Firm Characteristics of Two-Way Traders: Evidence from Probit Vs. Kernel-Regularized Least Squares Regressions

Firm characteristics in empirical models for margins of international trade usually enter these models in linear form. If non-linearities do matter and are ignored this leads to biased results. Researchers, however, can never be sure that all possible non-linear relationships are taken care of. A solution is provided by Kernel Regularized Least Squares (KRLS) that uses a machine learning approach to learn the functional form from the data. While in earlier applications the big picture revealed by standard empirical models and KRLS was identical this note presents a case where results from a standard approach and KRLS do differ considerably.