From Tweets to Trades: The Dynamic Dance of Investor Sentiment, Attention, and News Sentiment in ESG Stocks

2.1 Information Cascade and Signalling Theories

The Information Cascade Theory has been instrumental in understanding how investors, particularly in financial markets, make decisions based on the observed actions of others rather than their private information. Bikhchandani et al. (1992) established the foundational idea that in environments with high information asymmetry, investors may follow the crowd, assuming that others possess superior knowledge, leading to cascades that can cause significant market distortions. Duz Tan and Tas (2021) found that social media sentiment amplifies these cascades, particularly in international markets, by quickly spreading information (or misinformation), which can lead to irrational herding behaviour and increased market volatility. Doherty (2018) further highlighted how these cascades contribute to increased market volatility, especially in ESG-focused investments where investor decisions are often driven by perceived rather than intrinsic value.

Signalling Theory offers insights into how companies communicate their intrinsic quality and intentions through specific actions, such as ESG disclosures or financial decisions. The signals must be credible and costly to be effective, as highlighted by Connelly et al. (2011). The critical question today, especially in ESG contexts, is whether these signals are genuinely informative or merely symbolic gestures that appease stakeholders. Fu et al. (2022) explored this issue and suggested that broad or narrow stakeholder management strategies significantly affect the perceived credibility of ESG signals. Keleş et al. (2023) also highlighted the impact of corporate social responsibility (CSR) news on stock performance, underscoring the role of signalling in shaping investor expectations and market reactions.

When companies issue strong signals through ESG disclosures, they can initiate information cascades, particularly when they are amplified by media coverage. Barberis et al. (2020) discussed how investor reactions to these signals could lead to market overreactions or underreactions depending on the context and credibility of the information. The role of news sentiment in this process is critical, as noted by Tetlock (2007) and Smales (2021), in which media narratives can either stabilise or destabilise markets by influencing the direction and strength of these cascades. Cerqueti et al. (2021) added that ESG investing might also reduce systemic risk, particularly when signals are correctly interpreted and cascades are managed effectively. This study builds on these theories by examining how these dynamics specifically impact the Asia-Pacific ESG markets.

2.2 Investor Sentiment, Attention and News Sentiment

The effects of investor sentiment are particularly pronounced in markets such as China, India, and Singapore, where financial systems are still evolving and often exhibit higher levels of information asymmetry (Bouattour et al., 2024). Baker and Wurgler (2007) highlighted how shifts in investor sentiment can lead to mispricing and increased market volatility, particularly in markets with less developed institutional frameworks. Schmeling (2009) further corroborated these findings by showing that the impact of sentiment on stock returns is more significant in countries with lower market efficiency. Zouaoui et al. (2011) extended this line of research by demonstrating that elevated investor sentiment contributes to market instability, particularly in emerging markets, where regulatory oversight is less stringent. In the context of ESG-compliant stocks, recent studies by Sabbaghi (2023) and Tang et al. (2024) suggested that positive sentiment towards ESG factors can stabilise markets by reducing volatility, whereas negative sentiment exacerbates market fluctuations, especially in environments where ESG integration is still uneven. Given these dynamics, it is hypothesised that:

Hypothesis 1: Investor sentiment significantly impacts trading volume and market volatility in ESG-compliant stocks in China, India, and Singapore, with positive sentiment increasing trading volumes, reducing volatility, decreasing trading volumes, and increasing volatility.

Investor attention, driven by factors such as media coverage, social media activity, and public discourse, also plays a critical role in shaping market behaviour, particularly in less mature markets. Barber and Odean (2008) provided early evidence that stocks receiving heightened media attention tend to experience surges in trading volume as investors are more likely to act on widely disseminated information. This relationship was further explored by Da et al. (2011), who introduced the concept of the Google Search Volume Index and showed that increased search activity correlates with higher trading volumes and more pronounced price movements. Fang and Peress (2009) added to this by demonstrating that media coverage alone can explain significant variations in trading volume, particularly in less liquid markets. In the context of ESG investments, Khan et al. (2016) argued that heightened attention to sustainability issues often leads to increased trading volumes, as these stocks become focal points for both institutional and retail investors. Wan et al. (2024) further explored this dynamic by examining return and volatility connectedness across global ESG stock indices, finding that investor attention significantly influences these relationships. However, Tetlock (2007) and Smales (2021) pointed out that the volatility effects of investor attention are amplified when attention is driven by speculative news or rumours rather than fundamental analysis, leading to erratic market behaviour. This evidence suggests the following:

Hypothesis 2: Investor attention significantly influences trading volume and market volatility in ESG-compliant stocks in China, India, and Singapore, with higher attention increasing trading volume and volatility and lower attention decreasing trading volume and volatility.

News sentiment serves as a crucial moderating factor shaping the relationship between investor sentiment, attention, and market dynamics. Tetlock (2007) was among the first to quantify the impact of news sentiment on stock prices, showing that negative news tends to lead to stock price declines, while positive news bolsters investor confidence. Garcia (2013) expanded on this by demonstrating that the tone of media coverage plays a critical role in driving investor behaviour, especially during periods of economic uncertainty. Sabbaghi (2022) also demonstrated that news significantly affects the volatility of ESG firms, underscoring the importance of news sentiment as a key determinant of market volatility. Smales (2021) provided additional evidence that news sentiment has a more pronounced effect in markets with higher levels of information asymmetry, leading to more significant market movements in response to news events. This underscores the need to consider news sentiment when analysing market behaviour, leading to the hypothesis that:

Hypothesis 3: News sentiment moderates the relationship between investor sentiment and market dynamics in China, India, and Singapore, with positive news sentiment strengthening the impact of positive investor sentiment on trading volume and negative news sentiment amplifying the impact of negative investor sentiment on market volatility.

The differential impact of these factors on high versus low ESG-compliant stocks warrants particular attention as investors often perceive these stocks differently based on their ESG performance. Eccles et al. (2014) found that companies with strong ESG performance generally exhibit lower volatility and more stable returns, as they are perceived as lower-risk investments. This finding is consistent with Loang (2023), who showed that higher ESG ratings are associated with reduced risk and lower cost of capital, contributing to the stability of these stocks. By contrast, low ESG-compliant stocks may be viewed as more volatile and susceptible to market fluctuations, especially in markets where ESG standards are not well enforced, as noted by Khan et al. (2016). This variability in investor perceptions and market reactions leads to the following hypothesis:

Hypothesis 4: The interaction between investor sentiment, investor attention, and news sentiment has a more pronounced effect on trading volumes and volatility in low ESG-compliant stocks than in high ESG-compliant stocks in China, India, and Singapore.

3.1 Data and Sampling

This study spans the period from 2018 to 2023, focusing on listed companies in China, India, and Singapore. These markets were selected because of their economic significance and varying levels of ESG adoption, which makes them ideal for comparative analysis. A total of 5,054 companies were analysed, with 2,837 from China, 1,722 from India, and 495 from Singapore. ESG scores were categorised into quartiles, with the first quartile (scores ranging from 0 to 25) being classified as having low ESG scores. This categorisation allows for a detailed examination of the impact of ESG performance on market behaviour.

Investor sentiment and attention were measured using a substantial dataset comprising 5,634,742 tweets and 2,456,783 news articles. Tweets were collected via Twitter API, focusing on posts, comments, and engagement metrics related to the selected companies and relevant financial terms. Table 1 presents the variables, their descriptions, and corresponding data sources used in this study.

3.2 Investor Attention

Investor attention was quantified using the investor attention index, which has been previously employed to capture the level of engagement and focus investors place on companies via Twitter. This method aggregates three primary metrics—published posts, comments, and reading numbers (views)—which collectively represent varying levels of investor interaction and interest. To ensure the robustness of this measure, the weights for these components were determined based on a combination of empirical analysis and expert validation. Specifically, published posts were given the highest weight (0.6), as they are direct indicators of engagement; comments were weighted at 0.3 due to their role in reflecting more thoughtful interaction, and reading numbers were assigned a weight of 0.1, reflecting passive but broader engagement. These weights were calibrated to accurately reflect the importance of each type of interaction in measuring investor attention. The resulting Investor Attention Index (IAI) is calculated using the following formula:

where, IAI_i,t represents the Investor Attention Index for company i at time t, Post_i,t is the number of published posts, Comment_i,t is the number of comments, and Read_i,t is the reading number for company i at time t.

3.3 Garman and Klass Volatility Model

The Garman and Klass Volatility Model is a widely recognised method used to estimate the volatility of financial assets by incorporating detailed intraday price movements, specifically the opening, high, low, and closing prices on a trading day.

where σ is the GK estimator, N is the number of trading days in the period, h_i represents the highest stock price on day i, and l_i is the highest stock price on day i, C_i,t is the closing price of stock i and O_i,t is the opening price of stock i.

3.4 Panel Data Regression Model

All variables in this study were collected on a monthly basis because of data availability across markets. The use of monthly data helps smooth out short-term volatility and avoids noise that can distort the relationships between variables, providing a clearer analysis of trends over time. In addition, the study employs rigorous econometric techniques, including quantile-on-quantile analysis, Granger causality tests, and FGLS regression, to address potential issues related to non-stationary variables and ensure the robustness of the findings. These methods, along with robustness tests, help mitigate the risk of spurious regression results by capturing both short- and long-term relationships between variables. The models used in this study are specified as follows:

where Volume_i,t represents the trading volume for company i at time t, while GK_i,t denotes the Garman-Klass model. E_i,t, S_i,t, and G_i,t scores for environment, social and governance score of company i at time t. Additionally, Sent_i,t captures the investor sentiment, and Att_i,t reflects the investor attention directed towards company i at time t. NSent_i,t represents the news sentiment related to company i at time t. The control variables include MC_i,t (Market Capitalization), Lev_i,t (Leverage Ratio), DY_i,t (Dividend Yield), and ROA_i,t (Return on Assets), which account for various financial and operational characteristics of the companies being studied. All VIF values are well below the commonly accepted threshold of 5, indicating that multicollinearity is not a significant concern in the regression models.

5.1 Investor Sentiment and Attention on Trading Volume

Table 3 presents the results of the fixed-effect panel regression analysis examining the impact of Investor Sentiment (Sent), Investor Attention (Att), and News Sentiment (NSent) on trading volumes in China, India, and Singapore across the four models.

In China, the results show that Sent significantly increases trading volume, with a coefficient of 0.112 in Model 1, reflecting strong investor confidence that typically leads to heightened trading activities. This effect is further supported by Att, which also positively influences trading volume with a coefficient of 0.081 in Model 2. The introduction of NSent in Model 3 shows an additional positive effect on trading volume with a coefficient of 0.115, suggesting that favourable news coverage can further fuel trading activities, likely because positive news enhances investor optimism, leading to increased buying activity. Model 4, which includes interaction terms, reveals that the interaction between investor sentiment and news sentiment has a significant synergistic effect at the 5% level, where the co-occurrence of positive investor sentiment and positive news sentiment significantly boosts trading volume. Additionally, the E, S, and G factors in China are all significant, with the environmental component showing the highest impact in Model 4 (0.098, p < 0.05), indicating the increasing importance of ESG considerations in influencing trading behaviour.

In the Indian context, while the effects of Sent and Att are significant, they are weaker than those of China. Sent shows a coefficient of 0.077 in Model 1, and Att has a coefficient of 0.065 in Model 2, indicating that, while sentiment and attention drive trading volume, the impact is more moderate. The NSent in Model 3 also has a positive impact on trading volume, but the effect size is smaller than that in China, suggesting that news sentiment plays a role, but perhaps not as dominantly. The interaction effects in Model 4, although significant, are less pronounced. Furthermore, E, S, and G factors are all found to be insignificant in India across all models. This could be attributed to the evolving nature of India’s financial markets, where regulatory frameworks and investor sophistication are still developing, leading to tempered reactions to sentiment-driven factors.

Singapore’s results highlight the most substantial impact of Sent and Att on trading volume among the three countries analysed. Sent in Model 1 shows a high coefficient of 0.148, and Att in Model 2 exhibits a coefficient of 0.138, both significant at the 1% level. NSent also shows a strong effect on trading volume in Model 3, with a coefficient of 0.155, indicating that news sentiment plays a critical role in driving Singapore’s market behaviour. The interaction terms in Model 4 reveal that the combined influence of Sent and NSent, as well as Att and NSent, significantly increases trading volumes.

5.2 Investor Sentiment and Attention on Volatility

Table 4 presents the results of the fixed-effects panel regression models that evaluate the impact of Investor Sentiment and Investor Attention on Volatility (GK) across China, India, and Singapore. In China, Sent significantly influences GK in all models, with coefficients ranging from 0.117 to 0.129 at the 1% significance level, underscoring a strong relationship between investor sentiment and market volatility. The interaction term between Sent and NSent also shows a significant positive effect, particularly in Model 4 (0.129), indicating that news sentiment amplifies investors’ impact on volatility. This aligns with previous findings by Wan et al. (2024), who suggest that emerging markets such as China are particularly sensitive to sentiment-driven volatility.

In India, Sent also significantly impacts GK, but the coefficients are lower than those in China, ranging from 0.022 to 0.029. The interaction effect between Sent and NSent is significant in Model 4 (0.022), although the magnitude is smaller than that in China. However, the E, S, and G factors are insignificant in India, indicating that ESG considerations have less influence on market volatility in this context.

In Singapore, Sent is significant across all models, with coefficients between 0.144 and 0.154 (p < 0.01), highlighting the consistent impact of investor sentiment on volatility in developed markets. The interaction term is also significant, suggesting that news sentiment plays a moderating role, although the impact is less pronounced than that in China, reflecting the more stable financial environment in Singapore.

5.3 Volume and Volatility in High and Low ESG Groups

Table 5 shows the regression analysis for Volume and Volatility across different ESG value groups aggregated from China, India, and Singapore. The findings demonstrate that, in high ESG groups, Sent and Att exert a more substantial impact on both trading volume and volatility. In the high ESG group, Sent has coefficients ranging from 0.006 to 0.007 in Models 5 to 8, and Att coefficients range from 0.005 to 0.006, all statistically significant at the 1% level. These results are consistent with findings from recent studies, such as Sabbaghi (2023) and Wan et al. (2024), which highlight that firms with higher ESG performance tend to experience more pronounced market reactions to investor sentiment because of their enhanced reputational capital.

In contrast, the low ESG group exhibits weaker relationships between Sent, Att, and market dynamics, with Sent coefficients between 0.002 and 0.004, and lower statistical significance. This suggests that low-ESG firms are less responsive to changes in investor sentiment and attention, potentially because of their perceived higher risk or lower investor confidence. The interaction terms between Sent and NSent and Att and NSent also show stronger effects in high ESG groups, particularly on volatility, further emphasising the role of ESG factors in amplifying market reactions to sentiment.

5.4 Quantile-on-Quantile Analysis of ESG

The Quantile-on-Quantile (QQ) relationship (please refer to the Appendix on the Journal’s website) between aggregate ESG scores and two crucial financial metrics: trading volume and volatility, shows different pattern. As ESG scores increase, particularly in higher quantiles, trading volumes show a marked increase, suggesting heightened investor activity. This pattern indicates that investors may perceive high-ESG firms as more attractive, likely due to perceived lower risk or higher ethical standards, aligning with the signalling theory that suggests that firms with strong ESG performance signal long-term sustainability and stability in the market.

Conversely, it shows a clear inverse relationship between ESG scores and volatility. Firms with stronger ESG performance tend to experience lower price volatility, which indicates that robust ESG practices contribute to greater market stability. This supports the notion that ESG integration can act as a mitigating factor against market turbulence, potentially because of the more stable investor base and better risk management practices associated with high-ESG firms.

5.5 Robustness Test: Granger Causality and FGLS Regression

Table 6 presents the Granger causality test, which examines the causal relationships between variables. The findings indicate that Investor Sentiment and Investor Attention significantly Granger-cause both Volume and Volatility, demonstrating strong predictive relationships. Specifically, Investor Sentiment shows a significant causal effect on volume with an F-statistic of 4.521 (p = 0.004), and on volatility with an F-statistic of 3.945 (p = 0.050). Similarly, Investor Attention exhibits significant causality, with F-statistics of 5.103 (p = 0.002) for volume and 5.312 (p = 0.012) for volatility. News Sentiment also Granger causes both Volume and Volatility, with F-statistics of 4.873 (p = 0.003) for volume and 4.729 (p = 0.003) for volatility, indicating that media coverage plays a crucial role in shaping market behaviour. Notably, the results indicate that trading Volume and Volatility do not Granger-cause Investor Sentiment, Investor Attention, or News Sentiment, suggesting that these sentiment and attention metrics are exogenous to market movement.

Table 7 presents the FGLS regression results, addressing heteroskedasticity in the analysis of the impact of different news sentiments (positive, neutral, and negative) on trading volumes and volatility. The results indicate that Sent and Att exert a more substantial influence under positive and negative news conditions than under neutral news. Specifically, Sent demonstrates higher coefficients for trading volume and volatility under positive (0.425 and 0.278) and negative (0.452 and 0.289) news, all significant at the 1% level. This finding suggests that extreme news, whether optimistic or pessimistic, elicits more intense market reactions, leading to greater fluctuations in both trading activities and price volatility.

Att also shows a more pronounced impact under positive and negative news, with coefficients of 0.401 for volume and 0.267 for volatility in the negative news scenario significant at the 1% level. The interaction effects between Sent and NSent and the interaction effect between Att and NSent further amplify these impacts, especially under negative news conditions, where these interactions are statistically significant at the 1% level.

6.1 Implications: Theoretical, Managerial and Policy

This study offers several key implications that align with information cascade and signalling theories. In markets with high ESG values, the pronounced effect of investor sentiment suggests that positive ESG performance serves as a strong signal to investors, amplifying their confidence and influencing their trading volumes and volatility. This aligns with signalling theory, where companies with superior ESG performance send positive signals to the market, attracting more investment. Furthermore, the role of news sentiment as a moderator illustrates how the information cascade theory operates, where initial news reports can create a cascade effect that influences subsequent market reactions, particularly in volatile markets.

Managerially, firms with high ESG scores must be strategic in their communication efforts, recognising that their actions and disclosures can trigger significant market reactions. They should consider the potential cascading effects of news and investor sentiment to ensure that their ESG practices are communicated effectively to sustain investor confidence and mitigate market volatility. Policymakers should take concrete action by establishing mandatory ESG reporting standards that require companies to disclose consistent, detailed, and comparable ESG data across industries. This reduces information asymmetry and helps prevent information cascades that can lead to market volatility. Additionally, introducing robust auditing and verification mechanisms for ESG disclosures can ensure the credibility of the information provided, reduce the risk of greenwashing, and foster greater investor trust. Furthermore, incorporating ESG metrics into national and regional risk assessment frameworks would enable more accurate evaluations of long-term financial stability, helping markets respond better to ESG-related risks.

6.2 Limitations and Recommendations

This study is constrained by its reliance on secondary data, which may overlook the qualitative nuances critical for understanding market behaviours influenced by investor sentiment and ESG factors. The use of historical data limits the ability to capture real-time shifts, and a mix of stationary and non-stationary variables can result in spurious regression outcomes. Future research should address this by applying cointegration models or vector error correction models (VECM) to ensure robust long-term relationships between variables. While this study focused on China, Singapore, and India, comparisons with markets such as the US and EU, where ESG frameworks are more developed, would offer broader insights into how regulatory environments shape investor behaviour.