Credit Risk and Default Probability estimation: A study on structural, reduced-form and credit scoring models with applications to Apple Inc.

Abstract

The thesis explores the ever-present topic of corporate credit risk through the analysis of both structural and reduced-form models, casting light on their merits and drawbacks. The study begins with the pioneering work of Merton and its subsequent adaptation to real-world corporate aspects, like the cost of guarantees, interim payments, bankruptcy costs, taxes and so on. Mainly, the Leland model and the KMV model are inspected as more realistic variants. The thesis also presents the fundamentals of reduced-form models and distinguishes their probabilistic approach from credit-scoring models, Altman’s model and logit regression. Lastly, LASSO is presented as a technique that can select a conclusive set of bankruptcy predictors, determining the importance of both accounting-based and market-based variables to forecast default. This result fits into the debate following Altman’s model on whether accounting data have better discriminatory power than market data. The thesis’ empirical investigation centers around the application of credit risk models on Apple Inc. The goal is to produce an estimate of the company’s one-year default probability by combining adjustments of the Merton’s model, the Altman’s model and the LASSO regression. By comparing the results obtained, we propose a quantitative study on credit risk and on its practical implications.