Empirical Evidence on the Value of Group‐Level Financial Information in Insurer Solvency Surveillance

• DOI: http://doi.org/10.1111/j.1540-6296.2011.01195.x
• Published date: 01 March 2011
• Date: 01 March 2011

C

Risk Management and Insurance Review, 2011, Vol.14, No. 1, 73-88

DOI: 10.1111/j.1540-6296.2011.01195.x

EMPIRICAL EVIDENCE ON THE VALUE OF GROUP-LEVEL

FINANCIAL INFORMATION IN INSURER SOLVENCY

SURVEILLANCE

Steven W. Pottier

David W. Sommer

ABSTRACT

The existing empirical research on insurer insolvency relies almost exclusively

upon individual insurance company financial data, even though the insurance

industry is dominated by group-affiliated firms. This is the first study to eval-

uate the benefit of using group-level data to predict insurer insolvencies for

group-affiliated insurers. The study uses financial ratios from the NAIC FAST

scoring system, measured at both the company level and group level, as poten-

tial predictor variables. The results indicate that group-level financial informa-

tion substantially improves the predictive power of an insolvency prediction

model relative to a model that uses only the analogous company-level vari-

ables. In fact, the group-level variables are found to often be substantially more

powerful than company-level variables in predicting individual insurer insol-

vencies. These results suggest that future insolvency analysis should, whenever

feasible, include group-level information to obtain higher predictive accuracy.

INTRODUCTION

The risk of insurer insolvency is of utmost concern to insurance regulators and con-

sumers. A key focus of state regulators is the detection and prevention of insurance

company insolvencies. In the private sector, an entireindustry has developed to provide

insurer financial strength ratings to aid consumers and other interested parties in as-

sessing the insolvency risk of insurers. Academic research has reflected the importance

of this issue, with a large body of literature devoted to developing and evaluating insol-

vency prediction models. Previous articles on the topic have investigated a wide range

of potential predictor variables, as well as various statistical approaches to insolvency

prediction modeling. However, all previousresearch in this area, to our knowledge, has

had one thing in common. Specifically, these studies have all used predictor variables

based exclusively on individual insurance company data.1

Steven W. Pottier is with the Terry College of Business, University of Georgia; e-mail:

spottier@terry.uga.edu. David W. Sommer is at the Bill Greehey School of Business, St. Mary’s

University; e-mail: dsommer@stmarytx.edu. This article was subject to double-blind peer review.

1Browne and Hoyt (1995) and Browne et al. (1999) use macroeconomic industrywide data in

their analyses, but they are modeling the total number of insolvencies in the industry, and are

not attempting to predict individual insolvencies.

73

74 RISK MANAGEMENT AND INSURANCE REVIEW

The focus on individual insurer data in insolvency analysis is understandable, since

insurers are primarily regulated at the individual company level. For example, National

Association of Insurance Commissioners (NAIC) Insurance Regulatory Information Sys-

tem (IRIS) ratios, Risk-Based Capital (RBC) ratios, and Financial Analysis Solvency Tools

(FAST) scores are all calculated based on company-level data. However, the exclusive

focus on company-level data in insolvency prediction models misses potentially impor-

tant information related to the insolvency risk of many insurers, namely, the financial

condition of the group with which the insurer is affiliated. Previous research has indi-

cated that significant insurer decisions often occur at the group level rather than the

individual company level (Cummins et al., 1999; Gaver and Pottier, 2005), making the

financial condition of the group highly relevant. The majority of property–liability in-

surers are affiliated with one or more other property–liability insurers in an insurer

group. In 1999, insurer groups wrote over 90 percent of total property–liabilityindustry

premiums. The fundamental hypothesis tested in this article is that empirical insolvency

prediction models can be improved by including financial information about the group

with which an insurer is affiliated rather than relying exclusively on company-specific

variables.

The NAIC has recognized the inadequacy of regulators’ traditional focus on regulation at

the individual insurer level rather than at the holding company level. In 2002, the NAIC

Executive Committee adopted a report entitled, Framework for Insurance Holding Company

Analysis. Among the reasons the report cites for the need for a better system of holding

company analysis is “the realization that analysis of an individual company may not

be complete without understanding the context of the group and the markets in which

the company operates” (NAIC, 2002, p. 2). Therefore, regulators now look at a variety

of quantitative and qualitative information in performing solvency analysis, including

information about the group to which an insurer is affiliated. However, while regulators

review and analyze group holding company information, there is not at present any

uniform system of quantitative tools that use group financial information to prioritize

insurers for additional review or serve as an early warning system (NAIC, 2005).

Rating agencies also incorporate information about the insurer group when assigning

financial strength ratings to some individual insurers. The academic literature, on the

other hand, has not kept pace, with all existing empirical solvency studies focusing

exclusively on individual company data. This article seeks to address that gap in the

literature by providing evidence on the value of explicitly incorporating group-level

information in solvency analysis.

Initially, a standard insolvency prediction model is estimated for group-affiliated insur-

ers using financial ratios from the NAIC’s FAST system, measured at the individual

company level. This model is then re-estimated with the same variables but measured

at the group level rather than the individual company level. Finally, the model is es-

timated a third time, including variables measured at both the company level and the

group level. The predictive abilities of these three models are compared in three differ-

ent ways. The conclusion is that group-level variables have a significant impact on the

probability of insolvency for group-affiliated insurers, and the inclusion of such vari-

ables substantially increases the accuracy of the insolvency prediction model in each of

the 5 years tested. In fact, for 3 of the 5 years, the group-level variables are found to