Georges Dionne

Testing for Evidence of Adverse Selection in the Automobile Insurance Market: A Comment

We analyze jointly the distribution of automobile accidents and the choice of deductible. One prediction in the literature is that high risk individuals will choose small deductibles within risk classes, when there is asymmetrical information. We show, however, that risk classification is sufficient, in the sense that there is no residual adverse selection on risk types in the automobile insurance portfolio studied.

Insurance with Undiversifiable Risk: Contract Structure and Organizational Form of Insurance Firms

Previous explanations of the contract choice and organizational form of insurance firms do not explain, by themselves, the recent proliferation of mutuals and new contract designs. We first present risk-bearing arguments to address these phenomena. We present two forms of insurance. The first is a conventional transfer of risk whereas the second decomposes risk between idiosyncratic and nonidiosyncratic. We show that the latter form leads to more active trade in insurance markets with correlated exposures. Moreover, the decomposed form dominates the simple transfer. These results qualify and extend the work of Borch (1962) and Marshall (1974). Market responses to the recent “liability insurance crisis” are compatible with these predictions.

Adverse Selection in Insurance Markets

In this survey we present some of the more significant results in the literature on adverse selection in insurance markets. Sections 1 and 2 introduce the subject and section 3 discusses the monopoly model developed by Stiglitz (1977) for the case of single-period contracts and extended by many authors to the multi-period case.

Adverse Selection, Commitment, and Renegotiation: Extension to and Evidence from Insurance Markets

With asymmetric information, full commitment to long-term contracts may permit markets to approach first-best allocations. However, commitment can be undermined by opportunistic behavior, notably renegotiation. We reexamine commitment in insurance markets. We present an alternative model (which extends Laffont and Tiroles procurement model to address uncertainty and competition), which involves semipooling in the first period followed by separation. This and competing models (e.g., single-period models and no-commitment models) have different predictions concerning temporal patterns of insurer profitability. A test using California data suggests that some automobile insurers use commitment to attract selective portfolios comprising disproportionate numbers of low risks.

New Evidence on the Determinants of Absenteeism Using Linked Employer-Employee Data

This paper provides new evidence on the determinants of absenteeism. The authors extend the typical labor-leisure model used to analyze the decision to skip work to include firm-level policy variables relevant to the absenteeism decision and uncertainty about the cost of absenteeism. Estimates based on data from Statistics Canadas Workplace Employee Survey (1999–2002), with controls for observed and unobserved demographic, job, and firm characteristics (including workplace practices), indicate that work arrangements were important determinants of absence. For example, the authors find strong evidence that standard weekday work hours, work-at-home options, and reduced workweeks were associated with reduced absence, whereas shift work and compressed work weeks were associated with increased absence.

The value of a statistical life: A meta-analysis with a mixed effects regression model

The value of a statistical life (VSL) is a very controversial topic, but one which is essential to the optimization of governmental decisions. We see a great variability in the values obtained from different studies. The source of this variability needs to be understood, in order to offer public decision-makers better guidance in choosing a value and to set clearer guidelines for future research on the topic. This article presents a meta-analysis based on 39 observations obtained from 37 studies (from nine different countries) which all use a hedonic wage method to calculate the VSL. Our meta-analysis is innovative in that it is the first to use the mixed effects regression model [Raudenbush, S.W., 1994. Random effects models. In: Cooper, H., Hedges, L.V. (Eds.), The Handbook of Research Synthesis. Russel Sage Foundation, New York] to analyze studies on the value of a statistical life. We conclude that the variability found in the values studied stems in large part from differences in methodologies.

A Generalization of Automobile Insurance Rating Models

The objective of this paper is to provide an extension of well-known models of tarification in automobile insurance. The analysis begins by introducing a regression component in the Poisson Model in order to use all available information in the estimation of the distribution. In a second step, a random variable is introduced in the regression component of the Poisson model and a negative binomial model with a regression component is derived. The authors then present their main contribution by proposing a bonus-malus system which integrates a priori and a posteriori information on an individual basis. They then show how a multivariate net premium tables can be derived from the model. Examples of tables are presented. (A)

Replacement Cost Endorsement and Opportunistic Fraud in Automobile Insurance

Traditional insurance contracts do not offer protection against the replacement value of a vehicle. A replacement cost endorsement gives the opportunity to get a new vehicle in the case of a total theft or in the case of total destruction of the car in a road accident. This type of protection was introduced in Canada in the late 1980s. It is also offered in France and many insurers in the United States are going to move in that direction. We propose tests that separate moral hazard from adverse selection in the analysis of the effect of this additional protection on car theft. We show that holders of car insurance policies with a replacement cost endorsement have a higher probability of theft near the end of this additional protection (usually 24 months following the acquisition of a new car). Our tests indicate that this result is a form of ex post moral hazard or opportunistic insurance fraud.

An Empirical Analysis of Moral Hazard and Experience Rating

For many years, economists and actuaries have studied multiperiod insurance contracts independently and differently. The aim of this article is to reduce the gap between empirical studies on the determination of insurance premiums and theoretical studies on moral hazard and experience rating. The first objective is to verify empirically the proposition that past experience is a good predictor of risk by using data from a random sample consisting of 19,013 car drivers in Quebec. The second objective is to construct a multivariate, multiperiod pricing formula for automobile insurance in order to reduce the ill effects of moral hazard. Copyright 1989 by MIT Press.

Count data models for a credit scoring system

Credit scoring systems created for the evaluation of new applications are based on the available statistical information which is related to the behaviour of former clients with credit. Usually, financial institutions apply discriminant analysis techniques to create these systems but they lack of good properties due, for example, to the presence of non-normal variables. As an alternative, the future repayment behaviour is predicted by means of the expected number of unpaid instalments. The use of this latter variable suggests that appropriate models might be of interest, in which some covariant exogenous variables are included in order to specify the expected level of debt. At this point, prepayment is not explicitly considered. These models should be used as explanatory tools when evaluating the level of risk involved in personal credit transactions. Negative Binomial Distribution models are suitable when heterogeneity is taken into account. Some results related to prediction performance are shown for different model specifications in the case of data from a Spanish bank.