Additive Intensity Regression Models in Corporate Default Analysis

Abstract

This is the accepted version of the paper.This version of the publication may differ from the final published version. Permanent repository link AbstractWe consider additive intensity (Aalen) models as an alternative to the multiplicative intensity (Cox) models for analyzing the default risk of a sample of rated, non-financial U.S. firms. The setting allows for estimating and testing the significance of time-varying effects. We use a variety of model checking techniques to identify misspecifications. In our final model we find evidence of time-variation in the effects of distance-to-default and short-to-long term debt, we identify interactions between distance-to-default and other covariates, and the quick ratio covariate is significant. None of our macroeconomic covariates are significant. JEL Classification: G32, G33, C41, C52

Modeling Frailty Correlated Defaults With Multivariate Latent Factors

“…Our findings of a non-linear effect of the idiosyncratic stock volatility and log market value over total liabilities provide further evidence that the Merton model provides useful guidance for building default models but may need adjustments. As Lando et al (2013), Filipe et al (2016), Jensen et al (2017), we also find strong evidence of a time-varying coefficient for a size measure of the firm. In this regard we note that our size variable differs from the aforementioned papers by using the market value as in Shumway (2001).…”

“…Within the frailty literature it is common to assume that the coefficients for observable variables are constant through time, but Lando et al (2013) show that this assumption is too strict. Using non-parametric and semi-parametric models, they present evidence of non-constant coefficients, and not accounting for such effects may yield biased results and an invalid implied distribution for the default rate of a debt portfolio.…”

“…Using non-parametric and semi-parametric models, they present evidence of non-constant coefficients, and not accounting for such effects may yield biased results and an invalid implied distribution for the default rate of a debt portfolio. However, the models in Lando et al (2013) cannot directly be used for forecasting due to their non-parametric components. In this work we bridge these two approaches by presenting a frailty model that relaxes the assumption of constant coefficients, which is also able to forecast future default rates and properly account for conditional correlation.…”

Can Machine Learning Models Capture Correlations in Corporate Distresses?

“…This paper combines two strings of literature in the field of predicting corporate defaults. The first string focuses on frailty (and/or contagion) or time-varying effects (e.g., see Duffie et al 2009, Koopman et al 2011, Giesecke and Kim 2011, Duan and Fulop 2013, Lando et al 2013, Nickerson and Griffin 2017, Kwon and Lee 2018, Azizpour et al 2018. These papers generally show that models with a simple relationship between observable covariates and distress fail to capture the yearly fluctuations in default rates, i.e.…”

“…Some common macro variables in the existing literature are return of the S&P 500 index, 3-month treasury rate, 10-year treasury rate, inflation, GDP growth, and unemployment rate (e.g., see Duffie et al 2007, Das et al 2007, Duffie et al 2009, Chava et al 2011, Duan et al 2012, Lando et al 2013. We include the Danish equivalent of these variables in our models, except for the stock index return since the majority of firms in our sample are non-traded, and test if the models' predictions improve.…”