Deep Learning for Determinants of Default

Abstract

I identify the determinants of default and the incentives for strategic default using Deep Neural Network (DNN) methodology on proprietary Trepp panel data of commercial mortgages from 1998-2021. My results capture the elements of strategic default during the Great Financial Crisis (2008) and the COVID-19 pandemic (2020-2021). Net Operating Income (NOI), appraisal reduction amount, prepayment penalty clause, balloon payment amongst others co-determine the delinquency behavior in highly nonlinear ways compared to more statistically significant variables such as LTV. DNN results are the most stable even among non-parametric machine learning models. These findings have significant implications for investors, rating agencies and policymakers.

Keywords: Strategic default, CMBS, Machine learning, Stress Test 2008, COVID-19