Informed Trading in the WTI Oil Futures Market.

• Author: Rousse, Olivier
• Position: West Texas Intermediate

1. INTRODUCTION

   This paper provides evidence of suspicious trading patterns consistent with informed trading in the West Texas Intermediate (WTI) oil futures market on the days when the announcement of the U.S. crude oil stock level by the Department of Energy contrasts with the expectations of energy analysts collected in Bloomberg's inventory survey. Our results reveal significant order imbalances, with a majority of buyer-initiated trades in the two hours preceding the announcement of positive surprises--changes in inventories that are larger than expected--at 10: 30 a.m. on Wednesday. (1) The findings are robust to alternative definitions of the surprise and the measures of order imbalance considered. Our results have important implications, as the WTI futures market is the leading market with respect to price discovery (see the recent evidence in Elder et al. (2014)) and the most-traded futures commodity contract worldwide.

   The inventory level is known to be a central variable for the determination of oil prices, as highlighted in early theoretical contributions such as Deaton and Laroque (1996), Pindyck (1994, 2001), Geman and Nguyen (2005), Pirrong (2009) or more recently Kilian and Lee (2014), Kilian and Murphy (2014), Smith et al. (2015) and Knittel and Pindyck (2016). As noted in Kilian and Murphy (2014): "[...] any expectation of a shortfall of future oil supply relative to future oil demand not already captured by flow demand and flow supply shocks necessarily causes an increase in the demand for above-ground oil inventories and hence in the real price of oil." (p. 455) Inventories may also be related to precautionary demand, as highlighted by Kilian (2008, 2009) and Alquist and Kilian (2010).

   It is hence not surprising that weekly announcements about the level of oil stocks in the U.S. are eagerly anticipated by the financial community (2) Of course, changes in oil inventories are endogenous with respect to the economy. Hence, the relationship between changes in oil inventories and changes in the price of oil may not be causal. (3) Moreover, this relationship may be unstable over time (see, e.g., Kilian and Park (2009) and Fattouh et al. (2013)).

   This impact of the EIA announcement release on oil prices is illustrated in Figure 1, where we plot the transaction prices over the day of July 16th 2008. The Bloomberg median forecast for this day was a drop of 2.2 million barrels, with individual forecasts by oil experts going up to -3.9 million barrels. The actual reported value was instead a rise in stocks of 2.952 million barrels, which greatly surprised the market and led to a drop in price of almost 6 dollars in the few minutes following the release at 10: 35 a.m.

   Apart from inventory announcements, there has been substantial research on the impact of non-oil-specific news on oil prices. For example, using daily data, Kilian and Vega (2011) find no evidence of an impact of macro news on oil prices and, as such, oil prices are contemporaneously exogenous. This result has important implications for the ordering of the oil price variable in multivariate models such as vector-autoregressive models when the impact of shocks has to be investigated. Chatrath et al. (2012) confirm these findings using intraday data. Rosa (2014) and Basistha and Kurov (2015) analyze the impact of monetary surprises on oil prices using intraday data. As such, they can identify the effect of surprises at the exact time when they occur. Datta et al. (2014) provide evidence that events of various types can significantly affect on the conditional distribution of returns measured by the option-implied density. Berk and Rauch (2016) investigate the impact of Commodity Futures Trading Commision (CFTC) announcements on oil prices.

   It is well-known, however, that the largest impact on oil prices comes from the weekly inventory news release by the DOE-EIA about petroleum reserves in the U.S., which is the most anticipated piece of news in the oil market. Recent research has thus focused on the impact of this release of the stock level. (4) Bu (2014) is one attempt to assess the impact of inventory shocks on both oil prices and oil volatility. This work, however, uses daily data, a frequency at which a phenomena taking place over a few minutes might well not be visible. Halova Wolfe and Rosenman (2014) consider the impact of the oil-inventory news on both oil and gas prices and volatility. They use intraday data, but focus on the specific time of announcements, thereby ignoring the periods around the time of the release. Bjursell et al. (2015) also make use of intraday data to investigate the association between inventory shocks and price jumps, and provide evidence that these, mainly large jumps, often appear at the time of the news release. Elder et al. (2013) also use intraday data to detect intraday jumps in oil prices. The authors provide evidence that jumps often coincide with the release of either macro-news or oil-inventory news, showing the importance of fundamentals in the determination of oil prices beyond speculative motivations. (5) Finally, Ye and Karali (2016) consider the informational content of inventory shocks, looking at both EIA and API releases over the short August 2012-December 2013 time period. They find no significant, or at best a weakly significant, price impact of the API and that the market mover for oil is indeed the EIA release. In contrast to our study, the authors only focus on the impact of the news and not on potentially suspicious trading patterns before the official release time. (6) In what follows, we identify days with significant inventory surprises when the difference between the expectation (Bloomberg survey) and the realized value is "sufficiently large" on these days. (7)

   Our work takes a different approach, building on recent contributions by Irvine et al. (2007), Christophe et al. (2010), Blau and Wade (2012) and Bernile et al. (2016), where the focus is on the detection of information leakage before official announcements. In contrast with previous work such as Ederington and Lee (2002), these papers go beyond the simple analysis of return patterns to assess the possibility of leakage by taking into account order imbalance as a symptom of informed trading. We also base our analysis on order imbalance to draw conclusions about the likelihood that market participants trade on private information. To our best knowledge, this is the first time such an analysis is pursued for some commodity markets.

   Our paper is the first to focus on the possibility of information leakage before the inventory announcements, and to show that the trading patterns are consistent with informed trading. We specifically examine the trading activity around the weekly inventory announcements in the front-month WTI oil futures contract traded on the New York Mercantile Exchange (NYMEX) over the 2007-2014 period. The U.S. Department of Energy makes an announcement about the level of oil inventories each Wednesday at 10: 30 a.m. We investigate potential trading by informed investors in the hours preceding the official news release. We use intraday data to calculate order imbalances over short intervals (2 or 5 minutes) and show that there are significant order imbalances in days when the news release contains surprising inventory-level information. This pertains when the actual stock level is higher than expected (a positive surprise). The bulk of order imbalances occurs around the beginning of the open outcry trading session when liquidity is sufficiently high.

   Our results can be taken as providing preliminary evidence that the inventory level released by the DOE each Wednesday is known by some market participants who are able to benefit from their insider position to make money with the news. More generally, our results call into question the overall informational efficiency of the most liquid commodity market in the world. We should, nevertheless, be careful in interpreting these findings. While our results are consistent with the presence of informed trading, they do not explicitly demonstrate it. In particular, some traders may have superior ability in either predicting the inventory level to be released and/or in analyzing the ongoing information flow about oil supply and demand conditions. (8)

   The rest of the paper is organized as follows. The next section reviews the issue of insider trading in the specific case of commodity markets. Section 3 presents the price and inventory data and provides preliminary evidence of the diffusion of private information on the oil price before the official release time. Section 4 is then devoted to the empirical analysis, where order-imbalances are formally related to surprises. The last section concludes and discusses the implication of our findings, as well as avenues for future research.

2. INSIDER TRADING IN COMMODITY MARKETS

   Grossman (1986) discusses the potential interest in insider trading in futures markets. On the one hand, insider trading can be liquidity-enhancing but can also, on the other hand, reduce liquidity and markedly affect the viability of futures markets. Leland (1992) corroborates these effects, adding that insider trading does help to integrate information into prices. Outside investors, nevertheless, can lose significantly. John and Lang (1991), building on early work by Pettit (1972), analyze the impact of earnings announcements in the stock market. (9) More recently, Hirshleifer et al. (1994) and Brunnermeier (2005) have developed theories of the utilization of private information in trading and conclude that short-lived informational advantages can lead to significant profits for insiders.

   These high potential profits are naturally a source of concern for regulators. Indeed, "insider trading" is generally regarded as an illegal activity. However, this term covers both illegal and legal trading activities, depending on the nature of the...