How Rumor-Driven Irrationality Erodes the Pricing Efficiency of A-Shares

Chen Jiahe points out that in capital markets, investors are often more sensitive to rumors than to public data: From 2022 to 2023, the banking sector was generally labeled with ‘real estate loan risk explosion’ when it was at the bottom of valuation, but in reality, the relevant business accounted for less than 10% of listed companies’ total; the port industry was also surrounded by rumors that ‘Sino-US trade conflicts will collapse US-bound throughput’, but in fact, US-bound throughput accounted for only 5% of the total; the publishing industry experienced a short-term surge due to the rumor that ‘AI will significantly reduce editing costs’. These cases show how information salience and emotional responses create price distortions without substantial fundamental changes, thereby weakening the speed at which prices reflect true value [1].

From the perspective of behavioral finance, A-shares are also affected by biases such as attention bias, overconfidence, herding behavior, and loss aversion: Investors are more likely to respond to prominent negative/hot events, driving excessive price volatility in the short term; due to the dominance of retail investors and limited arbitrage conditions, rational investors cannot fully hedge against noise trading, so rumor-driven mispricing is not easily corrected quickly, causing the market to deviate from the weak-form efficiency hypothesis in information efficiency for a long time [2].

Evolutionary psychology suggests that the sensitivity of human ancestors to threat information (e.g., “There are enemies near you”) in tribal survival environments was amplified by natural selection—even if the risk probability was low, as long as the information was salient, it could trigger a mobilization mechanism. This means that modern investors automatically enter a state of high alert when facing rumors such as “real estate defaults”, “trade frictions”, and “AI replacement”, even if these pieces of information lack statistical support. Chen Jiahe thus proposes that the reason why current rumors can reshape market sentiment in a short time is precisely because this evolutionary residue makes humans tend to “react first and verify later” rather than make judgments calmly based on data [1].

1. Information Structuring and Fact-Checking Linkage
   • Establish a “multi-channel cross-validation” process: For the same topic, it must meet (1) company announcements/financial report data, (2) third-party industry data, and (3) regulatory data or authoritative media disclosure before being included in decision-making.
   • Set up a “fact-first” checklist: Define key indicators (such as loan proportion, customer concentration, main business revenue ratio, etc.) and allowable fluctuation ranges; when encountering rumors, first check whether the indicators actually deviate.

2. Emotion Recognition and Rhythm Control Mechanism
   • Quantify emotional states (e.g., through trading frequency, return volatility, public opinion heat index); once the graph shows a “suspected irrational peak”, trigger a “cooling-off period” or a fixed investment strategy to prevent impulsive trading.
   • Introduce “reverse opinion” checks: Require team members to play the roles of “rumor verifiers” and “fact watchers” respectively, record opposing opinions in writing, and avoid group cognitive biases.

3. Evolutionary Bias Compensation Mechanism
   • Introduce a “probability calibration” step in the strategy: Compare “salient but low-probability” events (such as rumor-driven systemic risks) with actual exposure coefficients (such as loan/operating income ratio) to prevent overreaction due to “evolutionary fear”.
   • Adopt “evolutionary self-awareness training”: Through regular reviews of missed opportunities and rumor-following behaviors, improve investors’ awareness of “intuitive reactions”, thereby reducing the “act first, verify later” path.

4. Public Opinion Fission and Capital Allocation Linkage
   • Combine public opinion data (search index, rumor heat) with capital flow data (large order net buy/sell, turnover rate) to build a “rumor heat-capital reaction” mapping model, identifying whether rumors are truly driven by capital or are short-term noise.
   • Set “rumor event trigger points”: When public opinion heat surges but fundamental indicators remain unchanged (or actual indicators are better), prioritize responses such as “increase holdings/maintain” or “wait and see” rather than “reduce holdings” to avoid locking in losses during irrational panic.

5. Governance and Institutional Enhancement
   • Promote funds or investment advisors to set up a “rumor filtering committee” to supervise high-frequency trading strategies and avoid algorithms automatically reacting to sudden public opinion.
   • Encourage regulators to launch an “information truthfulness disclosure platform” to disclose the authenticity of common misleading rumors and official clarifications, so that market participants have a sufficient factual basis before forming mainstream perceptions.

In the A-share market, the root cause of rumors’ ability to damage valuations lies in the intersection of investors’ “threat sensitivity” and “salience bias” from evolutionary mechanisms, and systematic cognitive biases in behavioral finance. To improve the market’s pricing efficiency, it is necessary to exert pressure simultaneously from multiple dimensions such as fact-checking, emotional self-control, evolutionary bias compensation, and institutional governance. By building a rational decision-making mechanism anchored on facts, shielded by data, and disciplined by processes, we can gradually reduce the impact of rumor-driven irrational trading and achieve a more precise match between price-value relationships and fundamentals.

[1] Xueqiu - “Chen Jiahe: Why Do We Believe Rumors?” (https://xueqiu.com/1340904670/365889265)
[2] Sina Finance - “A Brief Analysis of the Application of Behavioral Finance in Quantitative Investment” (https://finance.sina.com.cn/stock/zqgd/2025-04-24/doc-ineufhcf3801544.shtml?froms=ggmp)