🔬 physics

Effects of analyst sentiment on volatility dynamics in financial market

This paper analyzes analyst sentiment in the Chinese stock market using natural language processing and GARCH modeling to demonstrate that while both optimistic and pessimistic sentiments exhibit short-range memory, only pessimistic sentiment significantly drives and explains future market volatility.

Original authors: Xiongfei Jiang, Tao Cen, Ling Bai, Lifu Jin, Jiu Zhang, Long Xiong

Published 2026-06-02

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Xiongfei Jiang, Tao Cen, Ling Bai, Lifu Jin, Jiu Zhang, Long Xiong

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the stock market as a giant, bustling ocean. The waves (prices) go up and down, sometimes gently, sometimes violently. For decades, investors have tried to predict these waves by looking at the "weather reports" written by financial analysts. Usually, they've focused on the hard numbers in these reports—like earnings forecasts or buy/sell ratings.

But this paper asks a different question: What about the mood of the weather report itself?

The authors decided to look at the text of thousands of analyst reports from the Chinese stock market (2013–2015) and use computer technology to measure the "emotional temperature" of the writers. They wanted to see if the optimism or pessimism in these words could predict how wild the market waves (volatility) would get.

Here is what they found, broken down into simple concepts:

1. The Two Moods: Sunshine vs. Storm Clouds

The researchers built two "mood meters":

The Optimism Meter: How many happy, "bullish" words were used?
The Pessimism Meter: How many sad, "bearish" words were used?

They discovered that both moods are "short-lived." Like a sudden gust of wind or a brief rain shower, the mood in the reports doesn't last forever. However, pessimism tends to stick around a little longer (about 4–5 weeks) than optimism (about 3 weeks).

2. The Surprising Relationship with Waves

The most interesting part is how these moods relate to the "roughness" of the market (volatility).

Optimism acts like a party: When analysts are very optimistic, the market tends to get rougher. It's like a party getting louder and more chaotic; high optimism leads to more trading and bigger swings in price.
Pessimism acts like a quiet library: When analysts are very pessimistic, the market actually gets calmer. It's counter-intuitive, but the data shows that high pessimism leads to less volatility.

Why? The authors suggest it's about human behavior. When people are excited (optimistic), they rush to trade, creating chaos. When people are scared or pessimistic, they tend to freeze and stop trading, which actually quiets the market down.

3. Who is the Driver? (The "One-Way Street" Discovery)

The researchers used a mathematical tool called "Transfer Entropy" to figure out who is driving whom. Imagine two cars on a road: is Car A pushing Car B, or is Car B pushing Car A?

Pessimism drives Volatility: The study found a clear one-way street. Past pessimism predicts future calmness. If analysts were gloomy last week, the market is likely to be calmer this week. The gloominess provides useful information that helps settle the market.
Optimism is just a passenger: There was no clear direction between optimism and volatility. Optimism doesn't seem to reliably predict what the market waves will do next.

4. The "GARCH" Test: The Crystal Ball

To be sure, the authors ran a sophisticated statistical test (called a GARCH model), which is like a crystal ball for predicting market turbulence.

The Result: When they added "Pessimism" to their crystal ball, it got much better at predicting how calm the market would be.
The Catch: When they added "Optimism," the crystal ball didn't get any better. Optimism didn't help predict the waves.

5. Why Does Pessimism Calm the Market?

The paper offers a fascinating explanation for why bad news calms the market:

The "Rumor Mill" Effect: In the real world, bad news often leaks out as rumors before it's official. This causes panic and wild trading (high volatility).
The "Official Report" Effect: When the analyst finally writes the official pessimistic report, it confirms what people already suspected. The uncertainty is gone. The "speculation" stops because the truth is out. Once the truth is known, people feel more certain, they stop panicking, and the market settles down.

The Bottom Line

This paper tells us that in the Chinese stock market, bad news (pessimism) is actually a stabilizer, while good news (optimism) is a source of chaos.

If you want to know if the market is going to be calm next week, look at how gloomy the analysts were this week. If they were very negative, the market is likely to quiet down. If they were very happy, expect some turbulence. Optimism, however, doesn't seem to tell us much about what's coming next.

Technical Summary: Effects of Analyst Sentiment on Volatility Dynamics in Financial Market

Problem Statement
While analyst reports are established as indispensable references for investment decision-making, existing literature predominantly focuses on structured data (e.g., earnings forecasts, buy-hold-sell ratings). The qualitative textual data within these reports, which reflects investor sentiment, has been largely overlooked despite the feasibility of quantifying it via Natural Language Processing (NLP). This study addresses the gap in understanding how analyst sentiment, specifically extracted from unstructured text, influences the volatility dynamics of the Chinese stock market. The research seeks to determine whether sentiment acts as a systematic risk factor and to quantify the directional flow of information between sentiment and market volatility.

Methodology
The study employs a multi-stage quantitative approach utilizing data from the Chinese stock market:

Data Collection: The authors collected abstracts from 7,789 analyst reports covering 197 stocks between April 2013 and December 2015. To ensure reliability and predictive relevance, the sample was restricted to reports published prior to annual report releases and focused on widely watched stocks (excluding those with fewer than 20 reports per fiscal year).
Sentiment Extraction: Using the open-source Jieba text segmentation toolkit and a manually curated Chinese sentiment vocabulary, the authors identified emotional tendencies. Sentiment words were scored on a scale from -2 (very pessimistic) to +2 (very optimistic), accounting for negations and intensifiers.
Index Construction: Weekly sentiment indices were constructed for both optimistic ( $S^+$ ) and pessimistic ( $S^-$ ) sentiments, normalized by the total number of reports ( $N_t$ ) in a given week.
Statistical Analysis:
- Auto-correlation: Analyzed the memory properties of sentiment indices.
- Correlation Analysis: Defined a sentiment-volatility correlation function $G(\Delta t)$ to measure the linear dependency between sentiment and volatility (defined as the absolute value of log returns) at various time lags.
- Transfer Entropy: Applied to non-stationary financial time series to measure the net information flow ( $\delta_{XY}$ ) between sentiment and volatility, determining causal directionality.
- GARCH Modeling: Incorporated sentiment indices as exogenous variables into a GARCH(1,1) model to assess their explanatory power for conditional variance (volatility) after accounting for volatility persistence.

Key Results

Memory Characteristics: Both optimistic and pessimistic sentiments exhibit short-range auto-correlation. Pessimistic sentiment demonstrates a longer memory duration (approximately 5 weeks) compared to optimistic sentiment (approximately 3 weeks).
Correlation Dynamics:
- Pessimistic Sentiment: Shows a significant negative correlation with future volatility. High levels of pessimism are associated with reduced future volatility, while low pessimism correlates with higher volatility.
- Optimistic Sentiment: Shows a positive correlation with volatility. High optimism induces higher volatility.
- Directionality: The correlation is observed only when $\Delta t > 0$ (sentiment affecting future volatility), suggesting sentiment drives volatility but not vice versa.
Information Flow (Transfer Entropy): Analysis reveals a unidirectional net information flow from past pessimistic sentiment to future volatility. In contrast, there is no significant directional flow between optimistic sentiment and volatility.
GARCH Model Findings:
- When incorporated into the GARCH(1,1) framework, pessimistic sentiment serves as a statistically significant explanatory factor for volatility (specifically for the CSI 300 Index). The coefficient $\lambda$ is negative, confirming that high pessimism reduces future volatility.
- Optimistic sentiment is not statistically significant as an exogenous variable in the volatility model.
- The inclusion of pessimistic sentiment reduces the sum of coefficients ( $\alpha + \beta$ ), indicating it captures a portion of the volatility persistence.

Significance and Claims
The paper claims that pessimistic analyst sentiment provides independent, predictive information regarding market volatility dynamics, distinguishing it from optimistic sentiment which lacks such predictive power in this context.

The authors attribute the negative correlation between pessimism and volatility to two primary mechanisms:

Behavioral Factors: Investors exhibit bounded rationality and risk aversion. High pessimism leads to trading inactivity, thereby reducing volatility, whereas high optimism triggers active trading that increases volatility.
Information Disclosure: Analysts affiliated with investment banks are often cautious about releasing pessimistic reports early due to ties with listed companies. Consequently, pessimistic reports are often released after news has already circulated via rumors or inside information. Once the official pessimistic report is released, speculation ceases, market uncertainty decreases, and volatility drops.

The study concludes that pessimistic sentiment is a robust proxy for market sentiment that can improve the modeling of financial volatility, particularly within the GARCH framework, whereas optimistic sentiment does not offer similar explanatory value.