Predictive Analysis of Stock Returns using Volume-Weighted Average Returns and Time-Weighted Average Returns

In the pursuit of enhancing stock return predictability, this study explores the predictive capabilities of Volume-Weighted Average Returns (VWAR) and Time-Weighted Average Returns (TWAR). While VWAR and TWAR are traditionally used for trade execution, this research investigates their effectiveness as pre-execution indicators for return forecasting. Employing ten years of daily secondary data from NSE and BSE, various statistical and models SARIMA, and RNN-LSTM—were applied to predict and compare the performance of VWAR and TWAR. Findings consistently demonstrated that models based on VWAR, particularly from BSE over a 10-day horizon, exhibited superior predictive accuracy, lower error rates, and better model fit across all techniques. VWAR data showcased higher volatility persistence and seasonality, making it more suitable for forecasting applications than TWAR. The study highlights the untapped predictive potential of volume-based measures in financial markets, offering practical implications for institutional investors, algorithmic traders, and portfolio managers aiming to optimize return prediction models.