2014年-世界发展银行全球_Co-Movement_of_Major_Commodity_Price_Returns___Time-Series_Assessment_38页_1mb

Co-Movement of Major Commodity Price Returns: Time-Series Assessment Summary

Core Content

This paper analyzes the co-movement of price returns among 11 major energy, agricultural, and food commodities using monthly data from January 1970 to May 2013. The study employs three main econometric techniques: the uniform-spacings method, a multivariate dynamic conditional correlation (DCC) model, and a rolling regression approach. The goal is to assess both the unconditional and conditional correlations across commodity markets and to investigate the macroeconomic and financial factors that might drive these correlations.

Main Findings

1. High Cross-Sectional Correlation

• Commodity price returns, especially within energy and agricultural categories, show strong and positive correlations.
• The cross-section unconditional correlation is generally significant across all time periods, except for the 1979M11-1987M06 period.
• The most correlated pairs include:
  • Maize and sorghum (0.81)
  • Sorghum and wheat (0.53)
  • Barley and maize (0.49)
• These correlations are often categorized as "category-based" due to shared supply, demand, and input cost factors.

2. Increased Co-Movement in Recent Years

• The overall level of co-movement has increased, particularly between energy and agricultural commodities.
• Maize and soybean oil show the strongest co-movement in recent years, likely due to their role in biofuel production.
• This suggests a growing interdependence between energy and agricultural markets, especially since the rise of the biofuel industry.

3. Time-Varying Conditional Correlations

• A multivariate DCC model is used to estimate how the conditional correlations evolve over time.
• Conditional correlations between maize and crude oil, maize and natural gas, soybean oil and crude oil, and soybean oil and coal increased significantly between 2004 and 2009, then declined slowly.
• The correlation between maize and wheat increased starting around 2004, with a peak in 2009.
• Soybean oil and crude oil show a sharp increase in conditional correlation, reaching levels close to 0.2 by 2009.

4. Volatility Dynamics

• Volatility in commodity price returns is generally higher in recent years compared to the previous two decades.
• Natural gas and sugar are the most volatile commodities over the full sample period.
• Crude oil and soybean oil show the most significant increases in volatility between 2007 and 2013.

5. Rolling Unconditional Correlations

• Rolling correlations between commodity returns are computed using a 24-month window.
• These correlations are influenced by stock market volatility, especially after 2007.
• Real interest rates are not found to be a significant driver of co-movement in the data.

Key Information

Data and Methodology

• Commodities Studied: 11 major energy, agricultural, and food commodities.
• Time Period: January 1970 to May 2013.
• Techniques Used:
  • Uniform-spacings method: To assess cross-sectional unconditional correlation.
  • Multivariate DCC model: To estimate time-varying conditional correlations.
  • Rolling regression: To evaluate the impact of macroeconomic and financial factors on co-movement.

Structural Breakpoints

• Identified structural breaks in the price return dynamics at:
  • 1979M11: Associated with the Iranian revolution and the Iran-Iraq war.
  • 1987M07: Linked to severe U.S. droughts and crop failures.
  • 2007M05: Related to the global food price crisis.

Conclusion

The study concludes that:

• The co-movement of commodity price returns has increased significantly over time, particularly between energy and agricultural commodities.
• This increased co-movement is likely driven by the expansion of the biofuel industry, which has created stronger interdependencies between these markets.
• Stock market volatility is positively associated with the co-movement of commodity returns, especially after 2007.
• The use of monthly data allows for a more accurate assessment of dynamic relationships compared to lower frequency data.

The findings contribute to the understanding of commodity market behavior and provide insights into the factors that influence price co-movement, with implications for risk management, investment strategies, and policy-making in developing economies.