Basis‐Momentum

• Date: 01 February 2019
• Author: MARTIJN BOONS,MELISSA PORRAS PRADO
• Published date: 01 February 2019
• DOI: http://doi.org/10.1111/jofi.12738

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 1 •FEBRUARY 2019

Basis-Momentum

MARTIJN BOONS and MELISSA PORRAS PRADO∗

ABSTRACT

We introduce a return predictor related to the slope and curvature of the futures

term structure: basis-momentum. Basis-momentum strongly outperforms benchmark

characteristics in predicting commodity spot and term premiums in both the time se-

ries and the cross section. Exposure to basis-momentum is priced among commodity-

sorted portfolios and individual commodities. We argue that basis-momentum cap-

tures imbalances in the supply and demand of futures contracts that materialize

when the market-clearing ability of speculators and intermediaries is impaired, and

that it represents compensation for priced risk. Our findings are inconsistent with

alternative explanations based on storage, inventory, and hedging pressure.

IN THIS PAPER,WE INTRODUCE a characteristic derived from the futures term

structure, which we refer to as “basis-momentum.” Basis-momentum is directly

observable ex ante and is the strongest predictor of commodity returns to date,

in the cross section and time series as well as across maturities. A large liter-

ature studies stock and bond risk premiums along these dimensions. We show

that exposure to a basis-momentum factor is priced in the broadest cross section

of commodities studied to date. We further explore potential explanations and

argue that the basis-momentum effect is most consistent with priced risk that

derives from the role of speculators and financial intermediaries in commodity

∗Boons and Porras Prado are with the Nova School of Business and Economics in Portugal. A

previous version of this paper was circulated under the title “Basis-momentum in the futures curve

and volatility risk.” An Internet Appendix with supplementary results as well as replication data

may be found in the online version of this article. We thank WeiXiong (the Editor), two anonymous

referees, as well as Rui Albuquerque, Pedro Barroso, Magnus Dahlquist, Miguel Ferreira, Vincent

van Kervel, Antonio Moreno, Andreas Neuhierl, Robert Richmond, Frans de Roon, Pedro Santa

Clara, Ivan Shaliastovich, Kenneth Singleton, Marta Szymanowska, and Ke Tang for helpful

comments. We also thank seminar participants at the Bocconi University, Commodity Market

Workshop 2015 in Oslo, ECOMFIN 2016 in Paris, EFA 2016 in Oslo, FMA Annual Meeting 2016

in Las Vegas, Luxembourg School of Finance, NBER Commodity Workshop 2016, Nova School of

Business and Economics, Spanish Finance Forum 2016 in Madrid, Stockholm School of Economics,

Universit`

a Cattolica del Sacro Cuore, University of Kentucky,and 4Nations Cup in Lisbon. We are

grateful to Marta Szymanowska for providing us with data. This work was funded by Fundac¸˜

ao

para a Ciˆ

encia e a Tecnologia (UID/ECO/00124/2013 and Social Sciences DataLab, Project 22209),

POR Lisboa (LISBOA-01-0145-FEDER-007722 and Social Sciences DataLab, Project 22209) and

POR Norte (Social Sciences DataLab, Project 22209). We have read the Journal of Finance’s

disclosure policy and have no conflicts of interest to disclose. Corresponding author: Martijn Boons,

Address: Nova School of Business and Economics, Universidade NOVA de Lisboa, Campus de

Carcavelos, Rua da Holanda 1, 2775-405 Carcavelos, Portugal. Email: martijn.boons@novasbe.pt.

DOI: 10.1111/jofi.12738

239

240 The Journal of Finance R

markets. These asset pricing implications are also important for practitioners

because the recent financialization has induced large and increasingly active

institutional investment in commodities.1

Basis-momentum is measured as the difference in momentum between first-

and second-nearby futures strategies, and can be decomposed into average

curvature and changes in the slope of the futures curve. Given that the futures

curve is typically steeper on the short end, it is not surprising to find that

curvature positively predicts both nearby returns and spreading returns (from

a long-short position in both a nearby contract and a farther-from-expiring

contract). Likewise, persistence in the steepening of the slope predicts nearby

returns both in absolute terms and relative to farther-from-expiring returns.

As in bond markets, the dynamics of commodity futures curves are driven by

three factors: level, slope, and curvature (see Karstanje, van der Wel, and van

Dijk (2015)). Nevertheless, our evidence suggests that the basis-momentum

factor is the single best predictor of returns. In a similar spirit, Cochrane and

Piazzesi (2005) find that a single tent-shaped function of forward rates is the

best predictor of bond returns. More recent work on the term structure incorpo-

rates this tent-shaped factor as an additional characteristic to be matched by

the model, thus connecting the factor structure of expected returns and yields

(see, for example, Cochrane and Piazzesi (2008), Xiong and Yan (2010), and

Campbell, Sunderam, and Viceira (2016)).

Sorting 21 commodities since 1959, we find a large average annualized differ-

ence between the high and low basis-momentum portfolios of 18.38% (t=6.73)

in (first-) nearby returns and 4.08% (t=6.43) in (first-nearby minus second-

nearby) spreading returns.2These returns translate into Sharpe ratios of 0.9

and capture commodity spot and term premiums, respectively. In pooled re-

gressions that control for systematic differences across commodities, a one-

standard-deviation increase in basis-momentum predicts a large increase in

annualized nearby (spreading) returns of 10.23% (2.29%). Benchmark predic-

tors such as basis and momentum are considerably weaker in isolation and are

subsumed by basis-momentum in joint tests.3

Additional tests show that both curvature and changes in slope contribute to

basis-momentum predictability, but it is curvature that contributes the most.

This finding is important because basis and momentum are not directly re-

lated to curvature. Also, the restriction imposed by basis-momentum, namely,

that the difference between momentum measured at different points on the

curve outperforms a single momentum measure is supported in the data.

1For recent work on financialization, see, for example, Tang and Xiong (2012), Cheng, Kirilenko,

and Xiong (2015), Sockin and Xiong (2015), and Basak and Pavlova (2016).

2These returns are robust to using the estimates of transaction costs reported in Marshall,

Nguyen, and Visaltanochoti (2012) and Bollerslev et al. (2016) as well as to limiting various

subsamples.

3For empirical evidence on the basis (the difference between the futures and spot price) and

momentum, see, for example, Moskowitz, Ooi, and Pedersen (2012), Yang (2013), Szymanowska

et al. (2014), Bakshi, Gao, and Rossi (2017), and Koijen et al. (2018).

Basis-Momentum 241

Finally, evidence from first- to fourth-nearby contract returns suggests that

basis-momentum predictability is maturity-specific.

Documenting the strength of basis-momentum predictability across all of

these different dimensions is our first contribution. Our second contribu-

tion is to a recent literature that constructs commodity factor pricing mod-

els, in the spirit of Fama and French (1993). We construct basis-momentum

nearby and spreading factors. In time-series spanning regressions, the basis-

momentum factors generate large alphas relative to the three-factor models of

Szymanowska et al. (2014) and Bakshi, Gao, and Rossi (2017). These models

include commodity market, basis, and momentum factors. We run asset pricing

tests using as test assets the nearby and spreading returns of either a range

of portfolios (sorted on characteristics and sectors) or individual commodities.

We find that exposure to the basis-momentum nearby factor captures priced

risk that is orthogonal to the benchmark factors. The basis-momentum risk

premium is close to the sample average return of the factor and translates

into a Sharpe ratio ranging from 0.55 to 0.85 (depending on the specification).

In fact, a two-factor model that includes a commodity market factor and the

basis-momentum nearby factor provides a cross-sectional fit that is similar to

larger three- and four-factor models. In contrast to the two factors in this par-

simonious model, we find that the pricing performance of additional factors is

sensitive to the specification of the asset pricing test.

Our third and final contribution is in exploring the economic determinants

of basis-momentum. We argue that the classical theories of storage (Kaldor

(1939), Working (1949), and Deaton and Laroque (1992)), normal backwar-

dation (Keynes (1930)), and hedging pressure (Cootner (1960,1967)) are un-

likely to explain the effect.4For storage, we present three results. First, the

basis-momentum effect is similar for commodities with high or low inventory

and storability.Second, basis-momentum predicts returns controlling for basis,

momentum, and volatility, which are price-based measures of inventory risk

(Gorton, Hayashi, and Rouwenhorst (2012)). Third, basis-momentum predicts

returns of currencies as well as stock and bond indexes, that is financial assets

that can be stored costlessly. Although stronger in commodities, the existence

of an effect in other assets indicates that basis-momentum is of general inter-

est in asset pricing. Looking at hedging pressure, we note that the principal

ideas of Keynes and Cootner say little about maturity-specific effects. Moreover,

basis-momentum is empirically robust to controlling for hedging pressure.

We also analyze explanations that rely on the market-clearing ability of spec-

ulators, and of financial intermediaries more generally. We show that basis-

momentum predictability is substantially stronger when speculators have

many spreading positions. Such “spreading pressure” also predicts commod-

ity returns in isolation, which is a new result in the literature. The information

4The theory of storage assumes that holders of inventories receive a convenience yield that

declines as inventory increases, and that futures prices are set through cost-of-carry arbitrage.

Hedging pressure is a reinterpretation of the theory of normal backwardation and links futures

risk premiums to the net demand of producers and consumers relative to speculators.