Machine Learning Approach to Option Returns

Abstract: We employ machine learning techniques to examine the relationship between expected equity returns and option returns. While any option pricing model predicts a strong link between expected option returns and underlying stock returns, empirical analyses so far are limited due to difficulties in recovering forward-looking equity returns. Recent advances in the asset pricing literature via machine learning, however, offer new avenues to explore this relationship. Leveraging on these techniques we extract a prediction of expected equity returns and uncover an economic puzzle: While we can recover the link in simulation studies, we fail to recover it using the history of available option returns. We rule out that this is due to noise and explore market segmentation as an alternative explanation.