Academic journal article Risk Management and Insurance Review

Failure Risks in the Insurance Industry: A Quantitative Systems Analysis

Academic journal article Risk Management and Insurance Review

Failure Risks in the Insurance Industry: A Quantitative Systems Analysis

Article excerpt


We present in this article the findings from a study on insolvency in the property-casualty insurance industry that was commissioned by the Risk Foundation. The Risk Foundation contacted us for this work to draw from our experience in risk analysis based on systems analysis and probability. Therefore, we provide a different perspective on failure in the insurance industry by opening the "black box" to assess the contribution of different factors to the overall risk. Besides the development of a quantitative model for insolvency risk, our study for the Risk Foundation included insights from (1) unstructured interviews with 15 insurance industry experts and with six insurance regulators in different states, and (2) a statistical analysis of insolvency data (A.M. Best) covering the 1970 through 2005 period. Our focus here is centered on the practical insights that came out of the study, rather than on the technical details that led us to those insights.


In the last few years, the insurance industry has been exposed, worldwide, to large losses and disruptions, from the payments that had to be made following the attacks on the World Trade Center on September 11, 2001, to the costs of asbestos claims, the collapse of high-tech segments of financial markets, and government investigations of industry practices. Taken separately, these kinds of events do not seem to be the main causes of firms' insolvencies. Yet, the failures of insurance companies such as Kemper, Reliance, and Trenwick, among others, illustrate the reality of the exposure of insurance firms to a spectrum of business failure scenarios.

There is more to insurance companies' insolvencies than the usual suspects: catastrophes (natural or man made) and stock market collapses. The problem is that the different factors that weaken these firms are dependent and evolving. Statistics are useful but often insufficient because the industry may face new situations and "perfect storms," that is, rare conjunctions of dependent or independent events. For instance, a large catastrophe may affect the economy as a whole at the same time as it generates large claims. A system's analysis of the kind that is performed in engineering to compute (for example) the failure risk of an offshore platform can be helpful to combine in this case all available information, including statistical data and probability estimates. It starts with a systematic identification of possible combinations of nasty events and permits an assessment of the probabilities of failure given the current strategy. This kind of assessment does not yield precise, perfect numbers, but a result based on a logical combination of different pieces of information that is often better than simple guesses of the risk. The next question, of course, is: what can be done and when? Is there a reason to change the firm's strategy (choice of product, pricing, etc.)?

Using that approach, we performed a multifaceted study of property-casualty insurance firms' failures. First, we interviewed a spectrum of experts; second, we performed a statistical analysis of historical insolvency data; and third, we constructed a mathematical model based on systems analysis and probability to assess the failure risk of an insurance company given its current strategy in a specified time period. Therefore, this study is both qualitative and quantitative. Its objective is to provide a wide perspective on insurance companies' failures, to assess the risks to solvency, and to point to risk management solutions.


Research about insurance insolvency includes analytical models designed to estimate ruin probabilities (stochastic ruin process literature; e.g., Dickson, 2005; Rolski et al., 1999). Constrained by the need to obtain closed-form solutions of the ruin probabilities, such models are characterized by simplified models of claims and premiums processes and can estimate finite-horizon ruin probabilities only in a few specific cases. …

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


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.