关键词: 投资者关注, 波动率预测, 黄金期货, 异质自回归模型

Abstract: China listed gold futures in 2008 and implemented its night trading in 2013. Since then, the trading volume of Shanghai gold futures market has jumped to the third place in the world and ranked first in the Asia Pacific region. The gold futures market has gradually formed a “Chinese price”. With the continuous development and improvement of China’s gold futures market and its increasing influence, it is necessary to conduct a reasonable modeling and accurate prediction of the volatility of China’s gold futures.
As important participants in the gold market, investors’ attention to the market and their trading behavior will greatly affect the fluctuation level of gold futures prices. In view of this, this article starts from behavioral finance theory and adopts four classic heterogeneous autoregressive volatility models of realized volatility to explore the predictive abilities of institutional investors’ attention and individual investors’ attention on the volatility of China’s gold futures prices. Among them, the construction of individual investors’ attention proxy is based on the Baidu search index, and the construction of institutional investors’ attention proxy is based on regression decomposition of China’s gold futures trading volume, considering the time dimension of information. Specifically, based on expected and unexpected trading volumes, expected institutional investor attention and unexpected institutional investor attention are respectively constructed.
This study applies the 5-minute high-frequency prices for the main contracts of China’s gold futures from January 2, 2014 to June 30, 2023 to construct the realized volatility series. For each of the four classic heterogeneous autoregressive models, we consider the extension that only introduces individual investor attention, as well as the extension that simultaneously introduces individual investor attention, expected institutional investor attention and unexpected institutional investor attention, thus altogether twelve models. The analysis of fitting and prediction results for the twelve models shows that:
(1)Individual investors are easily influenced by market information due to cognitive limitations, which can lead to noise trading behavior and affect gold futures prices. Therefore, in the short term, individual investor attention is positively correlated with the volatility of gold futures.
(2)The expected attention of institutional investors based on historical trading volume reflects, to some extent, the trading conducted by institutional investors on the basis of analyzing market historical information. In addition, institutional investors will engage in zero sum games based on the limited attention of individual investors in the short term. Therefore, the expected institutional investor attention is positively correlated with the volatility of gold futures, while the unexpected institutional investor attention is negatively correlated with the volatility of gold futures in the short term.
(3)Introducing individual investor attention, expected institutional investor attention and unexpected institutional investor attention in the HAR class volatility models can significantly improve the fitting and forecasting performance for China’s gold futures, and this improvement is more significant during market turbulence.
The research conclusion of this article has practical significance. On the one hand, investors can more accurately grasp the volatility characteristics of China’s gold futures, and carry out more effective investment portfolio construction and risk management during periods of economic turbulence. On the other hand, the government and policy makers can monitor the market more effectively, formulate corresponding policies, improve risk prevention and control capabilities, and ensure the sustained and healthy development of China’s economy and social stability. As a prospect, the method proposed in this article to introduce investor attention variables can be extended to predict various financial and commodity futures price volatility, which is also the direction we are concerned about.

Key words: investor attention, volatility prediction, gold futures, HAR model