Elsevier

Energy Economics

Volume 32, Issue 2, March 2010, Pages 363-372
Energy Economics

Oil price dynamics and speculation: A multivariate financial approach

https://doi.org/10.1016/j.eneco.2009.08.014Get rights and content

Abstract

This paper assesses empirically whether speculation affects oil price dynamics. The growing presence of financial operators in the oil markets has led to the diffusion of trading techniques based on extrapolative expectations. Strategies of this kind foster feedback trading that may cause considerable departures of prices from their fundamental values. We investigate this hypothesis using a modified CAPM following Shiller (1984) and Sentana and Wadhwani (1992). First, a univariate GARCH(1,1)-M is estimated assuming the risk premium to be a function of the conditional oil price volatility. The single factor model, however, is outperformed by the multifactor ICAPM (Merton, 1973), which takes into account a larger investment opportunity set. Analysis is then carried out using a trivariate CCC GARCH-M model with complex nonlinear conditional mean equations where oil price dynamics are associated with both stock market and exchange rate behavior. We find strong evidence that oil price shifts are negatively related to stock price and exchange rate changes and that a complex web of time-varying first and second order conditional moment interactions affects both the CAPM and feedback trading components of the model. Despite the difficulties, we identify a significant role played by speculation in the oil market, which is consistent with the observed large daily upward and downward shifts in prices — a clear evidence that it is not a fundamental-driven market. Thus, from a policy point of view – given the impact of volatile oil prices on global inflation and growth – actions that monitor speculative activities on commodity markets more effectively are to be welcomed.

Introduction

Investment funds have recently poured profusions of money into the commodity markets and raised their holdings to $260 billions as of mid-2008 from $13 billions in 2003. During that period the price of crude oil, among other commodities, rose relentlessly, prompting debate on the role of speculation in oil prices.1

For many a decade commercial operators only were allowed to buy nearly unlimited amounts of oil. In 1991, however, the Commodity Futures Trading Commission (CFTC) granted a similar faculty also to financial firms.

The empirical evidence on the relevance of speculation in the oil market is not clear-cut. At the end of July 2008 a CFTC report concluded that speculators were not systematically driving oil prices.2 A few days later, however, data revision showed that just four swap dealers held 49% of all the NYMEX oil contracts that bet on oil price increases, providing clear evidence of concentration of power in the market. Indeed, it can prove difficult to distinguish between pure speculation and commercial trading whenever the latter leads to excess hedging of risk of adverse price shifts. The aim of this paper is to assess empirically whether speculation does affect oil price dynamics. While direct study is hampered by the lack of reliable data on speculative positions in the oil futures markets, the role of speculators can still be analyzed indirectly, with the help of heterogeneous agent models, based on the interaction between two stylized types of traders, viz. fundamentalists and noise/feedback/chartists. Oil supply is relatively inelastic and its price is mostly influenced by (excess) demand shifts that stem from the two categories of traders mentioned above.

The growing presence of financial operators in the oil markets has led to the diffusion of trading techniques based on extrapolative expectations, where a price trend is assumed to be lasting. Strategies of this kind tend to foster feedback trading: “positive” whenever investors buy when prices rise and sell when they fall, and “negative” if investors buy when prices fall and sell if they rise. The literature has mostly focused on positive feedback trading, seen as an irrational strategy that moves prices away from their fundamental-related values, raises uncertainty and contributes to market fragility. Its presence is typically associated with a negative autocorrelation of returns. Feedback trading seems to be a stylized aspect of stock market behavior. Cutler et al., 1991, Sentana and Wadhwani, 1992 find evidence of feedback trading in the US stock market whilst Koutmos, 1997, Koutmos and Saidi, 2001 detect its presence in, respectively, several European and emerging equity markets. Finally, the impact of feedback trading by specific groups of operators – such as foreign or institutional investors – is examined in Lakonishok et al., 1992, Hyuk et al., 1999, Nofsinger and Sias, 1999, among many others.

We first investigate the hypothesis that there are also some participants in the crude oil market who engage in feedback trading activities, using a behavioral CAPM that follows Shiller, 1984, Sentana and Wadhwani, 1992. We use a univariate GARCH(1,1) setting where the risk premium is a function of the conditional oil price volatility.

The single factor model, despite its attractiveness, misses some relevant aspects of financial market pricing and is outperformed by the multifactor ICAPM, which takes into account a larger investment opportunity set. Indeed, Scruggs' (1998) two-factor parameterization introduces an additional measure of risk and allows the covariance between the asset under investigation and the variable that proxies for the state of the investment opportunities to influence the behavior of returns over time. Such a framework can be used to model the role of oil in financial portfolio hedging decisions.

Oil price dynamics is often associated with both stock market and exchange rate behavior. A number of studies, based on different data and estimation procedures, find a negative financial linkage between oil and stock prices i.e. a large negative covariance risk between oil and a widely diversified portfolio of assets. A substantial body of literature, however, claims that there is a predominant real linkage between the value of equities and oil via production and the business cycle, expansionary periods (in turn related to stock price increases) being closely associated with oil price rises.

As for the dollar, it has traditionally influenced the price of oil and other commodities, including gold and base metals, which are mostly priced in the green currency. Here too we have two channels of transmission, one real, one financial. From the macroeconomic point of view higher oil prices lead to higher trade deficits which, weakening the dollar, bring about compensatory oil price increases. The financial channel has taken on greater weight in recent years, with the entry of hedge funds, banks and other financial institutions in the commodity markets. As noted by Roache (2008), commodities behave differently from stocks and bonds and offer diversification. Traders that are bearish on the dollar will sell a dollar labelled (stock) asset and buy oil (and vice versa if they are bullish on the dollar) in order to diversify their portfolio. Indeed, crude oil has attracted funds away from the financial markets during the recent bouts of turmoil.

This paper investigates the behavior, from October 1992 to June 2008, of weekly changes in the WTI oil price, in the Dow Jones stock index and in the US dollar effective exchange rate. The analysis improves upon previous work in several respects.

  • (i)

    It closely examines the relevance of feedback trading in the spot oil market using long and homogeneous time series which span more than fifteen years and encompass large shifts in market sentiment. The short run dynamics of oil price changes and its interaction with the corresponding futures price are parameterized with the help of models of growing complexity which identify a convincing common pattern. To the best of our knowledge, scant empirical work has been done on documenting the interaction between noise and informed trading by oil market participants.

  • (ii)

    While there is a large body of literature dealing with feedback trading in stocks and other types of assets in a univariate setting, very little research has been done in a multivariate framework. Our investigation builds on a bivariate approach originally set out by Dean and Faff (2008), which introduces feedback trading in a two-factor ICAPM model of stock and bond returns' interaction by Scruggs (1998). Oil prices, exchange rate and stock index rates of change are simultaneously modelled with the help of a GARCH-M approach which parameterizes their conditional covariance interactions. The complex dynamics of feedback trading behavior in periods of stress are carefully set out, first in a bivariate perspective, involving the WTI oil price and the Dow Jones stock market index, and, successively, adding the US dollar effective exchange rate, in a trivariate context.

Speculative behavior seems to have affected the crude oil and stock exchange pricing in the time period analyzed in this paper. Indeed, we find convincing evidence of positive feedback trading in the oil and stock markets. As expected, the corresponding price overshooting correction brings about serial correlation of the returns, the magnitude of which increases with the level of volatility within and across markets. Oil price shifts are negatively related to stock price and exchange rate variation, and our estimates unravel a complex web of time-varying first and second order conditional moment interactions that affect both the CAPM and feedback trading components of the model and justify the use of a multivariate approach.

The analysis is organised as follows. Section 2 introduces the theoretical framework, based on the multifactor inter-temporal CAPM developed by Merton (1973), where behavioral asset pricing mechanisms such as feedback trading can be accounted for thanks to the presence of noise traders. The empirical evidence is presented in Section 3 where the relevance of feedback trading in the crude oil market is investigated using GARCH parameterizations. Section 3.1 provides a basic estimation of the oil price dynamics in a univariate context. The analysis is then extended to a multivariate approach. Section 3.2 investigates the interaction between oil and stock prices via a two-factor ICAPM parameterized by a CCC bivariate GARCH-M with complex nonlinear conditional mean equations. Section 3.3 introduces the exchange rate in the previous model and provides a comprehensive picture of the dynamic interrelation between the conditional moments of the three time series. Section 4 concludes the paper.

Section snippets

A multifactor ICAPM with feedback trading

The relationship between returns and volatility is central to the pricing of an asset or a commodity. Indeed, as suggested by the Capital Asset Pricing Model (CAPM), the greater the uncertainty about the future price, which increases with its volatility, the higher is the return required in order to compensate for the non-diversifiable risk. In a major breakthrough Merton (1973) points out in the Intertemporal Capital Asset Pricing Model (ICAPM) that investors will price an asset in relation

The empirical evidence

Despite a large body of empirical evidence on the ICAPM, the focus has mainly been on equities and little has been done on the alternative asset class represented by commodities.

Our weekly data span from 6 October 1992 to 24 June 2008. The oil spot prices (St) are the WTI Spot Price fob (US dollars per Barrel), futures oil prices (Ft) are provided by the EIA database.6

Conclusions

This paper investigates the relationship between oil prices, stock prices and US dollar exchange rate using a behavioral ICAPM approach, where noise traders are allowed to influence asset demands.

A nonlinear model of the rate of change of spot oil prices is developed in a univariate framework and subsequently in a multivariate context, where the Dow Jones Industrial index return and the rate of change in the US dollar nominal effective exchange rate are assumed to account for changes in the

References (19)

  • KoutmosG.

    Feedback trading and the autocorrelation pattern of stock returns: further empirical evidence

    Journal of International Money and Finance

    (1997)
  • LakonishokJ. et al.

    The impact of institutional investors on stock prices

    Journal of Financial Economics

    (1992)
  • BollerslevT.

    Modelling the coherence in short-run nominal exchange rates: a multivariate generalized ARCH approach

    Review of Economic Studies

    (1990)
  • BollerslevT. et al.

    Quasi-maximum likelihood estimation and dynamic models with time-varying covariances

    Econometric Reviews

    (1992)
  • CutlerD.M. et al.

    Speculative dynamics and the role of feedback traders

    American Economic Review

    (1991)
  • DeanW.G. et al.

    Feedback Trading and the Behavioural CAPM: Multivariate Evidence across International Equity and Bond Markets

    (2008)
  • De LongJ.B. et al.

    Positive feedback investment strategies and destabilizing rational speculation

    Journal of Finance

    (1990)
  • EngleR.F. et al.

    Measuring and testing the impact of news on volatility

    Journal of Finance

    (1993)
  • GarbadeK.D. et al.

    Price movements and price discovery in futures and cash markets

    Review of Economics and Statistics

    (1981)
There are more references available in the full text version of this article.

Cited by (229)

  • Ambiguity and risk in the oil market

    2024, Economic Modelling
View all citing articles on Scopus

The authors are grateful to the anonymous referees for extremely useful suggestions.

View full text