1 Advanced Survival Modelling for Consumer Credit

Resumo do artigo ‘Advanced Survival Modelling for Consumer Credit’.1

1.1 Variáveis significativas

  • Applicant Variables (AVs):

    • gênero,
    • estado civil,
    • número de dependentes,
    • local de residência

  • Loan Variables (AVs):

    • valor do empréstimo,
    • valor das parcelas,
    • número de parcelas (tempo para pagamento do empréstimo),
    • local de residência

  • Behavioural Variables (BVs):

    • número de vezes que ficou inadimplente (number of delinquency spells),
    • período médio de inadimplência (average delinquency period) e
    • LTV (ratio of delinquent amount to loan amount)

  • Alone, the model with BVs was very competitive producing an AUROC curve (0.7345) in an acceptable threshold (more than 0.7).

  • This finding corroborate that of Bellotti and Crook, (2013) who found BVs, not only significantly improving model fit but also translating into better forecasting of credit card defaults in United Kingdom.

  • This underscores the fact that understanding the repayment behaviour of the obligors is key to the identification of those likely to recover from delinquency and when it is likely to happen.

Footnotes

  1. Chamboko, Richard (2018). Advanced Survival Modelling for Consumer Credit Risk Assessment: Addressing recurrent events, multiple outcomes and frailty. Thesis.↩︎