Discovering Fundamental Value

Abstract

Using machine learning methods we identify the fundamental value and noise components of quarterly stock prices. We show that 28% of stock price variation is attributable to noise, and that 40% of noise is attributable to mutual fund trading. We find spikes in noise around the dot-com bubble, the 2008 financial crisis, and the European sovereign-debt crisis. Noise is higher for small firms and firms with high R&D expenditures. In an application of our methodology, we show that corporate managers do not have private information about future changes in fundamental value nor can they identify noise in prices.

Keywords: Machine Learning, Fundamental Value, Noise

JEL Classification: G20, G14, L10