The cyclical behaviour of commodities

Abstract

Commodities are known to exhibit cyclical behaviour. This paper studies the dynamics of commodities regimes and their implications for portfolio diversification. Using an extension of the regime-switching model, we find that the 12 commodities studied can be clustered into four groups with different regime dynamics, demonstrating that the asset class behaviour of commodities is far from homogeneous. The existence of two regimes is transversal to the assets studied. One regime is marked by high volatility and the other by low volatility. In both regimes, most of the commodities exhibit returns that are not statistically significantly different from those of the stock market regime. The exceptions are oil and natural gas during the low-volatility regime. The analysis of regime synchronization shows that our stock market proxy has low synchronization with commodities, which suggests potential diversification value from adding commodities to an equity portfolio. Based on portfolio optimization, we find that commodities are included in the optimal portfolios in the bull and bear regime of the Standard & Poor’s 500 index. The benefits of diversifying into commodities are particularly strong in the bear stock market regime.

Keywords:

• commodities
• cycles
• diversification
• heterogeneity
• regime-switching model

JEL classification:

• E30
• G10

Acknowledgements

The authors would like to thank an anonymous referee for his/her constructive inputs, Chris Adcock (the editor), Dieter Gramlich and seminar participants at 2014 Portuguese Finance network for their comments, which helped us improve this manuscript. An earlier version of this manuscript was presented with the title ‘The cyclical behaviour of commodities and their investment benefits’.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed 10.1080/1351847X.2016.1205505.

Notes

1. For a recent survey on the literature on commodity futures, see Skiadopoulos (2013).

2. We thank an anonymous referee for making this point.

3. These models have been used to describe the cyclical behaviour of financial and economic variables such as stock returns (examples include papers by Ang and Bekaert 2002a; Pérez-Quirós and Timmermann 2000; Turner, Startz, and Nelson 1989), interest rates (see, e.g. Ang and Bekaert 2002b; Bansal and Zhou 2002; Garcia and Perron 1996; Gray 1996; Hamilton 1988; Kanas 2008) and volatility regimes (see, e.g. Hamilton and Lin 1996; Hwang, Satchell, and Pereira 2007; Ramchand and Susmel 1998).

4. We selected the most representative commodities in terms of economic importance and also those characterized by high liquidity, which are the basis for inclusion in commodity indexes. Agriculture commodities were removed because their price is affected by seasonality, and we did not want to conflate the issues of cycles and seasonality. These commodities are also pertinent in the context of the recent super cycle. The development of BRICS in the first decade of this century fuelled the demand for raw materials such as those in our dataset and is believed to be the foundation of the recent super cycle in commodities.

5. The US dollar is the basis of our analysis because most commodities are priced in US dollars. In the supplementary appendix, we present the results of the analysis using the euro and sterling. The results remained nearly unchanged. For a more detailed analysis of the benefits of commodity diversification from the perspective of a euro investor, we refer the reader to Belousova and Dorfleitner (2012).

6. These results are different from those in Gorton and Rouwenhorst (2006), which finds positive skewness for the distribution of commodity futures returns. While Gorton and Rouwenhorst (2006) used a sample period ranging from 1959 to 2004, ours encompasses the last 20 years. Therefore, our sample period is marked by the recent super cycle in commodity markets. Moreover, our sample covers the events of the 2008 credit crunch and subsequent recession, which produced deep negative returns for a few concentrated months in the commodity markets. These two events contributed decisively to the negative skewness we see in the return distribution of most of the commodities. Similarly, Miffre (2012) found negative skewness for several commodity indices in a sample period ranging from May 2008 to April 2012, attributing it to the fall in prices in the period July 2008–February 2009.

7. A detailed table with estimated prior and posterior probabilities is presented in the supplementary appendix.

8. It is acknowledged that one of the main problems of Markowitz optimal portfolios is their instability due to high sensitivity to changes in inputs. Estimation errors in the input parameter can lead to an inferior performance of mean variance optimization (DeMiguel, Garlappi, and Uppal 2009).

Additional information

Funding

Financial support from Fundação para a Ciência e Tecnologia (Portugal) is greatly acknowledged (PTDC/EGE-GES/103223/2008).