Scott E. Harrington

The Financial Crisis, Systemic Risk, and the Future of Insurance Regulation

This article considers the role of American International Group (AIG) and the insurance sector in the 2007–2009 financial crisis and the implications for insurance regulation. Following an overview of the causes of the crisis, I explore the events and policies that contributed to federal government intervention to prevent bankruptcy of AIG and the scope of federal assistance to AIG. I discuss the extent to which insurance in general poses systemic risk and whether a systemic risk regulator is desirable for insurers or other nonbank financial institutions. The last two sections of the article address the financial crisiss implications for proposed optional and/or mandatory federal chartering and regulation of insurers and for insurance regulation in general.

Insolvency experience, risk-based capital, and prompt corrective action in property-liability insurance

In December 1992, the National Association of Insurance Commissioners (NAIC) adopted a life-health insurer risk-based capital (RBC) formula and model law that became effective with the 1993 annual statement filed in March 1994. In principle, well-designed RBC requirements can help achieve an efficient reduction in the expected costs of insolvencies. They can provide incentives for insurers to operate safely in cases where market incentives are weak due to government mandated guarantees of insurer obligations or asymmetries regarding solvency between insurers and buyers. RBC requirements also may facilitate or encourage prompt corrective action by solvency regulators by helping regulators to identify weak insurers and giving regulators legal authority to intervene when capital falls below specified levels. RBC requirements may force regulators to act in amore timely manner when confronted with external pressure to delay action. However, RBC capital requirements have a number ofpotential limitations. Unavoidable imperfections in any meaningful RBC system will likely distort some insurer decisions in undesirable and unintended ways. RBC requirements by themselves will do little or nothing to help regulators determine when an insurer s reported capital (surplus) is overstated due to understatement of liabilities or overstatement of assets. A well-designed RBC system should minimize costs associated with misclassification of insurers. The system should be able to identify a high proportion of troubled companies early enough to permit regulators to take prompt corrective action and should identify as troubled only a minimal proportion of financially sound insurers. This study analyzes data on solvent and insolvent property-liability insurers to determine whether modifications in the NAIC s RBC formula can improve its ability to predict firms that subsequently fail without substantially increasing the proportion of surviving insurers that are incorrectly predicted to fail. It uses logistic regression models to investigate whether changes in the weight for the major components in the RBC formula and incorporation of information on company size and organizational form improve the tradeoff between Type I error rates (the percentage of insurers that later failed that are incorrectly predicted not to fail) and the Type II error rates (the percentage of surviving insurers that are incorrectly predicted to fail). The data analyzed were for 1989-91 for firms that subsequently failed and for firms that survived through the first nine months of 1993. The authors make four main conclusions. First, less than half of the companies that later failed had RBC ratios within the proposed ranges for regulatory and company action. Second, total and component RBC ratios generally are significantly different for failed and surviving firms based on univariate tests. Third, estimation of multiple logistic regression models of insolvency risk indicated that allowing the weights of the RBC component to vary and including firm size and organizational form variables generally produce a material improvement in the tradeoff between sample Type I and Type II error rates. And, fourth,the RBC models are noticeably less successful in predicting large firm insolvencies than in predicting smaller insolvencies. Regarding the estimated weights in the logistic regression models, a major conclusion is the reserve component of the NAIC risk-based capital formula, which accounts for half of industry risk-based capital, has virtually no predictive power in any of the tests conducted. Given the high costs associated with large failures and the inferior performance of the models in predicting large insolvencies, a higher payoff in terms of reduced insolvency costs is likely to be achieved by developing models that perform better for large firms.

Market Discipline in Property/Casualty Insurance: Evidence from Premium Growth Surrounding Changes in Financial Strength Ratings

Analysis of abnormal premium growth surrounding changes in financial strength ratings for a large panel of property/casualty insurers generally indicates significant premium declines in the year of and the year following rating downgrades. Consistent with greater risk sensitivity of demand, premium declines were concentrated among commercial insurance, which has narrower guaranty fund protection than personal insurance. Premium declines were greater for firms with low pre-downgrade ratings, and especially pronounced for firms falling below an A- rating. There is no evidence of moral hazard in the form of rapid commercial or personal lines premium growth following downgrades of A- or low-rated insurers.

A Regression-Based Methodology for Solvency Surveillance in the Property-Liability Insurance Industry

A new methodology for assessing property-liability insurer financial strength is suggested in this paper. The procedure involves using regression analysis to estimate the relationship between premium-to-surplus ratios and insurer characteristics including asset and product mix variables. Analysis of the regression residuals then can be used to identify insurers with ratios that are substantially higher than those for insurers with similar characteristics. The method is illustrated using data for solvent and insolvent insurers. Its ability to identify insurers that later became insolvent is compared to that of the NAIC Insurance Regulatory Information System.

Risk-Based Capital and Solvency Screening in Property-Liability Insurance: Hypotheses and Empirical Tests

For a fixed probability of wrongly classifying a strong insurer as being weak (Type I error), this paper examines the classification power (the probability of correctly identifying a weak insurer as being weak) for two potential solvency detection methods. The first is to classify insurers using ratios based on risk-based capital (RBC) standards and the second is to use the Financial Analysis Tracking System (FAST) solvency screening mechanism created by the National Association of Insurance Commissioners (NAIC). We test the hypothesis that the RBC system has at least as much power for identifying financially weak insurers as the FAST scoring system does. Our empirical results are largely inconsistent with this hypothesis: RBC ratios are less powerful than FAST scores in identifying financially weak property-liability insurers during our sample periods. We also provide limited evidence that RBC ratios and FAST scores are jointly more powerful in identifying weak insurers than FAST scores alone, which suggests that RBC ratios may convey new information about insolvency risk despite their relatively low power on a univariate basis.

Insurance Supply with Capacity Constraints and Endogenous Insolvency Risk

Negative shocks to industry capital and significant capital adjustment costs have been offered as an explanation of periodic “crises” in the property-liability insurance market. According to these capacity constraint models, in which post-shock production must meet a solvency constraint, increases in price can cause some or perhaps all of the cost of a negative shock to capital to be shifted to policyholders. This article develops a model of insurance supply with capacity constraints and endogenous insolvency risk that incorporates limited liability and potential loss of insurer intangible capital. If industry demand is inelastic with respect to price and capital, the model predicts that price will increase following a negative shock to capital, but by less than the amount needed to fully offset the shock. Equity value and the expected recovery by policyholders for post-shock production are predicted to decline. Elastic demand mitigates shock-induced price increases.

Volatility and Underwriting Cycles

This paper describes and illustrates the main ideas and findings of research on the volatility and cyclical behavior of insurance prices relative to those predicted by a perfectly competitive market in long-run equilibrium. After presenting evidence that insurance market prices indeed follow a second order autoregressive process, we examine several lines of research that have tried to explain the cyclical behavior of insurance prices. Particular emphasis is given to the theoretical developments of and empirical results supporting capital shock models, which primarily explain periods of high insurance prices. We then summarize the idea that moral hazard and/or winners curse effects can explain periods of low insurance prices. Finally, the potential effects of regulation on insurance price volatility are summarized.

Prices and Profits in the Liability Insurance Market

Commercial liability insurance premiums increased dramatically in 1985 and 1986. The growth for general liability insurance was especially pronounced: net premiums written increased from

The Relationship between Risk and Return: Evidence for Property-Liability Insurance Stocks

6.5 billion in 1984 to

A Note on the Impact of Auto Insurance Rate Regulation

20 billion in 1986. During this time, limits of coverage were shrinking for many of those insured, and cancellations and denials of renewal were widespread for some types of business in some states. Premium growth had moderated substantially by the end of 1986, but the large increases in the previous two years undoubtedly imposed high costs on many businesses and professionals. Insurers reported large accounting losses on operations in 1984 and 1985. The causes of the premium increases—whether rapidly escalating and unpredictable claim costs, severe underpricing of business written before 1985 because of cyclical influences in the industry, or both—have been vigorously debated, as has the extent to which industry operations actually were unprofitable. While widespread deregulation of commercial insurance rates occurred during the 1970s, a few states have enacted new forms of rate regulation in response to recent experience.