Momentum in International Commodity Futures Markets

• DOI: http://doi.org/10.1002/fut.21834
• Date: 01 August 2017
• Published date: 01 August 2017

Momentum in International Commodity

Futures Markets

Jangkoo Kang and Kyung Yoon Kwon*

This paper examines whether commodity futures momentum can predict business cycles in the

US, China, UK, Japan, and India. Momentum as a risk factor may play a role as a state variable

in the spirit of Liew and Vassalou (2000). We find significant and negative predictability of

commodity futures momentum, although the basis factor of the commodity futures markets

shows insignificant results. Moreover, we find that commodity futures momentum is an

independent factor that cannot be fully explained by traditional risk factors, macroeconomic

variables, or commodity sector momentum. © 2017 Wiley Periodicals, Inc. Jrl Fut Mark

37:803–835, 2017

1. INTRODUCTION

Liew and Vassalou (2000) document that the performance of a risk factor should be related to

future economic growth if that factor plays a role as a state variable based on the Merton

(1973) intertemporal capital asset pricing model. They examine whether size, book-to-

market, and stock momentum factors in the US stock markets have predictive power for

future GDP growth; they find that the size and book-to-market factors have positive and

significant predictive power.

The literature on commodity futures markets reports that the risk premium of

commodity futures is related to past returns (Erb & Harvey, 2006; Fuertes, Miffre, & Rallis,

2010; Miffre & Rallis, 2007; Narayan, Ahmed, & Narayan, 2015), and this positive relation

between past return and future return is generally called momentum. Erb and Harvey (2006)

show that momentum strategy with a 12-month ranking period and a 1-month holding period

is profitable in the US commodity futures markets. Miffre and Rallis (2007) examine the

profitability of momentum and contrarian strategies in the US commodity futures markets

and report that momentum strategies in commodity futures markets are profitable whereas

contrarian strategies are not. Asness, Moskowitz, and Pedersen (2013) also document that

momentum profits exist in the commodity futures market as well as other asset markets.

Likewise, studies consistently document the existence of momentum in the commodity

futures markets, but its role as a risk factor in the market has not been investigated

sufficiently.

Jangkoo Kang and Kyung Yoon Kwon are at the College of Business, Korea AdvancedInstitute of Science and

Technology, Seoul, South Korea.

*Correspondence author, College of Business, Korea Advanced Institute of Science and Technology, 85 Hoegiro,

Dongdaemoon-gu, Seoul 02455, South Korea. Tel: þ82-2-958-3693, Fax: þ82-2-958-3620,

e-mail: noldya@business.kaist.ac.kr

Received November 2016; Accepted November 2016

The Journal of Futures Markets, Vol. 37, No. 8, 803–835 (2017)

© 2017 Wiley Periodicals, Inc.

Published online 1 Feruary 2017 in Wiley Online Library (wileyonlinelibrary.com).

DOI: 10.1002/fut.21834

This paper examines whether commodity futures momentum predicts the business

cycle in the spirit of Liew and Vassalou (2000). We first confirm the existence of the

momentum effect in the commodity futures markets in five countries: the US, China, the UK,

Japan, and India; previous findings on commodity futures momentum focus mainly on the US

market. Using three risk factor models, the Fama and French (1993) three factor model, the

Carhart (1997) four factor model, and the Fama and French (2015) five factor model, we

confirm that momentum profits remain significant even after controlling for these risk

factors. Next, we examine the relation between momentum and future GDP growth. We find

that the commodity futures momentum factor negatively predicts GDP growth, from one

quarter to 1 year, and this predictability appears to be robust to other macroeconomic

variables.

1

In examining the predictive p ower of the momentum factor, we in vestigate the

predictability of the basis f actor for comparison. In the li terature, the significant relation

between commodity futures re turns and the basis, the differenc e between the spot price

and the futures price, has been consis tently reported (Fama & French, 19 87; Gorton,

Hayashi, & Rouwenhorst, 2012 ; Miffre & Rallis, 2007), and several studies report that the

profitability of momentum stra tegies is closely associat ed with this basis premium (Erb &

Harvey 2006; Gorton et al., 2 012; Miffre & Rallis, 2007; Szymanowska, de Roon, Nijman,

& Goorbergh, 2014). Gorton et al. (2012) suggest a model that provides theoretical

predictions on the relations among t he basis, past returns, and the risk premiu m of

commodity futures. They doc ument that the futures basis, pas t futures returns, past spot

returns, and spot price vola tilities reflect the current state of inventories, and thus are

related to the risk premium , whereas this varies with t he state of inventories. Mif fre and

Rallis (2007) document tha t the profits of momentum strateg ies can be generated by

buying backwardated contra cts and selling contangoed con tracts. Szymanowska et al.

(2014) also show that momentum returns base d on past 12-month returns can be

explained by the basis premium .

2

These studies consistentl y suggest the explanatory power

of the basis premium for momentum , though most of them are restricted to a small sample

of contracts or markets or focus on only one specific type of momentum strategy. Thus, we

investigate the predictabilit y of the basis factor, in addition to the momentum f actor, to

examine whether the momentum factor is independent of the basis factor. Surprisingly,

our results show that the ba sis factor does not have pre dictive power for future GD P

growth, as opposed to the mom entum factor. This shows tha t the momentum factor has

characteristics differen t from the basis factor. The momentum factor may be regarded as a

state variable or a risk factor t hat is closely related to futur e GDP growth, although the

basis factor is not related to it.

Finally, we examine momentum profit in the international commodity futures markets

using a risk-based approach and a commodity-characteristic approach. First, we test whether

momentum profit can be fully explained by the macroeconomic variable model. Following

Chordia and Shivakumar (2002), we predict the next-month return on individual commodity

futures contracts based on the time-series regression model with macroeconomic variables,

and examine whether momentum profits are predictable via macroeconomic variables. Our

results show that the macroeconomic model can predict only a portion of momentum profits.

1

For the test using macroeconomic variables, we restrict the sample country to the US because of data availability.

2

Szymanowska et al. (2014) show that returns sorted on the past 12-month returns can be explained by returns

sorted on the basis, but Shen, Szakmary, and Sharma (2007) report that momentum strategies in the commodity

futures markets generate the largest profits for short ranking and holding periods. Indeed, we also find that

momentum profits are largest for the 1-month ranking period and 1-month holding period. Thus, the explanatory

power of the basis premium for the most profitable momentum strategy, especially for these short ranking and

holding periods, remains unexplored.

804 Kang and Kwon

These results indicate that the risk-based approach using macroeconomic variables cannot

fully explain momentum profits. Next, we examine whether characteristics of the commodity

sector are the source of the commodity futures momentum. Belousova and Dorfleitner

(2012) note that the diversification benefits of commodities to a portfolio depend on

commodity type. Different diversification benefits can be related to differing relations with

the business cycle of each commodity sector; thus, we test whether commodity futures

momentum is related to the commodity sector. We categorize commodity futures contracts

into five sectors based on the characteristics of the underlying commodities: “Metals,”

“Softs,”“Grains,”“Meats,”and “Energies.”Then, we explore the role of sector momentum in

explaining individual commodity futures momentum in the same way that Moskowitz and

Grinblatt (1999) examine the relation between industry momentum and individual stock

momentum. Our results show that sector momentum appears to partially contribute only to

the 1-month momentum, but it cannot explain a substantial portion of the individual

commodity futures momentum. As both the risk-based approach using risk models and the

commodity-characteristic approach fail to explain commodity futures momentum, our

results suggest that a substantial part of commodity futures momentum stems from a

contract-specific component, which should be further examined.

The remainder of this paper proceeds as follows. Section 2 describes the data. Section 3

presents our empirical results. In Section 3.1, we first verify the existence of momentum in

the international commodity futures markets, and in Section 3.2, we examine predictability

of the momentum risk factor for future GDP growth. Section 3.3 presents the performance of

the macroeconomic variable model in explaining momentum, and Section 3.4 shows the

relation between individual commodity futures momentum and commodity futures sector

momentum. Finally, Section 4 sets forth our conclusions.

2. DATA

The data, obtained from Datastream, comprise daily settlement prices on 32 US commodity

futures contracts, 20 Chinese commodity futures contracts, 16 UK commodity futures

contracts, 16 Japanese commodity futures contracts, and 13 Indian futures contracts. We

include in our sample only countries that have more than 10 commodity futures. For each

country, the sample period starts with the year in which at least five commodity futures exist

and ends in June 2015. The sample period for each country and other details are reported in

Table I.

We exclude commodity futures that are delisted during the sample period in the

commodity futures markets.

3

To compile the time-series of futures returns, we assume that

we hold the nearby contract up to the end of the month prior to the maturity month.

4

At the

end of that month, we roll our position over to the second nearest-to-maturity contract and

hold that contract until the end of the month prior to maturity. This rolling procedure allows

us to minimize problems related to lack of liquidity and to compute the returns from holding

3

As this exclusion of delisted contracts in our sample may produce a look-ahead sample selection bias, we examine

whether including delisted futures contracts changes our results, but the results are qualitatively the same, whereas

the delisted futures have a rather short period compared to the whole sample period. We exclude these delisted

futures because prices of the nearest contracts do not change for several years before they are delisted, and

identifying the effective sample period of the contract is rather ambiguous as opposed to the case of delisted stocks.

Many studies use commodity futures data provided by the Commodities Research Bureau (CRB), and Gorton and

Rouwenhorst (2006) document that the CRB database contains data primarily for futures contracts that have

survived until today.

4

Some contracts are traded until the end of the maturity month. In these cases, we roll our position at the end of the

maturity month.

Momentum in International Commodity Futures Markets 805