The impact of US macroeconomic news announcements on Chinese commodity futures

Abstract

Using intraday data from 2013 to 2016, we examine the instantaneous response of eight Chinese commodity futures to 19 different types of scheduled US macroeconomic news announcements after the introduction of night trading in China. We provide robust evidence that the surprise components of a number of news announcements exhibit a significant effect on returns, trading volume, and volatility of a majority of Chinese futures contracts, with gold and silver futures being the most sensitive. Moreover, we observe an asymmetric effect between positive and negative surprise components. A further examination of the responses to the US macroeconomic news announcements using US gold and silver futures over the same sample period provides qualitatively similar results with larger magnitudes. This evidence suggests a possible channel through which the impact of macroeconomic news announcements transmits from the US to the Chinese commodity futures market.

Keywords:

• Trading activity
• Market quality
• Cross-border information transmission

JEL Classification:

• G10
• G13
• G14

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

† The Futures Industry Association annual survey reports the most traded commodity futures contracts in the world on the basis of the trading volume. For more information, see https://fia.org/articles/fia-releases-annual-trading-statistics-showing-record-etd-volume-2018.

† The night trading sessions run electronically during the following hours (Beijing time): 21:00–02:30 for gold and silver, 21:00–01:00 for aluminum and copper, and 21:00–23:30 for cotton, palm oil, soybean, and sugar. The trading venue for gold, silver, aluminum, and copper futures is the Shanghai Futures Exchange. Sugar and cotton futures are traded on the Zhengzhou Commodity Exchange, whereas soybean and palm oil futures are traded on the Dalian Commodity Exchange.

‡ The 5-min return estimation window provides a reasonable balance between blurring specific price reactions by sampling too infrequently and confounding market microstructure effects by sampling too frequently (Andersen et al. 2007).

§ We consider a longer examination window because the impact of news announcements on volatility lasts for a longer time period (see, for example, Balduzzi et al. 2001, Elder et al. 2012).

† Throughout this paper, by surprise or shock, we mean the standardized surprise.

† Due to the daylight saving adjustment scheme in the US, there can be two sets of Beijing time when macroeconomic news are released. For example, 08:30 in the US corresponds to 20:30 (without daylight saving) and 21:30 (with daylight saving) in Beijing. In this paper, we use the latter times consistently in the empirical analyses, i.e. 21:30, 22:15, and 23:00.

† The forward selection method starts with no variable in the model and adds a variable with the lowest p-value, followed by a variable with the next lowest p-value. The process stops when none of the remaining variables are statistically significant. The backward elimination method starts with all variables in the model and eliminates a variable with the highest p-value, followed by a variable with the next highest p-value. The process continues until no insignificant variable remains. The bi-directional elimination method combines the forward selection and the backward elimination methods to add or remove a variable at each step.

‡ Although the stepwise regressions are prone to a pretest bias, the results from these regressions are informative in revealing macroeconomic announcements with the greatest explanatory power.

§ Both the trading volume and the realized volatility exhibit significant persistence in our regressions. We do not report the coefficient estimates for lagged dependent variables in the paper because our focus is on the impact variables. The estimates for lagged dependent variables are available upon request from the authors.

† To conserve space, all results for agricultural commodity futures are delegated to the Appendix.

† As we need more 5-min intervals for constructing the 25-min realized volatility and lagged volatility to control for persistence, we abstain from examining the impact of news releases at 22:15. This mitigates possible contamination of the results by the announcements at 21:30.