Behavioral insurance and economic theory: A literature review

• Author: Jia Min Ng,Glenn W. Harrison
• Published date: 01 July 2019
• DOI: http://doi.org/10.1111/rmir.12119
• Date: 01 July 2019

© 2019 The American Risk and Insurance Association

Risk Manag Insur Rev. 2019;22:133–182. wileyonlinelibrary.com/journal/rmir

|

133

Received: 1 May 2019

|

Accepted: 1 May 2019

DOI: 10.1111/rmir.12119

INVITED ARTICLE

Behavioral insurance and economic theory:

A literature review

Glenn W. Harrison

1,3

|

Jia Min Ng

2

1

Department of Risk Management &

Insurance and Center for the Economic

Analysis of Risk, Robinson College of

Business, Georgia State University,

Atlanta, Georgia

2

Center for the Economic Analysis of

Risk, Robinson College of Business,

Georgia State University, Atlanta,

Georgia

3

School of Economics, University of Cape

Town, Cape Town, South Africa

Correspondence

Glenn W. Harrison, Department of Risk

Management & Insurance and Center for

the Economic Analysis of Risk, Robinson

College of Business, Georgia State

University, Atlanta, Georgia.

Email: gharrison@gsu.edu

Jia Min Ng, Center for the Economic

Analysis of Risk, Robinson College of

Business, Georgia State University,

Atlanta, Georgia.

Email: jiamin.ng@outlook.com

Abstract

Decisions to purchase insurance should be a perfect

place to see economic theory at work in general, and

behavioral economics at work in particular. We have

well‐developed theories of the demand for, and welfare

evaluation of, insurance products. These theories extend

relatively easily to the insights of behavioral economics.

Unfortunately, the empirical literature has not main-

tained this tight connection. In fact, much of the

empirical literature illustrates the dangers of the modern

passion with agnostic economics: avoiding theory at all

costs to focus on “what works.”We identify these

dangers and the implications in the literature.

1

|

INTRODUCTION

Decisions to purchase insurance should be a perfect place to see economic theory at work

in general, and behavioral economics at work in particular. We have well‐developed

theories of the demand for, and welfare evaluation of, insurance products. These theories

extend relatively easily to the insights of behavioral economics.

1

Unfortunately, the

empirical literature has not maintained this tight connection with theory. In fact, much of

the empirical literature illustrates the dangers of the modern passion with agnostic

------------------------------------------------------------------------------------

1

Behavioral economics is not defined here, nor should it be, as just the study of anomalies or irrationality.

economics: avoiding theory at all costs to focus on “what works.”We identify these

dangers and the implications for the behavioral insurance literature.

2

To keep the focus tight we limit discussions to empirical studies that use the methods of

experimental economics, whether it be in the laboratory or the field. This includes studies

that exploit naturally occurring data that offer many of the controls of experiments.

3

We

generally avoid any mention of studies using hypothetical surveys, because of the

overwhelming evidence of hypothetical biasacrossmostbehavioraldomains.

4

From a theoretical perspective, we can quickly identify several “behavioral moving parts”in

canonical insurance contracts. The first is obviously risk aversion, which can derive from

various psychological pathways. The second is, also obviously, subjective beliefs about loss

probabilities, as well as nonperformance risk and basis risk when applicable. The third concerns

time preferences, thinking of insurance as an explicitly time‐dated contract: in general, I give

you a known premium now in the expectation that if something happens to me over the coming

year you will honor that contract and help me mitigate the loss. The fourth then involves the

interaction of risk and time preferences, in the form of intertemporal risk aversion. As

explained below, this is not the same as atemporal risk aversion.

Sections 1 and 2 provide “helicopter tours”of key issues in theory and experiments that we

view as central to evaluating the literature. Section 3 reviews descriptive behavioral experiments

on insurance, and Section 4 reviews normative behavioral experiments on insurance.

2

|

THEORY

Insurance is a staple of any classroom discussion of risk attitudes and risk management. Indeed,

it is often used to immediately explain why we should be interested in knowing the risk

attitudes of an agent. The very definition of a risk premium, as the amount of money one is

willing to leave on the table, in expectation, in order to remove risk, defines willingness to pay

(WTP) for a full indemnity insurance contract with no deductible.

And the notion of a risk premium is one of the core concepts that different theories of risk

preferences actually agree on. Expected Utility Theory (EUT) posits a psychological pathway in

which aversion to variability drives a risk preference, where variability can be much more than

just variance. Rank‐dependent utility (RDU) posits an additional psychological pathway in

2

General discussions of the methodological implication of a theoretical behavioral research are provided by Harrison (2013, 2014a, 2014b) and Spiegler (2019).

3

Harrison and List (2004) carefully review the taxonomy of different types of experiments. The dictionary always provides a useful check on semantic

confusions. Consider the Oxford English Dictionary (Second Edition), and the definitions of the noun “experiment”in science: “An action or operation

undertaken in order to discover something unknown, to test a hypothesis, or establish or illustrate some known truth.”There is no direct mention of

randomization or causality. The verb “control”is defined in the following manner: “To exercise restraint or direction upon the free action of; to hold sway

over, exercise power or authority over; to dominate, command.”So the word means something more active and interventionist than is suggested by its

colloquial clinical usage. Control can include such mundane things as ensuring sterile equipment in a chemistry lab, to restrain the free flow of germsand

unwanted particles that might contaminate some test. When controls are applied to human behavior, we are reminded by this definition that someone’s

behavior is being restrained to be something other than it would otherwise be if the person were free to act. Thus, we are immediately on alert to be

sensitive, when studying responses from a controlled experiment, to the possibility that behavior is unusual in some respect. The reason is that the very

control that defines the experiment may be putting the subject on an artificial margin. Even if behavior on that margin is not different than it would

otherwise be without the control, there is the possibility that constraints on one margin may induce effects on behavior on unconstrained margins. This

point is exactly the same as the one made in the “theory of the second best”in public policy. If there is some immutable constraint on one of the margins

defining an optimum, it does not automatically follow that removing a constraint on another margin will move the system closer to the optimum.

4

See Cummings, Harrison, and Rutström (1995); Cummings, Ellot, Harrison, and Murphy (1997); and Harrison (2006a, 2014c) for evidence on hypothetical bias

across a range of domains and elicitation methods. There remains a limited role for hypothetical surveys, to explore tentative hypotheses quickly and cheaply,

and as statistical complements to incentivized responses that can be used to calibrate hypothetical responses and correct for biases. See Blackburn, Harrison,

and Rutström (1994) and Harrison (2006b).

134

|

HARRISON AND NG

which probability optimism or pessimism can augment, positively or negatively, any risk

premium due to an aversion to variability. And Cumulative Prospect Theory (CPT) posits yet

another psychological pathway on top of these, where sign dependence relative to some

reference point affects risk preferences. All agree on the same risk premium, and simply

decompose it differently.

Important extensions to these basic insights include considerations of downside risk

aversion that differs from the loss aversion of CPT, and is related to literature on “higher

order risk preferences”; considerations of “regret”or “disappointment”that can arise from

insurance decisions and outcomes; and allowance for multiattribute risk aversion, across

insurance product lines or between foreground and background risk.

Theories of time preference range from Exponential discounting to Hyperbolic and Quasi‐

Hyperbolic models. The differences can best be understood by thinking of the lender of money

as having some cost to not having her money for a time period. Exponential discounting

assumes a constant variable cost with respect to time and no fixed cost; Hyperbolic discounting

assumes a declining variable cost with respect to time and no fixed cost, and Quasi‐Hyperbolic

discounting assumes a fixed cost and a constant variable cost.

5

An alternative approach from

psychology is to view the perception of time horizon as subjective: if the agent perceives time

units contracting as the horizon gets longer, declining discount rates will arise.

Virtually all theories of time preference assume an additive intertemporal utility function, in

which utility over time is a discount factor weighted sumof utility for each distinct period.In this

respect, the alternative theories behind the discount factor tend to agree, and also use an additive

intertemporal utility function. This seemingly technical assumption, however, has dramatic

implications for behavior: it implies that agents are neutral toward risk over time, even if they are

averse to risk at a point in time. In words, agents might be temporally risk neutral to risk resolved

at a point in time but must be intertemporally risk averse to risk resolved over time.

6

Anasty

corollary is that atemporal risk preferences and time preferences are formally “tiedat the hip,”in

the sense that the intertemporal elasticity of substitution must be equal to the inverse of relative

risk aversion. This corollary sits uncomfortably with intuition and the stylized data one

encounters in aggregate data, forcing problematic calibrations in macroeconomic models. A

simple resolutionof this impasse is to allow nonadditive intertemporal utility functions, such that

interactions between atemporal risk aversion between time periods matter to the agent: see

Andersen, Harrison, Lau, and Rutström (2018) for a review of the theory.

The static theory of subjective beliefs is dominated by subjective expected utility (SEU),

which assumes that agents behave as if satisfying the reduction of compound lotteries (ROCL).

The effect is that nondegenerate subjective belief distributions can be replaced by the weighted

average belief, and then EUT applied as usual. It is noteworthy that SEU does not assume that

the subjective belief distributions that agents hold satisfy Bayes Rule when updated over time,

despite Savage being a staunch advocate for each. Bayes Rule is a separate model of (dynamic)

risk perception, which may or may not apply with SEU. Relaxations of ROCL that still assume

that the agent has a well‐defined subjective belief distribution characterize uncertainty, and

models of decision‐making that do not assume a well‐defined subjective belief distribution

characterize ambiguity: see Harrison (2011) for an exposition.

5

So stated, the Quasi‐Hyperbolic assumption is that these fixed costs are some constant fraction of the loan principal, which has the implication that it

stays important for all magnitudes of the loan. An alternative is to assume some fixed cost in terms of money, which of course declines in importance as

loan magnitudes increase.

6

This is a different matter than the agent having preferences over when risk is resolved.

HARRISON AND NG

|

135