Decision making with Expected Shortfall and spectral risk measures: The problem of comparative risk aversion

We analyze spectral risk measures with respect to comparative risk aversion following Arrow (1965) and Pratt (1964) for deterministic wealth, and Ross (1981) for stochastic wealth. We argue that the Arrow–Pratt-concept per se well matches with economic intuition in standard financial decision problems, such as willingness to pay for insurance and simple portfolio problems. Different from the literature, we find that the widely-applied spectral Arrow–Pratt-measure is not a consistent measure of Arrow–Pratt-risk aversion. Instead, the difference between the antiderivatives of the corresponding risk spectra is valid. Within the framework of Ross, we show that the popular subclasses of Expected Shortfall, and exponential and power spectral risk measures cannot be completely ordered with respect to Ross-risk aversion. Thus, for all these subclasses, the concept of Ross-risk aversion is not generally compatible with Arrow–Pratt-risk aversion, but induces counter-intuitive comparative statics of its own. Compatibility can be achieved if asset returns are jointly normally distributed. The general lesson is that these restrictions have to be considered before spectral risk measures can be applied for the purpose of optimal decision making and regulatory issues.     

A Value at Risk Approach to Background Risk

This paper studies the effects of an uninsurable background risk (BR) on the demand for insurance (proportional and with deductible). We study both the case of BR uncorrelated with the insurable one and the perfectly correlated one, in a Gaussian world. In order to perform our study, we exploit the new risk measure known as Value at Risk (VaR) and consider insurance contracts which are Mean-VaR efficient. We obtain results which depend on the parameters (moments) of both risks and on the magnitude of loadings charged by the insurance company, instead of depending on the risk attitudes of the insured, such as risk aversion and prudence.We demonstrate that, if loadings are not too high, the demand for insurance increases with positively correlated BR; it decreases with BR negatively correlated if the latter is less risky than the insurable one (in this case it can even go to zero, if loadings are too high); it goes to zero with BR which is negatively correlated and more risky than the insurable one.     

Changes in Background Risk and the Demand for Insurance

The demand for insurance against loss from a particular risky asset is likely to depend on other risks the decision-maker faces. For independently distributed other risks, referred to as background risk, Eeckhoudt and Kimball [1992] determine the effect on insurance demand of introducing background risk. Recently, Eeckhoudt, Gollier, and Schlesinger [1996] determine conditions on preferences such that first- and second-degree stochastic deteriorations in background risk lead to a decrease in the decision-maker's willingness to accept other risks. These results, although formulated in a general decision model, also apply to insurance demand. This article continues analysis of this question by determining the effect on insurance demand of several other general changes in background risk.     

Locally Greater Vulnerability to Background Risk

Willingness to take on risk is influenced by the presence of fair and unfair background risks for decision makers who are risk vulnerable as defined by Gollier and Pratt [1996], for these decision makers are more risk averse when they possess such an uninsurable background risk. We present an alternative derivation of the index of local vulnerability based on Diamond and Stiglitz [1974] compensated increases in risk, such that risk aversion increases with the introduction of any small fair background risk if and only if the index of local vulnerability is positive. We establish that the increase in risk aversion is greater for those who are more vulnerable as measured by the index of local vulnerability.     

Labor Income Risk and Car Insurance in the UK

Microeconomic theory shows that only under certain conditions higher background risk increases the propensity to insure against independent marketable risks. We provide empirical evidence for the case of labor income risk and car insurance in the UK. The main result is that households with higher labor income risk spend more on insurance. This finding is consistent with microeconomic theory if the utility function is of the HARA type. Moreover, we find that households spend more on insurance if they participate in the stock market.     

A Generalisation of the Mean-Variance Analysis

In this paper we consider a decision maker whose utility function has a kink at the reference point with different functions below and above this reference point. We also suppose that the decision maker generally distorts the objective probabilities. First we show that the expected utility function of this decision maker can be approximated by a function of mean and partial moments of distribution. This 'mean-partial moments' utility generalises not only mean-variance utility of Tobin and Markowitz, but also mean-semivariance utility of Markowitz. Then, in the spirit of Arrow and Pratt, we derive an expression for a risk premium when risk is small. Our analysis shows that a decision maker in this framework exhibits three types of aversions: aversion to loss, aversion to uncertainty in gains, and aversion to uncertainty in losses. Finally we present a solution to the optimal capital allocation problem and derive an expression for a portfolio performance measure which generalises the Sharpe and Sortino ratios. We demonstrate that in this framework the decision maker's skewness preferences have first-order impact on risk measurement even when the risk is small.     

The Spillover effect of compulsory insurance

The assumption usually made in the insurance literature that risks are always insurable at the desired level does not hold in the real world: some risks are not—or are only partially—insurable, while others, such as civil liability or health and workers' injuries, must be fully insured or at least covered for a specific amount. We examine in this paper conditions under which a reduction in the constrained level of insurance for one risk increases the demand of insurance for another independent risk. We show that it is necessary to sign the fourth derivative of the utility function to obtain an unambiguous spillover effect. Three different sufficient conditions are derived if the expected value of the exogenous risk is zero. The first condition is that risk aversion be standard—that is, that absolute risk aversion and absolute prudence be decreasing. The second condition is that absolute risk aversion be decreasing and convex. The third condition is that both the third and the fourth derivatives of the utility function be negative. If the expected value of the exogenous risk is positive, a wealth effect is added to the picture, which goes in the opposite direction if absolute risk aversion is decreasing.     

Relative Risk Aversion as an Arc Elasticity

Risk aversion is the central reason why individuals purchase insurance and undertake other forms of risk management. But deriving the Pratt–Arrow coefficient of relative risk aversion from a utility function requires familiarity with differential calculus—a level of mathematics beyond the prerequisites for most introductory risk management courses. Thus, students are not exposed to one of the most important and fundamental concepts in the field unless and until they take more advanced courses. The present article demonstrates that relative risk aversion can be obtained as an arc elasticity using only elementary mathematics. This approach highlights the relationship between risk aversion and the demand for insurance, and integrates concepts from the principles of economics course, helping to unify the business curriculum. Numerical examples are easily computed and graphed using electronic spreadsheets, providing students with a hands-on learning experience. For sufficiently small risks, the arc elasticity measure reduces to the Pratt–Arrow coefficient, providing a platform for discussing the difference between large-scale and small-scale risk aversion in upper-level courses.     

Background Risk, Demand for Insurance, and Choquet Expected Utility Preferences

This article deals with demand for insurance with a background risk in a nonprobabilized uncertainty framework, where preferences are represented by a nonadditive model of decision making. The Choquet expected utility model that we use generalizes expected utility and allows for a separation of the attitude towards uncertainty and the attitude towards wealth. When the insurable and the background risk are comonotone, the impact of the background risk on the demand for insurance is related to the attitude towards wealth. In contrast, when the two risks are anticomonotone, the attitude towards uncertainty is determinant. In this case, some of the resulting behaviors cannot be explained by the standard expected utility model.     

A Shortcut Way of Pricing Default Risk Through Zero-Utility Principle

Pricing the default risk is a hot challenge for every risk manager. The problem is tackled in the framework of the zero‐utility principle. According to Pratt (1964) , an approximation of the risk premium should be proportional to the Arrow–Pratt absolute risk aversion coefficient and the variance. Is that still true as a default risk is concerned? The answer appears to be negative, because the variance does not look to be an appropriate tool for asymmetrical risk. On the other hand, fear of ruin coefficient and probability of default are proved to be well‐tailored tools for a preliminary pricing. Bid and ask price approximations are both elicited and a necessary condition for risk exchange set out.     

Comparative precautionary saving under higher-order risk and recursive utility

Measuring and comparing the precautionary saving motive rest almost exclusively on the expected utility framework, and only focus on income risk or coefficients of the Arrow–Pratt type. We generalize the standard approach by characterizing comparative precautionary saving under recursive utility for increases in income risk and increases in risk on the saving return, including higher-order risk effects. We express the comparisons in terms of precautionary premia. In addition, we define a new class of preference coefficients, and derive the associated conditions to predict a stronger precautionary motive. The coefficients provide a detailed picture of the preferences sustaining precautionary saving and could be useful in applications.     

Optimal insurance contract and coverage levels under loss aversion utility preference

This study develops an optimal insurance contract endogenously and determines the optimal coverage levels with respect to deductible insurance, upper-limit insurance, and proportional coinsurance, and, by assuming that the insured has an S-shaped loss aversion utility, the insured would retain the enormous losses entirely. The representative optimal insurance form is the truncated deductible insurance, where the insured retains all losses once losses exceed a critical level and adopts a particular deductible otherwise. Additionally, the effects of the optimal coverage levels are also examined with respect to benchmark wealth and loss aversion coefficient. Moreover, the efficiencies among various insurances are compared via numerical analysis by assuming that the loss obeys a uniform or log-normal distribution. In addition to optimal insurance, deductible insurance is the most efficient if the benchmark wealth is small and upper-limit insurance if large. In the case of a uniform distribution that has an upper bound, deductible insurance and optimal insurance coincide if benchmark wealth is small. Conversely, deductible insurance is never optimal for an unbounded loss such as a log-normal distribution.     

Precautionary Insurance Demand With State-Dependent Background Risk

This article considers a zero‐mean background risk that is uncorrelated with insurable losses, but is not necessarily statistically independent. In particular, the size of the background risk can vary in different insurable‐loss states. We show how a prudent individual will buy either more insurance or less insurance than with no background risk, depending on the relative size of the background risk in the loss states vis‐á‐vis the no‐loss states. If we consider two individuals, with one more risk averse than the other, we need to compare the intensities of their precautionary motives, in addition to their measures of risk aversion, before we can determine who buys more insurance coverage in the presence of the state dependent background risk.     

A Gentle Introduction to Risk Aversion and Utility Theory

While the topics of risk aversion and utility theory have been discussed extensively in the academic literature on risk and insurance, this literature does not include a pedagogical discussion that is widely accessible for classroom use. This article provides a practical introduction to risk aversion that is designed for readers with little prerequisite course work in economics or statistics. We describe a simple model of insurance demand that can be applied to the property, liability, life, and health insurance markets. We also demonstrate how risk aversion affects a variety of real-life insurance decisions made under conditions of uncertainty, including how much the market will bear to pay for insurance administrative expenses and how demand varies for different types of auto insurance coverage. Exercises and practice problems are provided so that readers can test their mastery of the concepts presented in the article. An instructional note on using this article to teach risk aversion in the classroom is also provided.     

Optimal capital growth with convex shortfall penalties

The optimal capital growth strategy or Kelly strategy has many desirable properties such as maximizing the asymptotic long-run growth of capital. However, it has considerable short-run risk since the utility is logarithmic, with essentially zero Arrow–Pratt risk aversion. It is common to control risk with a Value-at-Risk (VaR) constraint defined on the end of horizon wealth. A more effective approach is to impose a VaR constraint at each time on the wealth path. In this paper, we provide a method to obtain the maximum growth while staying above an ex-ante discrete time wealth path with high probability, where shortfalls below the path are penalized with a convex function of the shortfall. The effect of the path VaR condition and shortfall penalties is a lower growth rate than the Kelly strategy, but the downside risk is under control. The asset price dynamics are defined by a model with Markov transitions between several market regimes and geometric Brownian motion for prices within a regime. The stochastic investment model is reformulated as a deterministic programme which allows the calculation of the optimal constrained growth wagers at discrete points in time.     

Background Uncertainty and the Demand for Insurance Against Insurable Risks

Theory suggests that people facing higher uninsurable background risk buy more insurance against other risks that are insurable. This proposition is supported by Italian cross-sectional data. It is shown that the probability of purchasing casualty insurance increases with earnings uncertainty. This finding is consistent with consumer preferences being characterized by decreasing absolute prudence.     

The Effects of Uncertainty on the Demand for Health Insurance

This article analyzes the effects of uncertainty and increases in risk aversion on the demand for health insurance using a theoretical model that highlights the interdependence between insurance and health care demand decisions. Two types of uncertainty faced by the individuals are examined. The first one is the uncertainty in the consumer's pretreatment health and the second is the uncertainty surrounding the productivity of health care. Comparative statics results are reported indicating the impact on the demand for insurance of shifts in the distributions of pretreatment health and productivity of health care in the form of first‐order stochastic dominance, Rothschild–Stiglitz mean‐preserving spreads, and second‐order stochastic dominance. The demand for insurance increases in response to a Rothschild–Stiglitz increase in risk in the distribution of the pretreatment health provided that the health production function is in a special class and the price elasticity of health care is nondecreasing in the pretreatment health. Provided also that the demand for health care is own‐price inelastic, the same conclusion is obtained when the uncertainty is about the productivity of health care.     

Variance Vulnerability, Background Risks, and Mean-Variance Preferences

An agent with two-parameter, mean-variance preferences is called variance vulnerable if an increase in the variance of an exogenous, independent background risk induces the agent to choose a lower level of risky activities. Variance vulnerability resembles the notion of risk vulnerability in the expected utility (EU) framework. First, we characterize variance vulnerability in terms of two-parameter utility functions. Second, we identify the multivariate normal as the only distribution such that EU- and two-parameter approach are compatible when independent background risks prevail. Third, presupposing normality, we show that—analogously to risk vulnerability—temperance is a necessary, and standardness and convex risk aversion are sufficient conditions for variance vulnerability.     

Exclusions and the Demand for Property Insurance

The paper examines property insurance contracts in which consumers choose the upper limit on coverage. Exclusions are of two types, and both reduce the demand for insurance of the included perils. A practical implication is that an insurer can raise the demand for fire insurance by offering an earthquake rider, and profit from the rider even when the premia are ceded in such a way that the rider does not raise profit directly. The results do not require assumptions about correlations between included and excluded losses, which is interesting because correlations are decisive in most of the other literature on background risk.     

Optimal Insurance Under Random Auditing

We provide a characterization of an optimal insurance contract (coverage schedule and audit policy) when the monitoring procedure is random. When the policyholder exhibits constant absolute risk aversion, the optimal contract involves a positive indemnity payment with a deductible when the magnitude of damages exceeds a threshold. In such a case, marginal damages are fully covered if the claim is verified. Otherwise, there is an additional deductible that disappears when the damages become infinitely large. Under decreasing absolute risk aversion, providing a positive indemnity payment for small claims with a nonmonotonic coverage schedule may be optimal.