Variable selection in proportional hazards cure model with time-varying covariates, application to US bank failures^*

* An oral presentation of this paper was given at the 10th International Conference of the ERCIM WG on Computational and Methodological Statistics, Senate House, University of London, UK, 16–18 December 2017

ABSTRACT

From a survival analysis perspective, bank failure data are often characterized by small default rates and heavy censoring. This empirical evidence can be explained by the existence of a subpopulation of banks likely immune from bankruptcy. In this regard, we use a mixture cure model to separate the factors with an influence on the susceptibility to default from the ones affecting the survival time of susceptible banks. In this paper, we extend a semi-parametric proportional hazards cure model to time-varying covariates and we propose a variable selection technique based on its penalized likelihood. By means of a simulation study, we show how this technique performs reasonably well. Finally, we illustrate an application to commercial bank failures in the United States over the period 2006–2016.

KEYWORDS:

• Bank failures
• survival analysis
• mixture cure model
• time-varying covariates
• penalized likelihood
• SCAD

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Alessandro Beretta  http://orcid.org/0000-0002-0427-8785

Notes

* An oral presentation of this paper was given at the 10th International Conference of the ERCIM WG on Computational and Methodological Statistics, Senate House, University of London, UK, 16–18 December 2017

1 Data are available through the FDIC Research Information System database, in the Statistics on Depository Institution (SDI) reports, which are based on the regulatory Call Reports filled by banks. Downloadable from the website https://www5.fdic.gov/sdi.

2 Conducted with a code developed by the authors, using C++ and the Armadillo library of Sanderson and Curtin [35]. Computational resources have been provided by the Consortium des Équipements de Calcul Intensif (CÉCI), funded by the Fonds de la Recherche Scientifique de Belgique (F.R.S.-FNRS) under Grant No. 2.5020.11.