Volatility Spillovers Across Petroleum Markets.

• Author: Barunik, Jozef
• Position: Report - Statistical data

1. INTRODUCTION, MOTIVATION, AND RELEVANT LITERATURE

   Research on the interdependencies observed on financial and commodities markets has led to analyzing not only returns and volatility, but also their spillovers (Dimpfl and Jung, 2012). The global financial and economic crisis, sharp fluctuations in commodity prices, the rapid financialization of petroleum commodities (1) and tight oil production from shale formations prompted a fresh surge of interest in how the dynamic links among commodities work (the relevant literature is shown presently). In this paper, we focus on petroleum commodities and analyze volatility spillovers across petroleum markets. In doing so we differentiate between spillovers due to negative and positive returns (negative and positive spillovers) as the asymmetry has been proven to play an important role in many economic and financial issues related to our analysis (Ramos and Veiga, 2013; Du et al., 2011; Nazlioglu et al., 2013; Bermingham and O'Brien, 2011).

   Why do we care about volatility spillovers, and what are the implications for investors, regulators, and facility operators? Since volatility serves as a proxy measure of risk, substantial changes in volatility and its spillovers across markets are able to negatively impact risk-averse investors. Hence, knowledge of volatility spillover dynamics has important implications for investors and financial institutions in terms of portfolio construction and risk management as these spillovers and their direction may greatly affect portfolio diversification and insurance against risk

   (Gorton and Rouwenhorst, 2005). Analyzing volatility spillovers also has important implications for the development of accurate asset pricing models, hedging strategies, and the forecasting of future equity and the volatility of oil price returns (Malik and Hammoudeh, 2007). Besides being used in risk management for a long time, volatility has recently become even more important as it is now directly tradable using swaps and futures (Patton and Sheppard, 2014). Further, volatility spillovers are closely associated with market co-movements and this phenomenon becomes quite pronounced during crisis events when, usually, financial market volatility sharply increases and spills across markets (Reinhart and Rogoff, 2008). Analyzing and measuring volatility spillovers enables providing "early warning systems" for dormant crises and to map the development of existing crises (Diebold and Yilmaz, 2012). Knowledge of volatility spillovers then becomes a segment of information useful for regulators, operators, and policy makers that may lead to the introduction of regulatory and institutional rules to reduce the cross-market impact of excessive price movements.

   Petroleum-based commodities form an asset class where spillovers historically play a prominent role (Haigh and Holt, 2002), given the importance of these commodities for the economy and economic development (Hamilton, 1983) and the fact that shocks' transmission into oil prices significantly affects the U.S. and the global economy (Kilian, 2008; Hamilton, 1996; Gronwald, 2012). However, the research on volatility spillovers among petroleum commodities is rather limited and the asymmetric aspect of spillovers is not adequately explored yet. In our paper we make two key contributions. First, we use high-frequency data to extend the literature on volatility spillovers among key petroleum commodities: crude oil, heating oil, and gasoline. Second, by augmenting the current methodology of Diebold and Yilmaz (2009, 2012), we are able to quantify negative and positive asymmetries in spillovers, including the directions and magnitudes over time. Among other results, we rigorously show that negative volatility spillovers are larger than positive spillovers across petroleum-based commodities. Such negative asymmetry is most visible before 2008 while later asymmetries in spillovers considerably decline.

   Petroleum-based commodities are essential to our economies primarily from an industrial perspective. (2) Accordingly, crude oil prices are driven by distinct demand and supply shocks (Kilian, 2008; Hamilton, 2009; Lombardi and Van Robays, 2011). Further, Kilian (2009) shows that shifts in the price of oil are driven to different extents by aggregate or precautionary demand related to market anxieties about the availability of future oil supplies. Kilian and Vega (2011) support this finding by showing that energy prices do not respond instantaneously to macroeconomic news but Mason and Charles (2013) argue that the spot price of crude oil and its futures prices do contain jumps. Finally, Sari et al. (2011) argue that global risk perceptions have a significantly suppressing effect on oil prices in the long run.

   Besides the above forces, oil prices might also be linked to large speculative trades (Hamilton, 2009; Caballero et al., 2008) and short run destabilization in oil prices may be caused by financial investors (Lombardi and Van Robays, 2011). These findings are in line with petroleum's increasing financialization after 2001 as shown in Fratzscher et al. (2013) and the expanding financialization of commodities in general (Mensi et al., 2013; Creti et al., 2013; Dwyer et al., 2011; Vivian and Wohar, 2012).

   Due to their real economic importance and their ongoing financialization, petroleum-based commodities are naturally sensitive to economic development as well as market volatility. The evidence in Vacha and Barunik (2012) indicates that during periods of recession there exists a much higher downside risk to a portfolio formed from oil-based energy commodities. The asymmetric risk and accompanying volatility spillovers are thus a feature one would like to measure and monitor effectively. The research related to volatility spillovers among energy commodities is surprisingly limited, though. On weekly data, Haigh and Holt (2002) analyze the effectiveness of crude oil, heating oil, and unleaded gasoline futures in reducing price volatility for an energy trader: uncertainty is reduced significantly when volatility spillovers are considered in the hedging strategy. Using daily data for the period 1986-2001, Hammoudeh et al. (2003) analyzed the volatility spillovers of three major oil commodities (West Texas Intermediate, heating oil, and gasoline) along with the impact of different trading centers. Spillovers among various trading centers were also analyzed by Awartani and Maghyereh (2012), who investigated the dynamics of the return and volatility spillovers between oil and equities in the Gulf region. The spillover effect between the two major markets for crude oil (NYMEX and London's International Petroleum Exchange) has been studied by Lin and Tamvakis (2001), who found substantial spillover effects when both markets are trading simultaneously. More recently, Chang et al. (2010) have found volatility spillovers and asymmetric effects across four major oil markets: West Texas Intermediate (USA), Brent (North Sea), Dubai/Oman (Middle East), and Tapis (Asia-Pacific).

   It is not surprising that different classes of petroleum commodities are affected by similar shocks given their potential substitution effect (Chevallier and Ielpo, 2013) or economic linkages (Casassus et al., 2013). However, the spillovers might evolve differently depending on the qualitative nature of the shocks. In terms of volatility spillovers, it is of key importance to identify how negative or positive shocks transmit to other assets. Changes in the volatility of one commodity are likely to trigger reactions in other commodities. We hypothesize that such volatility spillovers might exhibit substantial asymmetries and we aim to quantify them precisely.

   Much of the research studying volatility spillovers among markets have employed multi-variate GARCH family models, VEC models, etc. However, these methods have interpretative limitations as, most importantly, they are not able to quantify spillovers in sufficient detail. In our analysis we utilize more efficient techniques. Recently, Diebold and Yilmaz (2009) introduced a methodology for the computation of a spillover index (the DY index) based on forecast error variance decomposition from vector autoregresssions (VARs). (3) The methodology was further improved in Diebold and Yilmaz (2012) who introduced spillover direction and variable ordering in VARs. Another improvement of the original DY index has been introduced by Klo[beta]ner and Wagner (2014), who developed a new algorithm for the fast calculation of the index along with the computation of the minimum and maximum values of the index. Finally, based on the idea of realized semivariance due to Barndorff-Nielsen et al. (2010), Barunik et al. (2013) extended the information content of the DY index with the ability to capture asymmetries in spillovers that materialize due to negative and positive returns/shocks--negative and positive spillovers. We employ this methodology for our analysis.

   Our contribution is centered on finding substantial asymmetries in volatility spillovers across petroleum commodities, but our results are much richer. During the 1987-2014 period we document considerable volatility spillovers among petroleum commodities that substantially change their character after 2008: an increase in the magnitude of spillovers but a decline in their asymmetries. The increase in volatility spillovers seems to correlate with two important factors. First, the progressive financialization of the commodities has occurred since the beginning of the 21st century and the 2008 financial crisis deeply affected financial markets; the observed correlation resonates well with the findings of Tang and Xiong (2012), Creti et al. (2013), or Mensi et al. (2013). Second, the year 2008 is fundamentally important because it brought about much more activity in tight oil exploration and an increase in U.S. oil production that later resulted in a supply...