We document that, since the Fall of 2008, the benchmark West Texas Intermediate (WTI) crude oil has periodically traded at unheard of discounts to the corresponding Brent benchmark. We further document that this discount is not reflected in spreads between Brent and other benchmarks that are directly comparable to WTI. We utilize an empirical model linking inventory conditions to the term structure slope (Fama and French, 1988) to derive empirical implications about how time and quality spreads (prompt vs. first-deferred WTI; prompt Brent vs. WTI) should move over time and be related to inventory conditions. We then utilize a non-public CFTC dataset of trader positions in WTI crude oil futures to investigate whether trading activity helps explain the residuals after accounting for crude oil market fundamentals. < style="text-align: justify;" dir="ltr"> < style="text-align: justify;" dir="ltr">Methods
We carry out econometric analyses of the following four hypotheses:
Hypothesis 1: The Brent – WTI spot (and futures) price volatilities (both) experience structural breaks in Fall 2008 and again in January 2011; so does the U.S. domestic spot spread (LLS-WTI), but not the Transatlantic spot spread (Brent-LLS).
Hypothesis 2: The Brent – WTI spread (i.e., the commodity spread) volatility experiences a structural break in Fall 2008 and again in 2010; so does the volatility of the slope of the near-dated term structure of futures prices in Cushing (i.e., the WTI time spread or CL2-CL1 slope).
Hypothesis 3: The utilization rate of Cushing storage peaks in Fall 2008 and Winter 2009; this is not true for LLS.
Hypothesis 4: To relate commodity spreads to fundamental and trading (RHS) variables, we model relationships between (i) the commodity spreads and WTI slope and (ii) fundamentals; we then (iii) compute regression residuals and (iv) investigate whether residuals can be explained by trading activity.
To deal with possible endogeneity issues in item (iv) – deviations from traditional relationships might bring about greater activity by financial traders, and vice-versa – we develop an ARDL model. We use lagged values of the variables as instruments, and run it both ways: from spreads to trading activity, and from trading activity to spreads. < style="text-align: justify;" dir="ltr"> < style="text-align: justify;" dir="ltr">Results
Chow tests of the WTI, Brent and LLS return and volatility series show that the WTI-Brent commodity and the WTI calendar spreads both experienced two structural breaks – one starting in October 2008 and the second in early 2011. In contrast, the LLS-Brent spread shows little evidence of experiencing breaks. Preliminary econometric analyses indicate that the first structural break stemmed from storage conditions in Cushing, while the second break can also be linked to demand-supply imbalances in the Brent market. An analysis of the role of futures-market speculation on the term structure of WTI prices suggests that trading activity reacts to crude-oil market fundamentals. < style="text-align: justify;" dir="ltr"> < style="text-align: justify;" dir="ltr">Conclusions
Altogether, our (thus far preliminary) findings support the notion that fundamental factors – in particular, storage constraints –contributed to recent dynamics in Brent-WTI spreads.