We apply a selection of machine learning models to the task of forecasting future firm earnings. While estimates from financial analysts present the highest forecast accuracy, machine learning models are superior in terms of forecast bias and earnings response coefficient. We find models based on regression trees to perform best among the machine learning methods. The earnings forecasts are then used to estimate the implied cost of capital (ICC) which is the internal rate of return that equates the current stock price to the present values of future cash-flows to the shareholders. In our analysis, the ICC computed on the earnings forecasts from penalized linear models and tree-based methods show the strongest relationship with future realized returns.     «

We apply a selection of machine learning models to the task of forecasting future firm earnings. While estimates from financial analysts present the highest forecast accuracy, machine learning models are superior in terms of forecast bias and earnings response coefficient. We find models based on regression trees to perform best among the machine learning methods. The earnings forecasts are then used to estimate the implied cost of capital (ICC) which is the internal rate of return that equates...     »