The Question
Stock-specific sentiment helps you pick individual names, but what about calling the entire market? When Chinese financial media turns pessimistic—across all sectors, all topics—does the CSI300 follow? And more importantly, can you systematically measure aggregate sentiment to time your market exposure?
Traditional macro timing uses economic indicators that lag reality by weeks or months. PMI, credit data, and policy statements tell you what happened. News sentiment tells you what the market is feeling right now, and collective market emotion often precedes price action by days to weeks.
Can you construct a sentiment-based macro indicator that predicts broad market direction—not just which stocks to buy, but whether to be long at all?
The Approach
We aggregate individual stock sentiment from SmarTag News into a market-wide sentiment index by calculating the equal-weighted or cap-weighted average sentiment across all coverage. This captures the "mood of the market"—when positive news outweighs negative across the board, the index rises. When gloom dominates, it falls.
The key innovation: sentiment momentum matters more than sentiment level. Markets don't move on absolute optimism or pessimism—they move on changes in sentiment. We calculate the rate-of-change in aggregate sentiment over rolling windows (5-day, 20-day) to identify inflection points.
When aggregate sentiment shifts from negative to positive (crossing above zero from below), it signals emerging risk-on conditions. When sentiment rolls over from positive to negative, it warns of coming drawdowns. We test these signals against CSI300 forward returns at multiple horizons.
The Finding
The market sentiment index demonstrated significant predictive power for CSI300 direction, particularly at 5-10 day forward horizons. When sentiment momentum turned positive, the CSI300 delivered positive returns 58-62% of the time over the next 10 days—a meaningful edge over the baseline 51% win rate.
More importantly, extreme sentiment shifts flagged major market turning points. The late-2020 sentiment surge preceded a sustained rally. The mid-2021 sentiment collapse correctly warned of the correction that followed. The 2022 sentiment trough marked the October bottom within days.
The signal worked best as a risk on/risk off toggle rather than a linear return predictor. When sentiment momentum was strongly positive (top quartile), annualized CSI300 returns were 18-22%. When sentiment momentum was negative (bottom quartile), returns were near zero or negative with much higher volatility. Simply avoiding the market during negative sentiment periods would have improved Sharpe ratios dramatically.
Asset allocators and macro traders found the signal particularly useful: it provided a quantified, real-time proxy for the "market mood" that fundamental analysts try to gauge through anecdotal evidence. Instead of intuition, you had a systematic framework.
Try It Yourself
Building a macro sentiment indicator requires aggregating stock-level news data into a market-wide signal—then testing it rigorously to avoid overfitting to past cycles.
Start by calculating daily aggregate sentiment across your universe (equal-weighted or market-cap weighted). Smooth the raw signal with moving averages to filter noise—markets respond to sustained sentiment shifts, not one-day blips. Test different momentum lookback periods (5-day, 10-day, 20-day) to find what works for your holding horizon.
Use the signal for position sizing rather than binary in/out calls: scale up equity exposure when sentiment momentum is positive, scale down when it's negative. This avoids whipsaw from false signals while capturing the major regime shifts.
Combine with traditional macro factors for robustness: sentiment + credit spreads, sentiment + policy indicators, sentiment + cross-asset flows. Multi-factor macro timing reduces reliance on any single input.
Ready to integrate sentiment-based market timing into your asset allocation process? Book a call to discuss signal construction, regime detection, and portfolio implementation.
Source: JPMorgan《News Sentiment for China Macro》 (2020-11-25); 中金公司《量化配置系列（10）：如何利用市场主要矛盾辅助大势研判》 (2022-08-19).