Treasury bond pricing via no arbitrage arguments and machine learning: Evidence from China

This paper proposes a general framework that combines theoretical and empirical asset pricing research and applies the framework to formulate a novel bond return (price or yield curve) prediction methodology, integrating the classical and structural asset pricing theory with recent machine learning factor asset pricing models. The method is mathematically and theoretically rigorous, arbitrage-free, if no arbitrage pricing model is applied, and in the meantime enjoys the flexibility offered by the empirical asset pricing framework, i.e., a potentially rich factor structure, accurate function approximations and the ability to capture both cross-sectional and time-series asset price variation via machine learning regression. Real market backtesting studies show that our predictions are accurate, in the sense that the formulated equally weighted treasury bond portfolios in China exchange-based markets bear significant positive returns and outperforms major methods in the literature.