A1695
Title: Nuclear norm penalised non-linear modelling for asset pricing
Authors: Aristeidis Raftapostolos - Kings College London (United Kingdom) [presenting]
Ilias Chronopoulos - University of Essex (United Kingdom)
George Kapetanios - Kings College London (United Kingdom)
Abstract: A nonlinear and non-parametric estimator for interactive fixed-effect panel regressions is proposed. Further, we extend the nuclear norm penalized estimation in a nonlinear setup using neural networks and apply our procedure to a large factor data library containing risk factors discovered in the literature in the last 30 years. Our estimator achieves robust out-of-sample performance. Most importantly, we move elements of model interpretability and explainability to the foreground.
