The effect of information disclosure on demand for high‐load insurance

• Author: Benjamin L. Collier,Marc A. Ragin,Johannes G. Jaspersen
• Date: 01 March 2021
• DOI: http://doi.org/10.1111/jori.12308
• Published date: 01 March 2021

J Risk Insur. 2021;88:161–193. wileyonlinelibrary.com/journal/JORI

|

161

Received: 7 December 2019

|

Accepted: 5 February 2020

DOI: 10.1111/jori.12308

ORIGINAL ARTICLE

The effect of information disclosure on

demand for high‐load insurance

Marc A. Ragin

1

|Benjamin L. Collier

2

|Johannes G. Jaspersen

3

1

Department of Insurance, Legal Studies,

and Real Estate, Terry College of Business,

University of Georgia, Athens, GA

2

Department of Risk, Insurance, and

HealthcareManagement, TempleUniversity,

Philadelphia, PA

3

Munich Riskand Insurance Center(MRIC),

Ludwig‐Maximilians‐Universität, Munich,

Germany

Correspondence

Marc A. Ragin, Terry College of Business,

University of Georgia.

Email: mragin@uga.edu

Funding information

Terry College of Business; Center for

Insurance Education and Research;

Department of Risk, Insurance and

Healthcare Management in the Fox School

of Business; The University of

Wisconsin‐Madison BRITE Lab,

Grant/Award Number: PRJ68AP

Abstract

Economists, regulators, and consumer protection

agencies have highlighted the welfare losses for

consumers who purchase high‐load insurance against

modest stakes risks. Mandatory information dis-

closure is a potentially attractive public policy tool

that might improve consumers' choices, but has not

been widely tested in insurance settings. We conduct

an incentive‐compatible insurance demand experi-

ment, in which we manipulate the information dis-

closed to subjects. We test whether any of the three

most commonly suggested disclosures affect in-

surance demand, disclosing either (1) the true prob-

ability of loss, (2) the contract's expected loss, or (3)

the insurer's profit on the transaction. Similar to

consumers in naturally occurring insurance markets,

subjects in the laboratory demonstrate significant

demand for high‐load insurance against modest

stakes. However, we find no effect of any of the three

disclosure treatments on subjects' insurance choices.

We discuss the implications of our results for possible

public policy initiatives in insurance markets.

KEYWORDS

decision‐making under risk and uncertainty, disclosure, financial

advice, insurance demand

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

© 2020 The American Risk and Insurance Association

1|INTRODUCTION

From the perspective of policymakers, consumers frequently make ineffective insurance

choices. They purchase too much of some coverages (e.g., insurance with high loads or for small

stakes risks, Cutler and Zeckhauser, 2004; Sydnor, 2010) and too little of others (e.g., flood or

long‐term care insurance, Zhou‐Richter et al., 2010; Browne et al., 2015). When concurrently

choosing insurance against multiple risks, consumers choose their protection inconsistently

across risk domains (Barseghyan et al., 2011; Einav et al., 2012). In contract menus, they select

financially dominated contracts (Bhargava et al., 2017) and let their choices be guided by inertia

(Handel, 2013). How to improve consumers' insurance choices is an economically important,

but challenging, public policy question (Handel and Schwartzstein, 2018).

Information disclosure is an attractive public policy tool that might help consumers make

better insurance choices. Examples of information disclosures in other decision contexts in-

clude reporting mutual fund management fees, mortgage interest rates in annual percentage

rate (APR) terms, calories on fast food menus, and the harms of smoking on cigarette packages

(for more examples, see the review by Dranove and Jin, 2010). Information disclosures can help

consumers recognize potential mistakes without legislating their choices. Insurance decisions

are informationally intensive—a product's value depends on premiums relative to the prob-

ability and severity of a future potential loss—and so disclosures may be especially useful for

insurance consumers. However, while information disclosures have been shown to influence

consumers' choices in many contexts, their efficacy in insurance markets has not been widely

tested.

In this study, we use an incentive‐compatible laboratory experiment to examine how in-

formation disclosures affect insurance choices. Subjects in the experiment earn money ($20) in

a real‐effort task, and then face a potential loss to their earnings. Subjects incur a loss if they

draw a red chip from a clear jar of red and white chips. To cover this potential loss, we present

subjects with a menu of insurance options that have marginally increasing loads in the amount

of coverage. The “load”for an insurance product is the ratio of premium to the contract's

expected loss (similar to a markup). Loads in our experiment range from 1 (the actuarially fair

rate) to 5 times expected loss. While subjects in the control group must infer the probability of

loss (and consequently, the policies' loadings) from visual inspection of the jar, others receive an

information disclosure, such as the probability of a loss. We then assess the effects of this

disclosure on subjects' insurance decisions.

The insurance contract in our experiment was designed to mirror insurance products

available in consumer markets. The comparably low stakes of our experiment reflect many

types of “product insurance,”which tend to cover modest‐stakes risks (e.g., extended warran-

ties, credit insurance, and cell phone insurance). These policies often have high loads

(Baker and Siegelman, 2013).

1

The market for product insurance is substantial: American

consumers paid $30.5 billion for extended warranties and $7.8 billion for cell phone insurance

1

There is no threshold for what constitutes a “high load,”but insurance products in competitive markets often have

loads less than two—in 2016, the average loads were 1.9 for homeowners insurance and 1.4 for personal auto insurance

(NAIC, 2016). Studying product insurance coverages, Baker and Siegelman (2013) conducted several load calculations,

estimating that rental car insurance has a load of 11.3 and extended warranties have a load of 10.0. The average loads for

credit life and disability insurance (which pay a particular debt if the insured dies or is disabled) were 2.2 and 5.8,

respectively, in 2016 (NAIC, 2016). High loads do not necessarily equate to high profits earned by firms, as loading is

charged to cover nonloss expenses, such as overhead. However, higher loads decrease the benefit/cost ratio of insurance

products for consumers.

162

|

RAGIN ET AL.

in 2013 (Arnum, 2014). Further, the high loading of the insurance policies in our experiment is

also characteristic of certain insurance contracts. For example, Sydnor (2010) documents high

marginal loads when decreasing the deductible on homeowners insurance.

Academics, consumer advocates, and regulators alike have denounced product insurance

(e.g., Birnbaum, 1999,2001; Cutler and Zeckhauser, 2004; Ramsey, 2011) and low deductible

choices also seem similarly financially disadvantageous (Sydnor, 2010). Their concern reflects

the common public policy framework of viewing a consumer's welfare using the expected utility

of her lifetime wealth. Under this framework, purchasing high‐load insurance on modest risks

reduces welfare. Studies frequently recommend addressing this problem by mandating dis-

closure of potentially relevant information to consumers (e.g., Camerer et al., 2003; Bar‐Gill and

Ferrari, 2010; Schwarcz, 2010a).

In addition to whether information is disclosed, what information is disclosed can also

influence consumers' choices (Sunstein and Thaler, 2003). For example, Bertrand and Morse

(2011) found that describing the cost of a payday loan in dollar terms was most effective in

reducing the likelihood of borrowing, while providing information on the APR and repayment

schedule had smaller effects. Consequently, researchers have differed regarding what in-

formation about insurance might be most effective to nudge consumers toward better choices.

Drawing on common disclosure recommendations for insurance products, we provide one of

the following three information disclosures to subjects in our treatment groups:

1. the probability of loss (e.g., Camerer et al., 2003; Nalebuff and Ayres, 2003);

2. the contract's expected loss (e.g., Bar‐Gill and Ferrari, 2010; Schwarcz, 2010a); or

3. the insurer's expected profit on the contract (e.g., Birnbaum, 1999,2001; Kunreuther and

Pauly, 2004).

We also provide subjects with a simple formula relating the probability of a loss, expected loss,

and expected profit. So, while the type of disclosed information differs, subjects in all treatments

can calculate equivalent information. For example, subjects given the insurer's expected profit

can calculate the probability of a loss and vice versa. We examine the effects of each information

disclosure treatment on subjects' insurance decisions.

Insurance choices are some of the most challenging consumer decisions to understand:

while they have been a focus of research for many years, a consensus on the correct descriptive

model(s) has yet to be reached. In the meantime, much of the motivating literature takes a

pragmatic approach to public policy—how can we improve consumers' insurance choices? We

test recommended information disclosures to see whether they affect insurance decisions, in an

effort parallel to the ongoing fundamental research on the underlying decision process. Our

study aligns with a larger discussion on designing effective public policies despite the un-

resolved debate regarding how consumers make their choices (e.g., Chetty, 2015; Handel and

Schwartzstein, 2018).

In our experiment, subjects demonstrate strong demand for our modest‐stakes insurance

products. More than 96 percent of subjects chose to purchase at least one loaded insurance

product, paying an average load of 2.58. Approximately 35 percent of subjects paid a load of 5 at

least once. Subjects spent an average of 11.8 percent of their initial earnings as a risk premium.

Additionally, even though we consider a highly educated subsample of the overall population—

subjects were 134 students at the University of Wisconsin–Madison—12.7 percent of the par-

ticipants opted for a first‐order stochastically dominated (FOSD) option if it was offered.

Behavior in the experiment thus closely paralleled consumers' demand in modest‐stakes,

RAGIN ET AL.

|

163