The 259th Wenlan Financial Forum

Abstract:
This paper studies the pricing performance and forecasting effect of commodity-related sentiment in media information on the expected returns of Chinese commodity futures. Using the BERT+Attention-BiLSTM model, this paper extracts relevant sentiments from the daily media news, research reports and reviews of 13 major commodities from September 2019 to August 2021, and then constructs a tradable sentiment factor based on sentiment ranking. This paper finds that the risk exposure of the sentiment factor has an insurance-like effect and the sentiment feature can explain the near-term returns in the next period of China's commodity futures portfolio, while the diffusion returns are not significant. Adding a sentiment factor to the out-of-sample forecast improves the forecast accuracy of expected returns for coking coal, iron ore, lead, nickel, tin, corn, soybean, and cotton futures. The long-short portfolio constructed with sentiment factor can obtain an average monthly return of 1.6%, while the long-short portfolio without sentiment factor can only obtain a monthly return of -2.4%.

Speaker Introduction:
Xiang Zhang, associate professor, is currently the deputy director of the domestic cooperation and development department of Southwestern University of Finance and Economics, the deputy dean of the Big Data Research Institute, a doctoral tutor, and an expert of the Tianfu Ten Thousand People Program in Sichuan Province. He was a visiting professor at the Haas School of Business at the University of California, Berkeley, and the head of the Department of Financial Engineering at Southwestern University of Finance and Economics. His main research interests are asset pricing, financial risk management and financial technology.