The influence that investor sentiment can have on market movements is often overlooked by traditional financial models. This column analyses nearly three million stock-related tweets to investigate whether sentiment derived from such tweets can predict intraday stock market fluctuations. The findings suggest that tweet-based sentiment strongly predicts market trends in both developed and emerging markets. Recognising traders’ emotions has implications for traders, analysts, and regulators seeking to anticipate and interpret market behaviour.
Research on the accurate prediction of stock market movements is of interest to academics, economists, and financial analysts due to the profitability of accurately predicting the markets. The intersection of behavioural finance and Big Data has therefore become increasingly relevant in financial market analysis. Events such as the GameStop short squeeze in 2021 demonstrate the power of social media in influencing stock prices. This phenomenon aligns with behavioural finance theories that emphasise the role of investor sentiment in asset pricing (Baker and Wurgler 2006). While traditional asset-pricing models assume rational decision-making, behavioural biases and herd mentality often drive real-world trading decisions (De Long et al. 1990).
Previous research has explored investor sentiment using social media platforms such as Twitter (now X). Antweiler and Frank (2004) found that online messages contain valuable predictive information. Similarly, Bollen et al. (2011) demonstrated that Twitter mood correlates with stock market fluctuations. Van Wincoop and Gholampour (2017) considered the sentiment of opinionated tweets in predicting the euro-dollar exchange rate. Our study (Greyling and Rossouw 2022) builds on this research by analysing sentiment and emotions extracted from stock-related tweets to predict intraday market movements across multiple stock exchanges.
Analysing tweet sentiments and intraday market movements
We employ high-frequency intraday data covering both developed markets (France, Germany, Japan, Spain, UK, US) and emerging markets (India, Poland) and nearly three million stock-related tweets over a year. This permits us to derive investor sentiment from a set of global tweets (adjusted for time differences), allowing us to analyse the responsiveness (reactions) of both emerging and developed markets to the same set of tweets. We thus analyse the movements in developed and emerging markets simultaneously.
Our analyses use different machine learning classification algorithms – including naïve Bayes, K-nearest neighbours, and support vector machines – to classify market movements based on sentiment and emotion indicators.
Figure 1 illustrates the process of predicting market movements based on sentiment and emotions derived from extracted stock market-related tweets.
Figure 1 Systematic process of predicting market movements from tweet sentiments
Source: Authors’ compilation.
Findings: Social media sentiment predicts stock market movements
Our main results suggest that a keyword like “stock market” can accurately predict and explain the movements of stock markets in developed and emerging markets, with similar prediction accuracy shown in these markets. The sentiment and emotions contained in tweets appear to be significant predictors of stock market movements not only in a single market but also in multiple markets in developed and emerging markets. The accuracy of our prediction models exceeded 50% in all markets analysed, indicating that social media sentiment contains valuable information for traders and analysts.
For developed markets, we find that sentiment analysis alone produced strong predictive results. In the US market (S&P 500), machine learning models achieved an accuracy of over 55%, with recall measures reaching 92% in some cases. Conversely, we find that for emerging markets, a combined sentiment and emotion analysis of tweets predicted market movements better than sentiment alone. This suggests that investor emotions play a more prominent role in less mature markets.
Overall, we conclude that the most predictive emotions varied from stock market to stock market. However, the emotions found to be significant in predicting market movements most often were ‘fear’ and ‘trust’, which align with behavioural finance theory such as the prospect theory, for example (Kahneman and Tversky 1979).
Implications for policymakers and financial market participants
Our findings highlight the growing importance of social media sentiment in predicting stock market movements, with implications for both financial market participants and policymakers.
Sentiment shifts captured from platforms like Twitter can provide early warning signs of market disruptions, such as sudden sell-offs or speculative bubbles driven by viral posts. Regulators may therefore want to consider incorporating social media sentiment analysis into their financial stability assessments.
The rise of social media-driven market movements also calls for enhanced vigilance against potential market manipulation or misinformation campaigns. Regulators could develop frameworks to monitor and detect coordinated efforts to influence stock prices via social media, protecting investors from market distortions. To achieve this, regulators could use AI-driven platforms to track social media trends, while natural language processing models could help identify shifts in sentiment at scale.
Institutional investors can leverage sentiment analysis to refine their portfolio management strategies. By incorporating real-time social media sentiment into traditional financial models, they can gain a more comprehensive view of market trends and sentiment shifts before they materialise in stock prices. Sentiment analysis can also be used as an additional risk management tool, helping hedge funds and asset managers identify emerging risks before they impact asset valuations.
Retail traders may use real-time sentiment indicators derived from social media to inform their trading decisions. By tracking sentiment-driven movements, individual traders can anticipate shifts in stock momentum and make more informed buy or sell decisions. The accessibility of sentiment analysis tools has the potential to level the playing field between retail and institutional investors, offering retail traders a way to gauge market psychology alongside traditional technical and fundamental analysis.
In conclusion, our research underscores the predictive power of Twitter sentiment and emotions in stock market movements. We provide a novel approach to understanding market dynamics by integrating behavioural finance with Big Data analytics. Sentiment analysis of social media should become a key component of trading strategies and the refinement of models to predict market movements. As financial markets evolve, so too must the tools we use to inform trade and policy decisions.
Source : VOXeu
