Why did the price of oil fall after June 2014?

Lutz Kilian

25 February 2015

a

A

What played a role in the oil price decline in 2014

After a period of relative stability, the Brent price of crude oil – commonly considered a proxy for the global price of oil – recently experienced a sustained decline that rivalled some of the most dramatic oil price declines to date. Figure 1 shows that the cumulative decline between June and December 2014 alone was 44% (or $49), and the slide of the oil price has continued into January 2015. This price drop has put severe economic stress on oil producers around the world and has called into question the sustainability of alternative forms of energy production. There is growing concern that further steep declines in the price of oil may threaten the economic and political stability of oil-producing countries, but there is also hope that lower oil prices would add much needed strength to the global economy.

Many policymakers have been pondering the question of what caused this sudden decline, the severity of which surprised even industry experts, and whether the decline is likely to continue (see, e.g., Arezki and Blanchard 2015). Although sustained declines in the price of oil have occurred before, notably in 1986 and in late 2008, a natural question is whether this oil price decline is different and, if so, how.

Some observers have conjectured that factors specific to the oil market played an important role in causing the price of oil to fall. For example, Arezki and Blanchard (2015) suggest an important contribution of positive oil supply shocks after June 2014, highlighting the examples of Libya, Iraq, and the US. They also suggest that a major shock to oil price expectations occurred when OPEC in late November 2014 announced that it would maintain current production levels despite the increase in oil production in some non-OPEC countries. Both conjectures are perfectly reasonable ex ante, yet there is new quantitative evidence in a recent discussion paper by Baumeister and Kilian (2015) that suggests that neither explanation is supported by the data.

Figure 1. The Brent price of crude oil in 2013 and 2014

Was the decline foreseeable as of June 2014?

The central question in this debate is to what extent the decline in the price of oil between June and December 2014 was foreseeable as of June 2014 and to what extent it was associated with surprises triggered by oil demand and supply shocks taking place after June 2014. Answering this question requires the use of an oil price forecasting model. One such model that seems eminently suitable for this task – as it has been shown to generate reasonably accurate oil price forecasts in previous academic studies – relies on past data for global oil production, a measure of global real economic activity, a proxy for the change in global oil inventories, and the price of crude oil. Based on this four-variable oil price forecasting model, Baumeister and Kilian (2015) show that more than half of the observed cumulative decline, namely $27, was actually predictable using only the information publicly available to anyone at the end of June 2014 (see Figure 2).

Why was this $27 decline not also predicted by more conventional oil price forecasting methods? The most commonly used approach at policy institutions is to rely on Brent futures prices as a measure of the market’s oil price forecast. The Brent futures curve as of June 2014 was largely flat. This raises the question of why the model forecast proved so much more accurate. One reason is that oil futures prices are not oil price expectations, given the time-varying risk premium in the oil futures market, and hence should not be viewed as market forecasts. Recent research has shown that this risk premium may be as large as $26, so the failure of oil futures prices to predict the oil price should not be surprising (see Baumeister and Kilian 2014). Indeed, the oil price forecasting literature has documented the comparatively poor forecast accuracy of oil futures prices at the horizons considered in Figure 2.

It may seem that the higher accuracy of the forecasts generated in Baumeister and Kilian (2015) for the second half of 2014 could simply be due to good luck. This does not seem to be the case. In particular, it is not the case that the model always predicts large oil price declines and just happened to get it right in June.

Figure 2. Real-time forecast of the Brent price of oil as of June 2014

Table 1 shows that as recently as February 2014, the model actually predicted a slight increase in the price of oil. In fact, the model forecast as of February did not differ much from the oil futures curve. Only later in 2014, the predictions turned negative and only in June 2014 did the model predict a very large decline.

Table 1. Cumulative six-month percent changes in the Brent price of crude oil

One obvious question is why the forecasting model predicted a $27 decline in the price of oil.

  • One possibility is that this forecast was driven by predictable variation in global real economic activity.

However, only about $10 of the decline predicted as of June 2014 can be attributed to a predictable slowdown in global real economic activity. This result makes a lot of sense because unexpected changes in global real economic activity affect all commodity prices, and we do not see declines nearly as large in commodity market price indices for metals, industrial raw materials, and food as we see in the price of crude oil.

  • This means that the remaining predictable decline of $17 must have been related to oil-market specific shocks taking place prior to July 2014.

There are only two types of shocks in economic models of the oil market capable of generating a price decline that is specific to the oil market (see Kilian and Murphy 2014). One is a positive oil supply shock reflecting unexpected increases in oil production, and the other is a negative shock to the demand for oil inventories reflecting expectations of higher future oil production. Thus, it is safe to say that shocks of this type occurring in some combination prior to mid-2014 must have been the reason for the additional $17 predicted decline in the price of crude oil.

Another important question is what caused the remaining decline in the price of oil that was not predictable as of June 2014. We know that this additional $22 decline must be explained by oil market shocks occurring after June 2014, but when did these shocks occur and what was the nature of these shocks? This question can be answered by updating the four-variable forecasting model month by month for the remainder of 2014, using only information that was publicly available at the time.

It can be shown that there were only two major errors in forecasting the price of oil in the second half of 2014, one in July (accounting for a decline by $9) and the other in December (accounting for an additional decline by $13). In contrast, the model’s oil price forecasts for August, September, October, and November were typically quite close to the observed price. By evaluating the pattern of forecast errors for all four model variables (oil production, oil price, oil inventories, and real activity) in July and in December one can infer which economic shocks in the oil market are likely to explain these two forecast errors.

For example, there is no evidence of large positive oil supply shocks either in July or in December because the forecast errors for oil production were an order of magnitude smaller than the reductions in global oil production associated with the major oil supply shocks discussed in the literature. The July oil price shock instead appears to reflect a negative shock to the demand for oil inventories, reflected in a large negative forecast error for oil stocks amounting to 3.4% of OECD commercial oil stocks. Such a shock to storage demand can be explained by lower market expectations of future oil prices, reflecting expectations of a weakening global economy, the anticipation of higher global oil production, or both. The model does not allow one to separate the relative contribution of these explanations, but it does show that this shock to expectations explains an additional $9 drop in the price of oil in July 2014.

In contrast, the data for December suggest a large negative shock to the demand for oil associated with an unexpected weakening of the global economy, accounting for an additional decline in the price of oil by $13. It should be emphasised that the data do not appear consistent with the alternative hypothesis articulated in Arezki and Blanchard (2015) that the OPEC announcement in late November 2014 caused an important shift in expectations about future oil production and hence about future oil prices, prompting liquidation sales of oil stocks and a fall in the December price of oil. Such an explanation of the decline in oil prices based on shocks to the demand for oil storage is difficult to reconcile with the observed pattern of forecasting errors in the model.

Concluding remarks

The fact that the forecasting model as of December 2014 predicted a small additional decline in the price of oil to about $60 in January, followed by a gradual recovery to about $70 by June 2015, suggests that further declines in the price of oil in 2015 can only be explained by additional shocks. Indeed, the fact that the Brent price fell below $50 in January is evidence that the crude oil market experienced another large shock in January. Until data on the relevant oil market indicators are released, it will be difficult to determine the nature of this shock. Obvious candidates would be a further downward revision of oil price expectations or a further unexpected reduction in global real economic activity. Nevertheless, there are reasons to expect the recent decline in oil prices to end soon, as long as the global economy does not deteriorate further. One is that further price declines would make it unprofitable for some oil producers to continue oil production. The other reason is that that there already are clear indications of reductions in oil drilling activity that in due time will cause oil production to fall.

Author's note: The views expressed in this study are those of the authors and should not be attributed to the Bank of Canada.

References

Arezki, R, and O Blanchard (2015), “The 2014 Oil Price Slump: Seven Key Questions,” VoxEU.org, 13 January.

Baumeister, C, and L Kilian (2014), “A General Approach to Recovering Market Expectations from Futures Prices with an Application to Crude Oil,” CEPR Discussion Paper 10162.

Baumeister, C and L Kilian (2015), “Understanding the Decline in the Price of Oil Since June 2014”, CEPR Discussion Paper 10404

Kilian, L, and D P Murphy (2014), “The Role of Inventories and Speculative Trading in the Global Market for Crude Oil”, Journal of Applied Econometrics, 29(3), 454-478.

a

A

Topics:  Energy

Tags:  oil prices decline, oil market, demand and supply shocks

Professor of Economics, University of Michigan; and Research Fellow at CEPR.

Events