NBER Reporter: Spring 2001

Insurance

The NBER's Insurance Research Group met in Cambridge on February 16. Kenneth A. Froot, NBER and Harvard University and Howard C. Kunreuther, NBER and University of Pennsylvania, organized this program:

Martin F. Grace and Richard D. Phillips, "The Allocation of Governmental Regulatory Authority: Federalism and the Case of Insurance Regulation"

Discussants: Dwight Jaffee, University of California, Berkeley, and John Major, Guy Carpenter & Co.

Karen Epermanis, University of Hartford, and Scott E. Harrington, University of South Carolina, "Financial Rating Changes and Market Discipline in Property Liability Insurance"

Discussants: James Ament, State Farm Fire & Casualty Co., and Paul R. Kleindorfer, University of Pennsylvania

Jeffrey R. Brown, NBER and Harvard University, Olivia S. Mitchell, and James M. Poterba, "The Role of Real Annuities and Indexed Bonds in an Individual Accounts Retirement Program"

Discussants: James Garven, Baylor University, and Richard Meyer, Harvard University

J. David Cummins, University of Pennsylvania, and Olivier Mahul, INRA, "Managing Catastrophic Risk with Insurance Contracts Subject to Default Risk"

Discussant: Richard Derrig, Automobile Insurers Bureau of Massachusetts

Gordon Woo, "Risk Management Solutions: Territorial Diversification of Catastrophe Bonds"

Discussants: Paul Freeman, The World Bank, and Kenneth A. Froot

Howard C. Kunreuther and Mark Pauly, University of Pennsylvania, "Ignoring Disaster: Don't Sweat the Big Stuff"

Discussants: Steve Goldberg, USAA Property & Casualty Insurance Group, and Sendhil Mullainathan, NBER and MIT

Robert J. Willis and Lee A. Lillard, University of Michigan, "Cognition and Wealth: The Importance of Probabilistic Thinking"

Discussants: Olivia Mitchell, NBER and University of Pennsylvania, and Thomas Russell, Santa Clara University

Grace and Phillips investigate the states' incentives for providing insurance regulation in an efficient manner. Regulation of the U.S. insurance industry is unique because it is conducted primarily at the state level while the majority of insurance sales are interstate. Consistent with predictions from the federalism literature, the authors find evidence of trans-state externalities: states with small domestic insurance markets are less efficient producers of insurance regulation and appear to allow states that choose to expend the greatest resources to regulate for them. In addition, states with more profitable domestic insurers export greater levels of regulation, suggesting that extraterritorial regulation may erect barriers to entry. The authors find increasing economies of scale in the production of insurance regulation after they control for these regulatory externalities. Taken together, their results suggest that aggregating the production of regulation to a multistate or federal level may resolve a number of inefficiencies in the current system.

Epermanis and Harrington analyze growth in premiums surrounding ratings changes by the A. M. Best Co. during 1992-6 for a large panel of property liability insurers. Their analysis generally provides evidence of significantly lower revenue growth in the year of and the year after downgrades for insurers whose ratings were downgraded. The evidence of revenue declines is strongest for firms that had relatively low ratings (below A-) prior to being downgraded. The authors also find that rating upgrades were accompanied by increased growth in premiums. Overall, material market discipline appears to exist for rated insurers despite guaranty fund protection and other factors that dull consumer incentives to seek safe insurers, and despite insurer incentives for efficient risk management.

Brown, Mitchell, and Poterba explore four issues concerning annuitization options that retirees might use in the decumulation phase of an individual accounts retirement saving system. First, they investigate the operation of both real and nominal individual annuity markets in the United Kingdom. The widespread availability of real annuities in the United Kingdom dispels the argument that private insurance markets could not provide real annuities to retirees. Second, they consider the current structure of two inflation-linked insurance products available in the United States, only one of which proves to be a real annuity. Third, the authors evaluate the potential of assets such as stocks, bonds, and bills to provide retiree protection from inflation. Because real equity returns have been high over the last seven decades, a retiree who received income linked to equity returns would have fared very well on average. Nevertheless, the authors cast doubt on the "inflation insurance" aspect of equities, since it is mainly attributable to stocks' high average return and not to stock returns moving in tandem with inflation. Finally, they assess potential retiree willingness to pay for real, nominal, and variable payout equity-linked annuities. They conclude that people would value a variable payout equity-linked annuity more highly than a real annuity because the additional real returns associated with common stocks more than compensate for the volatility of prospective payouts.

Cummins and Mahul develop a model of optimal contracting for financing large, infrequent events in which the payoffs are subject to both default risk and basis risk. Full insurance above a deductible is optimal when both parties have the same perception of the seller's default risk. When this perception differs, the optimal insurance design depends on behavioral concepts that are standard in the literature of insurance economics, such as risk tolerance, and on actuarial-science concepts, such as hazard rates. Under some specific behavioral and actuarial assumptions, the first-best indemnity schedule should increase and be concave with loss. This could help to explain why in real-world reinsurance markets the proportion of the loss reinsured decreases with the size of the loss.

Woo analyzes the expanding territorial coverage of catastrophe bonds. With catastrophe bonds being under investment management as a distinct asset class, the addition of bonds that are geographically uncorrelated (or weakly correlated) with others comes as a welcome source of portfolio diversification. Apart from the most acute concentrations of earthquake risk in California and Tokyo, bonds have been issued to cover aggregations of earthquake risk in less active seismic zones, such as Monaco. Severe windstorms in Europe also have been covered, along with tropical cyclones in the United States and Japan. Although catastrophe bonds initially focused on a single peril and territory, they now have been structured with independent multiple event triggers, differing according to peril and territory. Woo reviews the territorial development of catastrophe bonds and explores the geographical horizon for new issues.

Individuals have difficulty in dealing with low-probability, high-loss events. Frequently they fail to obtain insurance against such losses, even when the insurance terms are very favorable. Kunreuther and Pauly suggest that one reason that many people do not purchase insurance is that the transaction costs and attention time of obtaining and processing information on protection is so high as to not justify this effort. While some rare events surely can cause enormous losses when they occur, the ex ante expected value of such losses may be small. But there is more to the problem than just comparing gains and losses. What individuals will choose to do depends on the assumptions they make about the functioning of insurance markets, as well as on their perceptions of the risk and their decisionmaking processes.

Willis and Lillard use a large set of subjective probability questions from the Health and Retirement Survey to construct an index measuring the precision of probabilistic beliefs. They then relate this index to household choices about the riskiness of their portfolios and the rate of growth of their net worth. The authors propose a theory of uncertainty aversion based on repeated sampling that resolves the Ellsberg Paradox within a conventional expected utility model; in this theory, uncertainty aversion is implied by risk aversion. They then propose a link between an individual's degree of uncertainty and his propensity to give "focal" or "exact" answers to survey questions. After constructing an index of the precision of probabilistic thinking, they show that it has a statistically and economically significant positive effect on the fraction of risky assets in household portfolios and on the rate of longitudinal growth of these assets. These results suggest that there is systematic variation in the competence of individuals to manage investment accounts; that variation should be considered in designing policies to create individual retirement accounts in the Social Security system.