The Post-Internationalization Evolution of the Price Discovery Pattern in China’s Iron Ore Markets

Abstract

Previous literature provided evidence of a negative impact of internationalization on the quality of China’s iron ore futures market within a 6-month window (i.e. June 1, 2018 to November 31, 2018), but little is known on the long-term impact. Using a longer, 4-year-7-month window from June 1, 2018 to December 30, 2022, we examine the market’s price discovery function, an important measure of market quality. We find significant structural break in China’s iron ore data series in August 2021, which largely coincide with a period of rising commodity prices to all-time high. Before the breaks, the price discovery function of China’s iron ore futures appears quite weak, with the spot market leading the futures market, which is consistent with the findings of Fan et al. During the post-break period, however, the futures market demonstrates a stronger price discovery function and takes over the leading role. The day-of-the-week effects in iron ore futures also seem to weaken post-break as well. Our GARCH model results exhibit significant bi-directional volatility spillover between the futures and spot markets.

Keywords:

• Iron ore futures
• information share
• price discovery
• volatility spillover

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Source: “China’s five-year plan to slash Australian iron ore imports”, May 21, 2021, by Michael Smith. Wire: Australian Financial Review (AFR). Available from Bloomberg Terminal. Retrieved on May 15, 2024.

2 Source: “World Steel in Figures 2023”, World Steel Association.

3 Source: “The 14th Five-Year Plan. Chapter 8 Implementing the Manufacturing Powerhouse Strategy”, China National Development and Reform Commission.

4 Source: “China’s NDRC to Hold Meeting After Commodity Prices Jump”, Nov 29, 2019, Bloomberg News.

5 See the timeline over a dozen actions by various government bodies from April to August 2021 in “China intervenes to manage commodity prices”, August 4, 2021, by National Post, Beijing.

6 Source: “Iron ore surges anew, Chinese exchanges seek to slow rally”, August 16, 2021, Bloomberg News.

7 Source: Statement by the National Development and Reform Commission, Nov 23, 2023.

8 Source: “China Regulators to Check Iron Ore Trading, Stock in Inspection”, February 11, 2022, Bloomberg News.

9 Source: “China’s five-year plan to slash Australian Iron Ore imports”, May 21, 2021, Australian Finance Review (AFR).

10 Ibid.

11 We roll the contracts over on the first day of the expiration month due to the typically low open interest and high price volatility during the last trading days. We apply the tests to data series with different rollover dates and the results are similar.

12 Tan (2021) reported that “in the third week of July, authorities unofficially told steel producers in Anhui, Gansu, Fujian, Jiangsu, Jiangxi, Shandong and Yunnan to limit this year’s output to that of last year’s, according to analysts,” and “in the following weeks, China’s Ministry of Finance said it would remove export tax rebates on 23 steel products, discouraging exports,” and “China Iron Ore and Steel Association (CISA) President Shen Bin also got in on the action with a ‘sucker punch’ to prices by vowing to accelerate and ensure China’s self-sufficiency in iron ore supply.”.

13 We further apply both the Lee and Strazicich (2003) and the Perron (1997) models to test possible structural breaks in the subsamples and the results do not indicate the presence of statistically significant breaks.

14 We follow Fan et al. (2020) by excluding observations in the same month as the break dates from the subsamples in order to avoid any potential influence by the structural breaks themselves.

15 Cointegration tests typically only apply to time series that are I(1) or higher orders.

16 Kolb (2003) discusses the weekend effect in futures markets as documented by numerous prior studies, and Lee et al. (2009) have found significant Friday excess returns in some Chinese futures markets.

17 A similar model is used in Baillie et al. (2002).

18 Hasbrouck (1995) and Baillie et al. (2002) information share results are available upon request.

19 See Tse (1999) for a more detailed discussion of this error correction process.

20 The structural breaks in Table 2 are in conditional and unconditional means but not in volatility. Thus we apply this model to the entire data sample to examine volatility spillover.

21 The GARCH models achieve convergence with a 95% winsorization.