The Dynamic Correlation between China’s Policy Uncertainty and the Crude Oil Market: A Time-varying Analysis

ABSTRACT

This research investigates the dynamic correlation between China’s policy uncertainty and the crude oil markets (i.e. domestic and international markets) using monthly time-series data from May 2003 to December 2018. To consider the non-linear and dynamic properties, we adopt the time-varying parameter structural vector autoregression model (TVP-SVAR) to estimate the dynamic correlations between the time series. By employing four categorical policy uncertainty indices, i.e. monetary policy uncertainty index (MYPU), fiscal policy uncertainty index (FLPU), exchange rate policy uncertainty index (EXER), and trade policy uncertainty index (TEPU), our results reveal that the correlation between China’s policy uncertainty and real crude oil returns is time-varying and non-linear. Specifically, the independence between policy uncertainty and oil returns varies more constantly and at a high degree while the time-varying interaction between policy uncertainty and global oil production (global economic activity) varies at a lower frequency and at a low level. Furthermore, regarding the associations between categorical policy uncertainty and global oil returns, the average correlation of EXER (negative) is the strongest one, followed by FLPU (negative), then MYPU (positive), and finally TEPU (negative). Moreover, it is noteworthy that WTI (Daqing) crude oil prices are utilized to increase the robustness of our conclusions. Finally, we generally confirm the dynamic and negative correlation between China’s geopolitical risk and crude oil returns. It is evidently clear from the results that investors should pay close attention to the policy uncertainty to avoid the adverse impact of specific policy uncertainty on crude oil returns.

KEYWORDS:

• Policy uncertainty
• crude oil returns
• dynamic correlation
• China
• Time-varying parameter structural vector autoregression model (TVP-SVAR)

JEL CLASSIFICATION:

• C32
• E44
• P18
• Q43

Acknowledgements

The authors are grateful to the Editor and the anonymous referees for helpful suggestions. Both authors contributed equally to this study and share first authorship.We acknowledge the financial support from the Natural Science Foundation of Jiangxi Province of China through Grant No: 20202BAB201006.

Notes

1. More detailed explanation can be found in Section 3: Methodology.

2. Daqing Field is the largest oil field in China, and Shengli Field is the second-largest oil field in China. Thus, their spot prices strongly represent domestic specific crude oil demand. The view is taken from the website of Wikipedia, https://en.wikipedia.org/wiki/Daqing_Oil_Field and https://en.wikipedia.org/wiki/Shengli_Oil_Field.

3. Of note here is that this seemingly unrelated regression model (SUR) with time-varying parameters is estimated by employing the Bayesian methods and Gibbs sampling algorithm. A more detailed characterization about this model is available in the online appendix of Akram and Mumtaz (2019).

4. The summary statistics of all raw time-series data can be provided upon request.

5. We also employ WTI crude oil price and Shengli crude oil price to check the robustness of our results.

6. The data of CPI are available at the website of China’s National Bureau of Statistics.

7. The personal website of Kilian is https://sites.google.com/site/lkilian2019/research/data-sets.

8. The definitions of specific policy categories for uncertainty are available from their personal website: https://economicpolicyuncertaintyinchina.weebly.com.

9. The personal website of Huang is https://economicpolicyuncertaintyinchina.weebly.com/

10. Following Kang and Ratti (2013), global crude oil production is transformed into log-returns by 100× the log differences.

11. We also present the average positive (negative) correlation between crude oil returns and categorical ECPU indices in the right side of Table 1.

12. Model d consists of GEPR, global oil production, global economic activity, Brent real oil returns, and real stock returns, whereas Model e encompasses GEPR, global oil production, global economic activity, Shengli real oil returns, and real stock returns.