The Implied Cost of Capital: A Deep Learning Approach

I exploit deep learning techniques trained on a set of common accounting items and constructed to mimic features of the human brain to predict future earnings. I show that this model offers incremental explanatory power in predicting future earnings and in estimating the associated implied cost of capital. My forecasting model exhibits less bias than human analyst forecasts and fits the data substantially better than linear regression models. In addition, the derived implied cost-of-capital estimates substantially outperform linear models in their ability to predict future returns. This study illustrates the power of machine learning techniques to improve the accuracy of accounting forecasting.