The Oil Price Crash In 2014/15: Was There A (Negative) Financial Bubble?
Moscow School of Economics, Moscow State University; National Research University Higher School of Economics (Moscow)
June 18, 2016
Energy Policy, Forthcoming
This paper suggests that there was a negative bubble in oil prices in 2014/15, which decreased them beyond the level justified by economic fundamentals. This proposition is corroborated by two sets of bubble detection strategies: the first set consists of tests for financial bubbles, while the second set consists of the log-periodic power law (LPPL) model for negative financial bubbles. Despite the methodological differences between these detection methods, they provided the same outcome: the oil price experienced a statistically significant negative financial bubble in the last months of 2014 and at the beginning of 2015. These results also hold after several robustness checks which consider the effect of conditional heteroskedasticity, model set-ups with additional restrictions, longer data samples, tests with lower frequency data and with an alternative proxy variable to measure the fundamental value of oil.
The Oil Price Crash In 2014/15: Was There A (Negative) Financial Bubble? – Introduction
The Brent and WTI prices of crude oil fell by 60 % between June 2014 and January 2015, marking one of the quickest and largest declines in oil history. This fall in oil prices is large but it is not an unprecedented event: the oil price fell more than 30 % in a seven-month sample already five times in the last three decades (1985-1986, 1990-1991, 1997-1998, 2001, 2008). Of these five episodes, the price slide in 1985-86 has some similarities with the fall in 2014/2015, because it followed a period of strong expansion of oil supply from non-OPEC countries and Saudi-Arabia decided to increase production and to stop defending prices. Several factors have been proposed to explain this latest price crash: Arezki and Blanchard (2014) suggested an important contribution of positive oil supply shocks after June 2014. For example, there was a faster than expected recovery of Libyan oil production due to a lull in the local civil war, as it is visible from the EIA estimated historical unplanned OPEC crude oil production outages:
Moreover, Iraq oil production was not affected by the civil war enraging in the west and in the north of the country, as initially feared. The success of US shale oil production (+0.9 million b/d in 2014) and the OPEC decision in November 2014 to maintain its production level of 30 mb/d, signalling a shift in the cartel’s policy from oil price targeting to maintaining market share, put additional pressure on oil prices.
Oil demand seems to have played a minor role compared to supply shocks: Arezki and Blanchard (2014) suggested that unexpected lower demand between June and December 2014 could account for only 20 to 35 percent of the price decline, while Hamilton (2014) found that that only two-fifths of the fall in oil prices was due to weak global demand. Baumeister and Kilian (2016) used the reduced-form representation of the structural oil market model developed in Kilian and Murphy (2014) and argued that, out of a $49 fall in the Brent oil price, $11 of this decline was due to adverse demand shocks in the first half of 2014, $16 to (positive) oil supply shocks that occurred prior to July 2014, while the remaining part was due to a “shock to oil price expectations in July 2014 that lowered the demand for oil inventories and a shock to the demand for oil associated with an unexpectedly weakening economy in December 2014, which lowered the price of oil by an additional $9 and $13, respectively”.
These and other potential factors which could have influenced the oil price decline are discussed in an extensive World Bank policy research note by Baffes, Kose, Ohnsorge, and Stocker (2015). Similarly to previous works, they also found out that supply shocks roughly accounted for twice as much as demand shocks in explaining the fall in oil prices. An alternative explanation is put forward by Tokic (2015) who suggested that the 2014 oil price collapse was partially an irrational over-reaction to the falling Euro versus the dollar. This seems to be consistent with a Bank of International Settlements report (Domanski, Kearns, Lombardi, and Shin, 2015), which shows that production and consumption alone are not sufficient for a fully satisfactory explanation of the collapse in oil prices. In this regard, Domanski, Kearns, Lombardi, and Shin (2015) advanced the idea that “if financial constraints keep production levels high and result in increased hedging of future production, the addition to oil sales would magnify price declines. In the extreme, a downward-sloping supply response of increased current and future sales of oil could amplify the initial decline in the oil price and force further deleveraging”.
Given this background, we want to propose a potential explanation for the part of the oil price decline which can not be explained using supply and demand alone, particularly in the last months of 2014, as highlighted by Baumeister and Kilian (2016). More specifically, we suggest that there was a negative financial bubble which decreased oil prices beyond the level justified by economic fundamentals. A negative financial bubble is a situation where the increasing pessimism fuelled by short positions lead investors to run away from the market, which spirals downwards in a self-fulfilling process, see Yan, Woodard, and Sornette (2012) for a discussion.
We employ two approaches to corroborate this proposition: the first approach consists of tests for financial bubbles proposed by Phillips, Shi, and Yu (2015) (hereafter PSY) and Phillips and Shi (2014) (hereafter PS). These tests are based on recursive and rolling right-tailed Augmented Dickey-Fuller unit root test, wherein the null hypothesis is of a unit root and the alternative is of a mildly explosive process. They can identify periods of statistically significant explosive price behavior and date-stamp their occurrence. The second approach consists of the log-periodic power law (LPPL) model for negative financial bubbles developed by Yan, Woodard, and Sornette (2012). This model adapts the the Johansen-Ledoit-Sornette (JLS) model of rational expectation bubbles developed by Sornette, Ledoit, and Johansen (1999), Johansen, Sornette, and Ledoit (1999) and Johansen, Ledoit, and Sornette (2000) to the case of a price fall occurring during a transient negative bubble, which they interpret as an effective random down payment that rational agents accept to pay in the hope of profiting from the expected occurrence of a possible rally. Despite the methodological differences between these bubble detection methods, they provide the same result: the oil price experienced a statistically significant negative financial bubble in the last months of 2014 and at the beginning of 2015. A set of robustness checks finally show that our results also hold with different tests, different model set-ups and alternative datasets.
The paper is organized as follows: the bubble detection methods are presented in Section 2, while the data employed in the empirical analysis are discussed in Section 3. The main results are described in Section 4, while robustness checks are reported in Section 5. Conclusions and policy implications are presented in