Academic journal article Journal of Risk and Insurance

A Comparison of Neural Network, Statistical Methods, and Variable Choice for Life Insurers' Financial Distress Prediction

Academic journal article Journal of Risk and Insurance

A Comparison of Neural Network, Statistical Methods, and Variable Choice for Life Insurers' Financial Distress Prediction

Article excerpt

ABSTRACT

This study examines the effect of the statistical/mathematical model selected and the variable set considered on the ability to identify financially troubled life insurers. Models considered are two artificial neural network methods (back-propagation and learning vector quantization (LVQ)) and two more standard statistical methods (multiple discriminant analysis and logistic regression analysis). The variable sets considered are the insurance regulatory information system (IRIS) variables, the financial analysis solvency tracking (FAST) variables, and Texas early warning information system (EWIS) variables, and a data set consisting of twenty-two variables selected by us in conjunction with the research staff at TDI and a review of the insolvency prediction literature. The results show that the back-propagation (BP) and LVQ outperform the traditional statistical approaches for all four variable sets with a consistent superiority across the two different evaluation criteria (total misclassification cost and resubstitution risk criteria), and that the twenty-two variables and the Texas EWIS variable sets are more efficient than the IRIS and the FAST variable sets for identification of financially troubled life insurers in most comparisons.

INTRODUCTION

Insurance company insolvency produces substantial losses to many stakeholders, and the identification of financially troubled firms is a major regulatory objective. Accordingly, there is a strong regulatory need for accurate prediction methods to signal financially impaired insurers in sufficient time to allow action to be taken to prevent insolvency or to minimize its cost.

In the context of warning of pending insurer insolvency, there are several sources of information available. These include, for example, the A. M. Best and other rating agency reports. In addition, the National Association of Insurance Commissioners (NAIC) developed the insurance regulatory information system (IRIS) and, following the extremely costly First Executive Life Insurance Company bankruptcy, the financial analysis solvency tracking (FAST) system to provide an early warning system. The NAIC also adopted risk-based capital (RBC) formula for insurance insolvency prediction. Some states have developed their own early warning systems. For example, the Texas Department of Insurance (TDI) implemented an early warning information system (EWIS) in early 1992 based upon their own model and data set.

There is substantial previous literature on insurer insolvency prediction: Barrese (1990) evaluated the adequacy of IRIS; Cummins, Harrington, and Klein (1995), and Grace, Harrington, and Klein (1998) provided evaluations concerning the accuracy of the RBC and FAST systems; and Cummins, Grace, and Phillips (1999) compared RBC and FAST using cash flow simulation.

Multivariate statistical approaches such as multiple discriminant analysis (MDA) and logistic regression (logit) have been explored in the literature. Trieschmann and Pinches (1973, 1974) reported that the six-variable MDA model outperforms all univariate models. BarNiv and Hershbarger (1990) demonstrated that logit and the nonparametric discriminant analysis outperform MDA in most situations. BarNiv and McDonald (1992) reviewed the literature and also show that qualitative response models such as Probit or logit can provide better predictions of both solvency and insolvency cases than does the MDA. Carson and Hoyt (1995) found that the logit model dominated the MDA and Recursive Partitioning (RP) models in terms of the number of correctly classified solvent insurers, and the RP model dominated the logit and the MDA model in terms of the number of correctly classified insolvent insurers. Carson and Hoyt (2000) use logistic regression to estimate the insolvency factors for the life insurers in the European Union. Baranoff, Sager, and Witt (1999) and Baranoff, Sager, and Shively (2000) constructed cascaded regression and nonlinear spline models, respectively, for insolvency prediction as well. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.