Catastrophes and the Demand for Life Insurance
Stephen G. Fier1 and James M. Carson2
Abstract: Prior research suggests that catastrophes may lead to increases in risk
mitigation, risk perception, and the demand for insurance. Given the extensive damage
inflicted by major natural disasters, such a phenomenon is intuitive for property risk.
However, theory and prior empirical evidence also suggest a broader behavioral
perspective and we therefore examine the possible link between catastrophes and
subsequent demand for insurance against mortality risk. Based on U.S. state‐level data,
we provide evidence of a significant positive relation between catastrophes and several
measures of life insurance demand. [Key words: catastrophes, insurance demand,
natural disasters.] JEL classification: C23, D80, G22, Q54.
INTRODUCTION
n April 2011, the United States experienced a record 875 tornadoes
(National Oceanic and Atmospheric Administration, 2011), resulting in
over $3 billion in property damage and hundreds of deaths.3 One month
I
1
The authors acknowledge helpful comments from seminar participants at Ludwig Maxi‐
milian University of Munich, Temple University, University of Nebraska, University of
Georgia, and University of Wisconsin; at meetings of SRIA and ARIA; from members of
Prof. Fier’s dissertation committee, including Patricia Born, Pamela Coates, Cassandra Cole,
Randy Dumm, and Kathleen McCullough; from colleagues at Florida State University, Uni‐
versity of Georgia, and University of Mississippi; and from two anonymous reviewers.
Stephen G. Fier is Assistant Professor and Holder of the Liberto‐King Professorship of
Insurance, Department of Finance, School of Business Administration, The University of
Mississippi, [email protected]
2
Daniel P. Amos Distinguished Professor of Insurance, Department of Insurance, Legal
Studies, and Real Estate, Terry College of Business, The University of Georgia,
[email protected]
3
Among those 875 total tornadoes, the southeastern United States (in particular, Alabama,
Arkansas, Georgia, Louisiana, Mississippi, and Tennessee) was assaulted with over 300 tor‐
nadoes occurring between April 25 and 28, with an estimated 321 deaths and insured losses
ranging from $3.7 to $5.5 billion for the period from April 22 to April 28 (Hemenway, 2011a).
125
Journal of Insurance Issues, 2015, 38 (2): 125–156.
Copyright © 2015 by the Western Risk and Insurance Association.
All rights reserved.
126
FIER AND CARSON
later, on May 22, the city of Joplin, Missouri, was struck by the largest
tornado in over 60 years, resulting in 134 total deaths (UPI, 2011). Total
insured losses from the Joplin tornado were expected to range from $1
billion to $3 billion, with approximately 2,000 buildings destroyed and over
5,000 buildings damaged by the tornado’s winds (Hemenway, 2011b).
Given the destruction that such catastrophes can cause both to health and
to property, one may anticipate that catastrophic events could impact
insurance markets in various ways. For example, Born and Klimaszewski‐
Blettner (2013) provide evidence that insurers’ supply decisions are
affected by natural disasters, and Thomann (2013) shows that insurer stock
price volatility increases following catastrophes. Prior literature also sug‐
gests that natural disasters such as hurricanes, floods, and tornadoes may
cause individuals to adjust their decision‐making processes and reassess
risks that could result in property damage, leading to an increased demand
for an assortment of property‐related insurance contracts (e.g., Browne and
Hoyt, 2000). While the question of how individuals respond to natural disasters has
been of theoretical and empirical interest for property risk, research thus
far has not focused on the potential impact that such catastrophes could
have on the demand for insurance against mortality risk. However, while
on the surface less intuitive, patterns in insurance buying behavior that are
exhibited with respect to property insurance may also occur in the life
insurance market and would support the existence of a relatively broad
behavioral impact. These patterns in the demand for life insurance may
occur as a result of changes in an individual’s cognitive processes and/or
increased marketing efforts and opportunities (e.g., while adjusting prop‐
erty claims) following a catastrophe. Empirical evidence of a link between catastrophes and the demand for
life insurance is important, as it would provide further support for results
in prior research that the effects of catastrophes may extend beyond the
very direct and observed increase in property insurance purchases (e.g.,
increased demand for flood insurance following a flood, as in Browne and
Hoyt, 2000) to a broader behavioral risk management response (e.g.,
increased household savings, as in Skidmore, 2001). In particular for this
study, catastrophic events may serve as catalysts for some consumers to
purchase life insurance rather than to merely consider it. To explore this
question, we examine the demand for life insurance following catastrophes
in the United States based on state‐level data for the period 1997 through
2009. As a preview to our findings, empirical evidence indicates signifi‐
cantly higher post‐catastrophe demand for life insurance in states affected
by catastrophes (CAT states) relative to states not affected by catastrophes
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
127
(non‐CAT states). More specifically, we report a positive and significant
relation between the occurrence of a catastrophic natural disaster in the
prior year and the demand for individual life insurance.
In the following sections, we first review the literature regarding
catastrophes and the demand for insurance, and we provide theoretical
and empirical support for why we might expect to see such a link for life
insurance demand. We then describe the data and empirical methods. Then
we present the model for the demand for life insurance as well as the
results. In the final section, we summarize and conclude the paper. CATASTROPHES AND THE DEMAND FOR INSURANCE
When catastrophes, particularly natural disasters, strike in the United
States, property losses often are large while loss of life is relatively minimal.
The potential losses associated with such catastrophes often can be reduced
through a combination of loss mitigation and the purchase of insurance.4
Although individuals may choose to insure these risks through the use of
various insurance contracts such as homeowners, earthquake, and/or flood
insurance, often the amount purchased is not sufficient to ensure indem‐
nification in the event of an insurable loss. For example, while property
losses from Hurricane Katrina totaled over $81 billion, only $44 billion of
the property damage was covered by insurance (Insurance Information
Institute, 2008).
A number of explanations have been provided as to why individuals
fail to purchase seemingly necessary insurance products or do not take
action to mitigate losses when they are at risk. Kunreuther (1976) argues
that the decision‐making process with respect to insurance purchasing
occurs in four distinct stages, and that if an individual fails to recognize
that a potential problem exists or assumes a probability of occurrence so
low that the event seems improbable, the steps necessary to reduce poten‐
tial losses will not be taken.5 Other researchers concur that if the probability
of loss is low (such as for some catastrophic losses), individuals may drive
their self‐assessed probability of loss to zero (even though the actual
probability is higher), ignore the potential loss, and not purchase insurance
(see Slovic et al., 1977).6 4
Although mitigation efforts may be useful in reducing the expected loss incurred from a
catastrophic event, Kleindorfer and Kunreuther (1999) note that individuals typically resist
adopting these mitigation measures unless the measures are extremely cost effective. Car‐
son, McCullough, and Pooser (2013) provide an empirical examination of several hypothe‐
ses related to the decision to mitigate. 128
FIER AND CARSON
Kunreuther (1984) further explores the failure to purchase insurance
against disasters. He suggests that individuals may rely on past experi‐
ences to determine if a risk is serious enough to warrant the purchase of
insurance, and that individuals will be more likely to purchase insurance
when they know someone else who has purchased coverage or when they
have had conversations with others regarding insurance purchases.7 Gan‐
derton et al. (2000) conduct a set of experiments and find that individuals
are more likely to purchase insurance for low‐probability events when the
cost of insurance is low, the expected loss is high, and the individual is less
wealthy (i.e., they are less likely to be financially secure enough to self‐
insure or retain the loss). Kunreuther and Pauly (2004) provide a theoretical
model to explain why individuals might not insure against low‐probability
events.8 The foregoing discussion indicates that, as a result of various decision‐
making behaviors and cognitive biases, individuals may not take actions
necessary to reduce potential losses and may not purchase insurance prior
to a large‐loss event.9 However, research suggests that some individuals
are moved to purchase property insurance after the occurrence of a cata‐
strophic event. Sullivan, Mustart, and Galehouse (1977) study the aware‐
5
According to Kunreuther (1976), individuals must first perceive the risk as an event that
may potentially result in a loss. Second, individuals must realize that insurance is a viable
coping mechanism. Once individuals realize that insurance is available to manage the loss
associated with potential risks, they can then (third) begin collecting and interpreting insur‐
ance‐related information (terms of insurance). Fourth, the individuals must determine
whether or not insurance is an attractive purchase.
6
Individual treatment of low‐probability events as zero‐probability events is also an impor‐
tant component of Tversky and Kahneman’s prospect theory, which attempts to explain
individual actions that are inconsistent with the axioms of expected utility theory presented
by Von Neumann and Morgenstern (discussed below). 7
Kunreuther also argues that individuals may not be aware of the existence of useful insur‐
ance products, or may perceive the cost of the insurance as too high (see also Sullivan, Mus‐
tart, and Galehouse, 1977; and Palm, Hodgson, Blanchard, and Lyons, 1990).
8
Kunreuther and Pauly argue that individuals are faced with explicit or implicit costs when
making insurance purchasing decisions. When individuals believe that the probability of a
catastrophic event occurring is relatively low, they may behave in a manner consistent with
the probability of occurrence equaling zero, similar to the threshold explanation presented
by Slovic et al. (1977). In addition to the aforementioned reasons provided for the failure of
individuals to mitigate against certain high loss risks, some researchers have posited that
the expectation of federal assistance following a disaster may reduce one’s mitigation efforts
and/or may impact the decision to purchase insurance (e.g., Kaplow, 1991). The concern is
that if citizens are provided with aid following a major disaster, individuals will anticipate
future governmental assistance and will thus discount the value of taking precautions prior
to the occurrence of a catastrophe. 9
For a thorough examination of these potential cognitive biases, see Meyer (2005). CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
129
ness and attitudes of individuals living near the San Andreas fault in
California. Among a variety of issues explored by the authors was whether
those surveyed had purchased earthquake insurance. The authors indicate
that in 1970, only 5 percent of respondents had purchased earthquake
insurance for their residence. In 1976 the authors again surveyed the area
and found that earthquake insurance purchases had increased from 5
percent to 22 percent. While the authors do not specifically provide a reason
for this increase, Lindell and Perry (2000) argue that the increase may be a
result of respondents experiencing the effects of the San Fernando earth‐
quake in 1971.10 Shelor, Anderson, and Cross (1992) examine the impact of the 1989
Loma Prieta earthquake on insurer stock prices. The authors find that stock
prices increased following the earthquake for two samples of insurers (one
sample consisting of property‐liability insurers only, and one sample con‐
sisting of property‐liability and life‐health insurers), particularly for the
two days following the earthquake’s occurrence. The authors conclude that
the positive market response was due to investor expectations of increased
demand for property‐liability insurance in the affected areas. However, the
authors do not discuss the potential relation between the increased insurer
stock prices and the inclusion of life‐health insurers in the second sample.
Browne and Hoyt (2000) assess the factors related to the demand for
flood insurance. The authors note that the occurrence of a disaster has been
shown to increase the awareness of insurance as a need. Evaluating the
period from 1983 through 1993, the authors find that flood insurance
purchasing behaviors are associated with the level of flood losses in a given
state during the prior year. The authors provide evidence that residents of
states that incurred larger flood‐related losses in the previous year had a
tendency to purchase a greater number of flood insurance policies with
greater levels of coverage than those individuals residing in states that had
not experienced such large losses. While property damage is large by definition in the event of a catas‐
trophe, deaths and injuries resulting from catastrophic events are relatively
minimal in the U.S., especially when compared to those that occur on an
international scale. As noted by Bourque et al. (2006), the number of deaths
attributable to natural disasters in the U.S. has declined over the previous
30 years. Despite the trend of declining deaths over time, a level of psycho‐
10
Similarly, Palm, Hodgson, Blanchard, and Lyons (1990) performed a survey‐based study
and evaluated the relationship between the occurrence of an earthquake and the insurance
purchasing behavior of residents in four counties located in California. They suggest that
earthquakes occurring in the early 1980s appeared to be associated with earthquake insur‐
ance purchases in three of the four counties surveyed.
130
FIER AND CARSON
logical distress is associated with the occurrence of a natural disaster. Such
distress may result in cognitive adjustments that lead to changes in the
demand for products that aim to secure the property, health, and financial
assets of individuals. Ganderton et al. (2000, p. 272) state, “The losses in
natural disasters can often be so severe and large that they dominate
people’s assessment of the risk they face.” For example, Skidmore (2001)
presents evidence of greater household savings rates following disasters.
Following a catastrophic event, both “availability” and “retrievability” (see
Tversky and Kahneman, 1974) may grow in importance, leading some
individuals to take “ameliorative measures” (see Sunstein and Zeckhauser,
2011) to manage risk and avoid regret (e.g., Samuelson and Zeckhauser,
1988). Prospect theory (Kahneman and Tversky, 1979; and Tversky and Kah‐
neman, 1992) provides further insight for the observed behavior of indi‐
viduals regarding various financial decisions, especially those involving
insurance purchasing decisions. Prospect theory posits an S‐shaped value
function that is generally concave for gains and is commonly convex for
losses, suggesting risk aversion with respect to gains and risk‐seeking
behavior with respect to losses. Because individuals generally are not able
to comprehend and properly evaluate extreme probabilities, individuals
tend to overweight low probability events and underweight high proba‐
bility events (McDermott, 1998). Accordingly, individuals exhibit a ten‐
dency to overweight the probability of large losses occurring in the future
and purchase insurance (and thus accept a certain small current loss) in
order to avoid the uncertain (low probability) large financial loss. Weinstein (1989) suggests that feelings of worry increase following the
personal experience of a traumatic event, which may then lead individuals
to attempt to protect themselves in various ways from future harm. While
such protection efforts have come in the form of property insurance pur‐
chases and increased mitigation efforts, it is possible that protection‐based
decisions also may come in the form of an increase in life insurance
demand. Those individuals without life insurance (or who do not carry a
“sufficient” amount) may be more likely to reassess their needs and pur‐
chase coverage after witnessing the destruction caused by a catastrophic
event. Furthermore, individuals may feel more inclined to proactively
protect themselves against mortality risk as a result of greater risk aware‐
ness.11 Toward this end, following a catastrophe that has caused significant
property damage, insurers and their agents may strive to increase life
insurance demand not only by virtue of increased marketing efforts in
general, but also as part of the increased contact with individuals during
the property claims process. Thus, individuals may be more likely to
recognize and respond to the need for life insurance at the same time they
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
131
purchase property insurance and/or engage in the property insurance
claim process. Some anecdotal evidence regarding the impact that catastrophes may
have on the demand for life insurance comes from the 1918–1919 influenza
pandemic that resulted in the death of hundreds of thousands in the U.S.
(and millions more around the world). Weisbart (2006, p. 11) notes, “In
1919, stories on the experience of major life insurers routinely reported
record sales in 1918, driven in part by people who came to have a fresh
appreciation of the value of owning life insurance.” While natural catas‐
trophes that affect the United States fortunately do not result in the number
of deaths attributed to the 1918–1919 influenza pandemic, catastrophes
may engender cognitive or market changes similar to those that led to the
increased demand for life insurance in the early twentieth century. Thus,
considering the theoretical literature and existing evidence of changes in
financial behavior following disasters, it would not be surprising to find
evidence that individuals exhibit changes in insurance purchasing behav‐
ior in areas beyond property insurance.
Zietz (2003) provides a survey of the various factors that have been
identified by prior research as determinants of life insurance demand. In
general, many of these factors are associated with significant life changes,
such as the birth of a child, a new job, or simply a change in age.12 In the
same vein, catastrophes may be events with enough significance so as to
be related to an increased demand for life insurance. Given the findings of prior research, which suggests that catastrophic
events and major life events are commonly associated with an increase in
11
That is, the link between a catastrophe and an increase in life insurance demand is not that
individuals necessarily decide to purchase life insurance because they believe they will die
as a direct result of a future catastrophe. Rather, it is likely the case that the catastrophe is an
event that motivates some individuals to more seriously consider the possibility that they
could die an untimely death, which then results in an increase in life insurance demand. 12
Among the factors identified by prior literature as having some influence on the demand
for life insurance are age, education, employment, income, population, life expectancy, mar‐
ital status, number of children, and a variety of psychographic traits (Zietz, 2003). The
important (albeit mixed) relation between many of these variables and life insurance
demand suggests some support for the notion that life insurance consumption is tied to an
individual’s life cycle, where the demand for life insurance is related to significant events
that occur over the span of an insured’s life (e.g., Liebenberg, Carson, and Dumm, 2012).
While we are able to control for some of these factors in our models, we are unable to control
for all of these factors. As discussed below, we are able to control for age, income, popula‐
tion, and dependency (among other controls). However, we do not control for education as
it is highly correlated with income (over 65 percent correlation between college education
and income). We also do not control for state‐specific life expectancy or marital status as we
are unable to locate this information at the state‐level for the entire sample period. 132
FIER AND CARSON
insurance purchasing activities, our primary hypothesis of interest, H1a,
is given as:
H1a: Demand for life insurance following a catastrophic event is signifi‐
cantly higher in CAT states than in non‐CAT states.
We also examine a related hypothesis, H1b, which is:
H1b: Demand for life insurance following a catastrophic event is signifi‐
cantly higher in states that are in close proximity to CAT states than
in states not in close proximity to CAT states.
We contend that consumers may be more likely to purchase insurance
coverage following a naturally occurring catastrophe (H1a). However, it is
also likely that the occurrence of such an event could impact the demand
for life insurance coverage in areas that neighbor affected states. If those in
neighboring states are more likely to receive regular media coverage of the
effects of a large natural disaster or are more likely to know others in the
affected states, then it is possible that the occurrence of a catastrophic event
in a neighboring state could cause non‐affected consumers to re‐evaluate
their needs and purchase insurance coverage. SAMPLE AND DATA
To examine the possible link between catastrophes and the demand
for U.S. life insurance, we focus on major catastrophes. For the purpose of
our analysis, we define a catastrophe as any event affecting the United
States and resulting in large insured property losses (i.e., greater than $1
billion) as identified by Swiss Reinsurance (SwissRe) Company.13 Table 1
lists the 21 catastrophes identified via Swiss Re that are used in the creation
of our binary catastrophe variable. For several reasons, the set of catastro‐
13
Catastrophes are identified via Swiss Re Sigma Reports (similar to Born and Viscusi, 2006).
Each year Swiss Re publishes a Sigma Report that explores the impact of catastrophes, both
in the U.S. and internationally. Included in each publication is a list of the 40 most costly
catastrophes since 1970, based on total insured property damage. Using available Swiss Re
reports, catastrophes are selected for inclusion in this study if at any point during the sam‐
ple period a catastrophe from the period was included on this list. It should be noted that
this list changes each year as catastrophes are added and removed. We include a catastrophe
once it is included on the list and keep it in the sample even if it is subsequently removed in
another year. Since we are interested in determining whether or not a relationship between
catastrophes and life insurance demand exists, we primarily focus on the largest catastro‐
phes for purposes of this particular study. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
133
Table 1. Large Catastrophes Included in the Sample
Date
Peril
Insured loss
States affected
09/05/1996
Hurricane Fran
$1.7B
MD, NC, OH, PA, SC, VA, WV
01/05/1998
Cold spell with ice
and snow
$1.3B
ME, NH, NY, VT
05/15/1998
Wind, hail, and
tornadoes
$1.4B
IA, MN
09/20/1998
Hurricane
georges
$3.6B
AL, FL, LA, MS
05/03/1999
Series of
tornadoes in
Midwest
$1.5B
AL, AR, FL, GA, IL, IN, KS, KY, LA,
MO, MS, NC, NE, OH, OK, SC, TN, TX
09/10/1999
Hurricane Floyd
$2.4B
CT, DE, FL, GA, MA, MD, ME, NC, NH,
NJ, NY, PA, RI, SC, VA, VT
04/06/2001
Hail, floods,
tornadoes
$1.9B
AR, KS, MO, MS, OK, TX
06/05/2001
Tropical storm
Allison
$3.2B
FL, GA, LA, MS, NC, NJ, PA, SC, TX
04/27/2002
Spring storm with
tornadoes
$1.7B
GA, IL, IN, KS, KY, MD, MO, NY, OH,
PA, TN, VA, WV
04/04/2003
Thunderstorms
and hail
$1.6B
AL, IL, IN, LA, MI, MO, MS, NY, TN,
TX
05/02/2003
Thunderstorms,
tornadoes, hail
$3.2B
AL, AR, CO, GA, IA, IL, IN, KS, KY,
MO, MS, NC, NE, OK, SC, SD, TN
09/18/2003
Hurricane Isabel
$1.7B
DE, MD, NC, NJ, NY, PA, VA, WV
08/11/2004
Hurricane Charley
$8B
08/26/2004
Hurricane Frances
$5B
FL, GA, NC, NY, SC
09/02/2004
Hurricane Ivan
$11B
AL, DE, FL, GA, LA, MD, MS, NC, NJ,
NY, OH, PA, TN, VA, WV
09/13/2004
Hurricane Jeanne
$4B
DE, FL, GA, MD, NC, NJ, NY, PA, SC,
VA
08/24/2005
Hurricane Katrina
$45B
AL, FL, LA, MS, TN
09/20/2005
Hurricane Rita
$10B
AL, AR, FL, LA, MS, TN, TX
10/19/2005
Hurricane Wilma
$10B
FL
08/26/2008
Hurricane Gustav
$4B
AL, AR, LA, MS
09/02/2008
Hurricane Ike
$20B
AR, IL, IN, KY, LA, MO, OH, PA, TX
FL, NC, SC
Source: Annual Swiss Re Catastrophe Sigma’s from 1995 through 2009. Insured loss
values are presented as initially reported (not inflation-adjusted) by Swiss Re in the year
of the event. Large catastrophes are identified via annual Swiss Re Catastrophe Sigma’s.
An event is classified as a large catastrophe for the purpose of this study only if it appears
on the list of the 40 most costly catastrophes since 1970 in a given year’s Swiss Re Sigma.
134
FIER AND CARSON
phes included in the sample is based on the size of property damage, as
opposed to the number of deaths or injuries associated with catastrophic
events.14 While the identification of catastrophes is based on the dollar
amount of property damage rather than deaths or injuries, we control for
death and injuries attributed to catastrophes in our models.15 We examine U.S. data for all states (and the District of Columbia) and
all (aggregated) insurers for the period from 1997 through 2009 for a total
of 663 state‐year observations. For purposes of this study, we utilize data
from the individual life insurance marketplace and do not include group
life insurance policies.16 Life insurer–specific data are obtained from the
National Association of Insurance Commissioners (NAIC), while addi‐
tional state‐specific economic and demographic data are from the U.S.
Census and U.S. Census Statistical Abstracts. Deaths and injuries attrib‐
uted to major catastrophes are from the National Oceanic and Atmospheric
Administration (NOAA). An overview of the dependent and independent
variables incorporated in our analysis is provided below. VARIABLES AND EMPIRICAL MODEL
Dependent Variables
To test our hypotheses, we construct five different dependent vari‐
ables, all of which are similar or identical to variables used in prior life
insurance demand literature. The dependent variables for our empirical models are defined as: • ForceAmt is the dollar amount of life insurance in force for a given state
in a given year in thousands, scaled by the state’s population in
thousands.
• ForceNum is the number of life insurance policies in force for a given
state in a given year, scaled by state population in thousands.
14
First, very few catastrophes occur within the U.S. that cause a significant number of
deaths. Second, a large catastrophe that results in many deaths and injuries presumably
should be associated with a large amount of property damage. Finally, because a change in
demand may be due to a change in the perception of risk, property damage resulting from a
catastrophe may be sufficient to induce an increase in the demand for life insurance. 15
We also estimated models using an alternative catastrophe measure equal to the dollar
amount of property damage, as reported through the Spatial Hazard Events and Losses
Database for the United States (SHELDUS) (Hazards and Vulnerability Research Institute,
2012). However, results on the alternative measure were statistically insignificant and thus
are not presented in this study to conserve space.
16
However, we do control for the group life insurance market in our models. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
135
• IssuedAmt is the dollar amount of life insurance issued for a given state
in a given year in thousands, scaled by state population in thousands.17
• IssuedNum is the number of life insurance policies issued for a given
state in a given year, scaled by the state’s population in thousands.18
• Premiums is the dollar amount of direct life insurance premiums
written for a given state in a given year, in thousands, scaled by
population in thousands.19
We create each dependent variable by aggregating data obtained from
the NAIC annual statement across each state.20 For each of the dependent
variables above, our focus is on the ordinary (individual) life insurance
market rather than the group life insurance market. We focus exclusively
on the individual market largely because the individual market should not
constrain the potential demand that follows a catastrophic event, while the
group market could. Specifically, many groups and businesses have eligi‐
bility periods which limit a group member’s ability to purchase coverage.
Given the existence of these periods, it is likely that consumers looking to
purchase coverage in the group market would be limited in their ability to
do so. Hence, our empirical analysis focuses on the individual market.21
17
Pauly et al. (2003) use the log of the face amount of new term life insurance purchased,
similar to our IssuedAmt variable. Similarly, Babbel (1985) utilizes the dollar amount of new
life insurance in force as a dependent variable.
18
IssuedNum and ForceAmt are similar to variables used in Browne and Hoyt (2000) in their
examination of the demand for flood insurance. In addition to Browne and Hoyt (2000),
other research has used these two variables as dependent variables for purposes of examin‐
ing life insurance demand. In particular, variables similar to our IssuedNum were used by
Chen, Wong, and Lee (2001). Browne and Kim (1993) and Beck and Webb (1997) use a vari‐
able similar to our ForceAmt variable. Beck and Webb (1997) use the ratio of the face of
amount of life insurance in force (plus dividend additions) scaled by GDP. Truett and Truett
(1990) use the quantity of life insurance demanded, proxied by the average dollar amount of
individual life insurance per U.S. family. 19
A similar variable is also used by Browne and Kim (1993). 20
The values used to create the dependent variables are obtained from the States Pages
found in the NAIC annual statements. Specifically, the data came from the following
locations in the NAIC annual statements: (1) The ForceAmt variable is constructed using the
dollar amount of coverage in force as of December 31 of the current year (Line 23, Column 2
of the NAIC States Pages using the 2009 annual statement, for example); (2) the ForceNum
variable is constructed using the number of policies in force as of December 31 of the current
year (Line 23, Column 1); (3) the IssuedAmt variable is constructed using the dollar amount
of ordinary coverage issued during the year (Line 21, Column 2); (4) the IssuedNum variable
is constructed using the number of ordinary policies issued during the year (Line 21,
Column 1); and (5) the Premiums variable is constructed using ordinary life insurance
premiums (Line 1, Column 1). 136
FIER AND CARSON
Table 2. Variable Definition and Summary Statistics for Dependent Variables
Variable
Definition
Mean
Std. dev.
ForceAmt
Dollar amount of life insurance in force for a given
state in a given year in thousands, scaled by the
state’s population in thousands
40147.24
12092.32
ForceNum
Number of life insurance policies in force for a given
state in a given year, scaled by the state’s population
in thousands
476.87
115.58
IssuedAmt
Dollar amount of life insurance issued for a given
state in a given year in thousands, scaled by the
state’s population in thousands
5179.87
1364.67
IssuedNum
Number of life insurance policies issued for a given
state in a given year, scaled by the state’s population
in thousands
36.52
13.30
Premiums
Life insurance premiums written for a given state in
a given year in thousands, scaled by the state’s
population in thousands
311.95
101.10
All dependent variables are winsorized at the 1st and 99th percentiles.
Consistent with prior literature (e.g., Babbel, 1985; Browne and Kim, 1993;
Browne and Hoyt, 2000; and Pauly et al., 2003), we estimate the models
using levels of the various dependent variables, described above.22 We posit
a positive relation between each of our dependent variables and the occur‐
rence of a catastrophe in a prior year. Variable descriptions and summary
statistics for our dependent variables are presented in Table 2.23 21
While the focus of this paper is on the individual market, we did estimate each model
using group data. Using the group data, we only find a significant relation between our
catastrophe variable and two of the demand variables (the number of policies in force and
the number of polices issued). Additionally, for those models showing a significant relation
between the CAT variable and demand, the coefficients on the CAT variable are appreciably
smaller than the corresponding coefficients from the models using the ordinary life insur‐
ance data that are present in this study. 22
We also estimated the models based on an alternative form of each of the dependent vari‐
ables (the percentage change of each of the level dependent variables). However, results are
much stronger and more robust based on the levels of dependent variables. Thus, consistent
with prior literature, the results presented are based on levels of the dependent variables.
23
All dependent variables are winsorized at the 1st and 99th percentiles to reduce potential
bias caused by extreme outliers. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
137
Independent Variables
Several independent variables are incorporated in our model in order
to examine the state‐specific demand for life insurance using both previ‐
ously utilized economic/demographic variables as well as our proposed
catastrophe variables.24 Key Explanatory Variables
Catastrophe Variables. The primary variables of interest are the CAT and
Contiguous variables. The CAT variable is a binary variable indicating the
occurrence of a major catastrophe (equal to one if yes, zero otherwise) in a
given state in the prior year, as identified by Swiss Re. The CAT binary
variable allows us to determine whether a significant difference in the
demand for life insurance exists in the year following a catastrophe
between states that did or did not experience a catastrophic event.25 Coef‐
ficients on the CAT variable should be significant and positive if the results
support Hypothesis 1a.26 The Contiguous variable is included in order to determine if those states
that are geographically close to a catastrophic event, but not directly
impacted by the event, experience an increase in life insurance demand,
related to Hypothesis 1b.27 As previously stated, individuals may rely on
past experiences to determine if a risk is serious enough to warrant the
purchase of insurance. The Contiguous variable represents those states that
do not experience a Swiss Re catastrophe but neighbor a state that experi‐
ences a Swiss Re catastrophe. Assuming that an indirect experience may
be sufficient enough to promote a change in the assessment of risk
24
The variance inflation factors (VIFs) were checked for each of the independent variables
employed in the models. None of the independent variables had a VIF greater than 5. Ken‐
nedy (1998) notes that a VIF greater than 10 may be cause for concern. 25
In addition to the CAT and Contiguous variables, we also considered the inclusion of an
interaction term, consisting of states that experience a catastrophe in succeeding years, in
order to determine if states that experience catastrophes in consecutive years display a dif‐
ferent demand for life insurance when compared to states that do not experience catastro‐
phes in consecutive years. The results suggest that states that experience catastrophes in
back‐to‐back years do not have a statistically different demand for life insurance when com‐
pared to other states. In order to maintain focus on the hypotheses of interest, we do not
include the interaction variable in the analysis presented here. 26
It should be noted that two phenomena potentially militate against finding a significant
catastrophe effect. First, several states reappear in Table 1 over time, rendering an ever‐
shrinking pool of people on which the catastrophes have an effect. Second, it is possible that
the states appearing multiple times may be more likely to experience a numbing effect in
relation to their potential response to catastrophe risk and the need for life insurance (see
Meyer, 2005). 138
FIER AND CARSON
(Hypothesis 1b), we anticipate the coefficient for the Contiguous variable
will be significant and positive.
Demographic and Economic Control Variables Although not the primary focus of the study, in addition to the catas‐
trophe‐related variables we also include state‐specific demographic and
economic factors that may be associated with life insurance demand. As
noted previously, control variables included in the regression models are
identified via prior life insurance demand literature. While we do not posit
expected signs for the following variables, the control variables included
in the models are discussed below. Income. Prior literature generally suggests that a positive relation exists
between the demand for life insurance and household income (Zietz, 2003).
Income may have a positive relation with life insurance demand given the
desire to protect the potentially higher level of income in the event of
premature death. Such a relation has been documented in numerous
studies, including in Truett and Truett (1990), Browne and Kim (1993), and
Showers and Shotick (1994). To control for the potential relation between
income and life insurance demand, we include the median level of house‐
hold income for a given state in a given year (Income).28 Homeownership. In addition to the average level of household income,
we also control for the level of homeownership exhibited in the state for
each year in our sample (Homeown).29 Using data from the U.S. Statistical
Abstracts, the Homeown variable represents the proportion of a state’s
population in a given year that owns a home. Population Density. Kunreuther (1984) argues that individuals will be
more likely to purchase property insurance when they know other indi‐
27
The Contiguous variable does not result in double‐counting if the state has already been
affected by a catastrophe. For example, if a hurricane strikes Florida but does not impact
Georgia, Georgia is considered a contiguous state. However, if a hurricane strikes Florida
and misses Georgia, but Georgia then experiences state‐wide flooding in the same year,
Georgia will not be considered a contiguous state. This coding approach is followed in order
to preserve the underlying purpose of the variable, which is to determine if non‐CAT states
experience an increase in the demand for life insurance when in close proximity to CAT
states. 28
We also considered the inclusion of household education and squared‐income as an inde‐
pendent variable. However, collinearity concerns prompted the exclusion of these variables
from the models. When we do include these variables in our models the variance inflation
factor (VIF) exceeds 10, which Kennedy (1998) states is indicative of a multicolliniarity prob‐
lem. However, the inclusion or exclusion of these variables does not influence the coefficient
on the binary CAT variable and the results on the other control variables generally remain
similar to those presented in this paper.
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
139
viduals who have purchased insurance or when they have had conversa‐
tions regarding the purchase of property insurance coverage.30 We control
for this potential by including state‐specific population density (Density)
in our models. The population density variable is constructed as the
population per square mile of land area, as reported in the U.S. Statistical
Abstracts. Child Dependency Ratio. To control for the impact that dependency may
have on the demand for life insurance, we include the child dependency
ratio (CDepend). This variable is constructed as the size of a state’s popula‐
tion under the age of 18 scaled by the state’s population between the ages
of 18 and 64.31 In addition to the child dependency ratio, we also include a variable
that accounts for the proportion of a state’s population over the age of 64
(Aged). Age may impact the demand for life insurance due to life‐cycle
factors or as a result of increased levels of dependency during certain
portions of one’s life.32 Market Competition. The extent of competition within a market can
directly affect the price of insurance as well as the quantity of coverage
available. If a greater degree of concentration exists within a market, the
potential exists for higher prices and, consequently, potentially lower
demand for the life insurance product. Conversely, a lesser degree of
concentration in a market has the potential to drive prices down and
29
Gandolfi and Miners (1996) discuss the potential relation between homeownership and life
insurance demand. Homeownership may lead to an increased demand for life insurance if
individuals hope to cover the cost of the mortgage in the event of death and if the home is
not considered a liquid asset. Conversely, homeownership could result in a reduced
demand for life insurance if individuals view the home as a liquid asset that could be used
as a substitute for life insurance.
30
A greater population density suggests there are larger numbers of individuals living by
one another. Individuals living in states with greater densities will thus have the ability/
opportunity to associate with a greater number of individuals who may have purchased life
insurance coverage, which could encourage them to purchase similar coverage. Addition‐
ally, the potential exists for individuals living in states with higher population densities to
know individuals impacted by catastrophes, which may induce the purchase of life insur‐
ance.
31
The child dependency ratio provides an estimate of the number of dependents that exist
within a given state population. Prior literature suggests that individuals may purchase life
insurance either to protect dependents from financial hardship in the event of premature
death or to simply provide the dependents with an inheritance or bequest (Campbell, 1980;
and Lewis, 1989). Browne and Kim (1993), in their international examination of determi‐
nants of life insurance demand, report a positive and significant relation between the
demand for life insurance and the dependency ratio. Li, Moshirian, Nguyen, and Wee (2007)
also present evidence of a positive relation between the demand for life insurance and the
dependency ratio in their analysis of life insurance demand within OECD countries. 140
FIER AND CARSON
increase insurance purchasing activity. We account for the level of concen‐
tration within a market by including a state‐specific Herfindahl‐
Hirschman Index (HHI). The HHI variable is calculated as the sum of
squared market shares within a given state, based on premiums written
within the state:
n
HHI s ,t =

i=1
Written i ,s ,t 2
 Premiums
---------------------------------------------------------- Premiums Written s ,t 
(1)
where i denotes each insurer writing business in a given state, s, in year t. Regulatory Stringency. In addition to the other control variables men‐
tioned above, we also control for the potential effect that regulatory strin‐
gency could have on the demand for life insurance. State regulatory
stringency could impact the demand for insurance if the regulatory envi‐
ronment affects the price of coverage or the type of coverage that is made
available to consumers. We follow prior literature and control for regula‐
tory stringency using a binary variable (Stringent) to account for prior
approval states (e.g., Barth and Feldhaus, 1999; Viscusi and Born, 2005;
Born and Klimaszewski‐Blettner, 2013).33 Gulf States. While we control for a variety of state‐specific factors, it is
possible that the location of a given state could also influence the demand
for life insurance. In particular, the events in our sample (see Table 1)
largely take place along the U.S. Gulf Coast, suggesting that these states
are more susceptible to naturally occurring catastrophic events. If this is
the case, then the potential exists that individuals living in states along the
U.S. Gulf Coast are more likely to demand insurance if they believe they
are more susceptible to catastrophic losses.34 We include a binary variable
to control for Gulf Coast states (Gulf), where we define Gulf Coast states
as Alabama, Florida, Louisiana, Mississippi, and Texas.35 Policy Surrenders. Prior literature suggests that life insurance lapse
activity is related to the demand for life insurance. If individuals lapse life
32
Lin and Grace (2007) find that the relation between age and life insurance demand changes
over time and varies based on the level of household financial vulnerability. In particular,
the authors find a positive relation between life insurance demand and financial vulnerabil‐
ity and that older individuals have a tendency to use less life insurance to protect against a
given level of financial vulnerability than do younger individuals. 33
We follow Barrese, Lai, and Scordis (2009) and classify the following 19 states as prior
approval states: Alaska, Arizona, Colorado, Delaware, Florida, Georgia, Kentucky, Louisi‐
ana, Missouri, Montana, Nevada, New Hampshire, New Jersey, New Mexico, North Caro‐
lina, Oregon, South Dakota, Vermont, and Wisconsin. 34
We thank two anonymous reviewers for this suggestion. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
141
insurance policies with the intent of replacing the policy with a new policy,
we anticipate a positive relation between surrender activity and life insur‐
ance demand (Outreville, 1990; Fier and Liebenberg, 2013). However, if
consumers are not replacing those policies, then one would anticipate a
negative relation between demand and surrender activity. To control for
the potential relation between the demand for life insurance and life
insurance policy surrenders, we include the ratio of total state‐specific
policy surrender benefits in a given year scaled by the total dollar amount
of life insurance in force in a given year. Group Life Insurance. The demand for individual life insurance policies
may also be a function of the demand for group life insurance. To account
for the potential relation between current year individual life insurance
policy demand and prior year group life insurance policy demand, we
obtain group life insurance data from the NAIC InfoPro database. For each
model, group life insurance demand is accounted for in a similar manner
to the corresponding dependent variable.36 Catastrophe‐Related Control Variables
Catastrophe‐Related Deaths and Injuries. In addition to economic and
demographic control variables, we also include additional catastrophe‐
specific control variables for which we have no a priori expectations. First,
we include both the natural logarithm of the number of deaths and the
natural logarithm of the number of injuries attributed to catastrophes in
each state (Death and Injury).37 These variables are included in each of the
regression models to account for the potential differential effects of catas‐
trophe‐related death and injury on the decision to insure mortality risk.
Prior empirical research that examines the demand for property insurance
following a catastrophe generally accounts for the size of insurable losses.
While such a proxy does provide some indication of the overall destruc‐
35
We also re‐estimated each of our models using a “coastal” variable in place of the Gulf
Coast variable. The “coastal” variable is a binary variable equal to 1 for all coastal states,
where coastal states include the following: Alabama, Alaska, California, Connecticut, Dela‐
ware, Florida, Georgia, Hawaii, Louisiana, Maine, Maryland, Massachusetts, Mississippi,
New Hampshire, New Jersey, New York, North Carolina, Oregon, Rhode Island, South Car‐
olina, Texas, Virginia, and Washington. The results obtained when replacing the Gulf Coast
variable with the coastal variable are both qualitatively and quantitatively similar to those
presented in this study. 36
For example, in a model that uses the dollar amount of individual life insurance in force
(ForceAmt) as the dependent variable, we also include the lagged dollar amount of group life
insurance in force as a control variable (GForceAmt). All group variables are winsorized at
the 1st and 99th percentiles to reduce potential biases induced by extreme outliers. 37
The total numbers of deaths and injuries come from the NOAA. 142
FIER AND CARSON
tiveness of the event, the number of deaths and injuries associated with a
catastrophic event may also be related to the population’s general insur‐
ance‐purchasing response.38 Descriptive statistics for all independent variables are reported in
Table 3. Empirical Model
An important consideration in estimating the model is the potential
impact of state effects and time effects.39 To address the potential bias, we
employ a fixed‐effects approach, as described in Petersen (2009).40 The
general form of the model is represented in Equation (2) as:
Demand i ,t =  +  1 CAT i ,t – 1 +  2 Contiguous i ,t – 1
+  3 Death i ,t – 1 +  4 Injury i ,t – 1 +  5 Income i ,t – 1 +  6 Homeown i ,t – 1
+  7 Density i ,t – 1 +  8 CDepend i ,t – 1 +  9 Aged i ,t – 1 +  10 HHI i ,t – 1
+  11 Stringent i +  12 Gulf i ,t – 1 +  13 Surrender i ,t – 1
+  14 GroupDemand i ,t – 1 +  i ,t
(2)
DEMANDi,t represents the various dependent variables (ForceAmt, Force‐
Num, IssuedAmt, IssuedNum, and Premiums) for each state i in year t. Each
of the dependent variables is scaled by state population in thousands. The
numerators of the ForceAmt, IssuedAmt, and Premiums variables are in
thousands. We next present the results of the regression analyses. 38
The relatively small number of deaths and injuries that occur in the U.S. resulting from
natural disasters suggests that the relation between these two variables (Death and Injury)
and the demand for life insurance may be insignificant.
39
In particular, if the residuals of the estimates between the states are correlated in a given
year or if the residuals of the estimates for a given state are correlated over time, the stan‐
dard errors will be biased downward, thus favoring a finding of statistical significance.
40
The model employed consists of yearly fixed effects with standard errors clustered by state
(see Petersen, 2009). We also considered a two‐way fixed effects model using year and state
fixed effects. However, a two‐way fixed effects regression would not be appropriate, given
our dataset, as it would eliminate too much of the cross‐sectional variation. As per Petersen
(2009), we also estimate the model presented in Equation (2) using standard errors that are
clustered by both state and year. The results on the primary variables of interest that are
obtained from the two‐way cluster models are generally consistent with those presented in
this paper and thus are not reported. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
143
Table 3. Variable Definition and Summary Statistics
for Independent Variables
Variable
CAT
Definition
Binary variable denoting a state that experienced a Swiss Re
catastrophe in the prior year
Binary variable denoting a state that borders a state directly
impacted by a catastrophe in the prior year
The natural logarithm of one plus the number of deaths
attributed to natural catastrophes for a given state in the
prior year
Mean
Std. dev.
0.1915
0.3938
0.1617
0.3685
0.1263
0.4881
0.2811
0.9354
Income
Median household income for a given state in the prior year 43124.93
8034.00
Homeown
Percent of homeowners in the population for a given state
in the prior year
Population per square mile of land area for a given state in
the prior year
Child dependency ratio, calculated as the proportion of a
state’s population under the age of 18 in the prior year scaled
by the proportion of a state’s population between the ages
of 18 and 64 in the prior year
Percent of the population over the age of 64 for a given state
in the prior year
State-specific Herfindahl-Hirschman Index, calculated
using premiums written
Regulatory stringency binary variable equal to 1 for prior
approval states
Binary variable equal to 1 for Gulf Coast states, including
Alabama, Florida, Mississippi, Louisiana, and Texas
Ratio of policy surrender benefits to the total dollar amount
of life insurance in force in the prior year
Dollar amount of group life insurance in force for a given
state in the prior year in thousands, scaled by the state’s
population in thousands
Contiguous
Death
Injury
Density
CDepend
Aged
HHI
Stringent
Gulf
Surrender
GForceAmt
GForceNum
GIssuedAmt
GIssuedNum
GPremiums
The natural logarithm of one plus the number of deaths
attributed to natural catastrophes for a given state in the
prior year
69.11
6.31
358.11
1261.60
40.32
3.93
12.67
1.80
0.0344
0.0396
0.3725
0.4838
0.0980
0.2976
0.0085
0.0024
25775.16
16464.77
29.54
51.57
3581.33
2303.03
Number of group life insurance policies issued for a given
state in the prior year, scaled by the state’s population in
thousands
4.41
25.60
Group life insurance premiums written for a given state in
the prior year in thousands, scaled by the state’s population
in thousands
89.33
74.85
Number of group life insurance policies in force for a given
state in the prior year, scaled by the state’s population in
thousands
Dollar amount of group life insurance issued for a given
state in the prior year in thousands, scaled by the state’s
population in thousands
All group life independent variables (i.e., GForceAmt, GForceNum, GIssuedAmt, GIssuedNum,
and GPremiums) are winsorized at the 1st and 99th percentiles.
144
FIER AND CARSON
EMPIRICAL RESULTS AND DISCUSSION
Univariate Comparisons
We first examine the relation between life insurance demand and the
occurrence of catastrophes in a univariate setting. We compare life insur‐
ance demand across CAT states and non‐CAT states in Table 4 and provide
some initial evidence that CAT states tend to have greater demand for life
insurance in the year following a catastrophe than do non‐CAT states. In
particular, states experiencing a catastrophe in the prior year have a signif‐
icantly greater number of life insurance policies issued (IssuedNum) and a
significantly greater number of life insurance policies in force (ForceNum)
than those states that did not experience a catastrophe in the prior year.
This finding holds for seven of the nine years in our sample characterized
by the occurrence of a catastrophe. In addition to the univariate tests in Table 4, we also illustrate life
insurance demand for states that did and did not experience a major
catastrophic event (identified via Swiss Re) in the prior year in Figure 1.
The figure illustrates that the variable used to proxy life insurance demand
is important in evaluating and interpreting the potential impact that a
catastrophe can have on the demand for life insurance. In particular, the
figure suggests a higher demand for life insurance in states impacted by
catastrophes when using the number of policies in force (ForceNum) or the
number of policies issued (IssuedNum) as a proxy for life insurance
demand. For other measures, the figure does not suggest (in general) a
significant difference in life insurance demand between CAT states and
non‐CAT states when using the dollar amount of life insurance in force
(ForceAmt), the dollar amount of life insurance issued (IssuedAmt), or life
insurance premiums (Premiums) as the proxy for demand. Both Table 4 and
Figure 1 provide initial support for the notion that states affected by a large
catastrophic event may experience higher demand for life insurance rela‐
tive to non‐affected states in the year following the catastrophe.41 41
The same figure for life insurance demand in the year of the event (as opposed to the year
following the event) shows very similar results. Throughout the rest of the paper, we focus
on life insurance demand for the year following catastrophes for three reasons: (1) Browne
and Hoyt (2000) analyze flood insurance demand for the year following flood losses, (2)
many catastrophes occur in the latter part of the calendar year and so life insurance pur‐
chases related to catastrophes likely are finalized in the following year, and (3) results for
both years (year of and year following) are similar.
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
145
Demand for Life Insurance
The results of various forms of the regression model in Equation (2)
using several dependent variables are presented in Table 5. With respect to
the primary variable of interest, the coefficients on the CAT variables are
positive and significant in four of the five models using each of the
independent variables. These results indicate that the demand for life
insurance is significantly higher in states that experienced a major catas‐
trophe (relative to those states that did not experience a catastrophe) in the
year following the catastrophic event (CAT). In terms of economic signifi‐
cance, the results suggest that states experiencing a catastrophic natural
disaster in the prior year issued roughly 4.8 additional policies per thou‐
sand individuals in the state. For a state such as Florida, with a population
around 18 million in 2005, our results suggest that the occurrence of a large
catastrophe would lead to an additional 86,400 life insurance policies in the
following year. Thus, the results presented in Table 5 provide evidence of
a relationship between catastrophes and increased demand for life insur‐
ance, consistent with Hypothesis 1a, and suggest that the increase is
economically significant. While the results generally support Hypothesis 1, the results presented
in Table 5 are mixed with regards to Hypothesis 1b. Specifically, the
coefficient on the Contiguous variable is positive and significant in three of
the five models, providing some initial evidence of a “neighbor effect.”
However, the general lack of statistical significance on the Contiguous
variable (or significance exceeding the five percent level of significance)
causes us to be cautious in interpreting this result. Overall, the results
suggest that there might be a neighbor effect, but the effect appears to be
statistically weak in our models. Beyond the primary variables of interest, we also find that a number
of the control variables have significant relations with the demand for life
insurance. First, we find that states with higher average incomes (Income)
tend to have higher policy face values and lower numbers of life insurance
policies. These results provide evidence that (1) individuals with higher
levels of income likely have a greater need to protect a higher level of
income, and (2) while policies in higher income states have (on average)
higher face values, these states tend to demand relatively fewer actual life
insurance policies than do other states characterized by lower levels of
income. Second, we find that states with a larger proportion of older
individuals (Aged) tend to exhibit a greater demand for life insurance for
all demand variables, excluding the IssuedNum dependent variable. Third,
we generally find a positive relation between the demand for group life
insurance (GDemand variables) and the demand for individual life
CAT
629.65
542.04
545.61
553.49
520.61
531.53
531.20
529.70
504.34
CAT
4195.82
4611.68
5406.29
5460.93
5255.66
5684.74
5834.11
5314.63
4721.55
Year
1997
1999
2000
2002
2003
2004
2005
2006
2009
Year
1997
1999
2000
2002
2003
2004
2005
2006
2009
5388.90
5643.55
5360.98
5597.26
5570.02
5603.43
4969.72
4619.04
3956.90
Non‐CAT
IssuedAmt
401.97
438.75
423.61
392.61
480.15
492.17
433.68
496.38
499.02
Non‐CAT
ForceNum
–667.35**
–328.92
473.13
87.48
–314.36
–142.50
436.57
–7.36
238.92
Difference
102.37***
90.95**
107.59***
138.92***
40.46
61.32*
111.93***
45.66
130.63***
Difference
2009
2006
2005
2004
2003
2002
2000
1999
1997
Year
2009
2006
2005
2004
2003
2002
2000
1999
1997
Year
298.41
288.14
353.62
328.46
301.54
299.70
324.59
292.39
293.82
CAT
39.27
50.93
44.28
43.41
40.00
49.60
44.24
46.85
53.73
CAT
336.63
328.54
303.45
304.85
317.80
313.38
276.09
290.26
274.28
Non‐CAT
Premiums
27.09
30.45
29.54
29.01
35.70
34.39
31.99
39.74
42.04
Non‐CAT
IssuedNum
–38.22*
–40.40*
50.17*
23.61
–16.26
–13.68
48.50**
2.13
19.54
Difference
12.18**
20.48***
14.74***
14.40***
4.30
15.21***
12.25***
7.11
11.69*
Difference
2009
2006
2005
2004
2003
2002
2000
1999
1997
Year
CAT
46156.06
41203.86
46526.48
42605.62
39088.54
37284.98
35291.27
31894.24
28855.86
Table 4. Comparison of Means Across the Sample Period
53639.90
47766.68
43408.66
42113.65
39602.88
39620.66
30763.53
31093.73
27841.24
Non‐CAT
ForceAmt
–7483.84**
–6562.82**
3117.82
491.97
–514.34
–2335.68
4527.74**
800.51
1014.62
Difference
146
FIER AND CARSON
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Variance comparison tests were conducted for each pairing
reported above, and Satterthwaite’s (1946) approximation for degrees of freedom are used in instances where observations exhibit unequal
variances. The annual mean values for each of the dependent variables are reported in Table 1. Mean values for states that experienced a large
catastrophic natural disaster during the prior year (as identified via annual Swiss Re Sigmas) are presented under the column titled “CAT.”
Mean values for states that did not experience a large catastrophic natural disaster during the prior year are presented under the column titled
“Non‐CAT.” The difference between CAT and Non‐CAT states is reported in the “Difference” column. Comparisons are omitted for 1998,
2001, 2007, and 2008 because catastrophic events as defined for purposes of this study did not occur in our sample for 1997, 2000, 2006, 2007,
or 2009. ForceAmt = dollar amount of life insurance in force in thousands, scaled by state population in thousands; ForceNum = number of life
insurance policies in force, scaled by state population in thousands; IssuedAmt = dollar amount of life insurance issued in thousands, scaled
by state population in thousands; IssuedNum = number of life insurance policies issued, scaled by state population in thousands; Premiums =
life insurance premiums written in thousands, scaled by state population in thousands. Dependent variables are calculated using individual
life insurance policies only and do not include group life insurance policy data.
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
147
148
FIER AND CARSON
Fig. 1. Comparison of demand measures across CAT and non‐CAT states.
Note: Catastrophic events as defined for purposes of this study did not occur in our
sample for 1997, 2000, 2006, 2007, or 2009. Each of the variables above is scaled by
state-specific population in thousands (000). The numerator of all dollar value
variables above (ForceAmt, IssuedAmt, and Premiums) is in thousands (000). Each
demand measure presented in Figure 1 is winsorized at the 1st and 99th percentiles.
Dependent variables are calculated using individual life insurance policies only and
do not include group life insurance policy data.
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
149
Table 5. Catastrophes and the Demand for Life Insurance
Variable
ForceAmt
ForceNum
IssuedAmt
IssuedNum
Premiums
CAT
1,828.93**
(758.06)
66.65***
(12.06)
333.03***
(119.05)
4.8317***
(1.14)
8.56
(6.97)
Contiguous
Death
Injury
Income
Homeown
1,650.09
43.04***
237.66*
2.689*
0.0476
(1,017.18)
(14.86)
(129.17)
(1.36)
(9.09)
–130.91
–12.64
58.34
–2.36**
6.83
(556.91)
(9.17)
(100.27)
(1.16)
(5.65)
–4.62*
–169.44
3.64
–54.85
0.90*
(277.63)
(5.64)
(54.05)
(0.5043)
(2.35)
0.8176***
–0.0022
0.1122***
–0.0006***
0.0047***
(0.1275)
(0.0016)
(0.0182)
(0.0002)
(0.0008)
–169.59
6.36***
–13.42
0.6057***
–1.18
(123.68)
(1.7825)
(18.04)
(0.1976)
(0.9544)
Density
–2.84***
0.0113*
0.10
0.0020***
–0.0093*
(0.7798)
(0.0065)
(0.0803)
(0.0006)
(0.0050)
CDepend
431.00**
–5.08
120.88***
–0.5223*
0.9708
(196.43)
(3.37)
(28.85)
(0.2700)
(2.3382)
Aged
1,851.39***
15.36*
175.74***
–1.04
13.1476**
(453.80)
(8.03)
(63.72)
(0.6533)
(5.1758)
HHI
12,652.20
529.01***
2,740.22
–4.25
59.78
(19,675.75)
(163.76)
(2,305.53)
(8.71)
(157.39)
Stringent
Gulf
1,310.07
–10.62
360.52*
1.21
18.14
(1,325.08)
(22.33)
(182.94)
(2.04)
(13.99)
1,914.48
39.90
517.31**
16.42***
12.20
(1,844.60)
(50.97)
(239.64)
(4.89)
(13.86)
–776851.17**
–15,903.34***
–161240.68***
–2,112.99***
–6,803.79**
(308,590.77)
(4,050.86)
(43,966.44)
(414.53)
(2,696.11)
GDemand
0.3643***
0.0031
0.1792***
0.7256***
0.9249***
(0.0545)
(0.0558)
(0.0286)
(0.2117)
(0.0630)
Constant
–27,608.11
237.93
–5,503.39**
64.42***
–46.85
(18,320.31)
(265.42)
(2,612.06)
(20.83)
(169.71)
663
0.8126
663
0.4464
663
0.6698
663
0.6637
663
0.7027
Surrender
Observations
R-squared
*, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Standard
errors are reported in parentheses under the coefficients and are clustered at the state
level. All dependent variables and all group life insurance independent variables are
winsorized at the 1st and 99th percentiles. GDemand represents group demand variables
(GForceAmt, GForceNum, GIssuedAmt, GIssuedNum, and GPremiums). Variable
definitions and summary statistics are provided in Tables 2 and 3.
150
FIER AND CARSON
insurance policies. This finding would suggest that states with a greater
demand for life insurance may exhibit larger demand both in the individ‐
ual and group insurance markets. Fourth, we find consistent evidence of a
negative relation between policy surrender activity and the demand for life
insurance, suggesting that policy surrenders result in decreased demand
for life insurance in the sense that consumers may terminate life insurance
without replacing the coverage. Finally, the results provide some evidence
of a positive and significant relation between Gulf Coast states and life
insurance demand. Specifically, there is some evidence that Gulf Coast
states tend to exhibit a greater demand for life insurance, even without the
occurrence of a natural disaster.42 DISCUSSION
The results from our analyses provide evidence that life insurance
purchasing behavior is related to recent state‐specific catastrophe activity.
Findings indicate that states affected by catastrophes exhibit a demand for
life insurance that is significantly greater than the demand in other states. Expected utility theory would suggest that an individual’s decision to
purchase life insurance should be unrelated to the occurrence of a cata‐
strophic event—the expected utility from the purchase should be equiva‐
lent both prior to and following the event, assuming the pricing of the life
insurance, the insured’s wealth, and the insured’s probability of death are
unaffected by the event. However, prospect theory allows for the potential
that the framing of the option (to purchase or not purchase life insurance)
impacts the individual’s decision making (Tversky and Kahneman, 1986).
Thus, the value that is assigned to the purchase of life insurance by the
insured may differ across pre‐ and post‐catastrophic periods, depending
on how the gains and losses from that decision are “framed.” Tversky and
Kahneman (1974, p. 1127) examine a variety of biases and heuristics that
affect judgment, including an “availability” heuristic and a corresponding
“retrievability” bias. The authors argue that individuals may assess the
probability of an event based on “the ease with which instances or occur‐
rences can be brought to mind.” Although availability may be useful in
42
A number of robustness checks were conducted to ensure the stability of the results, but
are not presented in order to conserve space. To ensure that the events of 9/11 and Hurricane
Katrina are not exclusively driving the results, we re‐estimate the models with the following
omissions: (1) the 2001 observation for New York; (2) the 2002 observation for New York; (3)
the 2005 observation for Louisiana; and (4) the 2006 observation for Louisiana. The results
from each of the four specifications detailed above are generally consistent with the results
presented in Table 5. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
151
determining probability, this heuristic may be biased by “retrievability,”
whereby an event that is easily retrieved appears more likely than an event
that is less easily retrieved, regardless of the actual probabilities of the two
events. It is possible that the ease of event recall increases following
catastrophes, which then leads to reassessment of the potential for loss and
increased life insurance purchasing behavior. The findings also may relate to an adjustment in risk perception
following an event characterized by a great deal of uncertainty. Sunstein
and Zeckhauser (2011, p. 436) argue that individuals have a tendency to
overreact to low‐probability high‐severity events, leading them to “exag‐
gerate the benefits of preventive, risk reducing, or ameliorative measures,”
particularly when faced by vivid and salient events. The occurrence of a
catastrophe may lead individuals to overreact to the event as a result of
what they witness both in person and through various forms of media, and
this overreaction may lead to increased insurance purchasing behavior. Another possible explanation for the higher demand for life insurance
in states affected by major catastrophes relates to the issue of regret (e.g.,
Samuelson and Zeckhauser, 1988). Braun and Muermann (2004) argue that
individuals may choose to purchase insurance not because they necessarily
believe they need it, but rather because they would regret not having the
insurance if an event occurred in which the insurance was needed. From
this perspective, the catastrophic event induces regret‐based concern
regarding the ownership of life insurance that prompts an individual to
make the purchase. The results may also suggest that life insurers direct their marketing
efforts toward those affected by the destruction caused by catastrophes.43
As noted in prior research, individuals have a tendency to purchase greater
levels of property insurance following the occurrence of a catastrophe (e.g.,
Palm et al., 1990; Browne and Hoyt, 2000). Our findings suggest that the
decision to purchase life insurance following a catastrophe may be made
jointly with decisions related to property insurance. Such a decision may
relate to the insurer’s increased marketing efforts in an affected area and/
or to increased interactions with the agent/adjuster during the property
insurance claims process.44 Similarly, given the heightened sensitivity to
43
This explanation seems rooted in the age‐old adage that “Insurance is sold, not bought.”
The findings also may be indicative of what Weisbart described as “a fresh appreciation of
the value of owning life insurance,” as exhibited in the early 1900s.
44
As noted previously, our identification of catastrophic events is based on insured property
losses. As such, the potential exists that insureds that interact with their agents to handle a
property‐related claim resulting from a catastrophe could be persuaded or encouraged to
purchase life insurance during the claims handling process. 152
FIER AND CARSON
natural disasters and their potentially devastating effects among consum‐
ers who have experienced a catastrophe, public service announcements to
encourage mitigation efforts and preparedness for future disasters may
experience greater impact following the occurrence of a catastrophe.
The decision to purchase insurance also may be related to the individ‐
ual’s reassessment of his or her financial position. For example, using a
sample of 15 different OECD countries, Skidmore (2001) finds that the
damage caused by natural disasters is positively correlated with household
savings rates. Thus, the occurrence of a catastrophic event similarly could
lead consumers to recognize potential weaknesses in their financial plan
that ultimately drives them to purchase life insurance. The findings also
are consistent with the tendency for some individuals to underinsure (e.g.,
Bernheim et al., 1999). Following a catastrophic event, underinsured indi‐
viduals may be more motivated to purchase new or additional life insur‐
ance so as to reduce the gap between their needs and their coverage. Given the higher demand for life insurance in CAT states, the results
suggest a number of managerial implications. First, if insurers are not
already increasing their general marketing efforts in catastrophe‐struck
states, the higher demand for insurance in these states suggests that life
insurers should consider such increased marketing efforts in affected
regions. Such an approach would be consistent with “striking while the
iron is hot,” and could further increase sales in affected areas.45 Second,
especially for those insureds with property insurance claims, agents /
adjusters should emphasize the importance of life insurance during the
claims process, thus capitalizing on the previously‐noted increase in the
demand for property insurance and maximizing the value of the agent/
adjuster contact time gained through the claims process.
SUMMARY AND CONCLUSIONS
Major natural disasters adversely impact individuals, firms, states,
and even national economies, causing substantial property damage and
loss of life. The large property losses caused by catastrophes have led prior
literature to focus on the relationship between catastrophes and property
insurance demand. We extend the scope of prior research by evaluating the
relationship between catastrophic events and the demand for life insur‐
45
One caveat to such a strategy is the potential for lapsed or surrendered policies. If individ‐
uals purchase the policies following the occurrence of a catastrophic event and then subse‐
quently decide coverage is unnecessary, insurers may face increasing expenses as the
policies are not held long enough to recoup expenses. CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
153
ance, and we provide empirical evidence that the demand for life insurance
in states directly affected by major catastrophes is significantly higher than
life insurance demand in non‐affected states in the year following the event.
Future research may consider the relation of catastrophes to life insur‐
ance demand in markets outside of the United States. As noted earlier, the
majority of natural disasters that occur in the U.S. do not result in substan‐
tial loss of life. However, outside of the U.S., natural disasters can and do
result in large losses of life, and thus the question arises as to whether a
similar relationship between catastrophes and life insurance demand in
other countries following non‐U.S. catastrophes might be observed. In
addition, future research could attempt to more definitively determine the
extent to which the greater life insurance demand in CAT states docu‐
mented here is due to a change in consumer risk perception, the identifi‐
cation of weaknesses in consumers’ financial plans, the increased market‐
ing efforts of insurers, and/or other factor(s). Finally, future research could
shed light on the characteristics of insurers that experience the greatest
post‐catastrophe increases in life insurance demand. REFERENCES
Babbel, D. F. (1985) “The Price Elasticity of Demand for Whole Life Insurance,”
Journal of Finance 40: 225–239. Barrese, J., G. C. Lai, and N. Scordis (2009) “A Measure of the Stringency of State
Insurance Regulation,” Journal of Insurance Regulation 27: 21–40.
Barth, M. M. and W. R. Feldhaus (1999) “Does Rate Regulation Alter Underwrit‐
ing?” Journal of Insurance Issues 22(1): 26–50. Beck, T. and I. Webb (1997) “Economic, Demographic, and Institutional Determi‐
nants of Life Insurance Consumption Across Counties,” The World Bank Economic
Review 17: 51–88. Bernheim, B. D., L. Forni, J. Gokhale, and L. J. Kotlikoff (1999) “The Adequacy of
Life Insurance: Evidence from the Health and Retirement Survey, National
Bureau of Economic Research Working Paper No. W7372, http://
papers.ssrn.com/sol3/papers.cfm?abstract_id=194613. Born, P. and B. Klimaszweski‐Blettner (2013) “Should I Stay or Should I Go? The
Impact of Natural Disasters and Regulation on U.S. Property Insurers’ Supply
Decision,” Journal of Risk and Insurance 80: 1–36.
Born, P. and W. K. Viscusi (2006) “The Catastrophic Effects of Natural Disasters on
Insurance Markets,” Journal of Risk and Uncertainty 33: 55–72.
Bourque, L. B., J. M. Siegel, M. Kano, and M. M. Wood (2006) “Weathering the
Storm: The Impact of Hurricanes on Physical and Mental Health,” The ANNALS
of the American Academy of Political and Social Science 604: 129–151.
Braun, M. and A. Muermann (2004) “The Impact of Regret on the Demand for
Insurance,” Journal of Risk and Insurance 71: 737–767.
154
FIER AND CARSON
Browne, M. J. and R. E. Hoyt (2000) “The Demand for Flood Insurance: Empirical
Evidence,” Journal of Risk and Uncertainty 20: 291–306.
Browne, M. J. and K. Kim (1993) “An International Analysis of Life Insurance
Demand,” Journal of Risk and Insurance 60: 616–634.
Campbell, R. A. (1980) “The Demand for Life Insurance: An Application of the
Economics of Uncertainty,” Journal of Finance 35: 1155–1172. Carson, J. M., K. A. McCullough, and D. M. Pooser (2013) “Deciding Whether to
Invest in Mitigation Measures: Evidence from Florida,” Journal of Risk and
Insurance 80: 309–327.
Chen, R. , K. A. Wong, and H. C. Lee (2001) “Age, Period, and Cohort Effects of Life
Insurance Purchases in the U.S.,” Journal of Risk and Insurance 68: 303–327. Fier, S. G. and A. P. Liebenberg (2013) “Life Insurance Lapse Behavior,” North
American Actuarial Journal 17: 153–167. Ganderton, P. T., D. S. Brookshire, M. McKee, S. Stewart, and H. Thurston (2000)
“Buying Insurance for Disaster‐Type Risks: Experimental Evidence,” Journal of
Risk and Uncertainty, 20: 271–289.
Gandolfi, A. S. and L. Miners (1996) “Gender‐Based Differences in Life Insurance
Ownership,” Journal of Risk and Insurance 63: 683–693. Hazards and Vulnerability Research Institute (2012) The Spatial Hazard Events and
Losses Database for the United States, Version 9.0. Online database. Columbia, SC:
University of South Carolina. Available from http://www.sheldus.org.
Hemenway, C. (2011a) “AIR: $3.7B to $5.5B Insured Losses from April Tornadoes,”
Property Casualty 360, May 9, www.propertycasualty360.com/2011/05/09/air‐
37b‐to‐55b‐insured‐losses‐from‐april‐tornadoes. Hemenway, C. (2011b) “EQECAT: $1–$3B in Insured Losses from Joplin, Mo.
Tornado,” Property Casualty 360, May 24, www.propertycasualty360.com/2011/
05/24/eqecat‐1‐3b‐in‐insured‐losses‐from‐joplin‐mo‐torna. Insurance Information Institute (2008) Catastrophes, U.S.: The Ten Most Costly Catas‐
trophes, United States, www.iii.org/media/facts/statsbyissue/catastrophes/. Kahneman, D. and A. Tversky (1979) “Prospect Theory: An Analysis of Decision
Under Risk,” Econometrica 47: 263–291. Kaplow, L. (1991) “Incentives and Government Relief for Risk,” Journal of Risk and
Uncertainty 4: 167–175. Kennedy, P. (1998) A Guide to Econometrics (4th edition), MIT Press. Kleindorfer, P. R. and H. Kunreuther (1999) “The Complementary Roles of Mitiga‐
tion and Insurance in Managing Catastrophic Risks,” Risk Analysis 19: 727–738.
Kunreuther, H. (1976) “Limited Knowledge and Insurance Protection,” Public
Policy 24: 227–261.
Kunreuther, H. (1984) “Causes of Underinsurance Against Natural Disasters,” The
Geneva Papers on Risk and Insurance 31: 206–220.
Kunreuther, H. and M. Pauly (2004) “Neglecting Disaster: Why Don’t People Insure
Against Large Losses,” Journal of Risk and Uncertainty 28: 5–21. Lewis, F. D. (1989) “Dependents and the Demand for Life Insurance,” American
Economic Review 79: 452–467. Li, D., F. Moshirian, P. Nguyen, and T. Wee (2007) “The Demand for Life Insurance
in OECD Countries,” Journal of Risk and Insurance 74: 637–652.
CATASTROPHES AND THE DEMAND FOR LIFE INSURANCE
155
Liebenberg, A. P., J. M. Carson, and R. E. Dumm (2012) “A Dynamic Analysis of
the Demand for Life Insurance,” Journal of Risk and Insurance 79: 619–644. Lin, Y. and M. F. Grace (2007) “Household Life Cycle Protection: Life Insurance
Holdings, Financial Vulnerability, and Portfolio Implications,” Journal of Risk
and Insurance 74: 141–173. Lindell, M. K. and R. W. Perry (2000) “Household Adjustment to Earthquake
Hazard: A Review of Research,” Environment and Behavior 32: 461–501.
McDermott, R. (1998) Risk‐Taking in International Politics, The University of Michi‐
gan Press. Meyer, R. J. (2005) “Why We Under‐Prepare for Hazards,” in R. J. Daniels, D. F.
Kettl, and H. Kunreuther, eds., On Risk and Disaster: Lessons from Hurricane
Katrina, University of Pennsylvania Press: 153–174. National Oceanic and Atmospheric Administration (2011) 2011 Tornado Informa‐
tio—Preliminary Tornado Statistics Including Records Set in 2011, www.noaanews
.noaa.gov/2011_tornado_information.html.
Outreville, J. F. (1990) “Whole‐Life Insurance Lapse Rates and the Emergency Fund
Hypothesis,” Insurance: Mathematics and Economics 9: 249–255. Palm, R., M. E. Hodgson, R. D. Blanchard, and D. I. Lyons (1990) Earthquake
Insurance in California, Westview. Pauly, M. V., K. H. Withers, K. Subramanian‐Viswanathan, J. Lemaire, J. C. Her‐
shey, K. Armstrong, and D. A. Asch (2003) “Price Elasticity of Demand for Term
Life Insurance and Adverse Selection,” NBER Working Paper Series, Working
Paper 9925. Petersen, M. A. (2009) “Estimating Standard Errors in Finance Panel Data Sets:
Comparing Approaches,” Review of Financial Studies 22: 435–480. Samuelson, W. and R. Zeckhauser (1988) “Status Quo Bias in Decision Making,”
Journal of Risk and Uncertainty 1: 7–59. Satterthwaite, F. E. (1946) “An Approximate Distribution of Estimates of Variance
Components,” Biometrics Bulletin 2: 110–114. Shelor, R. M., D. C. Anderson, and M. L. Cross (1992) “Gaining from Loss: Property
Liability Insurer Stock Values in the Aftermath of the 1989 California Earth‐
quake,” Journal of Risk and Insurance 59: 476–488.
Showers, V. E. and J. A. Shotick (1994) “The Effects of Household Characteristics
on Demand for Life Insurance: A Tobit Analysis,” Journal of Risk and Insurance
61: 492–502. Skidmore, M. (2001) “Risk, Natural Disasters, and Household Savings in a Life
Cycle Model,” Japan and the World Economy 13: 15–34. Slovic, P., B. Fischhoff, S. Lichtenstein, B. Corrigan, and B. Combs (1977) “Preference
for Insuring Against Probable Small Losses: Insurance Implications,” Journal of
Risk and Insurance 44: 237–258.
Sullivan, R., D. A. Mustart, and J. S. Galehouse (1977) “Living in Earthquake
Country,” California Geology 30: 3–8.
Sunstein, C. R. and R. Zeckhauser (2011) “Overreaction to Fearsome Risks,” Envi‐
ronmental and Resource Economics 48: 435–449. Thomann, C. (2013) “The Impact of Catastrophes on Insurer Stock Volatility,”
Journal of Risk and Insurance 80: 65–94.
156
FIER AND CARSON
Truett, D. B. and L. J. Truett (1990) “The Demand for Life Insurance in Mexico and
in the United States: A Comparative Study,” Journal of Risk and Insurance 57: 321–
328.
Tversky, A. and D. Kahneman (1974) “Judgment Under Uncertainty: Heuristics and
Biases,” Science, New Series 185: 1124–1131.
Tversky, A. and D. Kahneman (1986) “Rational Choice and the Framing of Deci‐
sions,” The Journal of Business 59: S251–S278.
Tversky, A. and D. Kahneman (1992) “Advances in Prospect Theory: Cumulative
Representation of Uncertainty,” Journal of Risk and Uncertainty 5: 297–323. UPI (2011) “Joplin Tornado Death Toll at 134,” www.upi.com/Top_News/US/2011/
06/01/Joplin‐tornado‐death‐toll‐at‐134/UPI‐29341306970369/?spt=hs&or=tn. Viscusi, W. K. and P. H. Born (2005) “Damages Caps, Insurability, and the Perfor‐
mance of Medical Malpractice Insurance,” Journal of Risk and Insurance 72(1):
23–43. Weinstein, N. D. (1989) “Effects of Personal Experience on Self‐Protective Behav‐
ior,” Psychological Bulletin 105: 31–50. Weisbart, S. (2006) “Pandemic: Can the Life Insurance Industry Survive the Avian
Flu?” Insurance Information Institute, http://server.iii.org/yy_obj_data/binary/
773486_1_0/Bird_Flu.pdf. Zietz, E. N. (2003) “An Examination of the Demand for Life Insurance,” Risk
Management and Insurance Review 6: 159–191.