Printer Friendly

POSTCLAIM UNDERWRITING AND THE VERIFICATION OF INSURED INFORMATION: EVIDENCE FROM THE LIFE INSURANCE INDUSTRY.

INTRODUCTION

Misrepresentation and fraud are important issues across all lines in insurance and as such have been the subject of a great deal of research (e.g., Picard, 1996, 2000, 2009; Crocker and Tennyson, 2002; Derrig, 2002; Tennyson and Salsas-Forn, 2002; Viaene and Dedene, 2004; Schiller, 2006; Krawczyk, 2009). Insurers use a variety of strategies to verify insured information in an effort to minimize misrepresentation and fraud costs. While significant research has been done on the impact of property--casualty claims handling on the insurers productivity and profitability (e.g., Cummins and Sommer, 1996; Epermanis and Harrington, 2006), the impact of claims-handling practices on life insurers has gone virtually unvisited, specifically as it pertains to the use of denied and resisted life insurance claims to verify information. To better understand the timing in which life insurers verify information, this article uses the requirement of life insurers to record all claims that are resisted or denied during the year. (1)

To decrease the time necessary to write policies and increase market share, life insurers are analyzing risks much faster than they have in the past and are, in some cases, significantly limiting initial underwriting (Schuman, 1995). When insurers limit initial underwriting, they likely take the risk of increased claim payments and/or use postclaim underwriting. Postclaim underwriting can be defined as the process whereby an insurer does not assess an insured's eligibility for insurance, according to the risk he/she presents, until after insurance has been purchased and a claim has been made (Cady and Gates, 1999). (2) Some insurers may be opting to forego or reduce the initial underwriting process in favor of postclaim underwriting life insurance in the form of denied or resisted life insurance claims.

Although previous literature has discussed, and in some cases condemned, postclaim underwriting by insurers (Schatz, 1986; Cady and Gates, 1999; Evans, 2011), there has been no empirical tests that have proven that insurers forgo the initial underwriting process in favor of postclaim underwriting. Only two known articles, Weisbart (1976) and Colquitt and Hoyt (1997), have analyzed denied and resisted claims. While this article uses a similar framework to that utilized by Weisbart, we include a theoretical model to explain the use of postclaim underwriting by some insurers. We also expand the sample and time period used, create yearly indices related to frequency and severity of denied and resisted ordinary life claims, as well as test for evidence of postclaim underwriting through the analysis of denied and resisted claims. (3) Thus, our focus is different from Colquitt and Hoyt (1997) who try to determine the validity of the reason given for the denial of the life insurance proceeds.

We first explore the insurer's incentives to use a postclaim underwriting strategy with a theoretical model based on Picard (1996). (4) Picard's model, which investigates the equilibrium of an insurance market where opportunists file fraudulent claims, is modified to apply to an insurer's decision regarding the extent to postclaim underwrite. We consider the impact of the insurer's cost structure and the insured's incentives to misrepresent to develop a framework that describes the cases in which insurers will utilize preloss underwriting, postloss underwriting, or both. The results of this model indicate that there are some equilibria where misrepresentation occurs and some level of postclaim underwriting is optimal.

The theoretical model gives rise to a series of hypotheses related to cases in which the insurers' cost structure would suggest incentives to postclaim underwrite. We analyze whether life insurers utilize the postclaim underwriting process to wait to verify information following the claim. Since there is no direct measure of postclaim underwriting, we use the reduction in initial underwriting expenses coupled with an increase in claims expenses and the recording of denied and resisted claims. (5) Since the insurer will be underwriting the policy following the claim and these postclaim underwriting expenses are recorded as expenses associated with claim investigation, the claim investigation expenses should be higher. (6) We acknowledge that the measure may bias against detecting all postclaims underwriting as a firm may decide to postclaim underwrite, reduce initial underwriting, then not detect a reason to deny or resist the claim during the investigation and not have an increase in claims costs.

This research is important for several reasons. First, it analyzes the insurer's decision to forgo or limit the initial underwriting process (information verification) in favor of a postclaim underwriting process. Second, it considers the asymmetric information implications of the application process, whereby the insured holds informational advantages to the insurer, which leads to the insurer's inability to price the risk appropriately (Schuman, 1994). In addition, this research can be useful to consumers who may prefer to select a company that uses due diligence in the initial underwriting process and eliminates postclaim underwriting for all reasons except blatant fraud. Finally, the ability to identify those insurers that reduce initial underwriting in favor of postclaim underwriting will allow regulators to focus more attention on the postclaim underwriting companies.

As a preview to the results, consistent with expectations, we find empirical evidence that companies with lower underwriting expenses during the initial underwriting process and insurers with higher levels of investigation/claim expenses have an increased number of and a larger dollar volume of denied and resisted claims. This indicates that the companies that deny and resist claims are reducing their initial underwriting and are choosing instead to focus on a postclaim underwriting process. There are also a variety of firm factors including commissions, distribution system, regulation, and firm size that affect the use of postclaim underwriting.

The remainder of this article is arranged as follows. In the next section, we describe the previous literature. This is followed by the theoretical framework. The hypotheses development, data, and methodology are presented in the next section, followed by a section that describes the results. In the final section, we provide the conclusions.

LITERATURE REVIEW

There are three streams of literature that specifically address the issues of this study. The first addresses the underwriting process in life insurance policy issuance and considers the concept of initial underwriting and postclaim underwriting. The second focuses on fraud detection and the process of claims auditing, which have similarities to the process of postclaim underwriting. (7) The third investigates the characteristics of denied and resisted claims.

Initial and Postclaim Underwriting

In the initial underwriting process, insurers typically use due diligence in analyzing and verifying the information provided by the insured and, if necessary, request any additional information prior to the issuance of the policy. The accuracy of the material provided is critical (Schuman, 1995). With the evolution of the Internet, some insurers are analyzing risks much faster than they had in the past and are issuing policies knowing less about whom they insure (Schuman, 1995). These

lower initial underwriting standards may lead to unanticipated losses. Once the policy is written, states limit the insurer's ability to contest claims to the incontestability clause (8) or to cases in which fraud is so blatant or atrocious that the payment of the death proceeds would violate public interest (Rejda, 2011). (9) There are two reasons that insurers will utilize the incontestability clause in order to review the applicant and the information obtained in the initial underwriting process: (1) to look for attempted deceit on the part of the insured (Colquitt and Hoyt, 1997) and (2) to make up for a lax initial underwriting process by using a system of postclaim underwriting.

The postclaim underwriting process is not limited to claims submitted during the period of incontestability. It can also include those claims that qualify to be contested after the period stated in the incontestability clause (as indicated previously). Experts vary on their assessment of postclaim underwriting with some insisting that postclaim underwriting is an "abomination," "opportunistic," and takes advantage of the insured's vulnerabilities (Cady and Gates, 1999). (10) Courts also have made decisions regarding the appropriateness of postclaim underwriting. (11) On the other hand, insurers must be able to rely on the honesty of the applicant in answering application questions (Schuman, 1995) and insurers have the right to utilize the opportunities provided by statute and case law to review life insurance policies for potential fraud and misrepresentation and, if found, to deny and resist coverage when permitted.

Fraud Detection and Auditing

One of the reasons that insurers may investigate life insurance claims is to look for misrepresentations or fraud that occurred in the initial underwriting process (Colquitt and Hoyt, 1997). (12) One way in which insurers can work to mitigate the impact of misrepresentation and fraud is through claims auditing. There are numerous studies that investigate various aspects of auditing insurance claims (e.g., Dionne and St-Michel, 1991; Kaplow, 1994; Schiller, 2006; Picard, 1996, 2000, 2009; Bond and Crocker, 1997; Boyer, 2000; Tennyson and Salsas-Forn, 2002; Dionne, Giuliano, and Picard, 2003; Krawczyk, 2009). Picard (1996) shows that insurers could benefit by transferring monitoring costs (i.e., audit expenses) to a budget-balanced common agency. This common agency would help to alleviate the inefficiencies found in audit strategies for participating insurers. Bond and Crocker (1997) determine that the optimal insurance contract is one that responds with an overpayment for easily verified losses and an underpayment for losses that are more difficult to confirm. (13) Regardless of the audit system in place, the purpose of the system is to act as a deterrence device (Dionne, Giuliano, and Picard, 2003) and the success of such a system is, at least in part, measured by the reduction in payment amounts on audited claims (Tennyson and Salsas-Forn, 2002).

When making a decision related to verification of information through claims auditing and postclaim underwriting, the insurer cannot differentiate between an honest policyholder and a dishonest or opportunistic one, ex ante. The insurer must expend costs to review identified claims, and these costs may be substantial. Ultimately, insurers have three options: (1) not to audit (postclaim underwrite) at all, (2) audit (postclaim underwrite) a few claims, or (3) audit (postclaim underwrite) all claims. Our theoretical model describes the insurer's decision based on the insurer's cost structure, the probability of detecting an opportunist, and the probability of an insured acting in an opportunistic manner.

Denied and Resisted Claims

When purchasing insurance, consumers may assume that the claims-handling processes of life insurers are equivalent (Weisbart, 1976). (14) However, there are three possible outcomes for the life insurance claim: (1) pay the claim in full, (2) deny the claim in its entirety, or (3) negotiate an amount less than the full amount of the policy (Weisbart, 1976). According to Weisbart (1976), there are six reasons that insurers either deny the entire claim or pay an amount less than the full face value of the policy: (1) the contract never went into effect, (2) there was a material misrepresentation or some other form of fraud, (3) the policy was not in force when the death occurred, (4) the claim made is for a benefit that the policy does not provide (i.e., the insured's death is the result of suicide during the first 2 years in force), (5) misstatement of age, and (6) the beneficiary designation is imprecise or contested by other potential beneficiaries. Any one of these six reasons can lead to a denied (no benefit is paid) or resisted (something less than the face value is paid) claim.

Weisbart (1976) and Colquitt and Hoyt (1997) investigate the role of denied and resisted claims in life insurance. Weisbart investigates the demographic qualities of insurers that deny or resist life insurance claims; however, he does not specifically address the issue of postclaim underwriting. His research is limited to life insurers doing business in the state of Georgia and only those policies written in the state of Georgia. His sample included 121 insurers that sold ordinary life insurance continuously from 1962 to 1972. Weisbart acknowledges that under ideal circumstances, any research regarding denied and resisted claims would differentiate between legitimate and illegitimate claims; however, this is not possible by simply reviewing the data on Schedule F of the financial statement. Although Weisbart's empirical results are statistically insignificant, he states that he believes underwriting levels are most likely the cause of the different levels of denied and resisted claims.

Colquitt and Hoyt (1997) look at denied and resisted claims documented in the 1994 annual statements of insurers licensed to do business in the state of Georgia. Specifically, this research looks to investigate the level of fraud in ordinary life and accidental death insurance. Unlike Weisbart (1976), Colquitt and Hoyt attempt to identify those claims that are legitimate. They indicate that although only 43 of the 7,596 denied and resisted claims were specifically identified as fraudulent; numerous others included reasons that "can be viewed as representing claiming behavior that is fraudulent." Although insurers may deny or resist legitimate claims, the incentive of fair dealing, the fear of reputational harm, or the costs of litigation would keep such events low (Colquitt and Hoyt, 1997).

With a more comprehensive and unique data set, this research expands on this prior work and attempts to empirically test the use of claims handling in individual life insurance as a means to postclaim underwrite the ordinary life insurance policy. This research does not specifically question the legitimacy of the claim, but looks to determine if insurer's opt to limit the initial underwriting process and utilize the ability to postclaim underwrite during the period set forth in the incontestability clause and after this period, when appropriate. Specifically, a severity index and a frequency index of denied and resisted claims are created in order to investigate the relationship between these claims, initial underwriting, and postclaim underwriting. If postclaim underwriting exists, a higher level of denied and resisted claims should be found in combination with a lower level of underwriting expenses (an indication of initial underwriting) and a higher level of the investigation/claims-handling expenses (an indication of postclaim underwriting). This would provide some evidence that insurers use these claims in order to reduce initial underwriting in favor of a postclaim underwriting process. This is the first known literature to attempt to empirically prove the existence of postclaim underwriting in life insurance.

THEORETICAL FRAMEWORK

Given the similarities between claim auditing and postclaim underwriting, we adapt Picard's (1996) model, which investigates the equilibrium of an insurance market where opportunists file fraudulent claims. Specifically, Picard's model considers an incomplete information model where the insurer must decide to audit or not to audit, on a postclaim basis. To develop an understanding of the issues surrounding an insurer's decision to postclaim underwrite in life insurance, Picard's property--casualty claims audit model is modified in this article and applied to an insurer's decision regarding the extent to postclaim underwrite life insurance claims. (15) The model presented here differs from Dixit and Picard (2003) in that the insured knows their risk type in this model rather than receiving a signal. It also differs from Picard (2009) in that there is no expectation of good faith on the part of the insured and the contract is not completely voided if misrepresentation is detected with postclaim underwriting. The model is similar to those two in that an insurer offers a menu of contracts (high- and low-risk contracts) with a Rothschild and Stiglitz (1976)-type separating equilibrium with high risks fully insured and low risks partially insured. The purpose of the theoretical model is to show that an equilibrium is possible where some high-risk opportunists choose to misrepresent while some insurers choose to conduct postloss underwriting of policies.

The Buyers

The general framework begins with risk-averse insurance buyers with initial wealth, W. (16) The buyers face the possibility of a loss, L. Unlike Picard (1996), and similar to Dixit and Picard (2003) and Picard (2009), there are both high-risk and low-risk insurance buyers with the probability of loss, [[delta].sub.H or L], with 0 < [[delta].sub.L] < [[delta].sub.H] < 1. Nature decides if an individual is high risk with probability, p (low risk with probability (1 - [pi])).

Following Picard (1996), we assume that the individual insurance buyers can be either an opportunist or honest, with probability determined by nature, [sigma], and (1 - [sigma]), respectively. An opportunist may choose to misrepresent on an application with probability [varies]. As long as the difference between the probabilities of loss between the high risks and low risks is large enough ([[delta].sub.H] is sufficiently larger than [[delta].sub.L]), there is no incentive for a low-risk opportunist to misrepresent (cannot improve their expected utility by misrepresenting as a high risk) so we end up with the distribution of risk and honesty, shown in Table 1.

The Sellers

Life insurance policies are sold by risk-neutral insurance companies. These companies can offer both high-risk and low-risk policies with premiums [P.sub.H] and [P.sub.L] ([P.sub.L] < [P.sub.H]) for a level of coverage [q.sub.H] or [q.sub.L], respectively. The insurance market is assumed to have free entry and be competitive. Insurance companies know the proportions of the population that are honest/opportunists and high/lowrisks, but cannot observe those distinctions at the individual level without underwriting. Insurers have the option of conducting underwriting at the time of application (preloss) underwriting or at the time of claim (postclaim) underwriting. We assume that underwriting always "catches" the misrepresenters. If an insurer decides to conduct postclaim underwriting, they do so randomly with probability p. Preloss underwriting costs r per policy and postclaim underwriting costs k(p) per policy. The postclaim underwriting costs are assumed to be a function of p. More frequent postclaim underwriting would increase the likelihood of being accused of bad faith claims practices, which could result in significant losses to the insurer. At this point, we assume all insurers face the same underwriting costs. That assumption will be relaxed later.

If an insurance buyer is a misrepresenting opportunist and the insurer is preloss underwriting, the buyer is "caught" misrepresenting and the insurer returns the premium (does not provide coverage) and the buyer is without insurance coverage for the remaining stages of the model. (17) If a misrepresenting opportunist is "caught" by postclaim underwriting the coverage amount [q.sub.L] is reduced by some amount m([q.sub.L]).

Insurers have no incentive to underwrite high-risk policies. Even if low risks had an incentive to misrepresent as high risks, this does not provide an incentive to insurers to "catch" them. To model the preloss versus postclaim underwriting decision as mutually exclusive, we assume that if an insurer preloss underwrites it does so for every low-risk policy. However, if the insurer opts for postclaim underwriting it does not have to underwrite every claim, but can choose to underwrite a claim with probability p. Opportunists are aware of the incentives for insurers to either pre- or postloss underwrite, but cannot determine which individual companies are following which underwriting strategy at application.

Stages of the Model

Given this framework, we can now delineate the stages of the model:

1. Stage 1: Nature determines high risk/low risk and honest/opportunist (exogenously determined).

2. Stage 2: Applications are submitted (this is where misrepresentation may take place); high-risk opportunists choose to misrepresent with probability a (endogenously determined).

3. Stage 3: Insurers decide underwriting strategy (preloss vs. postclaim and postclaim level p).

4. Stage 4: States of the world revealed.

High-Risk Policy

The easiest starting point is the high-risk policy. The probability of being high risk is p, with the high risk probability of loss [[delta].sub.H]. All honest high risks (1 - [sigma])[pi] and truthful high-risk opportunists (1 - [alpha])[sigma][pi] buy the high-risk policy. (18)

These policyholders maximize:

EU = (1 - [[delta].sub.H])U(W - [P.sub.H]) + [[delta].sub.H]U(W - [P.sub.H] - L + [q.sub.H]), (1)

subject to the participation constraint:

EU > (1 - [[delta].sub.H])U(W) + [[delta].sub.H]U(W - L). (2)

The insurer maximizes its profit (or minimizes its cost on a per policy basis):

[P.sub.H] - [[delta].sub.H][q.sub.H]. (3)

Since there is no incentive for any low risks to misrepresent, there is no incentive for insurers to underwrite high-risk policies and incur underwriting costs. Therefore, the cost per policy is simply the probability of loss multiplied by the coverage amount. With free entry into the market place, competition drives the profits to 0 so:

[P.sub.H] = [[delta].sub.H][q.sub.H]. (4)

With the premium being equal to expected loss, this yields that honest high risks and truthful high-risk opportunists, being risk-averse utility maximizers, will buy full insurance (L = [q.sub.H]) which brings the expected utility (Equation (1)) to Equation (5) with certainty.

U( W - [P.sub.H]). (5)

Low-Risk Policy

The low-risk policy is more complicated because of the potential misrepresenters. The probability of being low risk is (1 - [pi]), with the low-risk probability of loss [[delta].sub.L]. The low-risk policy may be purchased by all honest low risks (1 - [sigma]) (1 - [pi]), low-risk opportunists (1 - [pi])[sigma] who have no incentive to misrepresent, and misrepresenting high-risk opportunists [varies] [pi][sigma].

Low-Risk Expected Utility Function

To reach an equilibrium, low risks (both honest and opportunists) must participate in the marketplace. Otherwise, there is no low-risk policy offered. So low risks maximize their expected utility:

EU = (1 - [[delta].sub.L])U(W - [P.sub.L]) + [[delta].sub.L]U(W - [P.sub.L] - L + [q.sub.L]), (6)

subject to two constraints, the participation constraint:

EU > (1 - [[delta].sub.L])U(W) + [d.sub.L]U(W - L), (7)

and that low risks are better off not misrepresenting as high risks:

EU>U(W - [P.sub.H]). (8)

As noted earlier, as long as the difference between the probabilities of loss between the high risks and low risks is large enough (both exogenously determined) then:

(1 - [[delta].sub.L])U(W) + [d.sub.L]U(W - L) > U(W - [P.sub.H]). (9)

There is no incentive for a low-risk opportunist to misrepresent as they would just choose not to participate in the insurance market.

If we define [P.sup.max.sub.L] to be the maximum low-risk premium that will still have low risks participate, then:

(1 - [[delta].sub.L])U(W) + [[delta].sub.L]U(W - L) = (1 - [[delta].sub.L])U(W - [P.sup.max.sub.L]) + [[delta].sub.L]U(W - [P.sup.max.sub.L] - L + [q.sub.L]), (10)

and low risks will participate as long as [P.sub.L] [less than or equal to] [P.sup.max.sub.L], where [P.sub.L] is determined by the insurer cost structures discussed below.

Insurer Cost Structures

Insurers have two possible methods for "catching" the high-risk misrepresenters in the low-risk policy market. They can preloss or postclaim underwrite. The cost to the insurer for low-risk policies will then depend on which underwriting strategy they choose to follow.

Preloss Underwriting

Preloss underwriting will have the insurer underwriting every low-risk policy application at a cost of r per policy. (19) Once a misrepresenter is identified, the insurer refunds its premium and does not provide coverage. The preloss underwriting insurer profit is then:

[P.sub.L] - r - [alpha][pi][sigma][P.sub.L] - [[delta].sub.L][q.sub.L], (11)

which can be rewritten as:

(1 - [alpha][pi][sigma])[P.sub.L] - r - [[delta].sub.L][q.sub.L]. (12)

It can also be written in terms of cost (on a per policy basis):

r + [[delta].sub.L][q.sub.L] + [alpha][pi][sigma]r, (13)

where the first two terms in Equation (13) are the cost of the low-risk policyholders (both honest and opportunist) and the third term is the cost of underwriting the high-risk misrepresenting opportunists who are caught and the policy was never issued.

Postclaim Underwriting

The insurer underwrites a claim with probability p. The postclaim underwriting cost is k(p). (20) If a misrepresenter is caught, the coverage amount [q.sub.L] is reduced by m([q.sub.L]). The postclaim underwriting insurer profit on a per policy basis is then:

[P.sub.L] - [delta].sub.L](1 - [pi]) [pk(p) + [q.sub.L]] - [[delta].sub.H][alpha][pi][sigma][p(k(p) + ([q.sub.L] - m([q.sub.L]))) + (1 - p)[q.sub.L]], (14)

which can be rewritten as:

[P.sub.L] - p[[[delta].sub.L](1 - [pi])k(p) + [[delta].sub.H][alpha][pi][sigma](k(p) - m([q.sub.L]))] - [q.sub.L][[[delta].sub.L](1 - [pi]) + [[delta].sub.H][alpha][pi][sigma]], (15)

or just in terms of cost on a per policy basis:

p[[[delta].sub.L](1 - [pi])k(p) + [[delta].sub.H][alpha][pi][sigma](k(p) - m([q.sub.L]))] + [q.sub.L][[[delta].sub.L](1 - [pi]) + [[delta].sub.H][alpha][pi][sigma]]. (16)

The first bracket represents the policies that are underwritten on a postclaim basis. The first term in the first bracket of Equation (16) is the low-risk policies that suffer a loss multiplied by the postclaim underwriting cost; the second term is the high-risk misrepresenting opportunists multiplied by the postclaim underwriting costs minus the recovery. The recovery m([q.sub.L]) must be greater than k(p) for insurers to postclaim underwrite. The second bracket is the coverage amount multiplied by the low-risk policies with losses and high-risk misrepresenter policies with losses.

Therefore, insurers will have one of two cost structures (Equation (13) or (16)) depending on which underwriting strategy they follow (preloss or postclaim). With free entry into the marketplace competition will drive profits to 0, so insurers will need to pick an underwriting strategy (pre- or postclaim) that minimizes their cost, Equation (17): (21)

Min[(r + [[delta].sub.L][q.sub.L] + [alpha][pi][sigma]r), (p[[[delta].sub.L](1 - [pi])k(p) + [[delta].sub.H][alpha][pi][sigma](k(p) - m([q.sub.L]))] + [q.sub.L][[[delta].sub.L](1 - [pi]) + [[delta].sub.H][alpha][pi][sigma]])]. (17)

High-Risk Misrepresenting Opportunist Expected Utility Function

High-risk opportunists will never misrepresent if all insurers are preloss underwriting since they will always get caught. So for misrepresentation to take place, the postloss underwriting insurer cost must be less than or equal to the preloss underwriting cost. In this case, high-risk opportunists choose a to maximize their expected utility based on insurers using postclaim underwriting:

[mathematical expression not reproducible] (18)

Reaching an Equilibrium

An equilibrium with either deterministic preloss underwriting or random postloss underwriting with some degree of high-risk opportunists misrepresenting will be characterized by a system of equations, S([alpha], p), the joint definition of [alpha] and p such that:

1. high-risk opportunists are indifferent between misrepresenting or not (if there is postloss random auditing with probability p); and

2. insurers are indifferent between deterministic preloss underwriting and random postloss underwriting.

Define p so that high-risk opportunists are indifferent between honesty and misrepresenting:

[mathematical expression not reproducible] (19)

where the left-hand side of the equation is the utility from being honest and the right-hand side is the expected utility from misrepresenting.

Solving for p yields:

[mathematical expression not reproducible] (20)

Condition 1 implies that p = p, where the right-hand side of Equation (20) depends on [alpha] through the low-risk premium, [P.sub.L].

Define [alpha] such that insurers are indifferent between preloss underwriting and postloss underwriting; in other words, the costs under either underwriting strategy are equal. The left-hand side of the equation is the cost of preloss underwriting (Equation (13)) and the right-hand side is the cost for postloss underwriting (Equation (16)):

r + [[delta].sub.L][q.sub.L] + [alpha][pi][sigma]r = p[[[delta].sub.L](1 - [pi])k(p) + [[delta].sub.H][alpha][pi][sigma](k(p) - m([q.sub.L]))] + [q.sub.L][[[delta].sub.L](1 - [pi]) + [[delta].sub.H][alpha][pi][sigma]]. (21)

Solving for [alpha] yields:

[mathematical expression not reproducible] (22)

Condition 2 implies that [alpha] = [alpha].

Subject to the low risks being willing to participate ([P.sub.L] [less than or equal to] [P.sup.max.sub.L]).

With many of these variables being determined exogenously (W, L, r, k, [[delta].sub.H], [[delta].sub.L], [pi], [sigma]), there are sets of values that would generate the conditions necessary for an equilibrium S([alpha], p) defined by Equations (20) and (22) above to hold. (22) However, conditions 1 and 2 are restrictive and would likely not happen in practice. Allowing costs to differ by insurer relaxes this restriction.

Insurer-Specific Cost Structures

In the model specified above, the cost of preloss and postclaim underwriting, r and k(p), did not vary by insurer. It is likely, however, that insurers would have differing underwriting costs. For example, larger insurers could have an economy of scale that reduces r. Similarly, some insurers may have higher postclaim costs, k(p), due to a variety of reasons. Examples could include the value of their brand reputation, the nature of their distribution system, and value of the relationship with both their agents and the ultimate consumer. These companies could value these relationships highly and not be willing to jeopardize them with postclaim underwriting.

To reflect these concerns, the above model can be modified to allow for [r.sub.i] and [k.sub.i](p); these are insurer-specific costs of underwriting. Incorporating these two insurer-specific cost structures into Equations (13) and (16), respectively, alter Equation (17):

Min[([r.sub.i] + [[delta].sub.L][q.sub.L] + [alpha][pi][sigma][r.sub.i]), (p[[[delta].sub.L](1 - [pi])[k.sub.i](p) + [[delta].sub.H][alpha][pi][sigma]([k.sub.i](p) - m([q.sub.L]))] + [q.sub.L][[[delta].sub.L](1 - [pi]) + [[delta].sub.H][alpha][pi][sigma]]). (23)

The insurer specific costs of underwriting can then lead to some insurers preloss underwriting while other insurers have utilize postclaim underwriting, thus relaxing the equality restriction in condition 2 necessary to have insurers following differential underwriting strategies.

In summary, this theoretical model indicates that equilibria exist where some insurers opt for postclaim underwriting and that opportunists are willing to misrepresent. This result is consistent with Tennyson and Salsas-Forn (2002). In their investigation into the auditing of insurance claims, they find that the optimal auditing strategy would indicate that each claim be audited with some probability less than one. In other words, auditing and postclaim underwriting (as indicated by the model) should be undertaken at some level between reviewing all claims and reviewing none of the claims. In our sample, 35.4 percent of insurers deny and resist some claims. To understand the differences between insurers that deny and resist claims and those that do not and in order to investigate evidence of postclaim underwriting, an empirical analysis of denied and resisted claims is done.

HYPOTHESES DEVELOPMENT, DATA, AND METHODOLOGY

Following from the theoretical model, there are a series of testable hypotheses based on the insurers cost structure. This section outlines those hypotheses as well as the data utilized in the empirical study and initial observations related to the reasons for denied or resisted claims. The section concludes with a discussion of the measures for denied and resisted claims as well as the variables used in the analysis.

Hypotheses

Main Hypothesis: Presence of Postclaim Underwriting. As noted in Cady and Gates (1999), postclaim underwriting may be identifiable when, after policies are issued and claims have occurred, the insurer reviews the facts and denies or resists the claims based on information that was available or that could have easily been obtained during the initial underwriting process. (23) In order to measure the level of postclaim underwriting, we use a severity index and frequency index, consistent with Weisbart (1976), to measure the level of denied and resisted claims. (24) The severity index is the ratio of the total dollar amount of resisted or denied claims to the total amount of all incurred claims for ordinary life insurance during the year for the insurer. (25) The frequency index is the ratio of the number of resisted or denied claims to the total number of incurred ordinary life claims submitted in the given year for the insurer. For both of these index variables, for this analysis, denied and resisted claims made in previous years that remain open or were denied in the existing year are included.

In order to measure the impact initial underwriting expenses have on the level of denied and resisted claims, we include the general underwriting expenses, excluding commissions, for ordinary life policies divided by the number of new ordinary life policies written during the given year. This underwriting expense variable provides the average amount spent by the insurer to perform the initial underwriting of an individual, ordinary life policy. Therefore, we hypothesize that the likelihood of postclaim underwriting by a life insurance company (as measured by the index measures) is negatively correlated to the initial underwriting expenses (as proxied by the underwriting expense variable).

Further, insurers that postclaim underwrite may have increased expenses during the claim-handling process. The average unpaid claim, loss and loss adjustment expense per life claim submitted is used to measure how much an insurer spends, on average, to settle a claim (either through payment of the policy proceeds or through postclaim underwriting). (26) We hypothesize that the likelihood of postclaim underwriting by a life insurance company (as measured by the indices) is positively correlated with the expenses paid following a claim.

Other Hypotheses: Factors Impacting the Cost Structure of the Insurer. In addition to the presence of postclaim underwriting, it follows from the theoretical model that several aspects of the insurer may alter the insurer's cost structure and hence their decision to postclaim underwrite. First, firms with higher ratings are likely to be more harmed from the impact of bad publicity from postclaim underwriting. A.M. Best determines the rating based on a comprehensive evaluation of the insurer's balance sheet strength, operating performance, and business profile (A.M. Best, 2013). Companies that have a superior to excellent ability to meet their ongoing operations receive ratings of A++, A+, A, and A- (A.M. Best, 2013). Consistent with Pottier and Sommer (1997) and Regan (1997), a dummy variable indicating those insurers receiving one of the superior or excellent ratings in the previous year is included. We hypothesize that insurers that have an A.M. Best rating of superior or excellent in the previous year will be less likely to deny and resist claims and will postclaim underwrite less as indicated by the indices.

Similarly, financial institutions generally sell policies to their customers. This added relationship may result in the financial institution denying or resisting fewer claims. In other words, a negative correlation is expected between the use of financial institutions as a distribution channel and the choice to deny and resist claims and the level of postclaim underwriting.

Insurers writing in multiple lines of business may be hesitant to deny and resist claims, especially if the insured purchases multiple lines of coverage. Additionally, by writing in multiple lines of business an insurer may attain advantages in underwriting, through shared information, thereby reducing the costs associated with the initial underwriting process. This could lead to better initial underwriting of the life insurance policy and, therefore, less reason to postclaim underwrite when a claim occurs. On the other hand, companies that are more diversified may find that their expertise in personnel is spread thin through the underwriting departments of the various lines written. These companies may choose instead to investigate claims with greater scrutiny through a postclaim underwriting process. In order to control for business concentration, a line of business Herfindahl Index (27) (HHI) is included.

Companies that have been in business longer may have less concern with reputational risk associated with postclaim underwriting. In addition, these older firms are more likely to be the ones that have selected positive net present value investments and less risky investments (Krishnaswami and Pottier, 2001), which could lead to greater profitability and financial stability. For these reasons, the years of operation is expected to be positively related to the decision to deny and resist claims and the level of postclaim underwriting (as measured by the indices). (28) Similar to Pottier and Sommer (1997), we include the number of years the firm has been in business as a measure of age.

Larger insurers are considered safer and have a lower level of insolvency risk (BarNiv and Hershbarger, 1990). In addition, larger firms may have economies of scope and scale that may provide more efficient preloss underwriting. This may lead to lower levels of postclaim underwriting. We include firm size, defined as the natural logarithm of net total assets, which determines economies of scope and scale (Pottier and Sommer, 1997). A negative correlation between the size of the insurer and the use of postclaim underwriting (as measured by the indices) and the level of denied and resisted claims is expected.

Specific characteristics of the insurer also are likely to impact the use and level of postclaim underwriting. For instance, stock companies seek profits for their shareholders where mutual companies operate for the benefit of their owner--policyholder (Spiller, 1972). Due to this owner--policyholder relationship, mutual companies may feel added pressure to limit postclaim underwriting. A dummy variable is used to distinguish stock insurers from mutual insurers (Spiller, 1972; Pottier and Sommer, 1997; Greene and Segal, 2004). For the reasons specified above, it is expected that mutual companies would have less incentive to postclaim underwrite, and therefore, they should have fewer denied and resisted claims, both in number and dollar amount.

Insurers are often licensed in multiple states. It is possible that an insurer may be licensed in a state and yet writes no business in this state. Even though no business is written in a given state, the insurer is still subject to the state's laws and statutes. Due to the complexity and differences between the states regulations, insurers may be more likely to pay life insurance claim proceeds, leading to lower levels of the indices. The number of states the insurer is licensed in is included to control for this.

In theory, insurers in heavily regulated states may be more likely to follow a more stringent initial underwriting process, which in turn could lead to a lower level of postclaim underwriting. Previous literature has shown that life insurers domiciled in the state of New York are subject to more restrictive regulation than insurers domiciled in other states (Weisbart, 1976; Pottier and Sommer, 1997; Krishnaswami and Pottier, 2001). Therefore, insurers in New York would be less likely to utilize postclaim underwriting and would do so to a lesser degree, as measured by denied and resisted claims. For this reason, we hypothesize that the likelihood of postclaim underwriting by a life insurance company as well as the level of denied and resisted claims (as measured by the index measures) will be negatively correlated to the level of regulation.

Other Factors. In addition to the factors above, we include a series of controls for factors that are likely to impact the decision to postclaim underwrite. The use of agents, type of agent used, and level of commission the insurer pays to acquire business are potentially related to the insurer's ability to gain adequate information in the initial underwriting phase and therefore the likelihood of postclaim underwriting. A relatively close working relationship between the agent and the applicant and the monitoring capabilities of the agent should reduce the level of denied and resisted claims (Colquitt and Hoyt, 1997). However, the differences between independent agents and captive agents may impact postclaim underwriting. Life insurers whose products are distributed through independent agents display greater risk, suggesting perhaps that market discipline is reduced when policyholders rely on independent agents (Brewer, Mondschean, and Strahan, 1997). In addition, independent agents often have more than one insurer to consider for coverage and exert more information-gathering effort (Regan and Tennyson, 1996). With only one market, the captive agent may provide more precise details in their underwriting process in order to place the coverage. On the other hand, the insurer's use of captive agents has cost advantages, often in the form of lower commissions (Regan and Tennyson, 1996). The result is that a captive agent must sell more policies to reach his desired commission levels, which may reduce the level of due diligence in completing the application process. For these reasons, it is not possible to make an a priori prediction for these distribution channels as it will depend on what effect dominates. In order to address the distribution channel issues identified above, dummy variables are included for independent agents, and captive agents.

Commissions may impact agent behavior and the level of screening. Insurers that pay a higher commission may obtain a greater level of due diligence from the producer in completing the application properly by disclosing all known risk factors of the applicant (Regan and Tennyson, 1996). On the other hand, producers may try to place all risks with insurers that pay higher commissions even if it means limiting the level of disclosure during the application process. To control for these issues, the average commission paid by the insurer for a new ordinary life insurance policy is included. The level of commission will have an impact on election by the insurer to deny and resist claims and the level of denied and resisted claims; however, the direction of this impact is ambiguous.

Premium growth may impact postclaim underwriting by an insurer. According to Weisbart (1976), if two otherwise identical insurers write different levels of new business, then they would have a different percentage of policies subject to the incontestability clause, which could lead to larger levels of postclaim underwriting by the insurer. However, it is possible that a decrease in premium is a reflection of a decline in reputation due to performance, customer service, or possibly related to denied and resisted claims and therefore, due to higher levels of postclaim underwriting. For these reasons an a priori prediction is not made.

Utilizing postclaim underwriting may be a way of doing business for some insurers. Insurers that deny and resist claims in one year may have the propensity to do so in future years. A dummy variable is included to designate insurers that denied and resisted claims in the previous year. We hypothesize that there will be a positive correlation between the insurers choice to deny and resist claims in a previous year and the choice to do so in the subsequent year. (29)

According to Weisbart (1976) and Colquitt and Hoyt (1997), the insurer's lapse ratio is the ratio of the volume of ordinary life that lapsed or was surrendered during the year to the average total volume of ordinary life insurance in force during the year. A high lapse ratio can indicate that the relationship between the agent and the insured is weak and/or that there is poor management of the insurer's underwriting process (Colquitt and Hoyt, 1997). In addition, a high lapse ratio can indicate high costs to the insurer, especially when it occurs among new policies in which the heavy acquisition expenses have not yet been earned (Houston and Simon, 1970). For these reasons, there should be a positive relationship between an insurer's lapse ratio and the decision to deny and resist claims and the level of denied and resisted claims (Colquitt and Hoyt, 1997).

It is possible that insurers writing small life policies will forego the practice of postclaim underwriting. This may be due in part to the number of insurers offering low coverage, guarantee issue, life insurance products, often limited to $25,000 of coverage or less. (30) In the initial underwriting process for these smaller policies there is no medical information requested; therefore, there is little basis for a claim of misrepresentation against the insured (Schuman, 1995). Since these policies have no initial underwriting process, there is little reason for a postclaim underwriting process. (31) Weisbart (1976) supports this theory by stating that insurers that have larger average policy sizes are less likely to have higher levels of denied and resisted claims. On the other hand, the larger the average face amount of the policy, the larger the ultimate claim brought by the beneficiary. These larger claims may be selected for postclaim underwriting to a greater extent than smaller claims. For the reasons stated, it is not possible to make an a priori prediction as the results may depend on the insurer's level of large to small policies as well as the percentage of guaranty issue policies the insurer writes. (32)

In addition, insurers may benefit by being a part of a group of insurers, rather than a single insurer, and it is important to control for the benefits from this systematic relationship (Pottier and Sommer, 1997). Further, we also control for the changes in the market and the financial crisis with year dummy variables. Potential differences among states are captured with the percentage of premium written in each state.

Data

To test for the existence of postclaim underwriting, we create a sample of life insurer financial data from the NAIC and A.M. Best from 2002 to 2010 (2002 is included for some lagged variables). The initial sample includes all U.S. domiciled life insurers. Consistent with Weisbart (1976), we collect the resisted and denied claims data for individual, ordinary life policies from Schedule F of the statutory annual statement. (33) Insurers with no ordinary life insurance in force and those with missing observations for the needed variables are excluded.

The initial sample includes 57,562 observations of denied and resisted, individual, ordinary life claims from 2003 to 2010. (34) Ideally, this research would include only illegitimate or unjustifiable claim disputes. However, it is not always possible to determine which claims are legitimate (Weisbart, 1976). Therefore, we focus on those claims where the insurer identifies material misrepresentation as the reason for the denial or resistance of the claim. This eliminates claims noted as fraud or claims excluded by the contract for reasons such as suicide, murder, disappearance, alcohol, or age-related errors. It also excludes those claims where we were unable to categorize the claim under one of the categories just identified. (35) The focus of this research is to investigate the reasons that life insurers make decisions at the firm level regarding their underwriting and claims-handling process; therefore, we aggregate the denied and resisted claims to the company level. (36)

The total amount of denied and resisted claims for the life insurance industry, which is coded as material misrepresentation, has risen from $284.72 million in 2003 to over $376.95 million in 2010. During the period under review, 22.8 percent of life insurers denied or resisted at least one claim for material misrepresentation. (37) Table 2 shows the number of insurers that deny and resist claims for each year under review. It also shows the total amount of the denied and resisted claims. (38)

Table 3 shows the severity index for those insurers that deny and resist claims. The average fluctuates from alowof25.036 (dollars per $1,000 claims incurred)in2003to a high of 81.582 in 2004. The average for the period under review was 36.67. (39) Table 4 shows the frequency index for those insurers that deny and resist claims. The average for this index (rate per 1,000 policy claims) varies from a low of 16.74 in 2009 to a high of 25.166 in 2004. The average for the period under review was 21.46. (40) In addition, although it varies in a few years, our sample indicates that insurers that deny and resist claims have a larger average face value for the policies written. (41)

As discussed in the theoretical section, insurers must decide whether to preloss underwrite or postclaim underwrite at some level p. According to the theory presented, the probability of being subject to postclaim underwriting is optimal somewhere between 0 and 1 (0<p<1). As previously indicated, the recording of denied and resisted claims on Schedule F of the statutory annual statement is an indication of the outcome of the insurer's investigation of a claim. Not all claims investigated by the insurer will result in a recorded denied and resisted claim; in other words, the insurers' postclaim underwriting may not always catch the misrepre-senters. Therefore, some insurers may engage in the process of postclaim underwriting; however, they would not be identified through our model. Our goal is to determine if those insurers that deny and resist claims are postclaim underwriting as evidenced by a decrease in the initial underwriting expenses and an increase in investigation and claim settlement expense. To investigate this question, we utilize the following Tobit model:

SeverityIndex = [alpha] + [[beta].sub.1]underwritingexpense(lesscomm) + [[beta].sub.2]investigation / claimexpense + [[beta].sub.3]independentagent + [[beta].sub.4]captiveagent + [[beta].sub.5]financialinstit + [[beta].sub.6]commissionexpense + [[beta].sub.7]NYdomiciled + [[beta].sub.8]%preminstates+[[beta].sub.9]AMBest + [[beta].sub.10]increaseinpremium + [[beta].sub.11]previousdenied / resisted + [[beta].sub.12]yeardummies + [[beta].sub.13]lapseratio+[[beta].sub.14]avgpolicysize + [[beta].sub.15]stock + [[beta].sub.16]group + [[beta].sub.17]size + [[beta].sub.18]LOBHHI + [[beta].sub.19]licensedstates + [[beta].sub.20]yrsinoperation + [epsilon] (24)

The equations for this stage will be run for both the severity index and the frequency index. Refer to Table 5 for an explanation of the variables used. Additionally, the table summarizes the variables that have been defined in the section above and shows the expected sign for each variable.

RESULTS

The univariate results in Table 6 provide initial evidence of postclaim underwriting in that firms that deny and resist claims have lower underwriting expenses (an indication of lower initial underwriting) and higher investigation/claim expenses (an indication of postclaim underwriting). In addition, many of the hypotheses related to cost structure are supported. Lower levels of commissions may reduce their due diligence in the application underwriting process when the compensation from the insurer is lower. We also find that as expected, financially strong insurers are more likely to deny and resist claims as evidenced by a superior or excellent A.M. Best rating. In addition, inconsistent with expectations, insurers that deny and resist claims have a lower ordinary life lapse ratio. This will be further explored in a multi-variate framework. With respect to control variables, insurers that deny and resist claims also have a larger average policy size and are older and larger. They also are more likely to be domiciled in New York, which is generally considered to be a more heavily regulated state (Weisbart, 1976; Pottier and Sommer, 1997; Krishnaswami and Pottier, 2001). This result is in direct contradiction to the regulatory hypothesis and will be investigated further under the multivariate framework.

Next, we look at the results of our multivariate model to test the hypotheses outlined in the prior section. The results can be seen in Table 7. (42) We test for autocorrelation utilizing the Wooldridge test and there is no evidence of autocorrelation. We also test for multicollinearity. The variance inflation factors are below 4.0 for the variables in the models presented. We also test for heteroskedasticity, which is present and is corrected using robust standard errors.

In the multivariate framework, we find evidence of postclaim underwriting based on both the frequency and the severity indices based on the fact that the insurers that have lower underwriting expenses and those with higher investigation/claim expenses have a greater number and a greater average dollar amount of denied and resisted claims. The lower expenses in the initial underwriting process, coupled with an increased level of investigation/claim expenses, is an indication that these insurers are limiting the underwriting of policies during the application process and prior to policy issuance and are, instead, utilizing the postclaim underwriting process. A joint test of the significance of the two variables reveals that while they are both significant, the sum of the two variables is not significantly different from zero. In other words, while both are significant, it appears that the savings from the reduced preloss underwriting is offset by the increase in claims costs. This could indicate that one strategy does not dominate the other.

Similarly, our cost structure hypotheses are also supported. For example, financially stable insurers with a stronger reputation value are less likely to deny and resist claims. Further, insurers that utilize financial institutions deny and resist claims to a lesser extent (both in number and size of claims denied and resisted). This supports the theory that financial firms sell to their customers and in order to maintain this relationship, will deny and resist fewer claims. Additionally, factors related to insurers' size and business mix and organizational structure also matter, as suggested, with both the severity index and the frequency index indicating that smaller insurers and those that have been in business for fewer years deny and resist claims to a greater extent. In addition, the more concentrated the lines of business written by an insurer, the higher the percentage of denied and resisted claims. Interestingly, the frequency index indicates that the number of denied and resisted claims is higher for stock insurers that elect to deny and resist claims. The severity index indicates that the number of states the insurer is licensed in is positively correlated with the total amount of denied and resisted claims.

Other controls also are significant. We find that the type of agent matters as well as commission structure, with higher numbers of denied and resisted claims associated with the services of independent agents and insurers that pay lower commissions have a greater number and a greater average dollar amount of denied and resisted claims.

As previously indicated, there is significant literature that has shown that life insurers domiciled in the state of New York are subject to more restrictive regulation than insurers domiciled in other states (Weisbart, 1976; Pottier and Sommer, 1997; Krishnaswami and Pottier, 2001). Surprisingly, and in contradiction to the regulatory hypothesis, the results indicate that the number and dollar amount of denied and resisted claims increases for those firms domiciled in New York. This may indicate that the regulation seen by previous literature does not apply as strongly to this area of the life insurance industry, or more likely, it indicates that the insurers in New York that deny and resist claims feel with some certainty that they can justify their actions to the regulators. In addition, insurers may feel that the increased level of regulatory scrutiny in New York applies to their claims-handling practice. Therefore, they may investigate more of their claims, which could lead to increased levels of denied and resisted claims. (43)

Consistent with the expectation that the level of premium growth may impact the level of postclaim underwriting for an insurer, we find that insurers with higher premium growth have a greater number of denied and resisted claims. Premium growth indicates new business, which could be subject to the incontestability clause, which in turn could lead to larger levels of postclaim underwriting by the insurer (Weisbart, 1976). The frequency index also indicates that insurers that denied or resisted claims in the previous year will deny or resist a greater number of claims than those that did not deny or resist in the previous year. This is consistent with the theory that insurers that deny and resist in previous years will have the propensity to do so in future years.

The level of the ordinary life lapse ratio increases with both the number of claims and the amount of claims denied and resisted. The increasein lapse ratio may be a result of the damage to the insurer's reputation due to the denied or resisted claims. Certainly, a professional and attentive producer would be aware if a company within his or her sales area has a history of denying or resisting claims and that producer would carefully use this information during sales calls. This result implies that the insurer's reputation impacts the level of denied and resisted claims. In addition, the frequency index indicates that insurers that deny and resist a greater number of claims write larger average face value policies.

CONCLUSION

The purpose of this research is to investigate the information verification strategies of insurers as well as their potential use of postclaim underwriting through denied and resisted claims. A theoretical model is developed that shows that equilibria exist where opportunists may misrepresent and some level of postclaim information verification (postclaim underwriting) is optimal. Given the negative implication of postclaim underwriting (e.g., Schatz, 1986; Shuman, 1995, 1999; Cady and Gates, 1999), it is important to understand whether denied and resisted claims are being used in a postclaim underwriting strategy. Results based on the frequency and severity of denied and resisted claims support evidence of postclaim underwriting as these insurers also have a lower level of initial underwriting expenses and a higher level of investigation/claim settlement expenses. Combined with higher levels of denied and resisted claims, this provides evidence of a postclaim underwriting strategy. Results also indicate that there are financial, operational, and firm characteristic differences that alter the cost structure of insurers and make them more (or less) likely to postclaim underwrite.

The issues related to postclaim underwriting in life insurance impact policyholders, insurers, stakeholders, and regulators. Consumers generally would not expect the insurer to revisit the underwriting process after a claim is submitted. Furthermore, it is at claims settlement time where consumers want fair dealing with the insurer and postclaim underwriting techniques could appear to be bad faith of the insurer's fiduciary responsibility to the beneficiaries. In addition, and probably most critical to consumers is that denied and resisted claims are left to beneficiaries to deal with during, what is often, an emotional period of time.

Stakeholders may also want to understand to what extent the insurer denies and resists claims. The reputational risks associated with denying and resisting life insurance claims has already been addressed, but the impact of this on the shareholders of stock companies has not been overlooked. From a pure public policy standpoint, shareholders will need to determine if the benefits obtained through postclaim underwriting, as measured by the denied and resisted claims, outweigh the costs associated with the possible reputational harm.

Regulators should also be interested in this topic not only because of the social aspect of the implication of denying and resisting claims, but also because of the issue of insurers utilizing death proceeds as a way to reduce their initial underwriting responsibilities and duties.

REFERENCES

Best A.M., 2013, Guide to Best's Financial Strength Ratings. Availableathttp://www.ambest.com/ratings/guide.pdf (accessed October 31, 2013).

BarNiv, R., and R. A. Hershbarger, 1990, Classifying Financial Distress in the Life Insurance Industry, Journal of Risk and Insurance, 54(1): 110-136.

Bond, E. W., and K. J. Crocker, 1997, Hardball and the Soft Touch: The Economics of Optimal Insurance Contracts With Costly State Verification and Endogenous Monitoring Costs, Journal of Public Economics, 63: 239-264.

Boyer, M. M., 2000, Centralizing Insurance Fraud Investigation, Geneva Papers on Risk and Insurance Theory, 25: 159-178.

Brewer, E., III, T. S. Mondschean, and P. E. Strahan, 1997, The Role of Monitoring in Reducing the Moral Hazard Problem Associated With Government Guarantees: Evidence From the Life Insurance Industry, Journal of Risk and Insurance, 64(2): 301-322.

Cady, T. C., and G. L. Gates, 1999, Post Claim Underwriting, West Virginia Law Review, 102: 809-868.

Colquitt, L. L., and R. E. Hoyt, 1997, Determinants of Corporate Hedging Behavior: Evidence From the Life Insurance Industry, Journal of Risk and Insurance, 64(4): 649-671.

Crocker, K., and S. Tennyson, 2002, Insurance Fraud and Optimal Claims Settlement Strategies, Journal of Law and Economics, 45: 469-507.

Cummins, J. D., and D. W. Sommer, 1996, Capital and Risk in Property-Liability Insurance Markets, Journal of Banking and Finance, 20(6): 1069-1092.

Derrig, R. A., 2002, Insurance Fraud, American Risk and Insurance Association, 69(3): 271-287.

Dionne, G., F. Giuliano, and P. Picard, 2003, Optimal Auditing for Insurance Fraud, Working Paper, 03-29.

Dionne, G., and P. St-Michel, 1991, Workers' Compensation and Moral Hazard, Review of Economics and Statistics, 73(2): 236-244.

Dixit, A., and P. Picard, 2003, On the Role of Good Faith in Insurance Contracting, in: R. Arnott, B. Greenwald, R. Kanbur, and B. Nalebuff, eds., Economics in an Imperfect World, Essays in Honor of Joseph Stiglitz (Cambridge, MA: MIT Press), pp. 17-34.

Epermanis, K., and S. E. Harrington, 2006, Market Discipline in Property/Casualty Insurance: Evidence From Premium Growth Surrounding Changes in Financial Strength Ratings, Journal of Money, Credit and Banking, 38(6): 1515-1544.

Evans, D., 2011, Accidental Death Becomes Suicide When Insurers Dodge Payouts, Bloomberg Markets Magazine.

Fleming, D., 2013, Senior Insurance Reporting Analyst, Financial Regulatory Department, National Association of Insurance Commissioners, E-mail communication, November 12.

Gann, J., 2013, National Association of Insurance Commissioners, E-mail communication, February 12.

Goodman, O. R., 1968, Public Policy and the Age and Incontestable Clauses in Life Insurance Contracts, Journal of Risk and Insurance, 35(4): 515-535.

Greene, W. H., and D. Segal, 2004, Profitability and Efficiency in the U.S. Life Insurance Industry, Journal of Productivity Analysis, 21: 229-247.

Houston, D.B., and R. M. Simon, Economies of Scale in Financial Institutions: A Study in Life Insurance, Econometrica, 38(6): 856-864.

Ingalls v. Paul Revere Life Ins. Group, 561 N.W. 2d 273, 285 (N.D. 1997).

Kaplow, L., 1994, Optimal Insurance Contracts When Establishing the Amount of Losses Is Costly, Geneva Papers on Risk and Insurance Theory, 19(2): 139-152.

Krawczyk, M., 2009, The Role of Repitition and Observability in Deterring Insurance Fraud, Geneva Risk and Insurance Review, 34: 74-87.

Krishnaswami, S., and S. Pottier, 2001, Agency Theory and Participating Policy Usage Evidence From Stock Life Insurers, Journal of Risk and Insurance, 68(4): 659-683.

Lewis v. Equity National Life Insurance Co., 637 So. 2nd 183 (Miss. 1994).

McDowell, B., 1984, The Misrepresentation Defense in Insurance: A Problem for Contract Theory, Connecticut Law Review, 16: 513-540.

National Association of Insurance Commissioners, 2010 Life & Annuity Market Conduct Annual Statement. Available at http://www.naic.org/documents/industry_mcas_2010_la_data_call_elements_definitions.pdf (accessed December 31, 2013).

Picard, P., 1996, Auditing Claims in the Insurance Market With Fraud: The Credibility Issue, Journal of Public Economics, 63: 27-56.

Picard, P., 2000, Economic Analysis of Insurance Fraud, in: G. Dionne, ed., Handbook of Insurance (New York: Springer-Verlag), pp. 315-363.

Picard, P., 2009, Costly Risk Verification Without Commitment in Competitive Insurance Markets, Games and Economic Behavior, 66(2): 893-919.

Pottier, S. W., and D. W. Sommer, 1997, Agency Theory and Life Insurer Ownership Structure, Journal of Risk and Insurance, 64(3): 529-543.

Regan, L., 1997, Vertical Integration in the Property-Liability Insurance Industry: A Transaction Cost Approach, Journal of Risk and Insurance, 64(1): 41-62.

Regan, L., and S. Tennyson, 1996, Agent Discretion and the Choice of Insurance Marketing System, Journal of Law and Economics, 39: 637-666.

Rejda, G. E., 2011, Principles of Risk Management and Insurance, 11th edition (Upper Saddle River, NJ: Prentice Hall).

Rothschild, M., and J. E. Stiglitz, 1976, Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information, Quarterly Journal of Economics, 90: 630-639.

Salzman, G., 1969, The Incontestable Clause in Life Insurance Policies, Insurance Law Journal, 142-166.

Schatz, B., 1986, The AIDS Insurance Crisis: Underwriting of Overreaching? Harvard Law Review, 100(7): 1782-1805.

Schiller, J., 2006, The Impact of Insurance Fraud Detection Systems, Journal of Risk and Insurance, 73(3): 421-438.

Schuman, G., 1994, Misrepresentation of Smoking History in Life Insurance Applications, Tort and Insurance--Law Journal, 30(1): 103-140.

Schuman, G., 1995, Health and Life Insurance Applications: Their Role in the Claims Review Process, Defense Counsel Journal, 62: 225-243.

Spiller, R., 1972, Ownership and Performance: Stock and Mutual Life Insurance Companies, Journal of Risk and Insurance, 39(1): 17-25.

Tennyson, S., and P. Salsas-Forn, 2002, Claims Auditing in Automobile Insurance: Fraud Detection and Deterrence Objectives, Journal of Risk and Insurance, 69(3): 289-308.

Viaene, S., and G. Dedene, 2004, Insurance Fraud: Issues and Challenges, Geneva Papers on Risk and Insurance, 29(2): 313-333.

Weisbart, S. N., 1976, Contested Claims in Life Insurance: An Explanatory Study, Journal of Risk and Insurance, 53(4): 611-627.

Jill M. Bisco is at the Department of Finance, University of Akron, Akron, Ohio. Bisco can be contacted via e-mail: jbisco@uakron.edu. Kathleen A. McCullough and Charles M. Nyce are at Risk Management/Insurance, Real Estate & Legal Studies, Florida State University, Tallahassee, Florida.

DOI: 10.1111/jori.12189

(1) For the purpose of this research, denied claims are considered those claims where the insurer refuses to make any payment toward the proceeds of the life insurance policy. Resisted claims are those where the insurer makes a partial payment of the face value of the policy and denies the balance or those claims where the insurer is still disputing the payment of the claim and retains an amount as still outstanding on the financial records of the company. Denied and resisted claims are found on Schedule F of the statutory accounting statement. This will be discussed further in the "Data and Hypotheses Development" section. According to the National Association of Insurance Commissioners (NAIC), a claim is considered resisted when it is in dispute and not resolved on the financial statement date. A denied claim is one where the insurer has determined the claim will not be paid (NAIC, 2010). In addition, claims that are denied or resisted and close within the same year must still be reported as denied and resisted on Schedule F. Of the claims that appear on Schedule F of the statutory financial statements from 2003 to 2010, 42 percent are resisted claims and 58 percent are denied claims. Denied and resisted claims do not include life insurance claims being reviewed (not disputed) or those claims where the company is holding up payment for sufficient evidence or where a beneficiary has made a claim and then withdraws it. These claims are considered "in the course of settlement" (Fleming, 2013).

(2) All insurers may investigate claims for various reasons (i.e., when a person disappears and is assumed deceased, where suicide is suspected, or where there may be an error with the age of the insured). In addition, insurers may audit claims as a matter of practice. This type of investigation or audit does not necessarily indicate the presence of postclaim underwriting. Postclaim underwriting is the replacement of all or some of the initial underwriting with underwriting that occurs after a claim is submitted.

(3) The Statutory Annual Statement provides the following line of business categories for life and health insurers: industrial life, ordinary life, individual annuities, credit life, group annuities, group accident and health, credit accident and health, individual accident and health, and other. Ordinary life insurance refers to term insurance and all forms of permanent insurance (e.g., universal, variable, variable universal, whole) sold to individuals (Fleming, 2013). Ordinary life denied and resisted claims account for 63.5 percent of all denied and resisted claims reported on Schedule F and 89 percent of all funds being denied or resisted.

(4) The model presented in this article is similar (but more simplified) to Dixit and Picard (2003) and Picard (2009) in that there are both low-risk and high-risk insureds, low- and high-risk policies, and potential misrepresentation. This model does not delve into uncertainty of risk type by the insured (Dixit and Picard, 2003) or commitment to verification (Picard, 2009).

(5) The majority of the claims that are denied and resisted by the insurer are still within the period indicated in the incontestability clause. The incontestability clause is a provision which states a period, generally 2 years, during which insurers have the opportunity to dispute or contest the insurance in force. The incontestability clause limits the opportunity for insurers to utilize the postunderwriting process.

(6) Expenses associated with postclaim underwriting are recorded according to SSAP No. 55--Unpaid Claims, Losses and Loss Adjustment Expenses, paragraph 6d. It states that claim adjustment expenses include legal and investigative costs expected to be incurred in connection with the adjustment and recording of life claims. In other words, life insurers that revisit any underwriting practices during the claims-handling process would record the associated expenses under investigation and claim settlement expense.

(7) Although we draw some parallels between the functions of auditing and postclaim underwriting, they are not synonymous. Postclaim underwriting occurs when the practice of foregoing all or some underwriting processes at the time the policy is written is combined with underwriting after a claim has been submitted. Auditing does not depend on the level of initial underwriting. However, postclaim underwriting and auditing do have some similarities such as the fact that they are generally random and completed only on selected individuals or claims.

(8) The incontestability clause states a period during which insurers have the opportunity to dispute or contest the insurance in force. The clause is meant to protect the beneficiary from expensive settlement negotiations or litigation resulting from an insurer claiming fraud, mistake, or misrepresentation during the application process (Goodman, 1968). For life insurance, the incontestability clause is generally a period of 2 years (McDowell, 1984); however, some policies may contain a period of 1 year or less. This clause was first used in the United States in 1861 and is now required in all states within the United States. The states have statutes regarding the requirement of the incontestability clause. These state statutes and case law are used to balance the interest of the insured that in good faith relied on the coverage that was applied for and the interest of the insurer to avoid coverage they did not intend to undertake (Schuman, 1995). There is a vast array of literature that considers the use and legal aspects of the incontestability clause within life insurance policies (e.g., Goodman, 1968; Salzman, 1969; McDowell, 1984; Schuman, 1995).

(9) Specifically, there are three instances where the insurer can deny coverage after the period set forth in the incontestability clause. These include: (1) the beneficiary takes out apolicy with the intent of murdering the insured, (2) the applicant for insurance has someone else take a medical examination, and (3) an insurable interest does not exist at the inception of the policy (Rejda, 2011). With the exception of these reasons, policies that have exceeded the time frame set forth in the incontestability clause cannot be contested or denied.

(10) Some have claimed that insurer's utilize postclaim underwriting to deny payments claiming the insured committed suicide, even if that reason differs from one stated by law enforcement and medical examiners (Evans, 2011). In yet other cases, it has been charged that a postclaim underwriting process was used to deny the beneficiaries of AIDS victims life insurance claim proceeds (Schatz, 1986).

(11) In the 1994 Mississippi case of Lewis v. Equity National Life Insurance Co., the court addressed the issue of postclaim underwriting and stated that "an insurer has an obligation to its insureds to do its underwriting at the time a policy application is made, not after a claim is filed." In 2001, the Mississippi Supreme Court reaffirmed the determination that postclaim underwriting is illegal in Am. Income Life Ins. Co. v. Hollins. Mississippi is not the only state that has ruled regarding postclaim underwriting. In 1997, in Ingalls v. Paul Revere Life Ins. Group, the North Dakota courts found that the insurer was relying on postclaim underwriting, instead of looking to pay the claim; the insurer was "looking for all things in the application that it might be able to dig up...to rescind the policy" and as such, found in favor of Ingalls. Refer to Cady and Gates (1999) for additional cases specifically addressing the issue of postclaim underwriting.

(12) Misrepresentation and fraud are profound issues across all lines in insurance and as such have been the subject of significant research (e.g., Picard, 1996, 2000; Crocker and Tennyson, 2002; Derrig, 2002; Tennyson and Salsas-Forn, 2002; Viaene and Dedene, 2004; Schiller, 2006; Krawczyk, 2009). Much of this research has focused on the process of identification and detection of these claims.

(13) Some insurers have used the approach of trying to identify fraudulent or misrepresented policies when they are electronically submitted. This process utilizes comprehensive checks and up-front edits and audits to identify issues in procedures and details submitted (Viaene and Dedene, 2004). Another method of auditing is the "Red Flags Strategy," which has a system to score claims based on certain fraud indicators (Dionne, Giuliano, and Picard, 2003). Once flagged, claims identified as questionable are referred to an adjuster for further investigation.

(14) In property--casualty insurance, there is some subjectivity in determining values for claims settlement (i.e., value of property, liability settlements). In life insurance, consumers may believe that there is only one possible outcome for claims settlement when the insured passes away--full payment of death proceeds. In addition, there are currently no ratings that specifically rate the claims-handling practices of insurance companies. Consumers interested in obtaining any information on claims-handling practices would need to search out consumer complaints from claims for property--casualty or Schedule F denied and resisted claims for life insurers to obtain an opinion on the claims-handling practices of any given insurer.

(15) This model differs from Picard (1996) in that misrepresenting occurs with the application (representing as a low risk rather than a high risk) rather than with fraudulent claims. This is enabled by having both low and high risks with differing premiums. This allows this model to be more reflective of the life insurance market where misrepresenting on the application is more prevalent than filing of fraudulent claims (claiming death while still alive). Companies may postclaim underwrite a policy for several reasons, misrepresentation only being one. However, the data for this research indicate that misrepresentation is the leading cause that insurers provide for their reason to deny or resist life insurance claims. Therefore, for the purpose of simplifying the explanation of the model, the term "misrepresentation" is used to depict all reasons the insurer may postclaim underwrite.

(16) We assume U(R) represents a von Neumann--Morgenstern utility function with U'>0 and U"<0 and individuals maximize expected utility.

(17) Moving the misrepresenter to the high-risk contract and charging the appropriate premium [P.sub.H] does not change the substance of the model.

(18) Truthful high-risk opportunists only end up here after they maximize their expected utility and choose not to misrepresent, discussed in the "Low-Risk" Policy section.

(19) If insurers only underwrote a portion of the applications, some misrepresenters would slip through preloss underwriting and the insurer may still need to postclaim underwrite. In equilibrium, we would end up with a smaller p, but the problem remains unchanged in that postclaim underwriting may still occur in equilibrium.

(20) Where k(0) = 0, k'(p) > 0, and k"(p) > 0 for all p e [0,1].

(21) It is possible to have a corner solution where there is misrepresentation ([alpha] > 0) and insurers choose postclaim underwriting with a p = 0; that is, they choose to not underwrite. This will depend on the exogenously determined values in Equation (13) and in the second bracket of Equation (16). In other words, the costs associated with misrepresentation do not warrant spending the money to prevent it.

(22) Where k(p) > r.

(23) Refer to footnote 2.

(24) Unlike Weisbart (1976), who was limited in his data and therefore created one measure of each index for each company over the 10-year period of his study, the data in this study are substantial, and therefore, each index is created annually for each insurer within the sample.

(25) Refer to Table 5 for a definition of the calculation for the indices.

(26) According to the NAIC, any postclaim underwriting expenses are not included in the general underwriting category of the statutory accounting statements. In accordance with SSAP No. 55--Unpaid Claims, Losses and Loss Adjustment Expenses, paragraph 6d, claim adjustment expenses include legal and investigative costs expected to be incurred in connection with the adjustment and recording of life claims (Gann, 2013). Insurers' reports claim adjustment expenses, including any expenses related to postclaim underwriting under Unpaid Claims, Losses, and Loss Adjustment Expenses. Therefore, insurers that postclaim underwrite may see an increase in the expenses associated with this category.

(27) Consistent with Pottier and Sommer (1997), the lines of business used to create this variable are industrial life, ordinary life, individual annuities, credit life, group annuities, group accident and health, credit accident and health, individual accident and health, and other.

(28) Due to the fact that age may not be linear, all models are also analyzed using a dummy variable equal to 1 for all insurers in business at least 50 years and 0 otherwise as a robustness test. The results remain consistent.

(29) For robustness, we investigate the number of years an insurer has denied and resisted claims. During our period under review, 72 companies have such claims in each of the 8 years. Forty companies had claims in 1 year, 31 had claims in 2 years, 24 had claims in 3 years, 19 had claims in 4 years, 24 had claims in 5 years, 24 had claims in 6 years, and 20 had claims in 7 years. To test the impact of a habitual practice of denying and resisting claims, we created a dummy variable for insurers who denied and resisted for each of the 8 years under review. The main results were unchanged.

(30) These graded benefit plans are offered at various benefit levels for various age groups, depending on state of residence. No physical or health questions are required. For Colonial Penn's Guaranteed Acceptance Life Insurance see: https://www.cpdirect.com/products.aspx. For AAA Life Insurance Company's Guaranteed Issue Graded see: https://www.aaalife.com/our-products/life-insurance/whole-life-insurance/guaranteed-issue-whole-life-insurance/ and for Mutual of Omaha's Lifelong Protection Graded Benefit Life Insurance see: http://www.lifebymutual.com/?cc=V10&r=3.

(31) For robustness, those insurers that have an average policy size of $25,000 or lower are removed from the sample. The observation count is reduced to 1,001. The results remain consist with the full sample with the exception of financial institution, which becomes insignificant for the severity index.

(32) The level of guaranty issued policies is not available through the NAIC data.

(33) For the purpose of this research, a denied or resisted claim is evaluated and measured for each year that it remains on Schedule F of the statutory financial statement. In other words, a claim that is being disputed over the course of 4 years and appears on Schedule F for each of these years will be counted as four observations, one for each year it remains disputed. The amount under dispute will be adjusted each year to account for any amount paid to the beneficiaries during the year.

(34) Schedule F of the statutory annual statement provides a section for the insurer to provide a reason as to why a claim is denied and resisted. There is no standard format for the entry of this information. In an attempt to understand how insurers justify their action to deny and resist claims, we manually coded the main reasons identified. Of the total records included in the initial list of denied and resisted claims, 34,609 (60.13%) were denied or resisted for one of six reasons: (1) material misrepresentation, 32,088 (55.65%); (2) suicide, 1,835 (3.18%); (3) age, 539 (.93%); (4) alcohol, 70 (.12%); (5) disappearance, 44 (.08%); and (6) murder, 33 (.06%). The remaining 22,953 claims had various reasons why they were denied or resisted including a statement of the outcome of the review (e.g., "claim settled," "rescinded policy") or various other reasons (e.g., "health history," "questionable"), which may or may not be the same as one of the six reasons detailed above.

(35) For robustness, the models were run using indices of the claims coded as material misrepresentation and those that were uncoded (various explanations were provided) and the results of interest were unchanged.

(36) Not all members of an insurance group record denied and resisted claims.

(37) The total amount of all denied and resisted claims for ordinary life policies, regardless of the reason entered on Schedule F, has increased from $559.91 million in 2003 to over $925.28 million in 2010.

(38) For companies that denied or resisted claims, the average number of denied and resisted claims has increased from 21.65 in 2003 to 27.27 in 2010. Interestingly, the maximum number of these claims increased substantially from 2007, when there were 341, to 2008, when the number reached 717. It is possible that this level of denied and resisted claims was impacted by the financial crisis that occurred around this time. Through 2010, the numbers remained high compared to the 2007 numbers; however, the maximum has decreased slightly each year. The average size of the open disputed claim was $154,858.20 in 2003 and has grown significantly through the years, to $458,218.90 in 2010. There was also an increase in the average size of denied and resisted claims in 2008 and 2009. A comparison of the average amount under dispute for each denied and resisted claim, for each year under review and the average face amount of the insurer's policies indicates that the denied and resisted claims are larger policies (compared to the mean).

(39) Calculating the severity index across all insurers, even those that did not deny or resist a claim during the period under review, indicates the severity index to be 8.42.

(40) Calculating the frequency index across all insurers, even those that did not deny or resist a claim during the period under review, indicates the frequency index to be 4.91.

(41) In years 2003, 2004, and 2010, life insurers that do not deny and resist claims have a higher average policy size. On the other hand, for 2005-2009, insurers that deny and resist claims have a higher average policy size.

(42) For robustness, we ran a more parsimonious model with only the underwriting expense, investigation/claim expense, domiciled state, stock, and size variables. The main results were unchanged.

(43) It is also possible that insurers reduce their initial underwriting to avoid customer complaints that may be reported to New York regulators, leading to additional scrutiny. In other words, the increased regulation in New York may put added pressureon insurersto reduce the initial underwriting of life insurance applications.
TABLE 1
Distribution of Risk and Honesty

                     Opportunist [sigma]

Low risk (1 - [pi])  No incentive to misrepresent
High risk [pi]       Misrepresent with probability [varies]

                     Honest (1 - [sigma])

Low risk (1 - [pi])  Never misrepresent
High risk [pi]       Never misrepresent

TABLE 2
Number of Insurers and Total Amount of Denied and Resisted Claims

                             2003     2004     2005     2006

Number of life insurers       774      762      739      717
With denied and resisted      168      166      179      177
claims
Total material misrepresent   284.72   295.00   329.76   305.29
only (millions)

                             2007     2008     2009     2010

Number of life insurers       688      663      639      629
With denied and resisted      156      149      147      139
claims
Total material misrepresent   315.03   369.69   450.76   376.95
only (millions)

Notes: This table shows the total number of life insurers for each year
under review, the number of these insurers that have at least one
denied or resisted claim during the year, and the total dollar amount
of denied and resisted claims coded as material misrepresentation for
the year.

TABLE 3
Severity Index

Year  Obs.  Mean    Std. Dev.  Min.  Max.

2003  168   25.036   44.945  0.028    342.896
2004  166   81.582  564.204  0.157  7,145.500
2005  179   38.899  117.924  0.063  1,050.483
2006  177   26.984   54.507  0.006    385.831
2007  156   26.940   46.569  0.009    254.240
2008  149   25.635   41.198  0.042    194.024
2009  147   25.824   45.987  0.012    284.815
2010  139   40.576  148.228  0.007  1,666.667

Notes: This table shows the ratio of the dollar amount of denied and
resisted claims in any given year (includes denied and resisted claims
made in previous years that remain open or were denied in the existing
year) to the total dollars claimed for ordinary life in a given year.

TABLE 4
Frequency Index

Year  Obs.  Mean  Std. Dev.  Min.  Max.

2003  168  18.665  48.148  0.022  444.445
2004  166  25.166  57.781  0.014  333.333
2005  179  24.751  66.570  0.049  504.951
2006  177  19.289  44.479  0.045  327.684
2007  156  21.247  46.462  0.044  280.000
2008  149  22.498  52.162  0.020  333.333
2009  147  16.743  36.706  0.021  238.095
2010  139  22.978  58.422  0.013  500.000

Notes: This table shows the rate per 1,000 policy claims of denied and
resisted claims in any given year (includes denied and resisted claims
made in previous years that remain open or were denied in the existing
year) to the total number of ordinary life claims submitted in a given
year.

TABLE 5
Explanation of Variable Specifications

Variable                  Definition

Denied/resisted measures
  Severity index          Dollar amount of denied, resisted and
                          compromised claims/ (Incurred amount
                          of all claims/1,000)
  Frequency index         Number of denied, resisted and
                          compromised claims/ (Number of
                          incurred claims/1,000)
Underwriting
  Underwriting            (Underwriting expenses--life benefit
  expenses (less          losses--life commissions)/New
  comm)                   ordinary life policies issued
  Investigation/claim     Log of average investigation & claim
  expense                 settlement expense per ordinary life claim
Factors impacting cost
  structures
  AM Best rating          A dummy variable equal to 1 if the AM Best
                          rating is A--or better and 0 otherwise
  Financial institution   A dummy variable equal to 1 for the use of
                          financial institutions and 0 otherwise
  LOB HHI                 Line of business Herfindahl index
  Years in operation      Current year/year established
  Size (a)                Log of net total assets
  Stock                   A dummy variable assigned a value equal to
                          1 for stocks and 0 otherwise
  Licensed states         The number of states the insured is
                          licensed to do business
  NY domiciled            A dummy variable equal to 1 if the
                          insurer's state is NY and 0 otherwise
Additional controls
  Independent agent       A dummy variable equal to 1 for the use of
                          independent agents and 0 otherwise
  Captive agent           A dummy variable equal to 1 for the use of
                          captive agents and 0 otherwise
  Commission expense      Commissions on premiums/new ordinary life
                          policies issued
  Increase in premium     Percent increase in net premiums
  Previous denied/        A dummy variable equal to 1 if the insurer
  resisted                denied/resisted claims in the previous year
                          and 0 otherwise
  Lapse ratio             Ordinary life insurance lapse ratio
  Avg policy size (face   Log of the average face amount of ordinary
  amount)                 life policy written
  Group member            A dummy variable equal to 1 for group
                          members and 0 otherwise
  % Premium in state      Percentage of ordinary life premium written
                          in the state
  Year dummies            A dummy variable equal to 1 for the year
                          specified and 0 otherwise

Variable                               Expected

Denied/resisted measures               Sign
  Severity index
                                       --

  Frequency index                      --

Underwriting                           --
  Underwriting
  expenses (less
  comm)                                +
  Investigation/claim
  expense
Factors impacting cost
  structures
  AM Best rating

  Financial institution
                                       +/-
  LOB HHI                              +
  Years in operation                   +
  Size (a)                             +
  Stock
                                       -
  Licensed states                      -

  NY domiciled
                                       +/ -
Additional controls
  Independent agent                    +/ -

  Captive agent                        +/ -

  Commission expense                   +/-
                                       +
  Increase in premium
  Previous denied/                     +
  resisted                             +/-

  Lapse ratio                          +/-
  Avg policy size (face
  amount)
  Group member

  % Premium in state
  Year dummies

(a) Size is winsorized at the 1st and 99th percentiles.

TABLE 6
Summary Statistics and Univariate Comparisons

                                   Mean           Mean
                                   Without        With
                                   Deny/Resisted  Deny/Resisted

Percentage companies                0.7698         0.2302
Underwriting variables
  Underwriting expenses (less       0.0660         0.0259
  comm)
  Investigation/claim expense       2.0862         2.9135
Factors impacting cost structures
  AM Best rating                    0.3758         0.6840
  Financial institution             0.1235         0.2553
  LOB HHI                           0.7246         0.6228
  Years in operation               54.2059        72.8766
  Size                             11.8826        14.4992
  Stock                             0.9433         0.8657
  Licensed states                  25.4971        40.0617
  NY domiciled                      0.0992         0.1319
Additional controls
  Independent agent                 0.3988         0.5480
  Captive agent                     0.1324         0.2678
  Commission expense               16.9218         8.6274
  Increase in premium               4.5223         0.1627
  Previous denied/resisted          0.1314         0.9251
  Lapse ratio                       9.9058         9.2829
  Avg policy size (face amount)     9.9759        11.0493
  Group member                      0.7686         0.8447

                                   Difference  p-Value

Percentage companies
Underwriting variables
  Underwriting expenses (less        0.0401    0.0116
  comm)
  Investigation/claim expense       -0.8274    0.0000
Factors impacting cost structures
  AM Best rating                    -0.3082    0.0000
  Financial institution             -0.1318    0.0000
  LOB HHI                            0.1019    0.0000
  Years in operation                -18.6706   0.0000
  Size                              -2.6166    0.0000
  Stock                              0.0775    0.0000
  Licensed states                  -14.5646    0.0000
  NY domiciled                      -0.0327    0.0004
Additional controls
  Independent agent                 -0.1492    0.0000
  Captive agent                     -0.1354    0.0000
  Commission expense                 8.2944    0.0675
  Increase in premium                4.3596    0.2409
  Previous denied/resisted          -0.7936    0.0000
  Lapse ratio                        0.6229    0.0804
  Avg policy size (face amount)     -1.0734    0.0000
  Group member                      -0.0760    0.0000

TABLE 7
Tobit Model Output

                                     Severity Index
Variable                             Coefficient

Underwriting variables
  Underwriting expenses (less comm)   -6.1648 (**)
  Investigation/claim expense          7.9822 (***)
Factors impacting cost structures
  AM Best rating                      -7.8284
  Financial institution               -9.1841 (*)
  LOB HHI                              6.9183 (***)
  Years in operation                  -0.1341 (*)
  Size                               -12.3374 (***)
  Stock                               18.5008
  Licensed states                      1.0431 (***)
  NY domiciled                        26.4968 (**)
Additional controls
  Independent agent                   -6.0539
  Captive agent                        1.9139
  Commission expense                  -0.0248 (***)
  Increase in premium                 -0.4509 (*)
  Previous denied/resisted             7.0654
  Lapse ratio                          5.9132 (***)
  Avg policy size (face amount)        0.0001
  Group member                       -16.8230

Variable                             Std. Error

Underwriting variables
  Underwriting expenses (less comm)  [3.040]
  Investigation/claim expense        [2.528]
Factors impacting cost structures
  AM Best rating                     [7.533]
  Financial institution              [5.206]
  LOB HHI                            [1.841]
  Years in operation                 [0.075]
  Size                               [3.384]
  Stock                              [14.721]
  Licensed states                    [0.375]
  NY domiciled                       [12.038]
Additional controls
  Independent agent                  [8.114]
  Captive agent                      [6.135]
  Commission expense                 [0.009]
  Increase in premium                [0.270]
  Previous denied/resisted           [9.078]
  Lapse ratio                        [1.424]
  Avg policy size (face amount)      [0.000]
  Group member                       [22.989]

                                     Frequency Index
Variable                             Coefficient    Std. Error

Underwriting variables
  Underwriting expenses (less comm)  -2.5581 (**)   [1.089]
  Investigation/claim expense         3.4283 (***)  [1.034]
Factors impacting cost structures
  AM Best rating                     -9.5852 (***)  [3.520]
  Financial institution              -5.0121 (**)   [2.200]
  LOB HHI                            12.8015 (**)   [5.753]
  Years in operation                 -0.0238 (**)   [0.012]
  Size                               -7.9854 (***)  [1.350]
  Stock                              -4.2052 (***)  [1.482]
  Licensed states                     0.3670        [0.180]
  NY domiciled                        1.8530 (**)   [0.884]
Additional controls
  Independent agent                   5.2269 (*)    [3.066]
  Captive agent                       5.8882        [3.945]
  Commission expense                 -0.0273 (*)    [0.017]
  Increase in premium                 1.3128 (***)  [0.230]
  Previous denied/resisted            4.1549 (***)  [0.731]
  Lapse ratio                         7.0929 (***)  [1.876]
  Avg policy size (face amount)       0.0011 (**)   [0.000]
  Group member                       -5.9588        [7.062]

Notes: Number of observations:1,016. (*), (**), (***) correspond to
significance at the 0.1,0.05, and 0.01 levels, respectively.Due to
space constraints, the results for the premium percentage by state and
year dummies are omitted from the table.
COPYRIGHT 2019 American Risk and Insurance Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Bisco, Jill M.; McCullough, Kathleen A.; Nyce, Charles M.
Publication:Journal of Risk and Insurance
Date:Mar 1, 2019
Words:15231
Previous Article:A TERMINATION RULE FOR PENSION GUARANTEE FUNDS.
Next Article:CONVERGENCE OF CAPITAL AND INSURANCE MARKETS: CONSISTENT PRICING OF INDEX-LINKED CATASTROPHE LOSS INSTRUMENTS.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |