2014年-IMF国际货币组织全球_Oil_Price_Volatility_and_the_Role_of_Speculation_34页_715kb

Summary of "Oil Price Volatility and the Role of Speculation"

Core Content

This working paper by Samya Beidas-Strom and Andrea Pescatori explores the role of speculation in oil price volatility, building on the structural vector autoregression (SVAR) model introduced by Kilian and Murphy (2013). The paper aims to quantify the short-term impact of speculative demand shocks on oil prices and to distinguish between fundamental and non-fundamental (noise trading) shocks in the context of oil futures markets.

Main Points

1. Speculation and Oil Price Volatility

• The paper revisits the debate on whether speculation contributes to oil price volatility.
• It estimates a sign-restricted SVAR model to analyze the effects of speculative demand shocks on oil prices.
• Speculation is defined as the demand for oil inventories driven by expectations of future price changes, which can be either based on fundamental news or noise trading.
• The study distinguishes between short-term and long-term impacts of speculative shocks.

2. Key Findings

• Short-term impact: Speculative demand shocks contribute between 3% and 22% to short-term oil price volatility.
• This impact is smaller than flow demand shocks but larger than flow supply shocks.
• The paper argues that speculation is not a significant driver of long-term oil price movements, as its effects are temporary and revert to fundamental values.
• The short-run upper bound of speculative shocks is estimated using crude oil inventories, which are sensitive to speculative behavior.

3. Identification Strategy

• The authors use a storage model to derive the theoretical basis for identifying speculative demand shocks.
• They assume that oil prices are determined by the marginal productivity of oil, which is influenced by both fundamental and speculative factors.
• To distinguish between fundamental and non-fundamental shocks, they impose restrictions on the time horizon of speculative effects, assuming that noise trading anomalies are temporary.

4. News Shocks and Speculation

• News shocks (e.g., changes in oil discoveries, expectations of production disruptions) are treated as a form of speculative demand.
• These shocks have persistent effects on oil prices, while noise trading shocks are short-lived and do not contribute to long-term volatility.
• The paper emphasizes that speculation is a response to expectations about future market fundamentals, not just random behavior.

5. Empirical Analysis

• The model is estimated on quarterly data from 1983:Q1 to 2012:Q4.
• The real oil price is defined as the U.S. refiners' acquisition cost for imported crude oil, adjusted for inflation.
• Two proxies for global oil demand are considered: the Global Activity Index (GAI) and the global industrial production (IP) index.
• The IP index is preferred due to its superior forecasting performance compared to the GAI and other models.

6. Impulse Response Functions (IRFs)

• The paper examines the short-run responses of oil prices and inventories to different types of shocks.
• When speculative demand shocks are short-lived, the 2003–08 oil price surge is primarily attributed to flow demand shocks.
• After 2005, speculative demand shocks begin to play a more prominent role, with the same drivers re-emerging during 2011–12.
• If speculative shocks are allowed to have longer-lasting effects, flow demand shocks lose explanatory power, and speculative shocks contribute significantly to both short-run and long-run volatility.

Key Information

• Data Sources:

  • Crude oil production: International Energy Agency (IEA) monthly database.
  • Real oil price: U.S. Energy Information Agency (EIA) data, deflated by the U.S. consumer price index.
  • Global oil demand proxies: Global Activity Index (GAI) and global industrial production (IP) index.
  • Crude oil inventories: OECD total crude inventories from IEA.

• Model Specification:

  • The model is an adaptation of Kilian and Murphy (2013).
  • It incorporates a storage model and rational expectations.
  • The null hypothesis is that only fundamental shocks drive long-term oil price movements, and speculative shocks are temporary anomalies.

• Estimation Methodology:

  • The authors use sign-restricted SVAR to identify the effects of different shocks.
  • They impose additional restrictions to narrow the set of admissible models and estimate the short-run upper bound of speculative shocks.
  • The model is stationary, and no co-integration is found between oil production and crude oil inventories.

• Impulse Response Functions (IRFs):

  • The response of oil prices to speculative shocks is bounded between 3% and 22%.
  • The short-run effect of speculative shocks is less than flow demand shocks but greater than flow supply shocks.
  • The historical decomposition shows that speculative shocks become more significant after 2005, contributing to the 2003–08 oil price surge.

• Conclusion:

  • The paper contributes to the literature by showing that speculative demand shocks can be distinguished from fundamental shocks.
  • It proposes a novel method to bound the contribution of financial speculation to oil price volatility.
  • The main implication is that while speculation may influence short-term volatility, it is not a primary driver of long-term oil price trends.

Key Figures and Tables

• Figure 1: Evolution of the real oil price and political events.
• Figure 2: Admissible estimated models – impulse response functions.
• Figure 3: Narrowing the admissible estimated models – IRFs.
• Figure 4: Historical decomposition of the drivers of the real oil price.
• Figure 5: Absolute drivers of the real oil price.
• Table 1: Real oil price variance decomposition.

References

• Kilian, L. (2009)
• Kilian, L., & Murphy, A. (2013)
• Kilian, L., & Lee, S. (2013)
• Juvenal, S., & Petrella, S. (2011)
• Knittel, C. R., & Pindyck, R. E. (2013)
• Fama, E. F. (1998)
• Singleton, K. J. (2011)
• Alquist, R., & Kilian, L. (2010)
• Liu, L., & Tang, C. (2010)
• Tang, C., & Xiong, W. (2010)
• Masters, A. (2008)
• Buyüksahin, B., & Robe, M. (2012)
• Fattouh, B., Killian, L., & Mahadeva, G. (2012)
• IMF (2011b), (2012), (2013)
• Giese, M., Nixon, D., & Tudela, M. (2010)
• Baumeister, C. F., & Peersman, G. (2012)
• Aastveit, A., Bjørnland, H., & Thosrud, M. (2012)
• Dvir, G., & Rogoff, K. (2013)

Appendix

• Contains model equations, parameter definitions, and derivation of impulse response functions.
• The fundamental system is defined by equations (1)–(6), while the non-fundamental system is described by equation (7') and the associated deviations from fundamentals.