TX_1~AT/TX_2~AT International Journal of Energy Economics and Policy | Vol 6 • Issue 2 • 2016152 International Journal of Energy Economics and Policy ISSN: 2146-4553 available at http: www.econjournals.com International Journal of Energy Economics and Policy, 2016, 6(2), 152-158. Indexing Oil from a Financial Point of View: A Comparison between Brent and West Texas Intermediate Cem Berk* Department of Accounting Information Systems, School of Applied Sciences, Istanbul Arel University, Turkey. *Email: cemberk@arel.edu.tr ABSTRACT Brent crude and West Texas intermediate (WTI) are major indices for purchases of oil worldwide among with some others such as OPEC basket. Brent is traditionally a European index whereas WTI representing slightly sweeter and lighter crude is more applicable in USA. Until 2010, the spread between WTI and Brent hasn’t been more than few dollars. However in recent years, the spread is widening in favor of Brent and then returning to the mean. WTI which historically taken over Brent, has fallen below Brent which is now claimed to be the global oil index for the World. This is sometimes argued with the Shale production and over-supply in the U.S. and several macroeconomic events such as Libyan crisis. The aim of this paper is to analyze which of these indices is a better indicator for the energy industry. The variables from NYSE exchange traded funds namely energy select sector SPDR ETF (XLE), Teucrium WTI crude oil ETF (CRUD), and United States Brent oil ETF (BNO) for the period December 1994 and September 2014. The variables are analyzed for long-run and short-run relationships with unit root tests, vector autoregression models, and vector error correction models as well as cointegration and Granger causality tests. Keywords: Energy Modeling, Oil Indexing, Cointegration, Granger Causality JEL Classifications: C58, P48, Q37 1. INTRODUCTION For most trades and especially commodities, certain categorizations are required to see the quality of the goods. For oil trade, this is done through indices such as Brent, West Texas Intermediate (WTI), Dubai, Urals, Isthmus, LLS and OPEC. All of these oil have different characteristics, qualities, and market penetration and therefore have different prices. OPEC is a basket composed of Arab, Basrah, Bonny, Es Sider, Girassol, Iran, Kuwait, Marine, Merey, Murban, Oriente, and Saharan oil. The number of global indices used are over 150. These indices are used while pricing oil, so they have importance for international oil trade. Other crude oils are priced against major indices such as Brent, WTI, and Dubai. Technically WTI is the best quality oil among these. But this is just a slight difference in quality, which means WTI should trade a few U.S. Dollars premium to Brent. This is a light weight and low sulphur oil. This means when refined it could generate more gasoline. This is traditionally an American oil, however its production is decreasing. Brent represents a European index, and often characterized by the North Sea. The oil is in very different locations. The oil is still known to be light and sweet, however WTI is lighter and sweeter. So we know from law of one price that the price differential should be equal and otherwise arbitrage opportunities arise. This is true however, it is the supply and demand conditions, and the location differences as well as political risks (as in the case of Libyan crisis and many others) that could create spread between these two indices. Historically, Brent and WTI have traded very close to each other, spread almost mean reverted to zero level until 2010. There are many reasons, but to tell the result WTI has lost value against Brent, and nowadays recovered a bit. The most important considerations are supply related and geostrategic. U.S. also started to switch alternative and modern ways of using energy, Berk: Indexing Oil from a Financial Point of View: A Comparison between Brent and WTI International Journal of Energy Economics and Policy | Vol 6 • Issue 2 • 2016 153 such as Shale Gas. When WTI loses value, people producing and trading based on WTI lose money. This spread is very important for international trade, which is the research topic of this paper. In this paper, it is investigated whether any of these indices have explanatory power on energy industry. The remainder of this paper is organized as follows. In Section 2, some of the recent and important works in this research area are presented. Then the methodology and research model is given in Section 3. The information on data, as well as research results are available in Section 4. In Section 5, some of the important findings of the study are discussed. In Section 6, policy and financial implications are discussed. 2. LITERATURE REVIEW Liao et al., performed a unit root with structural breaks to test whether international crude oil markets are globalized or regionalized. Unit root is detected for lower quantiles however mean reversion is detected for upper quantiles. With Kolmogov- Sminov methodology it is proven that the price differential is mean reverting and thus globalization view supported. Oil traded in USA is more commonly used in WTI whereas out of USA Brent is used. WTI is also a higher quality with larger quality and less sulphur. It is argued that until 2010 WTI is traded with a premium and after 2010 there is a structural break such that WTI crude oil is traded at discount compared to Brent. Due to the non-normality and structural breaks, the method in Koenker and Xiao, and Enders and Lee, is used instead of conventional techniques. The spreads show unit root in the lower quantiles but mean reversion in the upper counterparts. The quantile Kolmogorov–Smirnov test statistic over the whole range rejects the null hypothesis of unit root which means that the differentials are globally stationary and supports globalization hypothesis (Liao et al., 2014). Creti et al., studies the relationship between oil price and stock market in oil importing and oil exporting countries. The long- run relationship with Engle-Granger causality are studied for this purpose. The short run co-spectral analysis of Priestley and Tong (1973) is also studied. The research period is 2000-2010. Brent oil is chosen as the oil index for this study. The relationship between oil index and stock market is found as a medium-term phenomenon. The relationship is more recognizable for oil exporting countries where oil shocks move together with stock market (Creti et al., 2014). Huang and Chao studies international and domestic oil prices and indices in Taiwan. The results are interesting; domestic oil prices don’t Granger cause international indices. Threshold vector error correction model (VECM) and threshold autoregression is used for this purpose. Brent crude oil is chosen to represent international oil index. Another conclusion is the mean reversion is faster when a small shock occurs than a big shock. Government intervention to the oil market is ineffective (Huang and Chao, 2012). Arouri analyzes the respond of European stock movements to oil changes. The power of this relationship varies according to the industry. The markets are analyzed between 1998 and 2010. Brent oil is used as an indicator for oil index. Zivot–Andrews is used for testing unit root. The study is a multifactor analysis including return of stocks, industry, oil, and a dummy variable to include whether there is a crisis. Furthermore Granger causality is short term variable. It is found that there is a relationship between oil price changes and stock markets. For the automotive industry there is a clear negative correlation between industry returns and oil. But the relationship is not such strong in other industries (Arouri, 2011). Wang et al., analyze oil price shocks with stock market activities. As expected the results state that there are different effects on oil exporting and oil importing countries. Also the dependence on oil, increases the negative effects on stock market in case of a price increase in oil. WTI is chosen as the benchmark oil for this study. Granger causality and vector autoregression (VAR) is used in this study. The results show that oil price shock explain 20-30% of global stock return variations (Wang et al., 2013). Lee et al., study stock market returns in G7 countries. The research period is 1999-2009. The research is interesting since it focuses on developed countries. Oil price changes don’t significantly affect the stock markets, however stock price changes lead oil prices. VAR, vector error correction and Granger causality are used in this study (Lee et al., 2012). Basher et al., study the relationship between oil price changes, exchange rates and emerging market stocks. The research method is structural VAR. Positive shocks of oil prices depress emerging market stock prices and US Dollar exchange rates. Most of this dynamic movements take place in the short run. Oil importers’ currency depreciate, whereas oil exporters’ currency appreciate in case of an increase in oil price (Basher et al., 2011). Tao et al., explain indexing in shale oil for industrial purposes for Bogda Mountain oil shale in China. The oil is classified according to petrological type, organic component content, hydrocarbon generating potential. The findings show that lithologic types and industrial classification of oil shales can be classified as follows: The content of organic component lower than 5%, between 5% and 15%, between 15% and 25%, and over 25% correspond to low-quality, medium-quality, and high-quality oil shale (Tao et al., 2010). Buyuksahin et al., has shown that starting from Fall of 2008, the benchmark WTI crude oil has traded at discount to Brent benchmark. However the same discount isn’t reflected to other oil indices. This spread is detected on oil futures positions when controlled macroeconomic and physical market fundamentals. WTI is historically a more reliable benchmark for U.S.A, where Brent is a European benchmark. The spread is also analyzed for several components both for WTI and Brent; such as WTI and Louisiana Light Sweet, Louisiana Light Sweet and Brent, and Brent for international oil and Brent. The macroeconomic events are considered in the analysis namely Libyan crisis and Arab Spring. The research period is between 2000 and 2012. There is clear evidence that WTI crude oil traded at discount compared to Brent. (Buyuksahin et al., 2013). Berk: Indexing Oil from a Financial Point of View: A Comparison between Brent and WTI International Journal of Energy Economics and Policy | Vol 6 • Issue 2 • 2016154 Kasibhatla studied whether there is a causal relationship between crude oil and U.S. dollar. The relationship is studied empirically with co-integration and error correction modeling. The study reveals that there is Granger causality from U.S. dollar to crude oil price. Over the past 15 years there wasn’t a stable correlation between S and P and crude oil ranging from plus or minus 20%. The data used in the study is U.S. Dollar index (usdx) and crude oil prices (coil) for the period January 1990-May 2010. The series are stationary with their first differences (I(1)) according to augmented Dickey-Fuller (ADF) and Kwiatkowski–Phillips– Schmidt–Shin. The series are then tested with trace test and maximum eigenvalue cointegration where one vector is found which is an indicator of long-run relationship. There is some doubt on short term relationship; however there is a tendency to restore equilibrium following a shock to the system. There is also proven causality, U.S. dollar index Granger causes the crude oil price (Kasibhatla, 2011). Gammara et al., study the Granger causality between the price of oil and integrated Latin American market index. The framework proposed by Hatemi (2012) is used as methodology. The result shows no significant causality. The authors further argue from the law of one price that there is no arbitrage opportunity between oil and index (Gamarra et al., 2015). Lee et al., study the relationship between stock prices and WTI oil index for the period January 1998 and March 2012. GARCH methodology is used for G7 countries’ stock market performance and WTI oil index. According to the results Canada has the highest hedge effectiveness and Japan has the lowest. Because of low correlation between the stock market index of Japan and the oil price, the optimal portfolio weight of Japan is higher (Lee, 2014). Wei and Chen examine the relationship between WTI oil spot returns and the S&P 500 energy index. Daily data is used for the period January 2000 and September 2009. Multivariate GARCH methodology is used in this paper. The result shows that WTI is significantly affected by energy index returns. Investors can also use energy index returns’ past volatility as the basis for WTI oil price forecasting (Wei and Chen, 2014). 3. RESEARCH MODEL In the study the variables are checked to see whether they are stationary. This is done first with ADF methodology. If any of the roots of the polynomial (1- ∂1L- ∂2L 2-…- ∂pL p) of an AR (p) stochastic process lie outside the unit circle, the process is said to non-stationary. The traditional ADF way of testing for non- stationarity of an AR (p) process involves testing for the null of one unit root in: ∆ = + ∆ + + +− − − − ∑y y y t ut t j t j t j p γ φ α β* 1 1 1 The stationary characteristics of the variables are tested also with Phillip-Perron (PP) methodology. PP test is a non-parametric modification to the standard Dickey-Fuller t-statistic to account for the autocorrelation that may be present if the underlying DGP is not AR (1). Instead of adding AR terms in the DGP to account for (possible) MA terms, they modify the test statistic. However, Schwert (1989) showed that PP test suffers from poor size properties if the MA term is large negative. Thus, ADF and PP tests suffer from quite opposite problems. While the ADF test does not suffer from as severe size distortions, it is not as powerful as the PP test. The other “problem” with the PP test is that of consistent estimation of the so called long-run variance or the variance of the sum of the errors: (Virmani, 2001). σ ε2 1 2 2 1 = − − ∑p T E j j T lim [( )] Since there are differenced variables the variables are tested for cointegration according to Johansen procedure. If the coefficient matrix Π has reduced rank r