International Oil Market Risk Anticipations and the Cushing Bottleneck: Option-implied Evidence.

• Author: Gagnon, Marie-Helene
• Position: Report

1. INTRODUCTION

   The integration of oil markets worldwide continues to attract significant research interest (e.g., Bachmeier and Griffin, 2006; Buyuksahin and Robe, 2011; Fattouh, 2011; Hammoudeh et al., 2008). The literature has found that the two most important indexes, WTI (US) and Brent (Europe), have not achieved complete integration due to macroeconomic, production, and transportation infrastructure frictions (Baumeister, Kilian and Zhou, 2018; Kilian, 2016; Liu, Schultz and Swerienga, 2015; Chatrath et al., 2016; Robe and Wallen, 2016). From a practitioner's point of view, it matters to what extent the US index (WTI) and the European index (Brent) are substitutes for one another, whether risk anticipations are similar in the two markets and how they are transmitted, or whether managers can hedge or diversify extreme risks (such as those linked to supply disruptions or economic slowdowns).

   This paper investigates the equilibrium dynamics in prices and risk anticipations in crude oil markets between WTI and Brent over the period 2006-2019. We introduce new metrics for crude oil markets by first computing daily time series observations of option-implied WTI and Brent volatility, skewness, and kurtosis (see Schema 1, and Birru and Figlewski, 2012, for more details). We analyze these variables pairwise using a bivariate time series system highlighting the channels of international linkages and risk spillovers.

   The model captures the international relationship between Brent and WTI risk expectations using a fractionally cointegrated VAR (FCVAR) model, which generalizes the classic cointegration framework (Johansen and Nielsen, 2010 and 2012). We use the FCVAR model because it accounts for two important features of the data, namely long memory in the cointegrating vector and in implied moments, and the equilibrium relationship between Brent and WTI, not only in prices but also in implied moments. We interpret the results from the cointegration analysis and the resulting hypothesis testing in order to assess the degree of oil market integration, as advocated by several contributions in international economics, finance (Dueker and Startz, 1998) and energy markets (e.g., De Vany and Walls, 1993; Hammoudeh, Ewing and Thompson, 2008).

   This framework allows for long- and short-run adjustments to be measured and analyzed with formal hypothesis tests, documenting comovement and spillovers between WTI and Brent (Johansen and Nielsen, 2012). First, we test whether there is an international long-run equilibrium (a cointegration relation) in investor anticipations of volatility, asymmetric risks, and tail risks. Second, we investigate whether significant WTI-Brent spreads in prices and risk metrics are supported in the long run (Dolotabadi et al., 2016). These tests document whether a difference in prices or in risk anticipations is likely to persist over time and therefore, whether meaningful differences exist between the two oil markets in terms of prices and risk expectations. The tests are helpful in detecting potential arbitrage possibilities between the two markets (Dolotabadi et al., 2016).

   Third, we perform weak exogeneity tests to assess speeds of adjustment in the model (Hammoudeh et al., 2008; Mann and Sephton, 2016). These tests have practical relevance because they highlight the adjustment process following disequilibrium between the two markets. The test results also show in which market conditions these adjustments are faster or slower, providing empirical evidence that can help oil producers, refiners, investors and traders develop more efficient risk managing strategies. Finally, we obtain Brent and WTI discovery (information) shares for prices and risk expectations (Hasbrouck, 1995; Gonzalo and Granger, 1995). This is a complementary measure to describe which index is the clear leader in the international oil market (e.g., Mann and Sephton, 2016).

   This paper therefore contributes to a better understanding of how Brent and WTI risk expectations are related, how they co-move, in which market they are formed (discovered) and transmitted, and how risk equilibria can become weakened or fragmented. The results deepen our knowledge of international oil market integration and have clear practical implications for portfolio diversification, cross-hedging, forecasting, and risk management.

   In economic terms, a crucial contribution of this paper is to focus on the impacts associated with a period of important production and transportation infrastructure changes in North America, linked to the shale oil revolution and the Cushing bottleneck (e.g., Borenstein and Kellogg, 2014; Kaminski, 2014). Indeed, until 2010, WTI prices typically traded at a premium of about two dollars over Brent until 2010 (Fattouh, 2011). The price spread subsequently inverted and reached lows of $29 under Brent. This paper documents further effects of the Cushing bottleneck and price reversal period on investor expectations of risk and on the equilibrium relationship itself.

   We document four main findings. First, long memory is a deep feature of crude oil market data, as we show persistence in the price cointegrating vector as well as in implied volatility, skewness and kurtosis. The results sharpen previous evidence of price cointegration as our new results cannot be captured by the classic cointegration model. These results suggest "limits to arbitrage" (Shleifer and Vishny, 1997), as deviations from equilibrium can be long-lasting and economically meaningful to market participants.

   Second, fractional cointegration in option-implied moments between Brent and WTI is supported, so there is at least partial integration in risk expectations. Our results underscore the importance of an international crude oil market equilibrium in investor anticipations of risk as well as in prices, and one that is considerably influenced by the persistence of the variables.

   Third, our results also document the importance of the storage constraints and economic conditions related to the Cushing bottleneck and US Midcontinent. For prices, these constraints imply slower speeds of adjustment to disequilibrium for both indexes, suggesting that oil price shocks in one market create a longer-lasting disturbance in the relationship than they did prior to the Cushing bottleneck. While there does not appear to be a long-term spread in prices, despite a price divergence linked to the Cushing bottleneck, the evidence supports a spread in implied volatility. This spread means that implied volatility is higher for WTI than Brent, and thus we expect risk management to be costlier for the US index.

   Fourth, our results suggest that oil markets are segmented in asymmetric and tail risks despite the existence of an equilibrium, and the equilibrium in higher moments is stronger at nearby than distant horizons. Investors do not anticipate extreme risks to be similar in both markets, although there is common comovement over the long run. This segmentation in risk anticipations is particularly acute during the Cushing bottleneck. Overall, asymmetric and tail risks are more locally/regionally driven and less affected by the international equilibrium than are prices and volatility. The upshot is that extreme risks are more easily diversifiable across markets precisely when diversification is more valuable for risk managers.

2. LITERATURE REVIEW

   This paper builds on and connects two strands of the literature. The first is the energy economics literature measuring the degree of market integration between different major crude oil indexes, in particular WTI and Brent (e.g., Bachmeier and Griffin, 2006; Buyuksahin et al., 2013; Fattouh, 2011). This literature investigates how international oil index prices are related by studying price differentials, cointegration (e.g., Hammoudeh et al., 2008; Mann and Sephton, 2016; Scheitrum, Carter and Revoredo-Giha, 2018), price discovery (e.g., Bachmeier and Griffin, 2006; Elder et al., 2014; Fattouh, 2010; Ghoshray and Trifonova, 2014; Liu, Schultz and Swieringa, 2015), and volatility spillovers (Arouri et al., 2011; Chang et al., 2010; Hammoudeh et al., 2003). Thus, market integration involves a common equilibrium shared by two (or more) markets reacting to external shocks, and a common stochastic trend to which all markets react and adjust.

   We contribute by investigating whether a similar linkage exists at the level of risk anticipations and higher-order moments, capturing extreme risks. We also assess whether this linkage weakened or strengthened during the Cushing bottleneck (see e.g., Borenstein and Kellogg, 2014). Moreover, our analysis accounts for an important stylized fact of volatility and higher order moment data: the presence of long memory (e.g., Baillie, 1996), whereby variables are fractionally integrated of order d with 0

   The second research strand concerns moments of option-implied distributions. The financial literature shows that option-implied moments are relevant ex ante risk metrics that relate to systematic risk and provide insightful economic interpretation (Dennis and Mayhew, 2002; Duan and Wei, 2009). For example, option-implied volatility (e.g., the VIX index in stock markets) is seen as an investor "fear gauge" (Whaley, 2000; Malz, 2013; Panigirtzoglou and Skiadopoulos, 2004). Option-implied skewness is linked to negative asymmetric returns in asset markets, and negative skewness in equity indexes is often interpreted as "crash-phobia" (Rubinstein, 1994). Intuitively, Brunnermeier et al. (2009) interpret implied skewness as the cost of insuring against crash risk (asymmetric risk in the case of crude oil). While positive skewness is rare in equity index option data, it is not unusual in the crude oil setting. Both positive and negative implied skewness suggest risk. Indeed, positive skewness could highlight the risk of price spikes due to supply disruptions or higher than expected demand. This is why we...