https://orcid.org/0000-0002-7631-0238
![Sample our Area Studies journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5922/TOC-Subject-Sample-2021-Area-Studies.jpg"})
ABSTRACT
We analyze the co-movement of a number of commodity markets in extreme financial episodes worldwide. More specifically, we provide extreme downside risk co-movement maps of these markets during six recent distress periods. We follow an expected shortfall-multidimensional scaling approach, which allows for an easy classification of markets according to their dynamics in risky episodes. No clear risk co-movement patterns are observed, nor spillover effects are detected. Financialization and speculation might have played some role in the dynamics of price and risk only in food commodity markets during the oil price increase 2007–2008.
Acknowledgements
We would like to thank the two anonymous referees and the Editor, for their comments and suggestions. We are also thankful to Jean-Francois Racicot and Khaoula Ghaiti (both from University of Ottawa), José Morales-Diaz (Ernst and Young) and Pablo Cousteau (Options and Futures Institute) for their help in collecting daily data, and Victórico Rubio, Deputy Director for Investments, Strategy and Asset Allocation of CAIXABANK, S.A., for his valuable help in the revision of the manuscript.
Disclosure statement
No potential conflict of interest was reported by the author(s).
ORCID
Lidia Sanchis-Marco http://orcid.org/0000-0002-7631-0238
Notes
1 Other alternatives are VaR and expectiles, but in subsection 2.1 it is shown that ES is preferred to both of these.
2 Time-series-based clustering is a more complicated multivariate alternative frequently used in financial studies.
3 It is of note that the EDR co-movement measure proposed is different from the contagion risk measures used in the literature on systemic risk to assess the contagion risk from commodity markets towards the economy as a whole and across sectors (delta conditional VaR, systemic risk index, Shapley value, systemic ES and distressed insurance premium, among others; see Bisias et al. Citation2012; and Billio et al. Citation2012, for a comprehensive survey). These measures aim to evaluate the contributions of individual commodity markets to contagion risk.
4 It only verifies subadditivity for elliptic distributions.
5 Note that if we take α = 0.5, AWLS actually reduces to ordinary least squares.
6 For risk management purposes, one might use a fat-tailed distribution not only in the volatility models but also in the ES framework itself.
7 We tested the performance of the GJR-GARCH specification for different p, r and q orders (see equation 5 and the text below for the interpretation of such orders), and the most parsimonious model provided the best results.
8 , where
represents excess kurtosis.
9 A model is adapted if everything required to model the mean at time t is known at time t −1. Standard examples of adapted mean processes include a constant mean or anything in the family of ARMA processes or any exogenous regressors known at time t – 1.
10 This set of restrictions is difficult to describe for a complete GJR-GARCH (p,r,q).
11 Liu and Hung (Citation2010) show strong evidence that modeling the asymmetric component is much more important than specifying the error distribution for improving the volatility forecasts of financial returns in the presence of fat-tails, leptokurtosis, skewness and the leverage effect. Furthermore, if asymmetric properties are neglected, the GARCH model with Gaussian distribution is preferable to those models with more sophisticated error distributions, suggesting that allowing for a flexible error distribution does not lead to significant improvements in volatility forecasts.
12 Under conditionally normal innovations, GJR-GARCH (1,1,1) is covariance stationary if, in addition, .
13 The ½ multiplying the leverage coefficient comes from the assumption of symmetric conditional distributions for the returns.
14 The estimates of the set of GJR-GARCH parameters for the commodity markets under study are listed in Appendix B.
15 Distances between points do not vary when variables are expressed in differences with regard to their means.
16 Notwithstanding, some researchers such as Munier (Citation2010) state that the financialization of commodities results in spillover effects from future to spot prices (he calls them the impact of informational externalities on actual spot prices). We refer readers to this paper and also to the interesting discussion in Bernard, Greiner, and Semmler (Citation2012).
17 In addition, monthly data have been proved not to satisfactorily capture the dynamic causal linkages between commodity prices (see Zhang and Reed Citation2008, for example). For this reason, Nazlioglu (Citation2011) uses weekly data (this data frequency has been very rarely used in the literature on the topic) to capture the dynamic causal linkages between oil and agricultural commodity prices.
18 Section 4 and Appendix I include details and literature on financialization, price behavior and co-movement in commodity markets.
19 In this context, spillover effect refers to the impact that seemingly unrelated events in one commodity market can have on other commodity markets. This definition follows that of the pioneering article by Masson (Citation1998).
20 For the sake of simplicity and ease of reading, we use absolute values for ES. Average, standard deviation, CV and other descriptives of interest of ES for each commodity market are listed in Appendix F.
21 The correlation matrices in values are shown in Appendix H.
22 See Appendix F. Notwithstanding, when interpreting the coefficient of kurtosis, it must be taken into account that most of the ES distributions are highly negatively skewed, which means a strong departure from normality.
23 According to Hong and Yogo (Citation2010), the levels of financial activity measured by open interest in commodity futures increased from $103 billion at the end of 2003 to $509 billion in July 2008, and the total value of commodity index-related instruments purchased by institutional investors rose from about $15 billion to $200 billion during the same period.
24 Which was preceded by a slowing of global growth in the two first quarters of 2008.
25 Note that the debate has focused on prices rather than in EDR.
26 Our results only support a weak degree of co-movement in the oil price increase 2007–2008 and during the global oil and food crisis 2008–2009, especially in the food markets.