Predicting and distinguishing bankruptcy: an application of a market and hybrid model to US publicly listed firms from 2008 to 2018

• Autor(a) principal: Silva, Bernardo Rui Vaz
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/21449
• topic: Bankruptcy; Credit risk modelling; ROC analysis; KMV Model; CHS Model; Falência; Modelização do risco de crédito; Análise ROC; Modelo KMV; Modelo CHS ii

Assessing the probability of bankruptcy has been a key topic approached by researchers and academics throughout the last half century. The bankruptcy of considerable firms, such as Enron or WorldCom, coupled with the rigorous regulatory environment triggered by Basel II guidelines, fostered even further the interest in the topic. Moreover, in the outcome of financial crisis, Credit Rating Agencies were criticized for addressing inflated ratings and not properly anticipating defaults. Besides, leading CRA’s do not assess the creditworthiness of all firms, and our intention is to provide to individual investor the best option available to autonomously estimate the probability of bankruptcy We analyse if either a market-based model, KMV, or a hybrid model, CHS, are able to properly anticipate the event of bankruptcy, and in case this is verified, which of them better distinguish between bankrupt and non-bankrupt firms. In order to do so, we resort to a sample of 354 US publicly listed firms, divided into bankrupt and non-bankrupt firms, and applied the ROC technique to assess our results, for a 10-year period. Our results prove that KMV model is slightly superior to the CHS model at maximizing the Area Under the Curve (AUC). Besides, it provided a higher optimal probability’s cut off point that distinguish both type of firms. Our results indicate that the KMV model is the best option available for an individual investor to assess the probability of default, given the results achieved and the easiness of application when compared to the CHS model.