ESG uncertainty, investor attention and stock price crash risk in China: evidence from PVAR model analysis

The paper investigates how ESG rating uncertainty interacts with investor attention and stock price crash risk in China’s A-share market. Stock price crash risk arises when hidden bad news is released abruptly, causing sharp price drops, and investor behavior plays a crucial role in such episodes. ESG, increasingly central to investment decisions, can reduce information asymmetry through disclosure but in China much ESG information is voluntary and ratings from domestic and international agencies often diverge, creating uncertainty for investors. The research question centers on whether ESG uncertainty directly influences stock price crash risk and how investor attention factors into this relationship. The authors posit that ESG uncertainty may attract investor attention, which can in turn magnify market reactions, potentially affecting crash risk. They contribute by constructing a comprehensive ESG Uncertainty Index from six rating agencies and employing a Panel Vector Autoregression (PVAR) framework to capture bidirectional dynamics among ESG uncertainty, investor attention, and crash risk in the Chinese context.

The study synthesizes three strands: (1) ESG and crash risk: Prior work links crash risk to agency problems and information asymmetry, with many studies finding that stronger ESG performance correlates with lower crash risk, though results vary by region, standards, and disclosure quality. ESG rating divergence (uncertainty) could exacerbate asymmetry and weaken signaling, potentially elevating crash risk, but the direct link remains underexplored in China and may be shaped by policy and disclosure practices. (2) ESG and investor attention: High ESG performance and better disclosures tend to attract investor attention; investor attention in turn can enhance ESG practices via external monitoring and engagement. The paper highlights the under-studied role of ESG rating uncertainty in driving attention: inconsistent ratings increase information search and scrutiny (information asymmetry and signaling theories), with voluntary disclosure potentially mitigating uncertainty while also drawing more attention. (3) Investor attention and crash risk: Elevated attention affects returns, volume, and volatility. Evidence (including for China’s retail-dominated market) generally shows a positive association between heightened attention and crash risk via amplified reactions and frictions in price discovery, though regulatory environment and transparency can modulate this relation. The literature lacks an integrated, bidirectional analysis encompassing ESG uncertainty, attention, and crash risk in China.

Data: A-share firms from Shanghai and Shenzhen main boards and GEM over 2011–2021. Exclusions: ST and *ST firms, financial industry firms, and observations with missing data. Variables winsorized at 1% and 99%. Data sources: CSMAR and Wind databases. Model: Panel Vector Autoregression (PVAR) with all variables treated as endogenous, estimated via system GMM using lagged instruments. Fixed effects handled via forward mean differencing (Helmert transformation). Optimal lag length selected by AIC/BIC/HQIC as one lag. Stationarity confirmed using panel ADF and Phillips–Perron tests. Variables: (1) ESG Uncertainty (Uncertainty): Constructed from six rating agencies (Hua Zheng, Bloomberg, Syn Tao Green Finance, Wind, FTSE Russell, MSCI). Agency scores are ranked to percentile [0,1], and the standard deviation across agencies per firm-year forms the uncertainty index. (2) Investor Attention (Attention): Based on Baidu search volumes for stock code, short name, and full name; sum plus one, then natural log to mitigate heteroscedasticity/outliers. (3) Stock Price Crash Risk (NCSKEW): Negative coefficient of skewness of firm-specific weekly returns W, where weekly returns are adjusted for market movements including two leads and lags of market return (Jin–Myers-style non-synchronous trading adjustment). Controls: profitability (net profit/avg net assets), asset structure (fixed assets/total assets), leverage (total liabilities/total assets), and firm age (ln(listing years + 1)). Main analyses: PVAR estimation, Granger causality tests, impulse response functions (Monte Carlo, 200 reps), and forecast error variance decomposition over 10 periods.

Descriptives and correlations: Mean Uncertainty=0.200 (SD=0.106), mean NCSKEW=−0.346 (SD=0.767), mean Attention=6.809 (SD=0.664). Correlations are small in magnitude; Uncertainty–NCSKEW ≈ 0.002 (significant), Uncertainty–Attention ≈ −0.002 (significant), NCSKEW–Attention ≈ −0.015 (significant). PVAR estimates (selected results): • In the Uncertainty equation: lagged NCSKEW (t−1) is not significant; lagged NCSKEW (t−2) is significantly positive; Attention (t−1) and (t−2) are significantly positive. • In the NCSKEW equation: Uncertainty (t−1) and (t−2) are not significant; Attention (t−1) significantly negative; Attention (t−2) significantly positive. • In the Attention equation: Uncertainty (t−1) and (t−2) significantly positive; NCSKEW (t−1) significantly negative; NCSKEW (t−2) not significant. Granger causality: • No causal relation between Uncertainty and NCSKEW in either direction (fail to reject). • Bidirectional Granger causality between Uncertainty and Attention (both directions rejected). • Bidirectional Granger causality between NCSKEW and Attention (both directions rejected). Impulse responses: • Positive shock to Uncertainty increases Attention, strongest initially and persisting. • Positive shock to Attention reduces NCSKEW initially (short-term information/transparency effect), while also increasing Uncertainty over time. • Positive shock to NCSKEW initially reduces Attention, with the negative effect persisting in the short run. Variance decomposition (10 periods): • For NCSKEW, own shocks dominate but decline from 99.8% (period 1) to 93.1% (period 10); contributions of Uncertainty and Attention rise gradually (to about 0.8% and 6.0% by period 10, respectively). • For Attention, own shocks explain ~94.8% at period 1 and ~87.8% at period 10; Uncertainty’s contribution rises to ~10.3% by period 10; NCSKEW’s share remains small (~1.8%). • For Uncertainty, own shocks dominate initially (100% at period 1) but Attention’s contribution grows over horizons (table indicates increasing role of Attention over time). Overall: There is no direct causal link between ESG uncertainty and crash risk; investor attention both responds to and influences ESG uncertainty and crash risk, implying a mediating role for attention in the ESG uncertainty–crash risk nexus.

The findings indicate that ESG rating uncertainty does not directly precipitate stock price crashes in China but significantly interacts with investor attention. Heightened ESG uncertainty attracts investor scrutiny; this attention, while initially improving transparency and potentially dampening crash risk, can over longer horizons intensify information processing frictions and market overreactions, thereby increasing volatility and crash propensity. Conversely, elevated crash risk draws investor attention in the short run but may deter attention subsequently as investors avoid high-risk names. These dynamics are consistent with information asymmetry and signaling theories and reflect the structure of China’s market (high retail participation, evolving ESG standards, and regulatory influences). The PVAR evidence and Granger causality collectively support a mechanism where investor attention mediates and amplifies the market impact of ESG uncertainty, indirectly influencing crash risk rather than establishing a direct causal pathway.

This study contributes by constructing a comprehensive ESG Uncertainty Index using six major agencies and applying a PVAR framework to capture bidirectional dynamics among ESG uncertainty, investor attention, and stock price crash risk in China. The main contribution is the identification of investor attention as a key mediating factor: ESG uncertainty and crash risk each increase attention, and attention, in turn, feeds back to amplify uncertainty and affect crash risk over time. Practical implications include standardizing ESG rating methodologies, enhancing disclosure quality and comparability, and strengthening investor relations and transparency to mitigate uncertainty and stabilize markets. For investors, integrating attention dynamics into risk management and avoiding overreaction to short-term ESG ambiguity can improve decisions. For policymakers, harmonized ESG disclosure requirements and enforcement can reduce uncertainty and attendant volatility. Future research should use more recent and cross-country data, consider industry and size heterogeneity, and employ complementary methods (e.g., case studies, surveys) to isolate ESG-specific attention and uncover micro-mechanisms.

The sample is restricted to Chinese A-share firms (2011–2021), which may limit temporal and cross-market generalizability. PVAR and proxy choices (Baidu searches for attention; NCSKEW for crash risk; constructed ESG uncertainty index) may introduce measurement bias and cannot capture all causal channels. The study does not separate attention specifically directed to ESG information from attention to other fundamentals due to data limitations. Model assumptions (stationarity, instrument validity) and parameter settings also constrain inference. Future work should incorporate more recent data, broader geographies, industry/size splits, and mixed methods to better identify mechanisms and ESG-specific investor attention.