Return Predictability of a Corporate Governance Neural Networks Trading System

Abstract

This article examines whether a corporate governance investment strategy, within the context of a full-fledged trading system, can generate economically significant returns. Using artificial neural networks to design buy/sell rules, we address the limitation in corporate governance literature by incorporating money management and risk control strategies in the trading system. In addition, the simulation considers realistic constraints which we observed also lacking in the literature, namely capital restriction, round lot, short selling and transaction costs. The trading system is developed in sample for out of sample forecasting, and we measure its performance using several performance metrics including the Sharpe and Sortino ratios. The overall results indicate superior performance of the corporate governance trading system compared to the benchmark buy-and-hold strategy, with significantly better returns and much greater Sharpe and Sortino ratios. This suggests that the market is inefficient at the semi-strong form. Consequently, this study has implications for public policy and the accuracy of previous studies.