Does Big Data Improve Price Discovery in Frontier Capital Markets? Evidence from African Frontier Exchanges Using Econometric and Machine Learning Models

Abstract

This study examines whether big data analytics improves price discovery in selected African frontier capital markets, using evidence from Nigeria, Ghana, Kenya, Rwanda, and Zambia. The study is motivated by the growing use of alternative digital datasets, artificial intelligence, and machine learning in global financial markets, alongside persistent informational inefficiencies in African frontier exchanges. An explanatory longitudinal time-series design was adopted, using daily data from Bloomberg, Refinitiv, Yahoo Finance, Google Trends, Twitter/X sentiment analytics, and financial news databases for the period 2015 to 2025. The econometric analysis employed Augmented Dickey-Fuller, Phillips-Perron, Johansen cointegration, Vector Error Correction Model, Hasbrouck Information Share, and Gonzalo-Granger Component Share techniques. The machine learning framework compared ARIMA, Random Forest, XGBoost, and Long Short-Term Memory models. The findings show significant long-run relationships between traditional market variables and big data indicators. Google Trends, Twitter/X sentiment, and news analytics improved price discovery efficiency and accelerated the incorporation of information into stock prices. Kenya and Nigeria showed stronger informational efficiency than Ghana, Rwanda, and Zambia. The machine learning results also indicate that LSTM and XGBoost outperformed ARIMA across the selected prediction metrics. The study concludes that big data analytics and machine learning can strengthen price discovery in African frontier capital markets by improving information processing, predictive accuracy, and market responsiveness.