm that the attainable out-of sample gains when adding

commodities to a stock-bond portfolio are much smaller than suggested by in-sample anal-

yses. Hence, in-sample analyses tend to portray a too optimistic picture in favor of commodi-

ties. We also observe that the portfolio effects vary for different commodities and sub-periods.

While aggregate commodity indices, industrial and precious metals as well as energy improve

the performance of a stock-bond portfolio for most asset-allocation strategies, we hardly find

any performance gains for agricultural and livestock commodities. An equally weighted

commodity index, excluding livestock and agriculture, generates a superior performance for

most asset-allocation strategies. Consequently, investments in food commodities are not es-

sential for efficient portfolio allocation. Our results hold for a variety of robustness checks

and alternative datasets.

The remainder of this study is organized as follows. In section 2 we present the em-

ployed asset-allocation models and performance measures. Section 3 describes our dataset of

commodity futures and financial assets. In Section 4 we discuss our empirical results and pre-

sent the robustness checks in section 5. Section 6 concludes.

2. METHODOLOGY

2.1. Review of Asset Allocation Models

For different asset allocation strategies, which were intensively analyzed in the literature and

that are commonly employed by asset managers, we evaluate the out-of-sample benefits for

-5-
different commodity groups when added to a stock-bond portfolio. The strategies include two

naive diversification rules ‘equally weighted’ (1/N) and ‘strategically weighted’ portfolios

(st.w.), the two simple asset allocation rules ‘risk-parity’ (RP) and ‘reward-to-risk-timing’

(RRT), as well as three portfolio optimization approaches: minimum-variance (MinVar),

mean-variance (MV) and the Black-Litterman (BL) model. Table 1 provides an overview of

the tested asset allocation approaches.

We group the investment strategies based on the number of input parameters they re-

quire. While the naïve asset-allocation strategies do not require any estimation of input pa-

rameters, the two risk-based investment strategies rely on the estimation of asset volatilities

(risk-parity) or volatilities and correlations (minimum-variance). The return-to-risk timing

(RRT), the mean-variance (MV) and the Black-Litterman (BL) approach additionally use es-

timates of future returns. If future returns and the future covariance matrix were certain, the

mean-variance (MV) strategy would dominate all other strategies. However, estimation errors

in the input parameter can lead to an inferior performance of MV (DeMiguel et al., 2009).

Because estimation errors in returns usually are higher than estimation errors in the covari-

ance matrix (Chopra and Ziemba, 1993), it might be beneficial to exclude return estimates and

to focus solely on risk estimates. This is one explanation for the popularity of the minimum-

variance and the risk-parity strategies. Another group of strategies tries to reduce estimation

errors in mean-variance optimization by using re-sampling techniques (Michaud, 1989) or

Bayesian estimation methods (Jorion, 1985, 1986; Black and Litterman, 1992). We implement

the Black-Litterman (BL) model as the most advanced approach in this class of models.

2.2. Out-of-Sample Estimation Procedure

To implement out-of-sample asset allocation strategies, we employ a rolling sample approach

as in DeMiguel et al. (2009). Portfolio weights are computed at every first trading day of each

-6-
month t based on data available up to month t (k observations). These weights are then used to

calculate the portfolio performance during the next month [t, t+1]. We repeat this process by

moving the sample period one month forward and computing the optimized weights for the

next month. This rolling sample approach is used to compute the out-of sample portfolio per-

formance for the different asset allocation strategies for the evaluation period from January

1986 to December 2013. To ensure the robustness of our results, we use different window

lengths ranging from 12 to 60 months and also apply alternative portfolio rebalancing fre-

quencies (monthly, quarterly, and yearly). To examine the impact of different levels of risk

aversion and different maximum desired portfolio volatilities, we distinguish between con-

servative and aggressive investor clienteles and analyze the benefits of commodities for each

investor type separately. As realistic investment constraints, we implement a budget re-

striction and exclude short selling. We distinguish between six sub-periods to investigate the

impact of economic cycles. Various robustness checks are performed to test whether our re-

sults are sensitive to the employed dataset, variations in the input parameters, estimation win-

dows, transaction costs, and short selling restrictions.

[Table 1 about here]

2.3. Asset Allocation Models

2.3.1. Naïve diversification rules

First, we analyze the portfolio benefits of commodities for the group of naïve diversification

strategies, which are particularly popular among private investors (Benartzi and Thaler,

2001). We implement a 1/N strategy which distributes wealth equally among the N selected

assets. We also use strategically weighted portfolios (st.w.) in which each asset has a strategic

weight that is constant over time. In contrast to 1/N, the strategically weighted strategy allows

-7-
us to set different strategic weights for different investor clienteles, reflecting their specific

risk aversion.

We distinguish between conservative and aggressive investor clienteles and set for

both investor types different strategic weights. Following earlier studies we set the strategic

weights for commodities to 5% and 15% for the conservative and aggressive investor clien-

teles, respectively (Anson, 1999; Erb and Harvey, 2006; Conover et al., 2010). Based on the

discussions with asset managers, the strategic weights for bonds are set to 80% and 20% and

for stocks to 15% and 65% for the conservative and aggressive investor, respectively. Table 2

summarizes the strategic weights for the different investor types. For the base case of a stock-

bond portfolio without commodities, the strategic weight for commodities is zero and more

wealth is allocated to stocks. We compute the naïve-diversified portfolios with rebalancing to

maintain the naïve 1/N or strategic weight for each asset over time.

[Table 2 about here]

2.3.2. Risk-parity

Recently, the risk-parity (RP) approach has attracted great interest from both academia and

practitioners alike. It is implemented by a growing number of funds and index providers (e.g.

Aquila Risk-Parity, Invesco Balanced Risk, MSCI) as well as by pension funds, endowments,

and other long term investors (Anderson et al., 2012; see Maillard et al. (2009) for a discus-

sion of this class of models). The idea of risk-parity is that each portfolio component contrib-

utes equally to portfolio risk. In the simple risk-parity approach correlations between asset

returns are neglected and assets are weighted anti-proportional to their sample variance :

1 / ˆ i2 ,
i  N
(1)

i 1
(1 / ˆ i2 )

-8-
Anderson et al. (2012) report that the risk-parity strategy performs well and usually outper-

forms 1/N, value-weighted, or 60/40 portfolios. This approach profits from the low-volatility

anomaly, which denotes the empirical finding that low volatility assets usually earn a higher

premium per unit of volatility than high-volatility assets (Baker et al., 2011; Frazzini and Ped-

erson, 2014). For our multi-asset portfolios including stocks, bonds and commodities, we ex-

pect that bonds obtain the largest portfolio weight, while commodities and stocks should re-

ceive similar but relatively lower weights due to their higher volatility. The relatively large

allocation to bonds should be preferable for conservative investors and be particularly advan-

tageous in a falling interest rate environment.2

2.3.3. Reward-to-risk timing

Kirby and Ostdiek (2012) propose a reward-to-risk timing (RRT) strategy that does not re-

quire optimization. This strategy sets portfolio weights based on the historical reward-to-risk

ratio, which is measured as the sample mean return of asset i ( ) divided by the sample vari-

ance ( ) of the asset. The reward-to-risk timing strategy overweighs assets with a high re-

turn and low variance in the sample period and, hence, is similar to a momentum strategy that

accounts for both risk and return. Formally, the weights of the RRT strategy are derived by

setting the off-diagonal elements of the covariance matrix equal to zero and solving for the

tangential portfolio. The portfolio weights of the RRT strategy are:

ˆ i / ˆ i2
i  N
. (2)

i 1
(ˆ i / ˆ i2 )

To prevent short sales, we set the weight of asset i equal to zero, if otherwise it would be

2
For the aggressive investor the strategy can be leveraged to achieve a larger expected return with a higher risk
level. We do not report the results for leveraged strategies, because the portfolio benefits of commodities do not
change when using leverage in the investment strategies.

-9-
negative by setting . In case all sample asset returns ( ) are negative, which

should not be a frequent problem in a multi-asset portfolio including stocks and bonds, we

assume that wealth is distributed equally among all assets. Kirby and Ostdiek (2012) find that

RRT outperforms MV in most cases after transaction costs due to its lower turnover. 3 In con-

trast to RP, the RRT strategy additionally incorporates expected asset returns. Consequently,

we expect that the RRT strategy allocates more wealth to stocks and commodities as these

asset classes typically provide higher returns than bonds. Therefore, the RRT strategy should

generate more risky portfolios than RP and might be a more appealing strategy for aggressive

investor clienteles.

2.3.4. Minimum variance portfolio

The MinVar approach is increasingly popular among investors and is implemented by various

quantitative investment funds and exchange-traded products. The objective of the minimum

variance (MinVar) strategy is to minimize portfolio risk measured as variance of portfolio

returns. The minimization problem is:

min '  , (3)


where ω is the vector of portfolio weights and Σ is the covariance matrix of the asset returns.

The advantage of MinVar is that it does not require any return estimates, which are usually

subject to large estimation errors. Just like RP, it benefits from the higher premium per unit of

volatility of low-risk assets (low volatility anomaly). In contrast to the simple RP approach,

the MinVar strategy also takes the correlation of assets into account. In the MinVar strategy,

3
The RRT strategy keeps turnover at a low level because it does not include optimization but directly computes
portfolio weights from the reward-to-risk of the assets during the sample period according to equation (2). Since
the reward-to-risk is estimated over the last k months, the change in portfolio weights from one month to the next
is relatively small because k-1 identical return observations are used to compute the portfolio weights in two
subsequent months.

- 10 -
we expect that commodities obtain a substantial portfolio weight if volatility is low and/or if

correlations with the traditional asset classes are small or even negative.

2.3.5. Mean-variance (MV) portfolio

The MV approach (Markowitz, 1952) builds on a tradeoff between risk and return. The mean-

variance optimization problem is defined as follows:


max U  '   '  , (4)
 2

where U is the investor’s utility, ω is the vector of portfolio weights, Σ is the covariance ma-

trix, μ is the vector of return estimates and δ is the risk aversion coefficient. To ensure that the

optimization strategies generate comparable risky portfolios, i.e. they have a similar volatility

as the strategically weighted portfolios, we use an upper volatility bound as constraint in the

optimization strategies according to equation (5).

'    ˆ C , (5)

Furthermore, Frost and Savarino (1988) and Jagannathan and Ma (2003) show that

imposing portfolio constraints is equivalent to shrinking the input parameters preventing ex-

treme portfolio allocations and enhancing the out-of-sample performance. Based on the his-

torical volatility of the strategically weighted portfolio before the evaluation period, we set the

volatility constraint to 5% and 15% for the conservative and aggressive investor, respectively.

Following Fletcher and Hillier (2005) and Daskalaki and Skiadopoulos (2011), we set differ-

ent risk aversion coefficients for the two investor clienteles to be 10 for conservative and 2 for

aggressive investors, respectively.

- 11 -
2.3.6. Black-Litterman portfolio

The Black-Litterman (BL) model has gained in popularity among quantitative portfolio man-

agers (Satchell and Scowcroft, 2000; Jones et al., 2007). It is an important extension of MV

and aims at reducing estimation errors in the input parameters. It combines two sources of

information: ‘implied’ returns that are derived from a market or benchmark portfolio using a

reverse optimization technique and ‘subjective’ return estimates – also referred to as ‘views’

(Black and Litterman, 1992). One major advantage of the BL approach is that the reliability of

each return estimate can be included. Black and Litterman (1992) compute the combined re-

turn estimates as:

ˆ
BL

   P'  1P
1
 
1 1

  P'  1Q , (6)

where П is the vector of implied asset returns, Σ is the covariance matrix, and Q is the vector

of the investor’s return estimates. We measure the reliability of each return estimate as the

variance of the historical forecast errors during the estimation window. The reliability

measures are written on the diagonal in the matrix Ω. P is the identity matrix and τ is a param-

eter that can be used to calibrate the tracking error to the benchmark portfolio (see Bessler,

Opfer and Wolff, 2015, for a detailed exposition). The combined return estimate is a matrix-

weighted average of ‘implied’ returns and ‘views’ (Lee, 2000) with respect to the correlation

structure. The posterior covariance matrix is derived as (Satchell and Scowcroft, 2000):


BL

      P'  1P
1

1
. (7 )

After computing combined return estimates and the posterior covariance matrix we

conduct a traditional risk-return optimization, maximizing the investor’s utility as presented in

equation (4). The utility function, risk-aversion coefficient, optimization procedure, and con-

- 12 -
straints are the same as for the MV approach. The BL model only differs in the input parame-

ters. While for MV the sample mean returns and the sample covariance-matrix are used,

the BL framework employs combined return estimates and the posterior covariance ma-

trix . Most importantly, the BL approach additionally incorporates the reliability of return

estimates. We implement a sample-based version of the BL model. 4

2.4. Performance measures

To evaluate the net benefits of commodities in complementing a stock-bond portfolio, we

compute several performance measures net of transaction costs. We assume proportional

transaction costs of 30 basis points of the transaction volume and vary transaction costs as

robustness check. As in DeMiguel et al. (2009) we compute the portfolio turnover PTi of

strategy i as the average absolute change of the portfolio weights ω over the T rebalancing

points in time and across the N assets:

PTi 
1 T N
 
 i, j ,t 1  i, j ,t  ,
T t 1 j 1
(8)

in which is the weight of asset j at time t in strategy i; is the portfolio weight be-

fore rebalancing at t+1; and is the desired portfolio weight at t+1, after rebalancing.

is usually different from due to changes in asset prices between t and t+1.

We compute the portfolio’s average out-of-sample net return and volatility as well as

the net Sharpe ratio as the average net excess-return (average return after transaction costs

less risk-free rate) divided by the volatility of out-of-sample net returns. Following Opdyke

4
For an application, see Bessler, Opfer and Wolff (2015) and Bessler and Wolff (2015). In the sample-based
version of the BL model ‘views’ are the sample means of the respective asset returns. The reliability of ‘views’
is measured as the variance of the historical forecast errors εi during the sample period ‘Implied’ returns are
computed based on the strategic weights presented in table 2. The parameter τ is set to 0.05. Earlier studies use
similar values ranging from 0.025 to 0.3 (Black and Litterman, 1992; He and Litterman, 1999; Idzorek, 2005).

- 13 -
(2007) we test whether the difference in Sharpe ratios of two portfolios is significant.5 As

alternative performance measure, we calculate the Omega measure, which is the ratio of aver-

age gains to average losses (Shadwick and Keating, 2002). We define gains as returns above

the risk-free rate and losses as returns below the risk-free rate. The advantage of the Omega

measure is that it does not require any assumption on the distribution of returns.

3. DATA
To analyze the diversification effects of commodities, we take the perspective of a US asset

manager already holding a portfolio consisting of stocks and bonds (similar to Daskalaki and

Skiadopoulos, 2011). Stocks are represented by the S&P 500 index and bonds by the Barclays

US aggregate government bond index. For the commodity investments, we rely on the S&P

Goldman Sachs commodity index family. These indices have a dominant position in the

commodity market because a large number of index funds track the performance of these in-

dices (Stoll and Whaley, 2011; Tang and Xiong, 2012). We concentrate on three different

commodity indices. The diversified S&P GSCI index (1) enables investors to participate in

the average price development of energy, industrial metals, precious metals, agriculture, and

livestock. Because the S&P GSCI is a production-based index, it has a strong overweighting

of the energy sector. Alternatively, we employ the S&P GSCI light energy index (2), which

has a lower exposure to energy and a more balanced weighting of different commodity

groups6 as well as a self constructed, equally weighted commodity index, which excludes ag-

riculture and livestock (3).

To evaluate the relative benefits of commodity groups, we examine energy, industrial

5
This test is applicable under very general conditions – stationary and ergodic returns. Most importantly for our
analysis, the test permits auto-correlation and non-normal distribution of returns and allows for a likely high
correlation between the portfolio returns with and without commodities.
6
The dollar weights of the GSCI (GSCI light energy) index on 31th of December 2012 were 69.0% (35.7%) for
energy, 6.9% (14.4%) for industrial metals, 3.6% (7.4%) for precious metals, 5.0% (10.3%) for livestock, and
15.6% (32.2%) for agricultural commodities.

- 14 -
metals, precious metals, livestock, and agriculture separately. Indices for these individual

commodity groups are part of the S&P GSCI index family. We also employ futures and spot

prices of individual commodities as robustness check. The shortest time series available starts

in January 1983 so that our dataset consists of monthly total return index data for stocks,

bonds, and commodities for the period from January 1983 to December 2013. Because several

years of historical data are required to calculate the first optimized portfolio, the evaluation

period starts in January 1986. All time series are denominated in US dollar and are from

Thomson Reuters Datastream. As risk-free rate we use the yield of a three-month US T-Bill.

4. EMPIRICAL RESULTS

The presentation of our empirical results is structured as follows: First, we provide descriptive

statistics (Section 4.1) and then discuss the results of our in-sample analysis, which provides

the maximum potential portfolio benefits of commodities in the absence of estimation errors

(Section 4.2). The analyses of the out-of-sample portfolio benefits of commodities for differ-

ent investment strategies demonstrate the attainable portfolio gains for investors (Section

4.3.). Then, we evaluate the portfolio shares allocated to commodities in the different asset

allocation strategies (Section 4.4). We extend our basic framework analyzing sub-periods

(Section 4.5) and proposing alternative equally weighted commodity indices excluding agri-

cultural and livestock commodities (Section 4.6). Finally, we calculate rolling Sharpe ratios

and examine the benefits of commodities over time (section 4.7). Section 5 presents the re-

sults of different robustness checks.

4.1. Descriptive statistics

In Table 3 we present descriptive statistics of the monthly asset returns for the evaluation pe-

riod from January 1986 to December 2013, consisting of 336 monthly observations. The aver-

- 15 -
age annualized returns are 11.42% p.a. for the S&P 500 index, 6.62% p.a. for the government

bond index, and 8.44% p.a. for the S&P GSCI. For the different commodity groups the aver-

age returns vary strongly, ranging from 1.99% for agricultural products to 12.64% for indus-

trial metals. Both risk measures annualized standard deviation and non-parametric value-at-

risk reveal that all commodity investments are substantially riskier than bonds and equally

risky as stocks. Energy, industrial metals and the S&P GSCI exhibit even higher return vola-

tilities and values-at-risk than stocks. During the period from 1986 to 2013, the average risk-

free rate is 3.65% p.a., which is larger than the average return of agricultural products, result-

ing in negative Sharpe ratios for these commodity groups. In addition, the Sharpe ratios of all

other commodity groups are lower than the Sharpe ratios of stocks and bonds, suggesting that

commodity indices are not very attractive as a stand-alone investment for long investment

horizons. 7 The Jarque-Bera statistics are significant at the 5%-level for all asset classes be-

sides the government bond index, rejecting the assumption of normally distributed returns for

all asset classes except bonds. This highlights the importance of using alternative risk and

performance measures.

[Table 3 about here]

Even if commodities do not appear to be an attractive asset class as a stand-alone in-

vestment, they might add value in a portfolio context by enhancing the risk-return profile. To

add value, the correlations with the traditional asset classes must be low or even negative. To

obtain some insights in terms of potential diversification benefits, we present the correlation

coefficients of the asset returns in table 4. Over the entire period, we find low but significantly

positive correlations between the S&P 500 stock index and the commodity indices. The only

7
For a shorter period from 1991-2009 Daskalaki and Skiadopoulos (2011) also find lower Sharpe ratios for the
commodity indices (S&P GSCI and DJ-UBSCI) and for most commodity futures than for stocks (S&P500) and
bonds (Barclays Agg Bond Index).

- 16 -
exceptions are precious metals, which do not show a significant correlation with stocks.

Hence, precious metals should be best suited to complement a stock-only portfolio. As we

analyze the benefits of commodities for stock-bond portfolios, the correlation of different

commodities with bonds are equally important and are presented in column two of table 4.

The data reveals that bond returns are significantly negative correlated with the aggregate

commodity indices and industrial metals. Energy, precious metals, livestock, and agricultural

products are insignificantly correlated with bond returns. Based on the correlation analysis,

commodities are a promising asset class to improve the out-of-sample risk-return structure of

stock-bond portfolios. In the next section, we investigate the maximum potential portfolio

benefits of commodities in an in-sample analysis.

[Table 4 about here]

4.2. In-sample benefits of commodities

We begin the empirical analysis by examining the in-sample benefits of adding commodities

to a stock-bond portfolio. ‘In-sample’ means that we use data available until month t to com-

pute the portfolio weight for the same month t. This implies that the investor has perfect fore-

casts for the expected returns of all assets. This does not reflect reality, but allows us to de-

termine the maximum potential portfolio benefits of commodities for a stock-bond investor.

Assuming that investors care only about return and variance and the investor’s utility function

is correctly specified by equation (4), than maximizing for U results in the optimal portfolio if

future asset returns μ and the covariance matrix Σ of future asset returns are given. Therefore,

in the in-sample analysis, the Markowitz mean-variance (MV) strategy dominates all other

strategies and there is no need for analyzing and comparing alternative asset allocation strate-

gies.

- 17 -
Table 5 reports the results of the in-sample MV analysis. Improvements relative to the

stock-bond portfolio are bold. In line with the in-sample analyses (e.g. spanning tests) report-

ed in the literature, we find that with perfect forecasts of expected returns all commodities

significantly increase the Sharpe ratio of a stock-bond portfolio. More interestingly, most

commodities increase portfolio returns rather than lowering portfolio risk. Only the S&P

GSCI light energy index and livestock commodities slightly reduce portfolio risk. Commodi-

ties also reduce portfolio tail-risk (VaR-99%) and maximum drawdown (unreported). Howev-

er, the portfolio benefits of different commodity groups vary substantially.

For both, conservative and aggressive investors, the largest increase in Sharpe ratio

and Omega measure is obtained when adding the aggregate S&P light energy commodity in-

dex to the stock-bond portfolio, followed by the ordinary S&P GSCI. Among the individual

commodity groups, precious metals and industrial metals provide the largest increase in the

Omega measure. So far, our results are consistent with other studies using in-sample analyses,

concluding that commodities and particularly precious metals add value to a stock-bond port-

folio. The portfolio effects of commodities in our in-sample analysis appear to be higher com-

pared to the analyses of ex post efficient frontiers (including spanning tests) reported in the

literature (Daskalaki and Skiadopolous, 2011; Belousova and Dorfleitner, 2012; Cao et al.,

2010). This stems from the fact that our in-sample analysis builds on monthly-rebalanced op-

timal portfolios given perfect forecasts of expected returns. In contrast, studies building on ex

post efficient frontiers (spanning tests) analyze shifts of the efficient frontier over a usually

long period of time. Implicitly, these studies analyze performance improvements of in-sample

MV optimal portfolios without rebalancing. Next, we analyze whether the gains from adding

commodities remain when we extend our analysis to an out-of-sample portfolio optimization

framework.

- 18 -
[Table 5 about here]

4.3. Benefits of commodities in out-of-sample optimization

In this section, we analyze the out-of-sample benefits of commodities and distinguish between

conservative investors (section 4.3.1) and aggressive investors (section 4.3.2). For the con-

servative (aggressive) investor the optimization strategies MV and BL include an upper vola-

tility bound of 5% p.a. (15%). We hypothesize that the minimum variance (MinVar) and risk-

parity (RP) strategies should be more appealing to conservative investors, allocating large

percentages of wealth to bonds (section 4.3.1), while in contrast, the risk-return-timing (RRT)

and the 1/N strategy should be more tempting for aggressive investors, offering higher expo-

sures to risky asset classes such as stocks and commodities (section 4.3.2.)

4.3.1. Conservative investor

Table 6 presents the results for the conservative investor for the evaluation period from 1986

to 2013. Performance measures are net of transaction costs for the out-of-sample optimized

stock-bond portfolios and the portfolios complemented with commodities using different asset

allocation strategies. Improvements relative to the stock-bond portfolio are bold.

The Sharpe ratio and Omega measure reveal that the S&P GSCI and the light energy

version of the S&P GSCI as well as industrial metals consistently enhance the risk-return pro-

file of a stock-bond portfolio for all asset allocation strategies. For all strategies industrial

metals offer the largest increase in Sharpe ratios, followed by the aggregate commodity indi-

ces. Energy commodities also provide substantial portfolio benefits with the BL, RP and the

strategically weighted allocation strategy, but do not provide portfolio benefits in the MV and

MinVar frameworks. Precious metals, livestock and agricultural commodities reduce portfolio

volatility in all asset allocation models besides MV but simultaneously reduce portfolio re-

- 19 -
turns resulting in lower Sharpe ratios and Omega measures compared to the stock-bond port-

folio. In general, all commodity groups are able to reduce portfolio volatility for all asset allo-

cation strategies besides MV. In contrast, only industrial metals, energy commodities and the

aggregate commodity indices improve portfolio returns in most strategies. Similar to

Belousova and Dorfleitner (2012), we find that the portfolio benefits of commodities depend

on the included commodity group.

Comparing the results for the different asset allocation models, it is evident that the

Black-Litterman portfolio performs best in most cases. In contrast to Daskalaki and

Skiadopoulos (2011), who do not find positive portfolio effects of commodities for a MV in-

vestor, our results suggest that industrial metals and the aggregate commodity indices (S&P

GSCI and S&P GSCI light energy) enhance the portfolio performance (Sharpe ratio; Omega

measure) of the MV portfolio. However, none of the commodities reduces portfolio risk.

[Table 6 about here]

4.3.2. Aggressive investor

Table 7 presents the benefits of commodities for an aggressive investor. Similar to the results

for the conservative investor, industrial metals offer the largest enhancement of the risk-return

profile for all asset-allocation strategies. The aggregate commodity indices (S&P GSCI and

S&P GSCI light energy), precious metals and livestock commodities also provide positive

portfolio effects for most asset-allocation models (BL, RRT, and strategic weights). As for the

conservative investor, we find no performance benefits from adding agricultural products to a

stock-bond portfolio, independent of the asset allocation strategy. While agricultural com-

modities help to reduce risk in most asset-allocation strategies (BL, RRT, and strategically

weighted portfolio), they simultaneously decrease portfolio returns, resulting in lower Sharpe

- 20 -
ratios and Omega measures. For the Black-Litterman model, the RRT strategy, and the strate-

gically weighted portfolios, all commodity groups besides agriculture improve the risk-return

profile (Sharpe ratio and Omega measure). For the MV optimized portfolio the aggregate

commodity index S&P GSCI, industrial metals, and energy have a positive effect on portfolio

returns but only industrial metals advance the risk-return profile (Sharpe ratio). Since MV is

most sensitive to estimation errors, this finding is probably due to estimation errors, which

outweigh the diversification benefits of commodities and thereby impede improving the out-

of-sample portfolio performance for most commodity groups. In line with DeMiguel et al.

(2009) we find that MV does not consistently outperform the naïve strategies (1/N and strate-

gically weighted portfolios). In contrast, the Black-Litterman model and the RRT strategy do

outperform the naïve strategies providing the highest performance. The superior performance

of the Black-Litterman model and the RRT strategy were already reported earlier (Kirby and

Ostdiek, 2012; Bessler, Opfer and Wolff, 2015; Bessler and Wolff, 2015).

For the equally weighted 1/N portfolio, none of the commodity groups enhances the

risk-return profile (Sharpe ratio) of a stock-bond portfolio. This finding suggests that the port-

folio weight of commodities is too large in the 1/N case (33.33%). Thus, the optimal com-

modity weight should be substantially lower than the one for stocks and bonds. In the superior

performing BL model, for instance, the average optimized portfolio weight for the S&P GSCI

is only 13.38%. For the strategically weighted portfolio, in which commodities have a portfo-

lio weight of 15% (table 2), all commodity groups consistently reduce portfolio risk and all

commodities, except livestock and agriculture, improve portfolio performance (Sharpe ratios).

The results for the conservative and aggressive investor clienteles can be summarized

as follows: First, all commodity groups and the aggregate commodity indices reduce portfolio

volatility for all asset allocation strategies besides MV and 1/N. Second, for livestock and

- 21 -
agricultural commodities, the reduction in volatility comes at the cost of lower portfolio re-

turns for virtually all asset allocation strategies. Hence, only the aggregate commodity indices,

industrial metals, energy and precious metals are able to enhance Sharpe ratios and Omega

measures compared to the stock-bond portfolio. Third, compared to our in-sample test (sec-

tion 4.2) and compared to many studies of ex post efficient frontiers (spanning-test) reported

in the literature (DeRoon, Nijman and Werker, 1996; Satyanarayan and Varangis, 1996; Jen-

sen, Johnson and Mercer, 2000; Scherer and He, 2008; Belousova and Dorfleitner, 2012), the

positive out-of-sample performance effects (increase in Sharpe ratio and Omega measure) of

adding commodities to a stock-bond portfolio are present but are relatively lower.

[Table 7 about here]

4.4. Analysis of commodity portfolio weights

So far, our analyses revealed that the portfolio benefits of commodities depend on the asset

allocation strategy. To gain more detailed insights and explanations why commodities are

more beneficial in some strategies than in others, we analyze the portfolio shares allocated to

commodities in different asset allocation models. Table 8 provides the average portfolio

weight of commodities (‘Mean’) and the respective standard deviation (‘Std.Dev.’) for vari-

ous optimization strategies and for the different commodities. The standard deviation indi-

cates how strongly the commodity portfolio shares fluctuate over time. Additionally, we re-

port the maximum portfolio weight of commodities during the 1986 to 2013 period. The re-

sults demonstrate that all asset allocation strategies allocate significant (1%-level) shares to

commodities. The standard deviation of commodity portfolio weights is the highest in the MV

approach. The immense instability is in line with the well known phenomenon of corner solu-

tions and large portfolio reallocations of the MV strategy (Broadie, 1993; Best and Grauer,

1991). The tremendous transaction costs are one explanation why commodities provide the

- 22 -
lowest portfolio benefits after transaction costs in the MV approach. The maximum portfolio

weight is 72.14% for the conservative and 100% for the aggressive investor, reflecting the

typical corner solutions in the MV approach. Such temporary dominance of commodities is in

sharp contrast to the primary motive of portfolio diversification.

In the BL and RP approaches, which perform substantially better than MV, the aver-

age commodity shares and their fluctuations over time are considerably lower than in the MV

portfolio. For the conservative investor the average allocation to commodities is between

4.37% and 6.26% in the BL model and between 2.03% and 13.01% in the RP approach. For

the aggressive investor the average allocation to commodities in the BL model is between

9.19% for agricultural commodities and 13.75% for the aggregate S&P GSCI light energy

index. Even if all asset allocation strategies allocate significant shares to agriculture, they do

not provide the desired performance benefits as reported in section 4.3.

[Table 8 about here]

4.5. Contribution of commodities in different market environments

Because the portfolio effects of commodities may depend on the market environment, we

separate the full evaluation period into expansionary and recessionary sub-periods on an ex

ante basis. For this, we combine monetary and stock market signals (Jensen and Mercer,

2003; Bessler, Holler and Kurmann, 2012; Bessler, Opfer and Wolff, 2015). Monetary signals

are the first change of the short-term interest rate by the central bank that runs counter to the

previous trend. Stock market signals are based on a simple moving average, assuming that the

stock market trend reverses if the 24-months moving average of the S&P 500 crosses the ac-

tual index from above (expansionary state) or below (recessionary state). For the transition

from one state to another both instruments must provide consistent signals. Figure 1 illustrates

- 23 -
the definition of sub-periods as well as the monetary policy and stock market signals. This

methodology determines six sub-periods of which three are characterized as expansionary and

three as recessionary.

[Figure 1 about here]

Table 9 presents the Sharpe ratios for the out-of-sample optimized stock-bond portfolios

and the portfolios including commodities for the six sub-periods. For brevity, we only present

the sub-period results for the conservative investor, but the results for the aggressive investor

are very similar. For both investor types, the benefits of commodities are time dependent. In

support of our previous results for the full period, the aggregate commodity indices (S&P

GSCI and S&P GSCI light energy), industrial metals, and energy generate the largest benefits

in sub-periods. The largest portfolio benefits of commodities are in the first two sub-periods

(1986-1990, 1990-1994) and in the two sub-periods between 2001 and 2008 (2001-2004,

2004-2008).

Industrial metals substantially improve the performance of stock-bond portfolios in

three sub-periods (1986-1990, 2001-2004, 2004-2008). This holds for virtually all asset-

allocation strategies. However, in the two recessionary periods from 1990 to 1994 and during

the most recent 2008 to 2013 period, industrial metals failed to enhance the performance of

stock-bond portfolios. The portfolio benefits of energy commodities are very similar to those

of industrial metals. This result is not surprising, given that the demand for energy and indus-

trial metals is pro-cyclical usually reflecting the current state of the economy. For most sub-

periods, we find the lowest or even no performance gain for agricultural commodities, con-

firming our full-sample results. There are a few exceptions, mostly in the first period and the

1994 to 2004 period, in which agricultural commodities, improve performance. However, the

performance effects of other commodity groups are usually larger. After 2004, there is no

- 24 -
gain from adding agricultural commodities to a stock-bond portfolio independent of the asset

allocation strategy. For livestock commodities, there are some portfolio benefits in the first

two sub-periods (1986-1990, 1990-1994), but there are basically no positive performance ef-

fects thereafter. Precious metals are particularly beneficial during the two sub-periods be-

tween 2001 and 2008. This covers the end of the new economy period and the subsequent

rebound of international stock markets. In the most recent period (2008-2013) including the

financial crisis, we hardly find any portfolio gains for any commodity group, except for the

RP portfolio where we observe some marginal enhancement of the performance (Sharpe ra-

tio). Overall, this might be due to the increasing correlations of commodities with the tradi-

tional asset classes stocks and bonds (Silvennoinen and Thorp, 2013; Büyüksahin and Robe,

2014).

[Table 9 about here]

4.6. Analyses of alternative commodity indices

So far, our results suggest that the S&P GSCI and the S&P GSCI light energy as well as cer-

tain individual commodities enhance the risk-return trade-off in a stock-bond portfolio. There

are lower or even no portfolio benefits of agricultural and livestock commodities so that in-

vestors might choose to exclude these commodities from their investment universe not only

for ethical but also for financial reasons. Because our results might depend on the construction

of the S&P GSCI indices, and possibly can be improved, we construct an alternative com-

modity index that equally weighs the three commodity groups ‘energy’, ‘industrial metals’

and ‘precious metals’ and excludes ‘livestock’ and ‘agriculture’ (EWCI ex LS&AG). We

compare the portfolio benefits of this index with the S&P GSCI indices and with a commodity

index that equally invests in all five commodity groups (EWCI).

- 25 -
The results in Table 10 indicate that a commodity index that excludes livestock and ag-

ricultural commodities and equally weighs ‘energy’, ‘industrial metals’ and ‘precious metals’

(EWCI ex LS&AG) would have improved the performance of a stock-bond portfolio consist-

ently for all asset allocation strategies. It even would have provided larger portfolio benefits in

comparison to the production weighted S&P GSCI index or the light energy version of the

S&P GSCI for all asset allocation strategies. The commodity index excluding livestock and

agricultural commodities (EWCI ex LS&AG) provides also larger portfolio benefits com-

pared to the index, which equally weighs all five commodity groups (EWCI) for most asset-

allocation strategies and independently of the investor type.

[Table 10 about here]

4.7. Rolling Sharpe ratios

To obtain a more detailed assessment of the time-varying portfolio gains from commodity

investments we compute rolling Sharpe ratios. The upper chart in Figure 2 presents rolling

Sharpe ratios with 60 months estimation windows for the BL stock-bond portfolio (dotted

line) and for the BL stock-bond portfolio with the aggregate commodity index S&P GSCI

(red line). The figure illustrates that the aggregate commodity index improves the Sharpe ratio

for most of the period. The lower chart in Figure 2 presents the marginal increase of the roll-

ing Sharpe ratio for the stock-bond portfolio when the S&P GSCI or the equally weighted

commodity index, excluding livestock and agricultural commodities (EWCI ex LS&AG), are

included. The figure indicates that the ‘EWCI ex LS&AG’ offers larger portfolio benefits than

the S&P GSCI since 2005 as well as a very similar performance increase as the S&P GSCI

index until 2005.

Figure 3 presents the increase in rolling Sharpe ratios relative to a stock-bond portfolio

for different commodity groups. In support of our findings for sub-periods, the results indicate

- 26 -
that the portfolio benefits of individual commodity groups are highly time-dependent and

were negative during some periods. For example, energy had a negative impact on the Sharpe

ratio between 1995 and 2000 but a positive effect between 2002 and 2009. In contrast, live-

stock offered only minor performance benefits during the 1991-1995 and 2003-2006 periods.

For agricultural commodities there was hardly any portfolio benefit over the entire period.

Consequently, we derive two main conclusions from our analysis so far. On the one hand,

including aggregate commodity indices in multi-asset portfolios positively affects perfor-

mance, but excluding livestock and agricultural commodities does not jeopardize perfor-

mance. On the other hand, commodities enhance the performance even further when a differ-

ent index-weighting scheme for commodities is used.

Moreover, in support of the literature (Cao et al., 2010), our analysis of rolling Sharpe

ratios provides evidence that for the most recent time period from 2008 to 2013, the portfolio

benefits of commodities diminished relative to earlier sub-periods for all commodity groups.

One explanation for this finding is the ‘financialization’ of commodity markets documented

in the literature (Domanski and Heath, 2007). The consequence is a higher correlation of

commodity returns with that of stock and bond returns (Silvennoinen and Thorp, 2013; Tang

and Xiong, 2012), which reduces the positive diversification effects. Whether this is an accu-

rate explanation or not, it seems that the benefits of investing in individual commodities as

part of a multi-asset allocation strategy are currently diminished.

Because it is a challenge for most investors to determine whether commodity invest-

ments offer future gains and to predict which commodity group will perform best, a reasona-

ble strategy is to invest in a commodity index including most commodity groups. Overall, our

empirical analysis suggests that a sensible strategy is to invest equally in different commodity

- 27 -
groups. Avoiding investments in livestock and agricultural commodities for ethical reasons

does not seem to reduce financial performance.

[Figure 2 and Figure 3 about here]

5. ROBUSTNESS CHECKS

To evaluate the sensitivity of our results to changes in the sample and input parameters we

perform various robustness checks. First, we use an alternative internationally diversified

stock-bond portfolio as basis portfolio to examine the portfolio benefits of commodities (5.1).

We then assess the value of alternative individual commodity investments such as commodity

futures and physical commodities (5.2). Moreover, we employ alternative estimation windows

(5.3), different optimization constraints (5.4), varying levels of transaction costs (5.5) and

alternative rebalancing frequencies (5.6).

5.1. Alternative basis portfolio

To check whether the portfolio effects of commodities differ for a more diversified portfolio,

we add international stocks and corporate bonds to our base portfolio so far consisting of US

stocks and bonds. The extended portfolio includes the MSCI North America, MSCI Europe,

and MSCI Pacific indices representing global stock markets as well as US government bonds

(Barclays US aggregate Government Bond index) and corporate bonds (Bank of America/

Merill Lynch US High Yield 100). Table 11 presents the results for the expanded portfolio. It

supports our previous findings and illustrates that the portfolio benefits of commodities are

very similar in an internationally diversified portfolio that includes corporate default risk. As

before, industrial metals and the aggregate commodity indices (S&P GSCI and S&P GSCI

light energy) offer the largest increase in portfolio performance. The contribution of energy,

precious metals and livestock commodities is lower. Agricultural commodities offer the low-

- 28 -
est gains and in most asset-allocation strategies even an inferior performance for diversified

international stock-bond portfolio.

[Table 11 about here]

5.2. Alternative commodity investments with futures or spot positions

To investigate whether our results depend on the employed S&P GSCI commodity indices,

we use commodity futures contracts and commodity spot prices of individual commodities as

robustness check. For each commodity sector, we select one representative commodity. Crude

oil as the world’s most traded commodity represents the energy sector. Copper, widely used

in the infrastructure and construction business, is the third most traded commodity represent-

ing industrial metals. Gold, which many investors view as a safe haven during crisis periods

and as a hedge against inflation, replaces the precious metals index. Live cattle and corn rep-

resent the livestock and agricultural commodity sectors, respectively. For each individual

commodity, we use the respective futures and spot returns. Because futures contracts do not

require an initial investment (except margins), future returns are excess returns over the risk-

free rate. The return is the sum of the futures return and the risk-free rate (Bodie and

Rosansky, 1980; Fortenbery and Hauser, 1990). In contrast, spot returns reflect the returns of

a physical investment in commodities without storage costs. We use the S&P GSCI spot re-

turn index for all commodities. Due to the availability of data, the analysis for individual

commodity futures and spot prices covers the period from 1993 to 2013.

Table 12 presents in Panel A and B the results for investments in individual commodity

futures and individual physical commodities, respectively. Overall, the results for individual

commodities confirm our base case results. As before, the industrial metal copper offers the

largest portfolio benefits followed by crude oil and precious metals. This supports the findings

of Agyei-Ampomah et al. (2014) who report that industrial metals, especially copper, tends to

- 29 -
outperform gold and other precious metals as hedging vehicles and safe haven assets. The

results for live cattle and corn are comparable to the results for the commodity group indices

and support our finding that investments in livestock and agricultural commodities provide

the lowest or even no improvement of the risk-return profile of a stock-bond portfolio.

[Table 12 about here]

5.3. Alternative estimation windows

We analyze different window lengths for estimating returns and the covariance matrix as an-

other robustness check. We vary the estimation window lengths between 12 and 60 months

and find that our base case results on the contribution of commodities to a stock-bond portfo-

lio are robust to these changes. Particularly, agricultural and livestock commodities have

hardly any positive effect on the performance of stock-bond portfolios, while positive effects

are present for both aggregate commodity indices (S&P GSCI and S&P GSCI light energy),

industrial metals, precious metals and energy commodities. As in Bessler, Opfer and Wolff

(2015), long estimation windows for returns of 48 months and longer reduce the out-of-

sample performance of all asset allocation models, as the models react too slowly to structural

breaks. In contrast, for very short estimation windows, transaction costs increase substantially,

thereby lowering the net performance.

5.4. Alternative optimization constraints

To check the robustness of our results with respect to the employed optimization constraints,

we compute all optimized portfolios without short-sale constraint (Table 13). As expected, the

out-of-sample benefits of commodities seem to be larger with short selling allowed, particu-

larly for the aggressive investor type. 8 As before, industrial metals provide the largest portfo-

8
Note that the results of the RRT, 1/N, and strategically weighted portfolio are unchanged compared to the base

- 30 -
lio benefits followed by the aggregate commodity indices S&P GSCI and S&P GSCI light

energy and energy, while agricultural and livestock commodities again offer the lowest bene-

fits. In addition, we optimize the portfolios without volatility constraint for varying estimation

windows. The results (unreported) are robust and do not change qualitatively when different

optimization constraints are used.

[Table 13 about here]

5.5. Alternative transaction costs

Transaction costs have a major impact on performance. To investigate the sensitivity of our

results to the assumed level of transaction costs, we vary the variable transaction costs from 0

to 50 basis points and we do not find that the results change qualitatively. As expected, a

higher level of transaction costs makes it less attractive to diversify with additional asset clas-

ses. However, this result is not restricted to commodities but holds for all asset classes. More-

over, risk-based (RP, MinVar) and naïve (1/N, st.w.) strategies are characterized by more sta-

ble portfolio weights and their net performance – with or without commodities – is less sensi-

tive to an increase in transaction costs.

5.6. Alternative rebalancing frequencies

So far, we compute all asset allocation strategies with monthly rebalancing. To investigate the

sensitivity of our results to alternative rebalancing frequencies, we compute quarterly and

yearly rebalanced portfolios. The results in table 14 (Panels A and B) confirm our base case

results. Particularly, aggregate commodity indices and industrial metals provide the largest

portfolio benefits. Interestingly, the portfolio effects of commodities tend to be higher with

results, since these strategies do not include any short-sales by construction.

- 31 -
less frequent rebalancing, because the higher diversification benefits due to more rebalancing

require additional transaction costs. For our dataset quarterly rebalancing results in a superior

performance (after transaction costs) for most asset-allocation strategies compared to monthly

or yearly rebalancing.

[Table 14 about here]

In sum, our results hold for an alternative internationally diversified basis portfolio, al-

ternative individual commodity investments, alternative estimation windows, different opti-

mization constraints, varying levels of transaction costs and alternative rebalancing windows.

6. CONCLUSION

We investigate the out-of-sample effects of adding commodities to a stock-bond portfolio by

analyzing a variety of asset allocation strategies including equally- and strategically weighted

portfolios, risk-parity, minimum-variance, reward-to-risk timing, mean-variance and Black-

Litterman. So far, most studies investigate the benefits of commodities based on MV and non-

MV spanning-tests or by visually analyzing shifts of the efficient frontier. These approaches

are in-sample analyses and limited to evaluate whether commodities improve the risk-return

profile for ex post efficient portfolios. However, this does not accurately reflect the decision

environment of asset managers and investors who are faced with the challenge of setting up

portfolios based on uncertain future returns.

Consequently, a more realistic assessment of the portfolio benefits of commodities re-

quires an out-of-sample analysis. However, the out-of-sample studies so far are contradicting

in their results and limited to the sample-based MV and non-MV approach. We add to the

literature by examining the out-of-sample benefits of commodities for a variety of asset allo-

cation strategies. Moreover, we analyze investments in individual commodity groups such as

- 32 -
energy, precious metals, and industrial metals but also focus on investments such as agricul-

tural and livestock commodities that for ethical reasons are critically discussed. We also ana-

lyze the portfolio effects of commodities for conservative and aggressive investor clienteles

and during bull and bear market environments.

Our empirical results suggest that the out-of-sample benefits of commodities are much

lower than previous in-sample analyses suggest. Therefore, in-sample analyses, including

spanning-tests, tend to overstate the actual portfolio gains from commodity investments. In

fact, we find that the out-of-sample portfolio benefits diverge for different commodities and

sub-periods. For most asset-allocation strategies, aggregate commodity indices and industrial

metals provide the largest performance enhancement followed by energy and precious metals.

Interestingly, we hardly observe any performance gains for agricultural and livestock com-

modities. An equally weighted commodity index, excluding livestock and agricultural com-

modities, generates a superior performance for most asset-allocation strategies. Consequently,

investments in food commodities appear redundant for efficient portfolio allocations.

Although the empirical findings of our analysis are quite appealing and may be applied

in portfolio optimization, asset managers and investors will have to use some caution when

implementing these approaches in asset allocation decisions. As usual, there are several po-

tential limitations as we analyze only a limited number of asset allocation strategies as well as

a specific set of assets. For other portfolio optimization approaches and other asset classes the

portfolio performance in general and the portfolio benefits of commodities in particular might

be different, although we feel confident that we have covered a wide array of models and as-

set classes. In addition, despite applying an out-of-sample portfolio optimization framework,

our analysis is still based on historical mean asset returns as return forecasts and historical

return variances and covariances for the covariance matrix. As it is well documented, histori-

- 33 -
cal returns can be poor estimates for future returns. However, the benefits might be higher for

investors that have superior skills in forecasting stock or commodity returns. Therefore, a

more objective and realistic approach is to first develop and estimate return prediction models

and then using these return forecasts in portfolio optimization. Testing asset allocation strate-

gies with return predictions models is left for future research.

- 34 -
Acknowledgements

We are grateful to two anonymous referees for helpful suggestions and comments that signifi-

cantly improved the quality of this research. The authors also thank Mikael Bergbrant, Mi-

chael Neumann, Terrence Hallahan as well as the participants at the 2014 Annual Meeting of

the Midwest Finance Association, the 2014 Paris Financial Management Conference, the Fi-

nance Conference of the Portuguese Finance Network (2014), and the Workshop on Determi-

nants and Impact of Commodity Price Dynamics at the University of Münster (2014) for help-

ful comments and suggestions.

- 35 -
References

Abanomey, W.S., Mathur, I., 1999. The hedging benefits of commodity futures in interna-
tional portfolio construction. Journal of Alternative Investments 2 (3), 51-62.

Agyei-Ampomah, S., Gounopoulos, D., Mazouz, K., 2014. Does gold offer a better protection
against sovereign debt crisis than other metals? Journal of Banking & Finance, 40, 507-
521.
Anderson, R.M., Bianchi, S.W., Goldberg, L.R., 2012. Will my risk parity strategy outper-
form? Financial Analysts Journal, 68 (6), 75-93.
Anson, M. J.P., 1999. Maximizing utility with commodity futures diversification. Journal of
Portfolio Management 25, 86-94.
Baker, M., Bradley B., Wurgler, J., 2011. Benchmarks as limits to arbitrage: Understanding
the low volatility anomaly. Financial Analysts Journal 67, 40-54.

Basak, S., Pavlova A., 2013. A model of financialization of commodities. London Business
School. Working Paper.
Beckmann, J., Belke, A., Czudaj, R., 2014. Does global liquidity drive commodity prices?
Journal of Banking & Finance, forthcoming.
Belousova, J., Dorfleitner, G., 2012. On the diversification benefits of commodities from the
perspective of euro investors. Journal of Banking & Finance 36 (9), 2455–2472.

Benartzi, S., Thaler R.H., 2001. Naive diversification strategies in defined contribution saving
plans. American Economic Review 91 (1), 79-98.
Bessler, W., Holler J., Kurmann P., 2012. Hedge Funds and optimal asset allocation: Bayesian
expectations and spanning tests, Financial Markets and Portfolio Management 26 (1),
109-141.
Bessler, W., Opfer H., Wolff, D., 2015. Multi-asset portfolio optimization and out-of-sample
performance: An evaluation of Black-Litterman, mean-variance and naïve diversifica-
tion approaches. European Journal of Finance, forthcoming.

Bessler, W., Wolff, D., 2015. Portfolio optimization with return prediction models: Evidence
for industry portfolios. Working Paper, University of Giessen.
Best, M.J., Grauer R.R., 1991. On the sensitivity of mean-variance-efficient portfolios to
changes in asset means: some analytical and computational results. Review of Financial
Studies 4 (2), 315.
Black, F., Litterman, R., 1992. Global portfolio optimization. Financial Analysts Journal 48
(5), 28-43.
Bodie, Z., Rosansky, V.I., 1980. Risk and return in commodity futures. Financial Analysts
Journal 36 (3), 27-39.
Broadie, M., 1993. Computing efficient frontiers using estimated parameters. Annals of oper-

- 36 -
ations research 45, 21-58.
Büyükşahin, B., Robe, M.A., 2014. Speculators, commodities and cross-market linkages.
Journal of International Money and Finance 42, 38–70.
Cao, B., Jayasuriya, S., Shambora, W., 2010. Holding a commodity index fund in a globally
diversified portfolio: a placebo effect? Economics Bulletin 30, 1842–1851.
Cheng, I.H., Xiong W., 2013. The financialization of commodity markets. NBER Working
Paper Series.

Cheung, C.S., Miu P., 2010. Diversification benefits of commodity futures. Journal of Inter-
national Financial Markets, Institutions, and Money 20, 451-474.

Chopra, V.K., Ziemba W.T., 1993. The effect of errors in means, variances, and covariances
on optimal portfolio choice. Journal of Portfolio Management 19 (2), 6-11.
Conover, C.M., Jensen, G.R., Johnson, R.R., Mercer, J.M., 2010. Is now the time to add
commodities to your portfolio? Journal of Investing 19, 10–19.
Croft H., and Norrish K., 2013. The commodity refiner: From an age of shortage to an era of
enough? Barclays Commodities Research Report April 2013.

Daskalaki, C., Kostakis, A., Skiadopoulos, G. 2014. Are there common factors in individual
commodity futures returns? Journal of Banking & Finance 40, 346–363.
Daskalaki C., Skiadopoulos G.S., 2011. Should investors include commodities in their portfo-
lios after all? New evidence. Journal of Banking & Finance 35, 2606-2626.
DeMiguel, V., Garlappi, L., Uppal, R., 2009. Optimal versus naive diversification: how inef-
ficient is the 1/N portfolio strategy? Review of Financial Studies 22, 1915-1953.

DeRoon, F.A., Nijman, T.E., Werker, B.J.M., 1996. Testing for spanning with futures con-
tracts and nontraded assets: A general approach. Working Paper, Tilburg University.
Domanski, D., Heath, A., 2007. Financial investors and commodity markets. BIS Quarterly
Review, 53–67.
Erb, C.B., Harvey, C.R., 2006. The strategic and tactical value of commodity futures. Finan-
cial Analysts Journal 62, 69-97.
Fletcher, J., Hillier, J. 2005. An examination of linear factor models in country equity asset
allocation strategies. The Quarterly Review of Economics and Finance 45 (4/5), 808–
823.
Fortenbery, T.R., Hauser, R.J., 1990. Investment potential of agricultural futures contracts.
American Journal of Agricultural Economics 72, 721–726.
Frazzini, A., Pedersen, L.H., 2014. Betting against beta. Journal of Financial Economics, 111
(1), 1–25.
Frost, P. A., and J. E. Savarino. 1988. For better performance: constrain portfolio weights.
Journal of Portfolio Management 15 (1), 29–34.

- 37 -
Galvani, V., Plourde, A., 2010. Portfolio diversification in energy markets. Energy Economics
32, 257–268.
Geman, H., 2005. Commodities and commodity derivatives: Modeling and pricing for
agriculturals, metals and energy. John Wiley & Sons Ltd., Chichester.

Gorton, G.B., Rouwenhorst, K. G., 2006. Facts and fantasies about commodity futures. Fi-
nancial Analysts Journal 62, 47-68.
Greer, R.J. 1994. Methods for institutional investment in commodity futures. The Journal of
Derivatives, 2, 28–36.
Gruber, J.W., Vigfusson, R.J., 2012. Interest rates and the volatility and correlation of com-
modity Prices. International finance discussion papers (Number 1065), Board of Gover-
nors of the Federal Reserve System.
Huberman, G., Kandel, S., 1987. Mean–variance spanning. The Journal of Finance 42, 873–
888.

Jagannathan, R., Ma T., 2003. Risk reduction in large portfolios: why imposing the wrong
constraint helps. Journal of Finance 58, 1651–1683.
Jensen, G.R., Johnson, R.R., Mercer, J.M., 2000. Efficient use of commodity futures in diver-
sified portfolios. Journal of Futures Markets 20, 489-506.
Jones, R., Lim, T., Zangari P. J., 2007. The Black–Litterman model for structured equity port-
folios. Journal of Portfolio Management 33 (2), 24–33.

Jensen, G.R., Mercer, J.M., 2003. New evidence on optimal asset allocation. Financial Re-
view 38, 435-454.
Jorion, P., 1985. International portfolio diversification with estimation risk. Journal of Busi-
ness 58, 259–78.
Jorion, P., 1986. Bayes–Stein estimation for portfolio analysis. Journal of Financial and Quan-
titative Analysis 21, 279–292.

Keating, C., Shadwick, W.F., 2002. A universal performance measure. Journal of perfor-
mance measurement 6 (3), 59–84.
Kirby, C., Ostdiek B., 2012. It’s all in the timing: simple active portfolio strategies that out-
perform naive diversification. Journal of Financial and Quantitative Analysis 47 (2),
437-467.
Lee, W. 2000. Advanced theory and methodology of tactical asset allocation. New Hope:
Frank J. Fabozzi Associates, 125-136.
Maillard, S., Roncalli, T., Teiletche, J., 2010. On the properties of equally-weighted risk con-
tributions portfolios. Journal of Portfolio Management 36 (4), 60-70.
Markowitz, H., 1952, Portfolio selection. Journal of Finance 7 (1), 77-91.
Michaud, R.O., 1989. The Markowitz optimization enigma: is the optimized optimal? Finan-

- 38 -
cial Analysts Journal 45 (1), 31-42.
Opdyke, J.D., 2007. Comparing Sharpe ratios: so where are the p-values? Journal of Asset
Management 8 (5), 308-336.
Pies, I., Will, M. G., Prehn, S., Glauben, T., 2013. Hungermakers? - Why futures market ac-
tivities by index funds are promoting the common good. Swiss Derivatives Review, 53,
12-16.
Satchell, S., Scowcroft A., 2000. A demystifcation of the Black Litterman model: managing
quantitative and traditional portfolio constructions. Journal of Asset Management 1 (2),
138–150.
Satyanarayan, S., Varangis, P., 1996. Diversification benefits of commodity assets in global
portfolios. Journal of Investing 5, 69-78.
Scherer, B., He, L., 2008. The diversification benefits of commodity futures indexes: a mean–
variance spanning test. Fabozzi, F.J., Füss, R., Kaiser, D.G. (Eds.), The Handbook of
Commodity Investing. John Wiley & Sons, Hoboken, 241–265.
Silvennoinen, A., Thorp, S., 2013. Financialization, crisis and commodity correlations dy-
namics. Journal of International Financial Markets, Institutions and Money, 24, 42-56.
Stoll, H.R., Whaley, R.E., 2011. Commodity index investing: speculation or diversification.
The Journal of Alternative Investments 4 (1), 50–60.
Tang, K., Xiong, W., 2012. Index investment and financialization of commodities. Financial
Analysts Journal, 68 (6), 54-73.
Welch, I., Goyal, A., 2007. A comprehensive look at the empirical performance of equity
premium prediction. Review of Financial Studies 21 (4), 1455–1508.
You, L., Daigler, R.T., 2012. A Markowitz optimization of commodity futures portfolios. The
Journal of Futures Markets, 1-26.

- 39 -
Figure 1: Definition of Sub-periods
1.800 Expansion Recession Expansion Recession Expansion Recession 12
01/1986 - 04/1990 - 03/1994 - 02/2001 - 07/2004 - 04/2008 –
1.600 03/1990 02/1994 01/2001 06/2004 03/2008 12/2012
10
1.400

1.200 8

1.000
6
800

600 4

400
2
200

0 0
Jan-86
Jan-87
Jan-88
Jan-89
Jan-90
Jan-91
Jan-92
Jan-93
Jan-94
Jan-95
Jan-96
Jan-97
Jan-98
Jan-99
Jan-00
Jan-01
Jan-02
Jan-03
Jan-04
Jan-05
Jan-06
Jan-07
Jan-08
Jan-09
Jan-10
Jan-11
Jan-12
Jan-13
MSCI World (LHS) Moving Average (LHS) Federal Funds Target Rate (RHS)

- 40 -
Figure 2: Rolling Sharpe ratios

Notes: This figure provides the rolling Sharpe ratio (60 months) for Black-Litterman stock-bond portfolios and
portfolios complemented with a commodity index. In the upper chart, the dotted line represents the rolling
Sharpe ratio for the stock-bond portfolio the bold line represent the stock-bond portfolio complemented with
commodities. The lower figure provides the increase in rolling Sharpe ratio when adding the indicated commodi-
ty index to a stock-bond portfolio.

- 41 -
Figure 3: Rolling Sharpe ratios: Increase compared to stock-bond portfolio

Notes: This figure provides the increase in rolling Sharpe ratio (60 months) for Black-Litterman out-of-sample
optimized portfolios due to adding the indicated commodity group to a stock-bond portfolio.

- 42 -
Table 1: Overview of the employed asset-allocation models

# Model Abbreviation

Naive asset allocation rules

1 1/N with rebalancing 1/N

2 Strategic weights (naive diversification) with rebalancing st.w.

Simple asset allocation strategies

3 Reward-to-risk timing RRT

4 Risk Parity RP

Portfolio optimization strategies

5 Minimum Variance MinVar

6 Sample-based mean-variance MV

7 Black-Litterman BL

Notes: This table lists the asset-allocation strategies employed in this study. The last column of the table pro-
vides the abbreviation used to refer to the strategy in the tables, which provide the empirical results.

Table 2: Strategically weighted portfolios

Strategic portfolio weights with (without) commodities


Optimization constraint: max.
Investor type Stocks Bonds Commodities portfolio volatility
conservative 15% (20%) 80% 5% (0%) 5.00% p.a.
aggressive 65% (80%) 20% 15% (0%) 10.00% p.a.

Notes: This table provides the strategic portfolio weights for the analyzed conservative and aggressive investor
clientele. In parenthesis, we provide the strategic portfolio weights for the portfolio without commodities. Based
on the discussion with practitioners, we assume that the two investor-types prefer a maximum expected portfolio
volatility of 5% and 15%, respectively.

- 43 -
Table 3: Descriptive statistics of asset returns (January 1986-December 2013)

Mean SD Skew Kurt Sharpe VaR99% JB (p-value) Obs.


S&P 500 11.42% 16.04% -0.74 5.58 0.48 14.83% 0.00% 336
US GOV 6.62% 4.60% 0.10 3.57 0.65 2.39% 6.60% 336
T-Bill 3.65% 0.70% -0.10 1.93 0.00 0.00% 0.33% 336
GSCI 8.44% 20.83% -0.41 5.03 0.23 15.01% 0.00% 336
GSCI LE 6.11% 14.77% -0.86 6.99 0.17 13.00% 0.03% 336
Energy 12.15% 32.24% 0.12 4.31 0.26 21.27% 0.06% 336
Ind. Metals 12.64% 24.50% 0.49 6.80 0.37 15.78% 0.00% 336
Prec. Metals 6.33% 16.46% 0.13 4.21 0.16 11.43% 0.15% 336
Livestock 4.18% 14.33% -0.04 4.26 0.04 9.13% 0.13% 336
Agriculture 1.99% 19.20% 0.23 4.79 -0.09 14.79% 0.08% 336

Notes: This table provides sample moments, Sharpe ratios, Value-at-risk and Jarque-Bera statistics of the stock
and bond indices, the risk-free rate and the seven commodity indices used in the empirical analysis. The evalua-
tion period covers 336 months from January 1986 to December 2013. ‘Mean’ denotes annualized time-series
mean of monthly returns while ‘Std.Dev.’ denotes the associated annualized standard deviation. ‘Skew’ and
‘Kurt’ represent the third and fourth moment of the return distribution. ‘Sharpe’ shows the annualized Sharpe
ratios of the respective asset classes. VaR99% shows the non-parametric 99%-value-at-risk for the respective asset
class during the period from January 1986 to December 2013 and ‘JB (p-value)’ is the p-value of the Jarque-
Bera statistic for testing normality of returns.

Table 4: Correlation matrix of asset returns (January 1986- December 2013)

Correlation Basis portfolio Commodities


(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(1) S&P500 1.00 -0.04 0.05 0.24* 0.32* 0.16* 0.23* 0.00 0.13** 0.23*
(2) US GOV -0.04 1.00 0.12** -0.13** -0.13** -0.10 -0.2* 0.00 -0.04 -0.05
(3) T-Bill 0.05 0.12** 1.00 0.08 0.09 0.08 0.11** -0.07 0.13** 0.03
(4) GSCI 0.24* -0.13** 0.08 1.00 0.91* 0.94* 0.37* 0.27* 0.18* 0.34*
(5) GSCI LE 0.32* -0.13** 0.09 0.91* 1.00 0.74* 0.48* 0.35* 0.29* 0.62*
(6) Energy 0.16* -0.10 0.08 0.94* 0.74* 1.00 0.26* 0.2* 0.09 0.14**
(7) Ind.Met. 0.23* -0.2* 0.11** 0.37* 0.48* 0.26* 1.00 0.26* 0.04 0.25*
(8) Prec.Met. 0.00 0.00 -0.07 0.27* 0.35* 0.2* 0.26* 1.00 -0.04 0.2*
(9) Livestock 0.13** -0.04 0.13** 0.18* 0.29* 0.09 0.04 -0.04 1.00 0.00
(10) Agriculture 0.23* -0.05 0.03 0.34* 0.62* 0.14** 0.25* 0.2* 0.00 1.00

Notes: This table provides the correlation matrix for the stock and bond indices, the risk-free rate and the seven
commodity indices used in the empirical analysis over the time-period from January 1986 to December 2013. *,
** indicate values significantly different from 0 at the 1% and 5% level, respectively.

- 44 -
Table 5: Contribution of commodities given perfect forecasts of expected return

stock-bond portfolio complemented with commodities


Asset Allo-
Performance/ stock-
cation
Risk measure bond GSCI Ind. Prec. Agri-
strategy
GSCI LightEn Energy Metals Metals Livestock culture
Panel A: Conservative investor

MV Return 11.24% 16.23% 16.29% 15.14% 16.38% 15.96% 15.13% 15.32%


Volatility 4.60% 4.80% 4.52% 4.78% 5.10% 4.74% 4.52% 4.81%
Sharpe 1.64 2.66* 2.83* 2.41* 2.50* 2.59* 2.52* 2.43*
Omega 3.30 7.59 8.83 6.07 7.19 7.21 6.58 6.31
Panel B: Aggressive investor

MV Return 19.55% 29.52% 27.69% 29.28% 29.53% 28.91% 27.62% 27.53%


Volatility 6.53% 7.48% 6.76% 7.67% 8.44% 7.52% 6.90% 7.64%
Sharpe 2.43 3.51* 3.59* 3.38* 3.09* 3.35* 3.48* 3.13*
Omega 6.89 20.40 21.14 17.93 20.09 19.91 18.20 16.00

Notes: This table reports the portfolio performance of in-sample optimized mean-variance (MV) portfolios for
stock-bond portfolios and portfolios complemented with commodities for the full sample from 1986 to 2013.
‘Return’ denotes the annualized time-series mean of monthly returns while ‘Volatility’ shows the associated
annualized standard deviation. ‘Sharpe’ represents the annualized Sharpe ratio and ‘Omega’ is the Omega meas-
ure for each portfolio. In this in-sample analysis the return forecasts for month t+1 are the realized returns in
month t+1 (perfect forecasts). The covariance-matrix is estimated using a 36-month rolling estimation window
up to month t. Two investor types are distinguished: A conservative investor (Panel A) with a maximum desired
volatility of 5% p.a. (risk aversion coefficient: 10) and an aggressive investor (Panel B) with a maximum desired
volatility of 15% p.a. (risk aversion coefficient: 2). The basis stock-bond portfolio consists only of US stocks and
US-bonds while the extended portfolio additionally includes the indicated commodity group. * indicates a signif-
icant higher Sharpe ratio of the extended portfolio in comparison to the stock-bond portfolio at the 1%-level.

- 45 -
Table 6: Out-of-sample performance benefits of commodities for conservative investor

stock-bond portfolio complemented with commodities


Asset
Performance stock-
Allocation
measure bond GSCI Industrial Precious Agri-
strategy
GSCI LightEn. Energy metals Metals Livestock culture
Conservative investor
BL Return 7.31% 7.68%⁺ 7.64%⁺ 7.65%⁺ 7.95%⁺ 7.40%⁺ 7.37%⁺ 7.30%
Volatility 4.53% 4.37%⁺ 4.35%⁺ 4.44%⁺ 4.45%⁺ 4.42%⁺ 4.35%⁺ 4.48%⁺
Sharpe 0.81 0.93⁺ 0.92⁺ 0.91⁺ 0.97⁺ 0.85⁺ 0.86⁺ 0.81
Omega 1.83 2.00⁺ 1.99⁺ 1.97⁺ 2.11⁺ 1.88⁺ 1.92⁺ 1.84⁺

MV Return 7.78% 8.47%⁺ 8.24%⁺ 7.95%⁺ 9.20%⁺ 7.60% 7.45% 7.82%⁺


Volatility 4.91% 5.32% 5.28% 5.31% 5.50% 5.41% 5.03% 5.34%
Sharpe 0.85 0.91⁺ 0.87⁺ 0.82 1.01⁺ 0.73 0.76 0.79
Omega 1.84 1.96⁺ 1.91⁺ 1.81 2.19⁺ 1.71 1.75 1.79

MinVar Return 6.71% 6.85%⁺ 6.71% 6.66% 7.31%⁺ 6.15% 6.42% 6.45%
Volatility 4.38% 4.33%⁺ 4.09%⁺ 4.46% 4.34%⁺ 4.11%⁺ 4.30%⁺ 4.22%⁺
Sharpe 0.70 0.74⁺ 0.75⁺ 0.68 0.85⁺ 0.61 0.65 0.67
Omega 1.68 1.75⁺ 1.76⁺ 1.66 1.91⁺ 1.57 1.63 1.64

RP Return 7.20% 7.48%⁺ 7.31%⁺ 7.30%⁺ 7.69%⁺ 6.94% 6.96% 6.98%


Volatility 4.53% 4.33%⁺ 4.13%⁺ 4.46%⁺ 4.35%⁺ 4.25%⁺ 4.37%⁺ 4.34%⁺
Sharpe 0.79 0.90⁺ 0.90⁺ 0.83⁺ 0.94⁺ 0.78 0.77 0.78
Omega 1.80 1.95⁺ 1.95⁺ 1.85⁺ 2.01⁺ 1.78 1.78 1.77

st.w. Return 7.40% 7.27% 7.15% 7.43%⁺ 7.47%⁺ 7.17% 7.06% 6.94%
Volatility 4.77% 4.44%⁺ 4.41%⁺ 4.58%⁺ 4.41%⁺ 4.37%⁺ 4.38%⁺ 4.47%⁺
Sharpe 0.79 0.82⁺ 0.80⁺ 0.83⁺ 0.87⁺ 0.80⁺ 0.79 0.74
Omega 1.82 1.85⁺ 1.83⁺ 1.87⁺ 1.95⁺ 1.83⁺ 1.82 1.75

Notes: This table reports the out-of-sample portfolio performance for stock-bond portfolios and portfolios com-
plemented with commodities for different asset allocation strategies during the time period from January 1986 to
December 2013. The table presents the results for a conservative investor with a maximum desired volatility of
5% p.a. (risk aversion coefficient: 10). The basis portfolio consists only of US stocks and bonds while the ex-
tended portfolio additionally includes the indicated commodity group. Improvements in comparison to the stock-
bond portfolio and are in bold and highlighted with +. ‘Return’ denotes the annualized time-series mean of
monthly returns while ‘Volatility’ shows the associated annualized standard deviation. ‘Sharpe’ represents the
annualized Sharpe ratio and ‘Omega’ is the Omega measure for each portfolio.

- 46 -
Table 7: Out-of-sample portfolio benefits of commodities for aggressive investor

stock-bond portfolio complemented with commodities


Asset stock-
Performance
Allocation bond
measure GSCI Industrial Precious Agri-
strategy portfolio
GSCI LightEn. Energy metals Metals Livestock culture
Aggressive investor
BL Return 10.58% 10.8%⁺ 10.6%⁺ 11.3%⁺ 12.6%⁺ 10.8%⁺ 10.55% 10.00%
Volatility 11.91% 10.5%⁺ 10.4%⁺ 11.0%⁺ 11.1%⁺ 10.8%⁺ 10.3%⁺ 10.9%⁺
Sharpe 0.58 0.68⁺ 0.67⁺ 0.69⁺ 0.81⁺ 0.66⁺ 0.67⁺ 0.58
Omega 1.57 1.68⁺ 1.67⁺ 1.69⁺ 1.83⁺ 1.65⁺ 1.66⁺ 1.56

MV Return 10.34% 10.9%⁺ 9.38% 11.2%⁺ 15.6%⁺ 9.68% 9.66% 9.06%


Volatility 10.70% 13.62% 12.37% 13.98% 13.87% 13.00% 11.04% 13.13%
Sharpe 0.63 0.54 0.46 0.54 0.86⁺ 0.46 0.55 0.41
Omega 1.66 1.53 1.45 1.55 2.07⁺ 1.44 1.52 1.42

RRT Return 7.31% 8.50%⁺ 8.09%⁺ 8.21%⁺ 10.6%⁺ 7.62%⁺ 8.05%⁺ 6.83%
Volatility 7.51% 6.83%⁺ 6.23%⁺ 7.63% 9.80% 7.28%⁺ 6.91%⁺ 7.20%⁺
Sharpe 0.49 0.71⁺ 0.71⁺ 0.60⁺ 0.71⁺ 0.54⁺ 0.63⁺ 0.44
Omega 1.56 1.78⁺ 1.77⁺ 1.67⁺ 2.03⁺ 1.58⁺ 1.71⁺ 1.46

1/N Return 8.58% 8.23% 7.60% 9.09%⁺ 9.53%⁺ 7.71% 7.01% 6.14%
Volatility 8.25% 9.75% 8.40% 12.68% 10.54% 7.75%⁺ 7.70%⁺ 9.27%
Sharpe 0.60 0.47 0.47 0.43 0.56 0.52 0.44 0.27
Omega 1.58 1.45 1.47 1.40 1.55 1.49 1.40 1.24

st.w. Return 9.55% 9.23% 8.92% 9.69%⁺ 9.83%⁺ 9.01% 8.67% 8.29%
Volatility 12.92% 11.6%⁺ 11.4%⁺ 12.2%⁺ 11.8%⁺ 10.7%⁺ 10.9%⁺ 11.4%⁺
Sharpe 0.46 0.48⁺ 0.46⁺ 0.49⁺ 0.52⁺ 0.49⁺ 0.45 0.40
Omega 1.42 1.46⁺ 1.43⁺ 1.47⁺ 1.50⁺ 1.47⁺ 1.42 1.37

Notes: This table reports the out-of-sample portfolio performance for stock-bond portfolios and portfolios com-
plemented with commodities for different asset allocation strategies for the period from January 1986 to Decem-
ber 2013. The table presents the results for an aggressive investor with a maximum desired volatility of 15% p.a.
(risk aversion coefficient: 2). The basis portfolio consists of US stocks and bonds only, while the extended port-
folio additionally includes the indicated commodity group. Improvements in comparison to the stock-bond port-
folio and are in bold and highlighted with +. ‘Return’ denotes the annualized time-series mean of monthly returns
while ‘Volatility’ shows the associated annualized standard deviation. ‘Sharpe’ represents the annualized Sharpe
ratio and ‘Omega’ is the Omega measure for each portfolio.

- 47 -
Table 8: Analysis of commodity portfolio weights

stock-bond portfolio complemented with commodities


Asset Commoditiy
Allocation portfolio
strategy weight GSCI Industrial Precious Agri-
GSCI LightEn. Energy metals Metals Livestock culture
Panel A: Conservative investor
BL Mean 5.34%* 6.26%* 4.37%* 4.75%* 4.65%* 5.19%* 4.46%*
Std.Dev. 2.90% 4.47% 1.72% 2.87% 2.65% 4.60% 3.23%
Maximum 11.96% 25.48% 9.05% 15.05% 12.96% 27.85% 25.80%
MV Mean 10.32%* 17.13%* 4.40%* 7.81%* 9.93%* 9.07%* 7.66%*
Std.Dev. 11.90% 18.36% 5.25% 8.95% 12.00% 12.30% 11.48%
Maximum 47.08% 72.14% 18.86% 32.22% 50.51% 53.46% 46.58%
MinVar Mean 7.60%* 14.55%* 2.80%* 6.63%* 9.56%* 9.79%* 8.05%*
Std.Dev. 6.20% 10.25% 2.25% 3.99% 8.44% 4.29% 6.35%
Maximum 27.32% 35.89% 11.59% 16.60% 29.15% 19.04% 22.74%
RP Mean 6.37%* 13.01%* 2.03%* 4.71%* 8.82%* 9.25%* 7.45%*
Std.Dev. 4.53% 7.84% 0.82% 2.69% 4.17% 3.29% 4.15%
Maximum 18.81% 29.73% 5.71% 14.60% 20.83% 20.00% 17.21%
st.w. Mean 5.00%* 5.00%* 5.00%* 5.00%* 5.00%* 5.00%* 5.00%*
Std.Dev. 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Maximum 5.00% 5.00% 5.00% 5.00% 5.00% 5.00% 5.00%
Panel B: Aggressive investor
BL Mean 13.38%* 13.75%* 12.68%* 12.01%* 12.33%* 10.96%* 9.19%*
Std.Dev. 11.22% 13.93% 8.25% 11.75% 12.14% 13.44% 10.22%
Maximum 43.72% 56.80% 35.21% 57.63% 49.57% 70.42% 54.54%
MV Mean 29.93%* 30.68%* 16.73%* 24.05%* 25.32%* 22.09%* 17.79%*
Std.Dev. 37.80% 42.56% 21.13% 32.91% 36.79% 36.60% 34.05%
Maximum 100.00% 100.00% 58.33% 100.00% 100.00% 100.00% 100.00%
RRT Mean 11.83%* 17.41%* 6.18%* 12.02%* 10.26%* 10.43%* 8.85%*
Std.Dev. 15.21% 19.18% 9.00% 19.06% 14.35% 13.76% 15.47%
Maximum 77.53% 100.00% 60.02% 100.00% 100.00% 69.38% 100.00%
1/N Mean 33.33%* 33.33%* 33.33%* 33.33%* 33.33%* 33.33%* 33.33%*
Std.Dev. 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Maximum 33.33% 33.33% 33.33% 33.33% 33.33% 33.33% 33.33%
st.w. Mean 15.00%* 15.00%* 15.00%* 15.00%* 15.00%* 15.00%* 15.00%*
Std.Dev. 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
Maximum 15.00% 15.00% 15.00% 15.00% 15.00% 15.00% 15.00%

Notes: This table reports the portfolio weights of commodities in the out-of-sample optimized portfolios. ‘Mean’
denotes the average portfolio weight of commodities during the 1986 to 2013 period. ‘Std.Dev.’ denotes the
associated standard deviation of the commodity portfolio weight. ‘Maximum’ refers to the maximum portfolio
share allocated to the indicated commodity group during the 1986 to 2013 period. Panel A reports the results for
a conservative investor and Panel B provides the results for an aggressive investor. * and ** indicate values
significantly larger than 0 at the 1% and 5% level, respectively.

- 48 -
Table 9: Out-of-sample contribution of commodities for conservative investor in sub-periods

Sharpe ratio stock-bond portfolio complemented with commodities


Asset SR stock-
Allocation bond GSCI Industrial Precious Agri-
Period strategy portfolio GSCI LightEn. Energy metals Metals Livestock culture

1986- BL 0.27 0.58⁺ 0.58⁺ 0.48⁺ 0.82⁺ 0.38⁺ 0.53⁺ 0.40⁺


1990 MV 0.23 0.77⁺ 0.84⁺ 0.18 0.99⁺ 0.13 0.55⁺ 0.28⁺
Exp
MinVar 0.29 0.63⁺ 0.76⁺ 0.31⁺ 0.92⁺ 0.29 0.55⁺ 0.38⁺
RP 0.39 0.69⁺ 0.83⁺ 0.42⁺ 0.85⁺ 0.41⁺ 0.63⁺ 0.50⁺
st.w. 0.49 0.58⁺ 0.57⁺ 0.60⁺ 0.80⁺ 0.48 0.61⁺ 0.49

1990- BL 1.35 1.61⁺ 1.61⁺ 1.63⁺ 1.33 1.29 1.53⁺ 1.25


1994 MV 1.31 1.23 1.09 1.35⁺ 1.30 0.79 1.28 1.20
Rec
MinVar 1.38 1.44⁺ 1.50⁺ 1.38 1.29 1.05 1.62⁺ 1.36
RP 1.49 1.75⁺ 1.75⁺ 1.58⁺ 1.44 1.40 1.62⁺ 1.49
st.w. 1.37 1.49⁺ 1.47⁺ 1.47⁺ 1.40⁺ 1.39⁺ 1.48⁺ 1.40⁺

1994- BL 0.75 0.75 0.83⁺ 0.68 0.83⁺ 0.68 0.66 0.86⁺


2001 MV 0.66 0.56 0.65 0.60 0.88⁺ 0.77⁺ 0.62 0.94⁺
Exp
MinVar 0.24 0.10 -0.04 0.19 0.27⁺ -0.09 -0.04 0.15
RP 0.64 0.57 0.48 0.62 0.67⁺ 0.36 0.38 0.66⁺
st.w. 0.73 0.59 0.56 0.62 0.63 0.54 0.52 0.54

2001- BL 0.68 0.98⁺ 0.89⁺ 1.04⁺ 0.79⁺ 0.79⁺ 0.79⁺ 0.76⁺


2004 MV 0.95 1.26⁺ 1.11⁺ 1.23⁺ 0.94 0.76 0.60 0.69
Rec
MinVar 0.96 1.07⁺ 1.14⁺ 1.01⁺ 1.08⁺ 1.00⁺ 0.90 0.99⁺
RP 0.93 1.07⁺ 1.09⁺ 1.03⁺ 1.01⁺ 0.98⁺ 0.93 0.87
st.w. 0.54 0.83⁺ 0.74⁺ 0.91⁺ 0.69⁺ 0.70⁺ 0.70⁺ 0.62⁺

2004- BL 0.88 0.93⁺ 0.92⁺ 0.93⁺ 1.56⁺ 1.34⁺ 0.71 0.84


2008 MV 0.86 0.88⁺ 0.75 0.94⁺ 1.20⁺ 1.15⁺ 0.59 0.81
Exp
MinVar 0.90 1.02⁺ 1.06⁺ 0.99⁺ 1.26⁺ 1.00⁺ 0.59 0.65
RP 0.93 1.04⁺ 1.07⁺ 1.01⁺ 1.47⁺ 1.37⁺ 0.79 0.65
st.w. 0.91 1.01⁺ 1.05⁺ 0.97⁺ 1.42⁺ 1.26⁺ 0.72 0.90

2008- BL 1.20 1.08 1.04 1.09 1.06 1.11 1.18 0.98


2012 MV 1.29 0.92 0.91 0.84 1.11 0.89 1.01 0.88
Rec
MinVar 0.90 0.76 0.76 0.77 0.82 0.86 0.76 0.87
RP 0.79 0.80⁺ 0.76 0.81⁺ 0.79 0.81⁺ 0.72 0.75
st.w. 0.98 0.84 0.85 0.83 0.83 0.96 0.94 0.81

Notes: This table reports the portfolio performance (Sharpe ratios) for different asset allocation strategies for six
sub-periods between 1986 and 2013 for a conservative investor with a maximum desired volatility of 5% p.a.
(risk aversion coefficient: 10). The basis portfolio consists only of US stocks and bonds while the extended port-
folio additionally includes the indicated commodity group. Improvements in comparison to the stock-bond port-
folio and are in bold and highlighted with +.

- 49 -
Table 10: Out-of-sample contribution of commodities for alternative commodity indices

Asset stock-bond portfolio complemented with commodities


Allocation Performance stock-bond GSCI EWCI ex
strategy measure portfolio GSCI LightEn. EWCI LS&AG
Conservative investor
BL Sharpe 0.81 0.93⁺ 0.92⁺ 0.96⁺ 0.96⁺
Omega 1.83 2.00⁺ 1.99⁺ 2.06⁺ 2.06⁺

MV Sharpe 0.85 0.91⁺ 0.87⁺ 0.96⁺ 1.03⁺


Omega 1.84 1.96⁺ 1.91⁺ 2.02⁺ 2.12⁺

MinVar Sharpe 0.70 0.74⁺ 0.75⁺ 0.79⁺ 0.79⁺


Omega 1.68 1.75⁺ 1.76⁺ 1.82⁺ 1.81⁺

RP Sharpe 0.79 0.90⁺ 0.90⁺ 0.92⁺ 0.90⁺


Omega 1.80 1.95⁺ 1.95⁺ 1.99⁺ 1.95⁺

st.w. Sharpe 0.79 0.82⁺ 0.80⁺ 0.82⁺ 0.85⁺


Omega 1.82 1.85⁺ 1.83⁺ 1.87⁺ 1.91⁺
Aggressive investor
BL Sharpe 0.58 0.68⁺ 0.67⁺ 0.75⁺ 0.74⁺
Omega 1.57 1.68⁺ 1.67⁺ 1.76⁺ 1.76⁺

MV Sharpe 0.63 0.54 0.46 0.59 0.67⁺


Omega 1.66 1.53 1.45 1.59 1.68⁺

RRT Sharpe 0.49 0.71⁺ 0.71⁺ 0.84⁺ 0.76⁺


Omega 1.56 1.78⁺ 1.77⁺ 1.94⁺ 1.85⁺

1/N Sharpe 0.60 0.47 0.47 0.59 0.62⁺


Omega 1.58 1.45 1.47 1.59⁺ 1.62⁺

st.w. Sharpe 0.46 0.48⁺ 0.46⁺ 0.49⁺ 0.52⁺


Omega 1.42 1.46⁺ 1.43⁺ 1.47⁺ 1.50⁺

Notes: This table reports the portfolio benefits of commodities for alternative equally weighted commodity indi-
ces for different asset allocation strategies between 1986 and 2013. Panel A reports the results for a conservative
investor and Panel B provides the results for an aggressive investor. Improvements in comparison to the stock-
bond portfolio are bold and highlighted with +. ‘Sharpe’ represents the annualized Sharpe ratio and ‘Omega’ is
the Omega measure for each portfolio.

- 50 -
Table 11: Out-of-sample contribution of commodities for alternative basis portfolio

stock-bond portfolio complemented with commodities


Asset stock-
Allocation Performance bond GSCI Industrial Precious Agri-
strategy measure portfolio GSCI LightEn. Energy metals Metals Livestock culture
Conservative investor
BL Sharpe 0.95 1.04⁺ 1.05⁺ 1.04⁺ 1.08⁺ 0.98⁺ 1.00⁺ 0.95
Omega 1.98 2.13⁺ 2.14⁺ 2.11⁺ 2.21⁺ 2.05⁺ 2.11⁺ 2.00⁺

MV Sharpe 0.94 1.03⁺ 0.94⁺ 0.96⁺ 1.03⁺ 0.89 0.90 0.81


Omega 2.00 2.17⁺ 2.01⁺ 2.05⁺ 2.28⁺ 1.93 1.97 1.84

MinVar Sharpe 0.78 0.83⁺ 0.85⁺ 0.77 0.87⁺ 0.70 0.75 0.73
Omega 1.80 1.87⁺ 1.88⁺ 1.79 1.94⁺ 1.68 1.77 1.72

RP Sharpe 0.86 0.92⁺ 0.91⁺ 0.88⁺ 0.92⁺ 0.85 0.84 0.83


Omega 1.92 2.01⁺ 2.00⁺ 1.94⁺ 2.00⁺ 1.89 1.90 1.87

st.w. Sharpe 0.79 0.82⁺ 0.81⁺ 0.83⁺ 0.86⁺ 0.82⁺ 0.81⁺ 0.76
Omega 1.80 1.87⁺ 1.85⁺ 1.87⁺ 1.92⁺ 1.85⁺ 1.85⁺ 1.78
Aggressive investor
BL Sharpe 0.66 0.72⁺ 0.71⁺ 0.77⁺ 0.75⁺ 0.68⁺ 0.73⁺ 0.68⁺
Omega 1.68 1.75⁺ 1.74⁺ 1.80⁺ 1.82⁺ 1.70⁺ 1.79⁺ 1.71⁺

MV Sharpe 0.65 0.63 0.61 0.67⁺ 0.78⁺ 0.58 0.64 0.44


Omega 1.66 1.63 1.62 1.69⁺ 1.92⁺ 1.56 1.64 1.43

RRT Sharpe 0.77 0.91⁺ 0.87⁺ 0.87⁺ 0.82⁺ 0.77⁺ 0.81⁺ 0.73
Omega 1.86 2.06⁺ 1.98⁺ 2.00⁺ 2.04⁺ 1.84 1.93⁺ 1.78

1/N Sharpe 0.46 0.46 0.44 0.46⁺ 0.50⁺ 0.46⁺ 0.45 0.37
Omega 1.43 1.43⁺ 1.42 1.44⁺ 1.48⁺ 1.43⁺ 1.41 1.34

st.w. Sharpe 0.38 0.40⁺ 0.38⁺ 0.41⁺ 0.44⁺ 0.40⁺ 0.39⁺ 0.34
Omega 1.34 1.36⁺ 1.35⁺ 1.38⁺ 1.41⁺ 1.36⁺ 1.35⁺ 1.30

Notes: This table reports the out-of-sample portfolio benefits of commodities for an extended basis portfolio
consisting of international stocks, government bonds, and corporate bonds and the basis portfolios complement-
ed with the indicated commodity group for different asset allocation strategies during the time-period from Janu-
ary 1986 to December 2013. Panel A reports the results for a conservative investor and Panel B provides the
results for an aggressive investor. Improvements in comparison to the stock-bond portfolio are bold and in ital-
ics. ‘Sharpe’ represents the annualized Sharpe ratio and ‘Omega’ is the Omega measure for each portfolio.

- 51 -
Table 12: Out-of-sample contribution of individual commodity futures and physical com-
modities

stock-bond portfolio complemented with commodities


Asset
Allocation Performance stock-bond
strategy measure portfolio Crude Oil Copper Gold Live Cattle Corn
Panel A: Individual commodity futures contracts (1993-2013)
BL Sharpe 0.91 1.01⁺ 1.04⁺ 0.99⁺ 0.94⁺ 0.89
Omega 1.98 2.13⁺ 2.22⁺ 2.10⁺ 2.03⁺ 1.94

MV Sharpe 0.94 0.87 0.99⁺ 0.91 0.70 0.80


Omega 1.98 1.89 2.08⁺ 1.95 1.67 1.78

MinVar Sharpe 0.74 0.74⁺ 0.79⁺ 0.70 0.76⁺ 0.73


Omega 1.74 1.75⁺ 1.83⁺ 1.69 1.77⁺ 1.73⁺

RP Sharpe 0.82 0.88⁺ 0.91⁺ 0.84⁺ 0.88⁺ 0.85⁺


Omega 1.85 1.93⁺ 1.99⁺ 1.88⁺ 1.92⁺ 1.88⁺

st.w. Sharpe 0.82 0.89⁺ 0.88⁺ 0.87⁺ 0.86⁺ 0.82


Omega 1.86 1.97⁺ 1.99⁺ 1.94⁺ 1.92⁺ 1.85
Panel B: Individual physical commodities (1993-2013)
BL Sharpe 0.91 0,98⁺ 1,01⁺ 0,94⁺ 0,92⁺ 0.85
Omega 1.98 2,09⁺ 2,18⁺ 2,03⁺ 1,99⁺ 1.89

MV Sharpe 0.94 0.84 0,97⁺ 0.81 0.78 0.79


Omega 1.98 1.85 2,06⁺ 1.83 1.76 1.76

MinVar Sharpe 0.74 0.72 0,78⁺ 0.64 0.71 0.71


Omega 1.74 1.72 1,81⁺ 1.61 1.70 1.70

RP Sharpe 0.82 0,86⁺ 0,89⁺ 0.79 0,83⁺ 0,83⁺


Omega 1.85 1,91⁺ 1,97⁺ 1.80 1,85⁺ 1,86⁺

st.w. Sharpe 0.82 0,85⁺ 0,85⁺ 0,83⁺ 0,83⁺ 0.78


Omega 1.86 1,91⁺ 1,94⁺ 1,89⁺ 1,86⁺ 1.80

Notes: This table reports the out-of-sample portfolio benefits of commodities for different asset allocation strate-
gies during the time-period from January 1993 to December 2013 for a conservative investor. Panel A reports
the results when investing in individual commodity futures and Panel B provides the results when investing in
physical commodities (spot prices) neglecting storage costs. Improvements in comparison to the stock-bond
portfolio are in bold and highlighted with +. ‘Sharpe’ represents the annualized Sharpe ratio and ‘Omega’ is the
Omega measure for each portfolio.

- 52 -
Table 13: Out-of-sample contribution of commodities with short selling

stock-bond portfolio complemented with commodities


Asset stock-
Allocation Performance bond GSCI Industrial Precious Agri-
strategy measure portfolio GSCI LightEn. Energy metals Metals Livestock culture
Conservative investor
BL Sharpe 0.82 0.94⁺ 0.95⁺ 0.91⁺ 0.97⁺ 0.85⁺ 0.87⁺ 0.83⁺
Omega 1.84 2.02⁺ 2.04⁺ 1.97⁺ 2.10⁺ 1.88⁺ 1.94⁺ 1.85⁺

MV Sharpe 0.78 0.94⁺ 0.92⁺ 0.83⁺ 1.01⁺ 0.72 0.78⁺ 0.79⁺


Omega 1.78 2.04⁺ 2.00⁺ 1.85⁺ 2.19⁺ 1.72 1.80⁺ 1.80⁺

MinVar Sharpe 0.67 0.71⁺ 0.72⁺ 0.65 0.80⁺ 0.59 0.62 0.63
Omega 1.65 1.71⁺ 1.71⁺ 1.63 1.85⁺ 1.55 1.60 1.59

RP Sharpe 0.79 0.90⁺ 0.90⁺ 0.83⁺ 0.94⁺ 0.78 0.77 0.78


Omega 1.80 1.95⁺ 1.95⁺ 1.85⁺ 2.01⁺ 1.78 1.78 1.77

st.w. Sharpe 0.79 0.82⁺ 0.80⁺ 0.83⁺ 0.87⁺ 0.80⁺ 0.79 0.74
Omega 1.82 1.85⁺ 1.83⁺ 1.87⁺ 1.95⁺ 1.83⁺ 1.82 1.75
Aggressive investor
BL Sharpe 0.56 0.68⁺ 0.70⁺ 0.68⁺ 0.74⁺ 0.61⁺ 0.66⁺ 0.59⁺
Omega 1.55 1.71⁺ 1.76⁺ 1.68⁺ 1.76⁺ 1.61⁺ 1.69⁺ 1.59⁺

MV Sharpe 0.47 0.62⁺ 0.59⁺ 0.49⁺ 0.75⁺ 0.46 0.51⁺ 0.47⁺


Omega 1.44 1.59⁺ 1.56⁺ 1.45⁺ 1.78⁺ 1.42 1.47⁺ 1.45⁺

RRT Sharpe 0.49 0.71⁺ 0.71⁺ 0.60⁺ 0.71⁺ 0.54⁺ 0.63⁺ 0.44
Omega 1.56 1.78⁺ 1.77⁺ 1.67⁺ 2.03⁺ 1.58⁺ 1.71⁺ 1.46

1/N Sharpe 0.60 0.47 0.47 0.43 0.56 0.52 0.44 0.27
Omega 1.58 1.45 1.47 1.40 1.55 1.49 1.40 1.24

st.w. Sharpe 0.46 0.48⁺ 0.46 0.49⁺ 0.52⁺ 0.49⁺ 0.45⁺ 0.40
Omega 1.42 1.46⁺ 1.43⁺ 1.47⁺ 1.50⁺ 1.47⁺ 1.42⁺ 1.37

Notes: This table reports the out-of-sample portfolio benefits of commodities when short selling is possible. The
evaluation period is from January 1986 to December 2013. Panel A reports the results for a conservative investor
and Panel B provides the results for an aggressive investor. Improvements in comparison to the stock-bond port-
folio and are in bold and highlighted with +. ‘Sharpe’ represents the annualized Sharpe ratio and ‘Omega’ is the
Omega measure for each portfolio.

- 53 -
Table 14: Alternative rebalancing frequencies

stock-bond portfolio complemented with commodities


Asset stock-
Allocation Performance bond GSCI Industrial Precious Agri-
strategy measure portfolio GSCI LightEn. Energy metals Metals Livestock culture
Panel A: Quarterly Rebalancing
BL Sharpe 0.86 0.92⁺ 0.91⁺ 0.88⁺ 1.03⁺ 0.85 0.86 0.80
Omega 1.90 1.98⁺ 1.97⁺ 1.92⁺ 2.24⁺ 1.88 1.91⁺ 1.81

MV Sharpe 0.84 0.85⁺ 0.84⁺ 0.77 1.00⁺ 0.76 0.72 0.76


Omega 1.84 1.87⁺ 1.86⁺ 1.75 2.18⁺ 1.74 1.71 1.77

MinVar Sharpe 0.71 0.74⁺ 0.74⁺ 0.69 0.87⁺ 0.63 0.64 0.67
Omega 1.69 1.75⁺ 1.74⁺ 1.68 1.95⁺ 1.60 1.62 1.64

RP Sharpe 0.80 0.91⁺ 0.91⁺ 0.84⁺ 0.97⁺ 0.78 0.77 0.78


Omega 1.81 1.98⁺ 1.96⁺ 1.87⁺ 2.06⁺ 1.78 1.78 1.78

st.w. Sharpe 0.81 0.84⁺ 0.82⁺ 0.86⁺ 0.90⁺ 0.82⁺ 0.80 0.76
Omega 1.85 1.88⁺ 1.85⁺ 1.92⁺ 1.99⁺ 1.85⁺ 1.84 1.77
Panel B: Yearly Rebalancing
BL Sharpe 0.81 0.85⁺ 0.80⁺ 0.81⁺ 0.91⁺ 0.79 0.82⁺ 0.75
Omega 1.81 1.88⁺ 1.81⁺ 1.83⁺ 1.99⁺ 1.79 1.85⁺ 1.74

MV Sharpe 0.88 0.77 0.65 0.72 0.91⁺ 0.80 0.87 0.75


Omega 1.90 1.77 1.62 1.71 1.98⁺ 1.78 1.90 1.75

MinVar Sharpe 0.72 0.75⁺ 0.75⁺ 0.69 0.92⁺ 0.63 0.64 0.65
Omega 1.71 1.75⁺ 1.75⁺ 1.67 2.03⁺ 1.59 1.61 1.61

RP Sharpe 0.79 0.88⁺ 0.86⁺ 0.82⁺ 0.99⁺ 0.75 0.77 0.74


Omega 1.79 1.93⁺ 1.90⁺ 1.84⁺ 2.13⁺ 1.73 1.77 1.72

st.w. Sharpe 0.81 0.84⁺ 0.82⁺ 0.86⁺ 0.92⁺ 0.81⁺ 0.80 0.76
Omega 1.84 1.87⁺ 1.85⁺ 1.90⁺ 2.02⁺ 1.84⁺ 1.84 1.76

Notes: This table reports the out-of-sample portfolio benefits of commodities for different rebalancing frequen-
cies during the time-period from January 1986 to December 2013. Panel A reports the results for a conservative
investor and Panel B provides the results for an aggressive investor. Improvements in comparison to the stock-
bond portfolio and are in bold and highlighted with +. ‘Sharpe’ represents the annualized Sharpe ratio and ‘Ome-
ga’ is the Omega measure for each portfolio.

- 54 -
Appendix:

Overview of asset allocation strategies

Asset Allocation Optimization function/


Constraint Required Parameters
Strategy Portfolio weight
Risk-Parity 1 / ˆ i2 not required volatility of asset- returns
i  N

 (1 / ˆ
i 1
2
i )

Reward-to-risk- ˆ i / ˆ i2 not required returns and volatility of


i 
timing N returns
 (ˆ
i 1

i / ˆ i2 )

Minimum- variances and covariances


min '  N
Variance 
 i  1 und i  0
i 1
(covariance-matrix) of
asset-returns

 N
Mean-Variance max U  '   ' ,
 2 
i 1
i  1 und i  0 risk-aversion-parameter,
covariance-matrix and risk-
estimates of all assets
'    ˆ C ,

 N
Black-Litterman max U  '  BL  '  BL
 2 
i 1
i  1 und i  0 risk-aversion-parameter,
reference-portfolio,
with: covariance-matrix ,
'    ˆ C ,
ˆ
BL

   P' 1P
1
 
1 1
  P' 1Q , risk-estimates of all assets,
reliability of return

BL

      P' 1P
1

1
estimates

  

with:   portfolio weight of asset i


σ i  volatility of asset i
U  utilityfunction of investor
ω  vector of portfolio weights
μ  vector of return estimates
δ  risik - aversion - coefficien t
Θ  estimate for covariance matrix
Q  vector of views
Π  vector of imploed return estimates
Ω  reliability of views
τ  parameter that refelcts reliability of views
  vector of portfolio weights of reference portfolio
P  unity matrix (in case of one absolute return estimate for each asset)

- 55 -

You might also like