Climate change, particularly extreme weather events, poses significant challenges to businesses, governments, and stakeholders. News media coverage of these events influences investor sentiment and market behavior. Existing research demonstrates the impact of climate change on asset pricing, showing, for instance, that green stocks often outperform brown stocks when climate concerns rise. However, the predictive power of extreme climate concerns for equity premiums remains an under-researched area, particularly within the context of the Chinese equity market. This study aims to address this gap by constructing an extreme climate concern indicator (ECC) using big data on climate news coverage and exploring its ability to predict equity risk premiums in China. The study's importance lies in its potential to provide valuable insights for investors and policymakers in navigating the financial implications of extreme climate change and informing investment strategies and climate policies.

Prior research highlights the under-reaction of small-cap stocks to climate news and the outperformance of green stocks relative to brown stocks during periods of heightened climate concern. Studies using climate change news indices have shown a relationship between stock returns and climate change shocks. The aversion to climate model uncertainty amplifies the negative price of climate risk. Climate risk significantly impacts the bond market as well, with bonds exhibiting higher climate change news betas demonstrating lower future returns. Short-term weather forecasts and long-term warming trends also affect the prices of financial derivatives. Companies with high climate risk betas earn a risk premium, with stock prices falling as risks materialize. Institutional investors increasingly demand high-quality climate risk information disclosure. Research has also demonstrated that increased extreme heat exposure negatively impacts corporate financial performance. This existing literature provides a foundation for the current study, which examines the specific implications of extreme climate concerns for asset pricing in the Chinese equity market.

The authors construct an ECC indicator using big data from the CSMAR (China Stock Market & Accounting Research) database, analyzing 7,816,916 news items between January 1, 2008, and December 31, 2021. An extensive dictionary of extreme climate keywords (58 keywords such as gale, rainstorm, blizzard, etc.) is employed to measure monthly news coverage of extreme climates. The ECC is calculated using a logarithmic expression to enhance stationarity: ECCₜ = ln(1 + nₜ₊₁) − ln(1 + nₜ), where ECCₜ is the indicator at month t, and nₜ is the number of keywords in news reports in month t. The study employs univariate and bivariate regression models to assess the predictive power of ECC for equity premiums (excess returns of the stock market). The univariate model is R^m_t+1 = α + βZ_t + ε_t+1, where R^m_t+1 is the excess return at t+1, and Z_t is the ECC or an alternative predictor. The bivariate model incorporates both ECC and alternative confidence/economic predictors. Four existing climate risk indices (CPR, CAI, CRI, CRP) are also compared with the ECC. Out-of-sample predictions are conducted using a recursive prediction model to evaluate the timeliness and accuracy of ECC. The economic value of ECC is assessed using mean-variance portfolio optimization, calculating certainty equivalent return (CER) gains and Sharpe ratios. Robustness checks involve using alternative methods to construct the ECC, testing across different industries, stock markets (Shanghai and Shenzhen), high/low concern periods (April-October vs. November-March), and out-of-sample evaluation periods.

The study's key findings are as follows: 1. **Significant Negative Prediction:** The ECC significantly and negatively predicts subsequent monthly stock market returns (in-sample β = -1.815%, R² = 4.012%). This predictive ability is robust across univariate and bivariate models and outperforms several alternative predictors. 2. **Paris Agreement Impact:** The predictive power of ECC is significantly stronger after the Paris Agreement (November 2016), suggesting that increased policy attention to climate change enhances its impact on market sentiment and returns. 3. **Low Concern Period Superiority:** The ECC's predictive accuracy is higher during periods of low climate concern (November-March), implying that mispricing might be more prevalent when investor attention is less focused. 4. **Industry Heterogeneity:** ECC demonstrates significant predictive power for most industry premiums, although the impact varies significantly across sectors, with some industries being more sensitive to climate-related news than others. Specifically, ECC significantly improved return prediction accuracy in the hydropower, construction, IT, real estate, and scientific research industries. 5. **Out-of-sample Performance:** Out-of-sample results confirm the significant negative predictive power of ECC for stock market returns (R²_os = 2.151%), again outperforming many alternative predictors. Combining ECC with other predictors (like BCI, CLI, DY) further enhances the predictive ability. 6. **Economic Value:** ECC generates substantial economic gains for mean-variance investors, with positive CER gains and Sharpe ratios that surpass the benchmark of simply investing in the market itself. 7. **Robustness:** Robustness tests, employing alternative methods (3PRF) to construct the ECC and analyzing various periods and market segments, consistently support the findings.

The study's findings support the hypothesis that extreme climate concerns influence equity premiums in the Chinese stock market. The negative predictive power of ECC suggests that increased concern about extreme climate events leads to decreased stock market returns. The enhanced predictability after the Paris Agreement highlights the role of climate policies in shaping investor sentiment and market dynamics. The greater accuracy during periods of low concern underscores the impact of limited attention bias on market efficiency. The heterogeneous effects across industries reflect sector-specific vulnerabilities to extreme climate events. The out-of-sample and economic value assessments strengthen the overall significance of the findings. The results have implications for investors, policymakers, and researchers, underlining the importance of considering climate risks in investment strategies and policy formulation.

This study makes several key contributions. First, it introduces a novel ECC indicator, adding to the existing literature on climate risk and equity markets. Second, it demonstrates the significant and negative predictive power of ECC for equity returns, identifying it as a pricing factor. Third, it highlights the differential impact of extreme climate concerns before and after the Paris Agreement, revealing the influence of climate policies. Finally, it shows that periods of low climate concern can result in greater mispricing. Future research could explore the interaction of ECC with other financial and economic variables, examine the effects of specific extreme weather events on individual sectors, and investigate cross-country comparisons of climate risk's impact on asset pricing.

The study's limitations include the reliance on news coverage as a proxy for true investor sentiment, the potential for omitted variable bias, and the focus on the Chinese market, limiting the generalizability of the results to other countries with different market structures or climate characteristics. The specific keywords used to define extreme weather events could affect the index and therefore limit the generalizability. Further research should consider these limitations to refine the methodology and broaden the scope of the analysis.