Commodity Futures Returns: What Drives The Raw Material And Energy Markets?
TU Dortmund University
Michael Mauboussin: Here’s what active managers can do
University of Dortmund – Faculty of Business, Economics and Social Sciences
December 1, 2015
Applying a Generalized Dynamic Factor model we identify one latent common factor in a sample of 32 daily commodity futures returns. This factor is driven by five market shocks and accounts for 12% of total return variation. We find evidence for the cross section of individual commodity futures returns to be significantly exposed to a latent energy specific component. Our findings shed new light on the question of how commodities are related and supports the importance of the energy sector for the other commodities. In light of our findings we suggest reconsidering the heterogeneity hypothesis of the commodity markets.
Commodity Futures Returns: What Drives The Raw Material And Energy Markets? – Introduction
Since the 2000s, booms and downturns in commodity prices illustrate the tendency of commodities to co-move, also across different groups of commodities. The development of commodity markets in general is of vital importance not only on academic grounds but also from an economic and political point of view. Especially energy commodities, as they are an important input parameter for most other commodities, stand at the heart of the commodity markets. Hence, the question which arises is how different commodity markets are related and if and how they move together.
In this paper we extract latent common factors for a sample of 32 commodities traded on futures markets using a Generalized Dynamic Factor model (GDFM, hereafter) proposed by Forni et al. (2000). The GDFM allows, in contrast to other factor models, for serial correlation and idiosyncratic cross-sectional dependencies in the data, cf. Barigozzi & Hallin (2014). These elements are crucial in modelling commodity returns and the methodology enables us to extract market shocks which drive the common components of the commodity return series. We use this common factor to explain the cross-sectional relation between individual commodities.
Next we investigate if the common components can explain the cross-sectional behavior for different periods of time, namely during the period before and after the global financial crisis. During that time commodity markets’ correlation to other markets changed dramatically. As the current literature is still undecided on asset pricing models able to explain the cross-section of individual commodity futures returns, see Daskalaki et al. (2014), we contribute towards the development of a working asset pricing model for commodities.
Our findings and contribution to the literature is threefold. First, we show that a common component for a sample of 32 commodity futures returns accounts for 12% of total return variation. We find this common component to be driven by five latent common shocks. This finding sheds light on the discussion of how many factors drive the commodity markets. As this common component fails to explain the cross-sectional returns of our set of individual commodities we continue and identify common components for the subgroups of agriculture, metals, precious metals, energy, and livestock. The explained variation of the common component for these subgroups varies between 24% and 56% of their total return variation. Third, for the period after the financial crisis we find a common energy component which explains the cross-section of individual commodity futures returns in a three factor model. This indicates an increased importance of the energy sector for all other commodities and opens up new vista on the relation between different groups of commodities.
Applying factor models to understand commodity markets was done as early as Pindyck & Rotemberg (1988) who shows prices of unrelated raw commodities to have a persistent tendency to move together, even in excess to macroeconomic variables such as inflation, industrial production, interest rates, and exchange rates. This finding was both rejected and confirmed in following investigations, e.g. Deb et al. (1996) or Karstanje et al. (2013) who examine co-movement of factors driving commodity futures curves in price levels and in futures curve shapes. Based on the dynamic Nelson-Siegel model they conclude individual futures curves to be driven by common components, whereas the commonality mostly is sector specific. Byrne et al. (2013) and Vansteenkiste (2009) extract common unobserved factors from individual non-fuel commodity prices using principal component techniques. Vansteenkiste (2009) finds various periods of changing co-movement. She suggests that supply, global demand, exchange rate and real interest rate are important factors when describing the co-movement. This is in line with Frankel (2006), Calvo (2008), and Wolf (2008). These authors identify real interest rates, excess liquidity, and shifts in global supply and demand driving commodity prices. Based on returns Christoffersen et al. (2014) find evidence of a factor structure in daily commodity futures returns and volatilities. Comparing commodity and equity markets they conclude that commodity market returns are again segmented from equity markets since 2010, whereas commodity volatility shows a nontrivial degree of market integration.
See full PDF below.