In behavioural finance, the relationship between investor sentiment and stock returns has long been recognised. However, most studies rely on sentiment data at the same frequency as the returns being forecast—such as daily, monthly, or quarterly. With the rise of digital platforms, high-frequency intraday sentiment data has become accessible, yet its potential to improve low-frequency return forecasts remains underexplored. Against this backdrop, China Finance Review International (CFRI) brings you a study titled “Does intraday high-frequency investor sentiment help forecast stock returns? Evidence from the MIDAS models”, which investigates whether high-frequency sentiment extracted from Chinese online stock forums can enhance the predictability of daily stock returns.
Methodology and Scope
The authors employ Mixed Data Sampling (MIDAS) models to integrate intraday high-frequency investor sentiment with daily stock returns of Chinese A-shares. Sentiment is constructed from over 6.7 million posts on the Eastmoney stock forum between 2014 and 2022, using a tailored Chinese financial sentiment dictionary. The study distinguishes between sentiment during trading hours (TS) and non-trading hours (PS, LS, AS), and compares the performance of various MIDAS specifications—including U-MIDAS, Beta, and Almon lag models—against a daily sentiment (DS) baseline.
Key Findings and Contributions
- High-frequency intraday sentiment significantly outperforms daily aggregated sentiment in forecasting daily stock returns.
- Sentiment during non-trading hours has stronger predictive power than sentiment during trading hours.
- Among MIDAS-class models, the U-MIDAS model delivers the best forecasting accuracy, both in-sample and out-of-sample.
- The study also introduces a novel, high-frequency sentiment proxy tailored to the Chinese market, filling a gap in the existing literature.
Why It Matters
China’s A-shares market is the world’s second-largest by capitalisation and is dominated by retail investors, who are more prone to sentiment-driven trading. Improving return predictability in such a market has significant practical implications for both domestic and international investors. This research demonstrates that intraday sentiment—especially from non-trading periods—can capture nuanced market dynamics that daily measures miss, offering a more timely and granular tool for forecasting.
Practical Applications
- Investors and Fund Managers: Can use intraday sentiment signals, particularly from non-trading hours, to refine trading strategies and asset allocation.
- Financial Analysts: May incorporate U-MIDAS models with high-frequency sentiment data for more accurate equity return forecasts.
- FinTech and Data Providers: Can develop sentiment tracking tools that segment data by trading vs. non-trading periods to enhance predictive analytics.
- Academic Researchers: The methodology and findings offer a framework for applying high-frequency sentiment analysis in other emerging markets.
China Finance Review International: “Does intraday high-frequency investor sentiment help forecast stock returns Evidence from the MIDAS models”. DOI: 10.1108/CFRI-12-2023-0344