Understanding Oil and Commodity Prices

The Money, Macro and Finance Research Group held a workshop, jointly organised with the Bank of England and the Centre for Applied Macroeconomic Analysis (ANU), at the Bank of England on 25 May 2012, entitled Understanding Oil and Commodity Prices. This report was compiled by Professor Simon Price.1

It hardly needs saying why we may be interested in this topic. Charts 1 and 2 show the recent history of broad commodity indices, the real level and volatility (12 month moving SD). Commodity prices have always been volatile and consequently have a large impact on the economy — in terms of the direct impact on inflation, on exchange rates and on supply. And to some degree commodity prices have tended to move up and down together. Some have argued that this can in part be explained by the increase in financialisation of commodity prices ie, the increase in trading of financial products based on commodity prices. Others have argued that no, it is simply that the shocks driving everything are inevitably common. The workshop managed to shed light on these views.

First though, notwithstanding what the fundamental drivers are, many players, whether policymakers or market participants, would like to be able to forecast movements in commodity prices.

Forecasting price movements

The workshop opened with a paper presented by Ron Alquist of the Reserve Bank of Canada, entitled ‘Forecasting the Price of Oil’ . This is a chapter in the forthcoming 2nd volume of the Handbook of Forecasting. There is a view that the efficient markets hypothesis implies that because all information is summarised in the spot price, oil prices are unforecastable (they follow a random walk). Ron presented convincing evidence that it is in fact possible to beat that by large margins and forecast oil prices in real and nominal terms at short horizons using various methods. Interestingly, using the futures price is not among them.

Is this contrary to efficient markets? No. The current price probably aggregates all information, including — especially, in fact — expectations about where the price will go next. Otherwise there would be arbitrage possibilities in the two markets (traders in future prices, and holders of current inventories). But that doesn’t mean that the price is necessarily expected to remain constant. Financial arbitrage implies that the futures price is affected by not only the expected future price but also by risk premia. However risk premia are unobservable and we may not know what drives them. So the futures curve alone is not necessarily informative about where the future spot price will go — you need to understand risk premia, and that’s a tough call. The physical spot market is equilibrated by the arbitrage condition that the price of a barrel today must equal the future price less the interest foregone plus the convenience yield (benefits of holding a physical stock less storage costs). And that convenience yield can change over time (for instance if inventories are low the yield is likely to be high). So although it’s widely believed as an empirical fact that you can’t beat a random walk forecast, that isn’t implied by the theory. And in fact it is now well understand that this does not follow, even for purely financial variables (even if markets are efficient, future movements can be forecastable). Likewise the fact that futures alone don’t forecast the spot better is not evidence against market efficiency. It’s now easy to see that publicly available information — for example, about inventories — can be used to forecast future oil prices. So in related work Kilian and Murphy (2010) find that global activity and inventories help forecast the real price of oil at short horizons, while Baumeister and Kilian (2011) find this would have worked in real time.

In the same forecasting session, Francesco Ravazzolo (Norges Bank) presented work in progress on ‘Oil Price Density Forecasts: Exploring the Linkages with Stock Markets’ (with Marco Lombardi, ex ECB and now BIS). Increasingly, attention among forecasters has been focusing on density rather than point forecasts. The Norges Bank is in the forefront of this. One rationale is that a particular class of point estimate is only optimal conditional on a particular loss function — eg, the mean for quadratic loss. But if we know the density then that is sufficient no matter what the loss. And in general densities communicate useful information — as the bank of England’s forecast ‘fancharts’ reveal. In this case the question arises from comovements between oil price and equity volatilities, where there is plenty of evidence of increasing and time varying correlation. A VAR turns out to produce forecast densities that are marginally superior to the unconditional densities. And moreover, that predictability could have been used profitably for an investor with power utility (parameterised at relative risk aversion equal to two) and reasonable transaction costs. The money would have been largely made between September and November 2008, in the commodity price boom.

The effect of system-wide shocks

The question then might be where those correlations come from. For the macroeconomist, the natural answer is a set of fundamental shocks hitting all sectors of the economy. This was explored by Tamarah Shakir, Bank of England in ‘The impact of oil shocks on the UK economy’ (with Stephen Millard). It is now totally uncontroversial that the impact of oil price movements on any economy differs depending on the underlying source of the shock. In particular, whether the driver is a supply, or demand, disturbance. Originally put forward as a proposition by Jim Hamilton (1983), it has subsequently been explored by several authors, most recently with Lutz Kilian to the fore. In this current paper, the empirical framework has time-varying parameters, which may be particularly relevant for the UK (given the changes in the structure of oil production and labour markets over recent decades). In line with earlier studies on larger economies, they find that that the source of the shock does indeed affect the size and nature of the eventual impact on the UK economy. Oil supply shocks typically lead to larger negative impacts on output and slightly higher increases in inflation relative to oil shocks stemming from shocks to world demand, which typically have small and frequently positive impacts on UK output. The nature of shocks in the world oil market has changed over time, with the oil price becoming more sensitive to changes in oil production, consistent with evidence from Gert Peersman (Ghent, also at this workshop) in Baumeister and Peersman (2011) that demand and supply elasticities have become smaller over time. There is also evidence that the impact of oil shocks became much smaller from the mid-1980s onwards, although the impact has risen slightly since around 2004.

Gert himself presented work in progress on ‘The US dollar exchange rate and the demand for oil’ (with Selien De Schryder). Most recent structural work on oil prices approaches the problem from an SVAR, identifying shocks. From these, it is possible to infer demand and other elasticities. But in this paper an older structural tradition is adopted, identifying elasticities from the conventional identification approach. They are able to do this because in their 65 country panel data set the price might be legitimately considered to be exogenous to the country. Macro panels are characteristically roughly square — the cross-section N and time period T the same order of magnitude — and characterised by dynamics. Under these circumstances conventional techniques designed for large N and small T are inappropriate, and indeed unnecessary. Instead, the problems arise from cross sectional heterogeneity and dependence. There are well-known solutions to these problems (eg Pesaran, Shin and Smith, 1999 and Pesaran, 2006). Remarkably, there are no published panel studies that apply these well known techniques. Apart from this, the novel aspect is that rather than use a domestic currency real price, the price drivers are the real dollar price and a world effective dollar rate. In the long run one might expect equal coefficients, but the short run elasticities examined here differ — the exchange rate elasticity is more than three times the price elasticity.

Financialisation of commodities trading

In the afternoon attention turned to aspects of financialisation, meaning the process whereby commodities and derivative products have become increasingly widely traded on financial markets. The session began with an examination of the evidence for bubbles in commodity prices: ‘The recent behaviour of commodity prices: fundamentals, speculative bubbles and relation to the global economic environment’ by Rod McCrorie, St Andrews (with Isabel Figuerola-Ferretti Garrigues and Christopher L Gilbert, Trento). A bubble here is defined as explosive behaviour. They test key commodities — crude oil, gold, silver, aluminium and copper — over the last decade or so. They use the ADF-type tests of Phillips, Wu and Yu (2011) and Phillips and Yu (2011), but modified so that critical values are made robust to allow for possibly different data spans and sampling frequencies. There is evidence of bubble behaviour in the copper, gold and silver markets in the first half of 2006. Results are less conclusive for the aluminium market, and there was no evidence for a 2007-08 crude oil bubble.

Speculative bubbles are one aspect of financialisation that may exist at medium frequencies, but Nicholas Maystre (UNCTAD) presented some evidence about very high frequency behaviour. In ‘The Synchronized and Long-Lasting Structural Change on Commodity Markets: Evidence from High Frequency Data’ (with David Bicchetti), he examined intraday co-movements between returns on several commodity markets and the stock market in the United States between 1997 and 2011. He computed various rolling correlations at 1-hour, 5-minute, 10-second and 1-second frequencies. It seems there was a clear synchronized break which starts in the course of 2008 and continues thereafter. This is consistent with the idea that recent financial innovations on commodity futures exchanges, in particular the high frequency trading activities and algorithm strategies have an impact on these correlations. So this is a call for research into the causes of this break.

These two papers were essentially presenting evidence of phenomena requiring explanations. Moving back to identifying causes, Claudio Morana (University of Milan - Bicocca) presented ‘Oil Price Dynamics, Macro-Finance Interactions and the Role of Financial Speculation.’ The issue was the extent of the role of financial speculation in determining the real oil price. He explored this using a novel framework that allowed many variables to enter a multi-country model, partly by blocks of country variables and partly through factor augmentation, and with an usually large number of shocks. As ever, identification is the key, and in this case a Cholesky structure is used, ordering by blocks of shocks and within blocks. Subject to this, the paper concludes that while macroeconomic shocks were the major upward driver of the real oil price since the mid 1980s, financial shocks also sizably contributed since the early 2000s, and to a much larger extent since the mid 2000s. The third oil price shock was a macro-financial episode: macroeconomic shocks largely accounted for the 2007-2008 oil price swing.

Ivan (Petrella, Birkbeck) also explored speculation, although in a different concept, in ‘Speculation in the Oil Market’ (with Luciana Juvenal). The paper was motivated by the fact that the run-up in oil prices since 2004 coincided with growing investment in commodity markets and increased price comovement among different commodities, and then asking whether speculation in the oil market played a role in driving this. However, in this context, following Kilian and Murphy (2011) a ‘speculative’ shock is not so much to do with financialisation as with beliefs about future demand which affect inventory holdings. From an econometric point of view, this enables identification by sign restrictions on impulse responses, including on inventories. The novelty in this paper is to augment the VAR with factors, a useful way of incorporating large amounts of information while still restricting the dimensionality of the problem. In other applications, augmenting VARs in this way is often argued to improve dynamic properties, for example by removing well-known ‘puzzles’ in VAR impulse responses. The main results were: that while global demand shocks account for the largest share of oil price fluctuations, speculation is the second most important driver; the comovement between oil prices and the prices of other commodities is explained by global demand and speculative shocks; and the increase in oil prices over the last decade is mainly driven by the strength of global demand. However, speculation played a significant role in the oil price increase between 2004 and 2008 and its subsequent collapse.

In the final paper, 'Index Funds Do Impact Agricultural Prices' Simone Pfuderer and Christopher L Gilbert (both from Trento) performed a simple exercise using Granger-causality to re-examine the data analyzed in Sanders and Irwin (2011), who concluded there was no effect from the rise in commodity index trading. They found support for Sanders and Irwin's conclusion that no impacts are discernible for the four grains markets they consider. However, Granger-causality is established in the less liquid soybean oil and livestock markets. That seems to suggest that index investment does also have price impact in liquid markets but that market efficiency prevents the detection of this impact using Granger-causality tests.


Alquist, R, L Kilian and R Vigfusson (2011) ‘Forecasting the price of oil’, prepared for G Elliott, C Granger, and A Timmerman (eds), Handbook of Economic Forecasting.

Baumeister, C, and L Kilian (2011) ‘Real-time forecasts of the real price of oil’, CEPR Discussion Paper No. 8414.

Baumeister, C and G Peersman (2011) ‘The role of time-varying price elasticities in accounting for volatility changes in the crude oil market’, Working Papers 11-28, Bank of Canada.

Hamilton, J D (1983) ‘Oil and the macroeconomy since World War II’, Journal of Political Economy, Vol. 91, pages 228-48.

Kilian, L, and D Murphy (2011) ‘The Role of Inventories and Speculative Trading in the Global Market for Crude Oil’, CEPR Discussion Paper No. DP7753.

Nixon, D and T Smith (2012) ‘What can the oil futures curve tell us about the outlook for oil prices?’ Bank of England Quarterly Bulletin, Q1, pages 39-49.

Pesaran, M H (2006) ‘Estimation and inference in large heterogeneous panels with a multifactor error structure’, Econometrica, Vol 74, pages 967-1012.

Pesaran, M H, Y Shin, and R P Smith (1999) ‘Pooled Mean Group estimation of dynamic heterogeneous panels’, Journal of the American Statistical Association, Vol 94, pages 621-34.

Phillips, P C B, Y Wu and J Yu (2011) ‘Explosive behavior in the 1990s Nasdaq: when did exuberance escalate asset values?’, International Economic Review Vol 52, pages 210-26.

Phillips, P C B and J Yu (2011) ‘Dating the timeline of financial bubbles during the subprime Crisis,’ Quantitative Economics Vol 2, pages 455-41.

Sanders, D R and S H Irwin (2011) ‘New Evidence on the impact of index funds in U.S. grain futures markets’, Canadian Journal of Agricultural Economics, Vol 59, pages 519-32.


1. Bank of England, City University and Centre for Applied Macroeconomic Analysis, Australian National University.

From issue no. 158, July 2012, pp.10-12

Page Options