Abstract
This study investigates the predictive power of sentiments, expressed on X (formerly Twitter), over stock market returns across G7 and BRICS countries, while considering differing degrees of market efficiency. Using the Autoregressive Distributed Lag (ARDL) model, the analysis incorporates 33,946 tweets, aggregated into 120 daily average compound scores, collected between January and June 2022 and processed using the VADER sentiment analyser. The results reveal a mixed landscape: less efficient markets, such as Brazil (Hurst exponent = 0.605), demonstrate stronger sentiment-driven price movements, while more efficient markets like the US and UK (Hurst exponents ≈ 0.5) show weaker but still significant impacts. These findings challenge the Efficient Market Hypothesis by highlighting sentiment's influence in markets across the efficiency spectrum. By demonstrating the utility of social media sentiment analysis for enhancing investment strategies and market predictions, this study contributes to behavioural finance literature and underscores the importance of integrating sentiment as a key factor in stock market dynamics.