Autoencoder asset pricing models and economic restrictions - international evidence

We evaluate the performance of the Conditional Autoencoder (CAE) model by Gu et al. (2021) in an international context and under economic constraints, such as the exclusion of microcap and illiquid firms, and accounting for transaction costs. The CAE model leverages latent factors and factor exposures dependent on asset characteristics, modelled as a flexible nonlinear function while adhering to the noarbitrage condition. The original study showed significant reductions in out-of-sample pricing errors from both statistical and economic perspectives in the U.S. context. We replicate these results on the U.S. dataset and extend the analysis to international data with a different set of firm characteristics, achieving consistent outcomes that demonstrate the model's robustness. However, the economic benefits after accounting for transaction costs are limited, even after the exclusion of illiquid firms, highlighting the importance of considering these constraints.