Contingent Commissions - Conflict of Interests.pdf

8/12/2019 Contingent Commissions - Conflict of Interests.pdf

1/28


2/28

1 Introduction

Insurance prices are based on the expected future claim payments and thevolatility of the risk being insured. Actuaries have developed models thatallow insurance companies to evaluate the expected future claim paymentsfrom past loss experience. However, because customers know that their pricewill depend on their past claim history and their current risk profile, theyhave the incentive to hide or obscure information in the insurance purchaseprocess. Also, because it is costly for an insurance purchaser to negotiateprices with several insurance companies simultaneously, the customer mayuse an intermediary that can provide a number of prices at the same time.

Cummins and Doherty (2005) identify three major insurance intermedi-ary models. First, in the direct model, insurance companies sell their policiesdirectly to the customer through employees in a professional sales force ora call center. In this arrangement, because no intermediary exists, there isno intermediary conflict of interest, although standard employer-employeeagency issues may remain. The second is an exclusive (also known as a cap-tive agent or tied agent), in which the agent contracts to sell insurance for asingle company or an affiliated group of insurance companies. For example,Allstate Insurance Group is a holding company for at least twenty-eight dif-ferent writing companies. Each is a separate company that holds licensesto sell insurance in various states, but is held within a single publicly-tradedfirm. Allstate retains exclusive agents, who have the right to sell insuranceonly for the Allstate companies, and no other firms. The agent may be

employed by the insurance company or may operate independently, but isusually compensated largely on a commission basis. Since the agent cannotsell insurance products from a different company, the intermediary conflictsof interest are limited.

Finally, the most common type of intermediary is an independent agentor broker. As in our sample, DArcy and Doherty (1990) also show thatthe majority of property-liability insurance market share is sold throughindependent agents and brokers. This type of intermediary sells insurancefor several different insurers, soliciting quotations based on information thatthey collect from potential customers. There are numerous conflicts of in-terest, both between the intermediary and the customer, and between the

intermediary and the insurer. These are the relationships we explore indepth.

On October 14, 2004, New York State Attorney General Elliot Spitzer(NYAG) announced a lawsuit against three large insurance companies andone broker, alleging anti-competitive behavior. In the announcement, the

2


3/28

NYAG said that the industrys fundamental business model needs major

corrective action and reform, referring to the practice of contingent com-missions and the subsequent bidrigging that it encouraged. Following theannouncement, stock prices of the majority of property/casualty insurancecompanies and publicly-traded insurance brokerage firm fell significantly.After an in-depth article about the lawsuit appeared in the New York Timeson October 17, 2004, stock prices for most property and casualty insurancecompanies fell further1. However, within 16 trading days after the event,most of the stocks had regained the value lost on the announcement date.In this paper, we explore the characteristics of companies that lost marketvalue, and document that the valuation loss was recovered for firms notnamed in the lawsuit within days after the announcement.

We review the agency issues inherent in the current insurance distribu-tion models and suggest three main hypotheses. First, the companies namedin the lawsuit will have negative and significant abnormal returns if the mag-nitude and gravity of the NYAGs lawsuit was not fully anticipated by themarket. Second, insurance firms with distribution channels similar to thenamed firms will also experience negative returns following the announce-ment. Finally, for firms not named in the lawsuit, negative excess returnswill be reversed shortly after the announcement. We examine seventy-onepublicly-traded property and casualty insurance companies, and analyze thechanges in their stock prices following the NYAGs announcement in rela-tion to their corporate premiums, commission practices and distributionmethods.

Consistent with the hypotheses, we find that during the days surround-ing the event, both firms named in the lawsuit as well as the portfolio ofother insurance companies had negative returns; also, returns from insur-ance firms that use independent and exclusive agents were negative whilethe firms employing other distribution methods did not have significant neg-ative returns. Finally, we observe that firms named in the lawsuit recoveredonly part of the cumulative loss in the days following the announcement,while the portfolio of firms not named in the suit recovered nearly all of itsvalue within a month.

This paper is structured as follows. In the next section, we describe thevarious forms of insurance intermediaries. The third section reviews liter-

ature and discusses the theory describing the incentive for bid-rigging un-der contingent commission arrangements. In the fourth section, we provide

1In a companion paper, we document and analyze a similar valuation effect for a sampleof life and health insurance companies.

3


4/28

background and timelines related to the NYAGs lawsuit. Then, we develop

hypotheses related to stock returns following that announcement. The sixthsection presents our empirical results and the seventh section concludes.

2 Insurance Intermediaries

Insurers sell policies through two primary means: direct sales (either throughcaptive agents or a salaried sales staff) and independent agent/broker rela-tionships. Regan and Tennyson (1996) show that the distribution methodoptimally depends upon the type of risk involved and the level of difficultyinvolved in categorizing and sorting risks. When the risks are relativelystandard and easily observed, the insurer tends to use a direct sales method

(call center or internet sales), because the sorting and categorization servicesperformed by agents are worth less than the cost of commissions. However,when an agent can gather and provide valuable information for pricing andservicing policies, independent agent and broker distribution is preferred.

Independent agents are typically seen as agents for the insurer and bro-kers are typically seen as agents for the insurance buyers. However, as shownby Cummins and Doherty (2005), these definitions are somewhat arbitraryand not completely accurate. For example, although brokers have fiduciaryresponsibility to the insurance buyer, they provide a number of services onbehalf of the insurer. Similarly, although independent agents have fiduciaryresponsibility to the insurer, they provide consulting, pricing and risk man-

agement advice for the insurance buyer. Both agents and brokers can bebetter described as insurance market makers, convening buyers and sellers,gathering and disclosing important information to complete the transaction.As such, they demand compensation for their services, and this compensa-tion is typically provided by the insurer. Commissions to brokers are paidby insurance companies, and this creates conflicts that may prevent themfrom serving adequately in a fiduciary role for the customers.

Agent and broker commissions fall into one of three categories. The first,discussed by Wilder (2004), is a fixed percentage commission, in which theinsurer pays the agent some portion of the premium generated by the newor renewed policy, according to market prices and the level of sophisticationof the policy. While the agent or broker may have an incentive to undertakeextra effort to sell a policy with a higher commission payment (resultingeither from a higher premium, a higher commission percentage, or both),if only a fixed fee arrangement is in place, the agent will only do so if thepresent value of all expected future commissions resulting from this partic-

4


5/28

ular business exceeds the difference in commission payments between two

insurers. Since an agent receives commissions not only from initial busi-ness, but repeat business, reputation effects are important considerations inchoosing between two insurers when only a fixed commission is offered.

The second type of commission offered by insurance companies is profit-based contingent commission. Under these plans, agents receive a com-mission in addition to the base commission when certain profit targets areachieved from the business placed with the insurer. This provides an incen-tive to the agent to uncover as much information as possible to classify therisk of the new policy correctly. For example, if an agent or broker sells apolicy for $1,000 and the losses resulting from that policy, along with salesand underwriting expenses, do not exceed $1,000, then the policy is prof-

itable from the insurers perspective. When the insurer agrees to share someof that profit with the agent, the agents incentives are more fully alignedwith those of the insurer. While it is not apparent that this arrangementwill create an incentive for bid-rigging, it may create an incentive for theagent to encourage the customer to purchase higher-priced policies for thesame level of risk.

The third type of commission offered by insurance companies is volume-based contingent commission. Insurers offer volume-based contingent com-missions for several reasons. First, they can achieve economies of scale byincreasing the concentration of business from a particular agent or brokerage.Second, as an insurer increases the number of uncorrelated risks assumed,the standard deviations of the losses (that is, the associated volatility of the

loss pool) should decrease. This means that increasing the number of policiesmay decrease the insurers risk as long as the insurer is selling the policiesat a profit. However, as shown by Wilder (2004), when a broker is close to avolume-based contingent commission threshold, the marginal revenue froma policy may far exceed the standard or profit-based commissions that couldbe earned by placing the business with another insurer. As Wilder explains,if a broker has placed $499,000 of business with one insurer in a given year,and has a target of $500,000 before the next 2.5% volume-based contingentcommission is triggered, a single $1,000 policy has a marginal commissionof $500,000x0.025=$12,500 or 1,250%. With such strong incentives, agentsmay be inclined to make otherwise less-appropriate policies appear more

attractive to purchasers. They could do this by ignoring or not seekingcompetitive quotes from other companies, over-emphasizing the benefits ofthe policy that would trigger the contingent commissions, or colluding withother insurers to provide false, higher bids to make the current bid looklike a good deal. When the contingent commission arrangements are not

5


6/28

disclosed to the clients, the incentives are even stronger, because monitoring

by the client has been effectively mitigated.There are similarities between the insurance intermediary market and

the real estate broker market. In fact, similar to the two functions of re-altors observed by Miceli, Pancak, and Sirmans (2006), insurance agentsand brokers provide two key services to insurance customers: matching andbargaining. In the matching service, the agent determines the risk profile ofthe customer and identifies insurers likely to accept the customers risk. Inthe bargaining service, the broker negotiates prices with the chosen insurers,seeking to find the best price and policy for the customer. As Miceli, et al,observe, a commission paid by the insurance company generates a perverseincentive and may prevent the agent from exerting effort on behalf of the

customer.

3 The Incentive for Bid-Rigging

Insurance is the business of selling risk management to customers. Becauseeach policy entails the transfer of some risk from the insured to the insurer,a transfer of wealth also must be made to compensate for that risk. The re-quired payment (premium) is a function of the expected loss associated withthe policy, the variance of the loss and the additional expenses associatedwith transferring the risk (including adequate return on capital, overheadexpenses and commissions or salaries to insurance salespeople). While the

customer typically knows more about his own risk, the insurer knows moreabout how to evaluate and price the risk. Therefore, in an effort to achievea risk transfer at the lowest possible price, a potential customer is likely toattempt to withhold as much information as possible from the insurer.

The agent, seeking to generate business, will normally attempt to pro-vide as many feasible policies as possible to the customer. That agent willgather as much information as necessary to generate quotes from a numberof insurers and thus help to mitigate the information asymmetry betweenthe customer and the insurer. However, armed with this information, theagent enjoys an information advantage over both the insurer (more com-plete knowledge of the customers risk profile) as well as the customer (morecomplete knowledge of the available products, and their commissions), andcan exploit that information advantage to his own benefit. This informationasymmetry can impose costs on both insurers and customers.

Also, in the agents work on behalf of the insurer, the agent has anincentive to shirk in identifying all of the pertinent risks. This is possible

6


7/28

because the agent knows more about the insurance products available, and

their prices, as well as the commissions they may be able to garner fromvarious products. The agent may also be in a position to withhold productsthat are superior for the customer but not as lucrative for the agent. Inthis way, the agent can contribute to an adverse selection problem in whichpolicies are not priced correctly or customers do not know the extent ofproducts available to them.

Without regulations stipulating full disclosure of all commissions, theremay be more incentives for the agent to take advantage of the informationasymmetry on either side of the transaction. In this section, we review theextant literature on the possible agency issues in the insurance distributionsystem.

3.1 Information Asymmetry

Generally, as shown by Akerlof (1970), the agent may represent quality orfail to represent quality. If the principal chooses not to, or is unable to(due to asymmetric information) observe the true quality, the agent is in aposition to expropriate wealth from the principal. This potential informationasymmetry leads to other problems, including adverse selection and moralhazard, discussed in following subsections.

In their study on private information in the insurance industry, DArcyand Doherty (1990) note that an insurance agent has an information ad-vantage over insurers and may benefit by selling some of that information

to another insurer. Lengwiler and Wolfstetter (2005) show that when theauctioneer is an agent for the seller, as is the case in insurance agency, thereare incentives to engage in bid-rigging to benefit the auctioneer. Finally,Crawford and Sobel (1982) argue that an agent will only engage in perfectcommunication with a principal when its incentives are completely alignedwith those of the principal. As noted earlier, the fiduciary responsibility ofan agent or broker is not always clear, and there are some incentives thatare tied to customers objectives (primarily reputation and repeat business)while others are tied to insurers objectives (commission payments of thetype described in the previous section). Under direct commission arrange-ments, it is not clear which incentive will dominate.

In order to prevent the sale of information anticipated by DArcy andDoherty, the insurer may choose to share the rent on information by cre-ating a contingent commission, making it costly for the agent to place thebusiness elsewhere. Consistent with Crawford and Sobels predictions, thisaction more closely aligns the agents incentive with the insurers interests

7


8/28

and compels a higher level of information disclosure. Under Lengwiler and

Wolfstetters predictions, an agent may engage in bidrigging or other anti-competitive behavior to entice the customer to place or keep their businesswith the insurer providing the most attractive contingent commission. Onepossibility is that employees of an insurer may provide false bids if theyshare in the contingent commission payoff in the form of kickbacks or mayprovide the bids as part of a working arrangement and a promise of futurebusiness. When customers are not aware of the volume-based and/or profit-based contingent commissions, they are not likely to monitor the behaviorof the agents.

The agent also has an information advantage over the insurer. Havinggathered as much of the customers risk profile as possible, the agent may

sell that information to a competitor, in order to earn a higher commission.DArcy and Doherty (1990) show that insurers can mitigate this problemby paying a contingent commission, sharing the rent from the informationwith the agent.

3.2 Adverse Selection

As discussed above, the agent has an information advantage over the in-surer. Besides the information asymmetry issues, such as the incentive tosell information to competitors, the agent can also contribute to an adverseselection problem. When the incentive to sell a policy dominates the incen-tive to place only profitable business with the insurer, the agent may have an

incentive to withhold information or not invest sufficient effort in gatheringinformation about the customers risk profile. DArcy and Doherty (1990)show that some insurers mitigate this adverse selection problem by payinga profit-based contingent commission, where part of the commission is paidretroactively, after losses are fully recognized by the insurance company.This profit-based commission provides an incentive for agents to gather andreveal as much information as possible to price policies correctly and thusreduce the likelihood of adverse selection related to information asymmetry.

3.3 Moral Hazard

Moral hazard is the result of an incentive to not invest appropriate effortfollowing the formation of a contract. In strict insurance parlance, moralhazard is reflected in the incentive for a policyholder to take on more riskafter he has purchased an insurance policy. In our examples, moral hazardtakes the form of the agent investing less effort on behalf of either the cus-

8


9/28

tomer or the insurer after agreeing to sell insurance. In interactions with

the customer, this means the incentive to not search for the best productand price for the customer. In interactions with the agent, this means notexerting sufficient effort to identify all the risk characteristics and/or to closethe sale of an insurance policy.

In his discussions about contingent fees in the legal services industry,Hay (1996) argues that attorneys and clients can contract a fee structurethat optimizes the attorneys quantity and quality of effort in litigating thecase. Levitt and Syverson (2005) show that in the real estate industry, theagent works hardest for the party that compensates him. However, in theinsurance industry, it is not always clear which party is the principal: theinsurer or the insured. Since almost all commissions are paid by insurers

and not insured parties, we would expect insurance agents to work on behalfof the insurers and to respond to commission-paying insurers rather thancustomers.

As discussed earlier, the commission received by the agent takes severalforms. With direct commissions, the payment to the agent depends on theamount of premium written. The agent can seek to increase the amount ofinsurance sold by providing estimates for, and convincing the customer ofthe need for, more robust insurance policies, or by convincing customers tochoose an insurer with a higher premium charge than its competitors. Thus,a direct commission creates a moral hazard opportunity for an insuranceagent to invest more effort in selling a higher premium policy, or sellingunnecessary coverage. This moral hazard, however, is offset by the agents

investment in reputation and desire to enhance the relationship with theinsured to continue the stream of revenue.

In order to more fully align the incentives of the agents and brokerswith those of the insurers, specifically in making it attractive for an agentto sell the policies of one insurer over another, some insurance companiesoffer contingent commissions based on the volume of business placed withthat insurer. As shown in the last section, this incentive may be a relativelysmall part of overall premium generated, but as premium levels approachthe thresholds for volume-based contingent commissions, the marginal com-mission for the last policy before the threshold becomes very large.

It is this type of contingent commission, and the anti-competitive be-

havior encouraged by it, that captured the interest of the NYAG in 2004.

9


10/28

4 Background and Impact of NYAG Intervention

On October 14, 2004, the NYAG announced a lawsuit against Marsh &McClennan Companies, a leading broker of insurance products. 2 Thelawsuit alleged abuse of certain industry marketing practices and indicatedthat other major insurance carriers would also be affected by the lawsuit.the NYAG said that the action could force significant change in insurancemarketing practices. Concurrent with the lawsuit announcement, two se-nior executives of American International Group (AIG), one of the largestinsurance carriers, pleaded guilty to charges of anti-competitive practices.Marsh & McLennan, Aon Corporation, AIG and The Hartford subsequentlysettled by agreeing to place hundreds of millions of dollars into a fund tocompensate victims of the alleged fraud.

This announcement was not a complete surprise. In fact, in February2004, the NYAG was contacted by a Washington, D.C.-based think tank,which asked him to investigate brokerage commission agreements, and he is-sued over a dozen subpoenas by May of the same year. Pursuant to Securitiesand Exchange Commission rules, these subpoenas qualified as significantevents that must be reported by the subpoenaed companies; a flurry ofpress releases were published by companies in this time period. Shortly be-fore the NYAG announcement, a lawsuit was filed by an insurance consumerrights group on the issue of contingent commissions. However, the gravityand breadth of the lawsuit apparently surprised investors, as stock pricesfor both named and most un-named property and casualty insurance firms

fell dramatically in the days following the announcement. A chronology ofkey events is provided in table 13.

The NYAGs announcement may be considered both a regulatory com-pliance event and also a signal of changing regulatory enforcement. As aregulatory compliance event, we expect that the three firms named in thelawsuit would incur costs in defending themselves in the lawsuit, as wellas an uncertain reduction to earnings and cash flow in the event of fines,penalties and disgorgement of profits. The reaction to stock prices for theU.S.-based insurers named in the lawsuit is shown in figure 1, panel A. Thepattern shows a significant negative return on the value-weighted portfolioof these three stocks, with an incomplete recovery over the following trading

2See Investigation Reveals Widespread Corruption in Insurance Industry, Of-fice of New York State Attorney General Elliot Spitzer, October 14, 2004.http://www.oag.state.ny.us/press/2004/oct/oct14a 04.html

3For a complete summary of events, see the reports by the Insurance InformationInstitute. http://www.iii.org/media/hottopics/insurance/brokercompensation/.

10


11/28

Table 1: Chronology of EventsThe series of events leading up to the October 14, 2004 announcement

by New York Attorney General Elliot Spitzer. On this day, Spitzer an-

nounced a lawsuit against four insurers and one brokers for alleged bid-

rigging and other anti-competitive practices. He also indicated that the

insurance industry would need to closely examine and correct its busi-

ness practices with respect to use of contingent commission arrangements.

Date Event

February 2004 Washington Legal Foundation asks New York At-torney Elliot Spitzer, as well as the New York andCalifornia Insurance Commissioners, to investigate

commission practices of major insurers.May 2004 Spitzer issues subpoenas to brokers and insurers, re-questing information on contingent commission ar-rangements.

July 2004 An Illinois circuit court judge certified a class actionlawsuit against insurance broker Aon Corporation,alleging that they accepted contingent commissionsfrom insurers without disclosing them to customers.

August 2004 Insurance policyholder rights group United Policy-holders sued three brokers over contingent commis-sion disclosure violations.

October 14, 2004 Spitzer filed a civil lawsuit in New York againstMarsh and McLennan and four insurers, allegingfraud and anti-trust violations. Simultaneously, twoAIG executives and one ACE executive pleadedguilty for criminal behavior uncovered in the courseof the investigation

October 17, 2004 A feature article in the New York Times outlinedSpitzers investigation and lawsuit, and indicatedthat more insurers were likely to be investigated orsued in the near future.

Sources: Timeline and Chronology of Events, Insurance Information Institute,

http://www.iii.org/media/hottopics/insurance/brokercompensation/; AttorneyGeneral Elliot Spitzers website,

http://www.oag.state.ny.us/press/2004/oct/oct14a 04.html and corporate press

releases.

11


12/28

month, as shown in 1, panel B.

As a signal of changing regulatory enforcement, the NYAG clearly statedthat the insurance industry would be more broadly affected by the announce-ment. To illustrate, we quote from the NYAGs press release:

The insurance industry needs to take a long, hard look atitself, Spitzer said. If the practices identified in our suit are aswidespread as they appear to be, then the industrys fundamen-tal business model needs major corrective action and reform.

There is simply no responsible argument for a system thatrigs bids, stifles competition and cheats customers, he added. 4

Facing a change in regulatory environment for the insurance industry,we expect similar firms not named in the suit to have negative returns aswell, especially if the changes would affect their profitability and cash flowposition in the future.

5 Hypotheses

Binder (1985a) asserts that stock prices of affected firms will change onthe event date only when that information was unanticipated. Baucus andBaucus (1997) note that following discovery and consequences of illegal cor-porate behavior, offending firms achieve lower accounting returns and slower

sales growth and show that markets react to this change in expectations withsharp reductions to stock prices of these firms. In research about litigationrelated to securities fraud, Griffen, Grundfest, and Perino (2000) show thatstock prices of firms named in federal class-action litigation react negativelyto announcement of the litigation. Based on this evidence, we state a hy-pothesis:

Hypothesis 1 The NYAGs announcement of the lawsuit about the anti-competitive pricing strategy of three publicly-traded U.S. property casualtyinsurance companies (AIG, Ace and The Hartford) was unanticipated by themarket and would induce negative valuation effects in these stocks. Thus,the null hypothesis is no statistically significant negative valuation effects

following the announcement for the firms named in the suit. If we reject thenull hypothesis, we find support for our hypothesis.

4From http://www.oag.state.ny.us/press/2004/oct/oct14a 04.html.

12


13/28

Figure 1: Cumulative Average Abnormal ReturnsValue-weighted cumulative average abnormal returns relative to a modified Fama-

French Four-Factor return generating model. Day 0=October 14, 2004, the date

of the announcement of a lawsuit by New York Attorney General Elliot Spitzer.

Panel A: Firms Named in Lawsuit

Panel B: Firms Not Named in Lawsuit

13


14/28

If we observe a significant negative change in the three named insur-

ance company stock prices on the day of the announcement, the marketdid anticipate the announcement, or at least its breadth and gravity. Asshown in table 1, however, the lawsuit was not entirely unanticipated. Thesubpoenas and prior lawsuits indicated some legal activity before the an-nouncement. However, it is possible that the market underestimated thepotential penalties and sweeping changes suggested by the announcement.A negative return for the firms named in the lawsuit would indicate that atleast part of the announcement was unanticipated.

Regarding the other industry firms that were not named in the lawsuit,Prince and Rubin (2002) show that some lawsuits have a broad and uniformimpact on the industry, while other lawsuits impact the named firm nega-

tively and the competitors positively. They suggest that when the industryinvolves common design parameters, lawsuits affecting one firm will even-tually affect the competing firms, eliciting a negative reaction from manystock prices in the industry. In Prince and Rubins model, common designparameters were defined as technologies or factors of production that werecommon across several firms. For example, the auto industry generally de-signs cars with relatively common parts. When Ford Motor Company wassued for using unsafe parts in the Pinto, other companies that used similarparts in their vehicles experienced negative returns as well. The commonparts were the common design parameters. In the insurance industry, partof the technology of production is the distribution method. Since the lawsuitrelated to commission practices, we would expect to see negative abnormal

returns for companies using the same distribution method. In essence, thisis a contagion effect.

Conversely, when the industry involves heterogeneous design parameters,such as in the pharmaceutical industry, a lawsuit against one company willdiminish that companys market position and improve those of competitors.In Prince and Rubin, when a pharmaceutical company is sued for negativedrug reactions, firms that sell a substitute chemical compound to treat thesame illness will have increased cash flows as patients switch to that drugand away from the drug that is the subject of the lawsuit. In our case, iffirms using independent agents had a competitive advantage generated bytheir use of contingent commissions and risked losing it as a result of the

NYAGs actions, firms that did not rely on that distribution method willhave a relative advantage. Consequently, companies whose primary distri-bution methods are through independent channels (agents and independentbrokers) will have negative stock price reactions, while companies whose pri-mary distribution methods are through direct channels (captive agents and

14


15/28

employee sales forces) will have positive stock price reactions. In aggregate,

since independent agents and brokers are the most common form of distri-bution in the property and casualty insurance sector, this is likely to impactour overall portfolio negatively. In short, this is a competitive effect.

Based on these notions, we propose our second hypothesis:

Hypothesis 2 The NYAGs announcement will have an overall negativeimpact on the portfolio of property and casualty firms not named in the law-suit. However, companies whose primary distribution methods are throughindependent channels (agents and independent brokers) will have negativestock price reactions, while companies whose primary distribution methodsare through direct channels (captive agents and employee sales forces) willhave positive stock price reactions. Thus, the null hypothesis is no statis-tically significant negative valuation effects following the announcement forthe firms not named in the suit. If we reject the null hypothesis, we findsupport for our hypothesis.

Antweiler and Frank (2005) demonstrate that markets over-react andusually reverse losses within a few days of the publication of negative news.In the event under study, there are two possible scenarios. First, after theinitial announcement, more information may be released, either explicitlyor through market transactions, to reassure investors that there is a verylow probability many of the non-named companies will be implicated in asubsequent lawsuit. The release of this information will cause the stock

price losses to reverse over time. The overall effect for the portfolio willbe mean-reverting. Second, as the competitive effects of the announcementbecome apparent, investors will shift their holdings from firms that are likelyto be negatively impacted by the regulatory change to the firms that standto benefit from the change. The overall result for the portfolio should bemean-reverting.

Given these predictions, we propose our third hypothesis:

Hypothesis 3 Because some of the firms not named in the lawsuit are likelyto gain a competitive edge from the possible regulatory change, the valuationloss sustained by these companies at the announcement will be reversed soonafter the market recognizes that potential. Thus, the null hypothesis is nostatistically significant negative valuation effects that persist fol lowing theannouncement for the firms not named in the suit. If we fail to reject thenull hypothesis, we find support for our hypothesis.

15


16/28

6 Empirical Results

6.1 Data

We constructed our sample from the universe of U.S. publicly-traded com-panies selling property and casualty insurance and for which premium andcommission data was readily available. We began by choosing all compa-nies from the Mergent On-line database that were traded on U.S.-basedexchanges and with primary and secondary two-digit SIC codes of 63 (In-surance Carriers). For these companies, we verified that the company isan insurance company by reviewing the companys narrative entry in theMergent database.

Next, we cross-referenced this list with entries in A.M. Best Aggregates

and Averages, 2004, including only companies for which net premium written(NPW), direct commission and contingent commission data were available.We eliminate firms for which return data exist for less than 200 days in theestimation period. 75 firms entered the sample: 3 U.S.-based property andcasualty insurers that were named in the lawsuit and 72 U.S.-based propertyand casualty insurers that were not named in the lawsuit.

Descriptive statistics for selected variables are presented in table 2. Wecalculate means, medians and standard deviations for the entire sample, forthe firms named in the lawsuit, and then for each group of firms using one ofthe standard distribution channels. We analyze net premium written for theyear ending 2003, direct commission to net premium written (NPW) ratio,

contingent commission to net premium written (NPW) ratio and marketcapitalization, defined as the product of common shares outstanding andper share price as of September 1, 2004. From these descriptive statistics,we can see that the firms named in the lawsuit have larger market sharesand market capitalization than the average for the portfolio, but pay lowercommission and lower contingent commission ratios than the average for theportfolio. We also see that means are higher than medians for market shareand market capitalization and contingent commission ratios for all subsetsof the sample, suggesting that these data are skewed. Furthermore, meansand medians for direct commissions appear to be distributed symmetrically.

6.2 Methodology

In a pair of companion papers, Bhagat and Romana (2001b) and Bhagatand Romana (2001a) show that event studies are an effective method for de-termining the effect of lawsuits on corporate law and corporate governance

16


17/28

Tab

le2:DescriptiveStatistics

Descriptivestatisticsforpropertyandcasualtyin

surancefirmsinthesample.

FirmCharacteristics

Allfirms

Namedfirms

Broker

IndependentAgent

ExclusiveAgent

DirectMarketing

N

75

3

24

58

44

27

NetPremiumWritten

Mean

2,534,395

13,426,843

4,202,019

2,717,155

1,610,066

3,45

4,711

Median

718,507

8,876,260

958,168

754,542

705,310

774,854

StandardD

eviation

5,274,686

12,887,368

7,137,790

5,128,448

2,249,459

7,01

3,462

Dire

ctCommission/NetPremiumW

ritten

Mean

12.12%

9.04%

12.4%

13.46%

13.86%

9.83

%

Median

13.7%

9.36%

13.94%

14.07%

14.43%

7.96

%

StandardD

eviation

4.12%

0.97%

6.58%

5.8%

6.22%

8.94

%

Contin

gentCommission/NetPremium

Written

Mean

1.16%

1.02%

1.49%

1.29%

1.05%

1.48

%

Median

0.52%

0.87%

0.68%

0.61%

0.87%

0.03

%

StandardD

eviation

2.5%

1.15%

3.6%

2.76%

1.27%

3.83

%

MarketCapitalization

Mean

7,183,818

70,426,091

15,703,218

8,180,623

262,826

14,1

38,553

Median

1,196,223

17,696,888

2,375,537

1,238,409

1,191,679

1,28

2,357

StandardD

eviation

24,602,196

97,490,482

41,987,177

27,643,789

3,802,561

39,8

15,367

Note:somefirmsusemorethanonedistribution

method,andthusareincludedinmorethanonecolumn.

17


18/28

issues. They note, however, that event studies may underestimate the neg-

ative impact of a lawsuit or regulation change, because information leakedto the market ahead of the announcement may already be impounded intostock prices. They also show that while a one-day event window is preferred,a three-day event window does not lose significant statistical power.

Henderson (1990) lists a number of issues in event studies, includingthe problem of calendar clustering, where the events for a number of firmsoccur on or near the same day. In such a case, we cannot assume thatreturns are not cross-correlated. As suggested by Binder (1985b), Malat-esta (1986), Karafiath (1988), Ingram and Ingram (1993) and others, weuse a joint generalized least squares (commonly called Seemingly UnrelatedRegression (SUR)) approach to generate estimates that are robust to these

cross-correlation problems. This is the method used by Fenn and Cole (1994)in a similar insurance industry study focused on a common event date, byCornett and Tehranian (1990) in their review of regulatory events on bank-ing returns and by Bastin and Hubner (2006) in their review of the effectsof presidential announcements about federal funding for genetic research onthe returns of biotech firms.

One issue in specifying event studies is the choice of the benchmark re-turn. We generate the benchmark return using a four-factor model inspiredby Fama and French (1992) and Fama and French (1993), with the first threefactors specified as the traditional excess market return, small-minus-big as-set portfolio return and high-minus-low book-to-equity portfolio return. Weadapt their measure to the specifics of the insurance industry by adding a

fourth factor that takes into account the relative amount of risk assumedby the firm, the net premium written (NPW). NPW is equivalent to salesin other industries, and is defined as the total premium written by a firm,less ceded reinsurance. We included this factor because preliminary cross-sectional models on both the Fama and French three-factor model and thesingle-factor market model abnormal returns consistently identified net pre-mium written as a significant factor in both estimation period and abnormalreturns.5

We conducted event studies for three windows. The standard window of(-1,1) captures the excess returns following the NYAG announcement. Onthe third trading day after the announcement, a feature article about the

lawsuit appeared in the New York Timesdescribing the nature and extentof the bid-rigging schemes and the likely fallout for the insurance industry(Morgenson (2004)). To capture the effect of this announcement, we test

5Net premium written was taken from A.M. Best Aggregates and Averages, 2004.

18


19/28

an event window of (-1,3). Finally, to test whether the insurance industry

losses were reversed, consistent with our third hypothesis, we test an eventwindow of (-1,16).

6.3 Model

To estimate the impact of the NYAG announcement on the returns ofproperty-casualty insurance companies, we estimate the modified Fama-French Four-Factor Model. Instead of the momentum factor used by Famaand French, we used a factor reflecting the portion of stock returns attributedto the market share of the company as measured by Net premium written(direct premium written less ceded reinsurance), HiNPWmLoNPW. Togenerate this variable, we sorted the companies by NPW as stated in their

2003 statutory financial reports (the reports that companies make to thestate insurance departments), taken from A.M. Best Aggregates and Aver-ages, 2004 and divided them into three equally-sized, value-weighted port-folios. We calculated the return on high NPW portfolio and the low NPWportfolio for each day in the estimation period and the event periods. Wethen subtracted the low NPW portfolio return from the high NPW portfolioreturn to create this return-generating factor. We also included the threefactors suggested by Fama and French: an excess market return factor, afirm size variable and a book-to-market variable.

We use SUR to estimate the following equation for each firm:

Rjt = j+jRmt +sjSML+hjHML+pjHiNPWmLoNPW+jtDt +jt(1)

whereRjt is the return on the jth firm on dayt,Rmt is the excess return

of the market portfolio over the three-month treasury rate on day t, SMBtis the average return on three small market capitalization portfolios minusthe average return on three large market capitalization portfolios on day t,HMLtis the average return on two high book-to-market portfolios minus theaverage return on two low book-to-market portfolios, HiNPWmLoNPWtis the return on a high net premium written (NPW) portfolio minus thereturn on a low NPW portfolio on day t and Dt is an indicator variabletaking the value of one during the event period and zero otherwise. jt is

assumed to be zero and uncorrelated with other return-generating variables;however, since we are using SUR, the error terms between models is allowedto be correlated without affecting the consistency of our results.

We calculate the abnormal returns from the regression using the follow-ing model:

19


20/28

ARj =Tt=1

Rjt

j+ jRmt+ sjSMBt+ +hjHMLt+ pjHiNPWmLoNPWt

(2)Finally, we estimate the effect of several vectors of potential explanatory

variables on abnormal returns, attempting to identify significant factors thatmay explain the abnormal returns.

6.4 Results

We first graphed the cumulative average abnormal returns over the (-30,30)window, with 0=October 14, 2004, to observe the impact of the NYAG an-nouncement. This graph is presented as figure 1. Panel A shows cumulativeaverage abnormal returns for the three companies named in the lawsuit andpanel B shows returns for all other property and casualty insurance compa-nies in our sample. Panel A of our figure shows that the portfolio of firmsnamed in the lawsuit experienced extreme negative returns in the days fol-lowing the announcement, but that those returns were partially reversedwithin a few days following the announcement. In panel B, we see that theportfolio of companies not named in the lawsuit also experienced markednegative returns in the days following the announcement, but that thosenegative returns were essentially reversed within several days following theannouncement.

To test our first hypothesis, we conducted the event study describedin the previous section for the property and casualty firms named in theNYAGs lawsuit over two event windows, (-1,1) and (-1,3). Summary pa-rameter estimates from the estimation period are presented in table 3, panelA. Event period results are reported in table 4, panel A. We find supportfor hypothesis 1 with significant negative returns for named firms in bothwindows (-11.77% and -14.97%, respectively).

To test our second hypothesis that other insurance firms will have neg-ative stock returns in the event window, we conducted the event study de-scribed in the previous section for firms not named in the lawsuit over twoevent windows, (-1,1) and (-1,3). The results are reported in panel B of ta-

ble 4. We find support for hypothesis 2 with significant negative returns inboth the (-1,1) and (-1,3) windows. The mean cumulative abnormal returnin the (-1,1) window is -2.4% and the mean cumulative abnormal return inthe (-1,3) window is -3.84%.

To test the second part of our hypothesis that the negative stock returns

20


21/28

Table 3: Estimation Period Parameter EstimatesMean and median parameter estimates from a four-factor modified Fama-Frenchstock return model:Rjt = j+ jRmt+ sjSML + hjHML +pjHiNPWmLoNPW+ jtDt+ jtwhere Rjt is the return on the jth firm on day t, Rmt is the excess return of the

market portfolio over the three-month treasury rate on day t, SMBt is the average

return on three small market capitalization portfolios minus the average return on

three large market capitalization portfolios on day t,HMLtis the average return on

two high book-to-market portfolios minus the average return on two low book-to-

market portfolios,HiNPWmLoNPWtis the return on a high net premium written

(NPW) portfolio minus the return on a low NPW portfolio on day t. Parameter

estimates are for the estimation period only.


Intercept Market SMB HML NiNPWmLoNPW

Mean 0.00027 1.09 -0.28 0.17 0.06Median 0.00026 1.2 -0.29 0.25 0.05


Intercept Market SMB HML NiNPWmLoNPW

Mean 0.00035 0.68 0.17 0.19 -0.04

Median 0.00029 0.74 0.12 0.19 0.02

21


22/28

Table 4: Event Period Parameter EstimatesCumulative Average Abnormal Returns Results of a seemingly unrelated regres-

sion (SUR) model analyzing returns over a 255-day estimation period and two

event windows. Window (-1,1) captures the effect of the NYAGs announcement

of a lawsuit against three insurance companies and one insurance broker. Window

(-1,3) captures the effect of a feature New York Times article outlining the lawsuit

and its potential to affect other firms. The benchmark is a four-factor return gen-

erating model including the excess market return, the difference in returns between

a set a of portfolios of small asset value firms and a set of portfolios of large asset

value firms, the difference in returns between a set of portfolios of portfolios of

high book-to-market firms and a set of portfolios of low book-to-market firms and

the difference in returns between a portfolio of high net premium written (NPW)

insurance companies and a portfolio of low NPW insurance companies. The cumu-

lative average abnormal return (CAAR) for each window is reported, along with

the generalized sign Z-test, the F-statistic for the significance of the SUR model

and the p-value for the hypothesis that the CAAR=0 in the event period.

Panel A: Firms named in lawsuit, Four-Factor Model

Window N CAAR F-statistic

p-value(numerator, denominator d.f.)

(-1,1) 3 -11.39% 84.825

0.0001(1,759)

(-1,3) 3 -14.24% 81.606

0.0001(1,750)

Panel B: Firms not named in lawsuit, Four-Factor Model



(-1,1) 72 -2.39% 11.232

0.0008(1,17204)

(-1,3) 72 -3.83% 22.638

0.0001(1,17204)

22


23/28

Table5:Cross-SectionalResults

Cross-sectiona

lresultsfromabnormaleventperiodreturnsforfirmsnotnamedintheNYAGlawsuitofOctober14,2

004.

Parameterestimatesfromthemodel:

ARj

=+Xj

+

WhereARjis

thecumulativeabnormalreturnonsecurityjandXj

isavectorofe

xplanatoryvariablesrelatedtosecurity

j,witht-statisticsinparentheses.

Parameter

(-1,1)

(-1,3)

(-1,1)

(-1,3)

(-1,1)

(-1,3)

Intercept

0.01543

0.012

-0.00183

-0.01029

-0.03715

-0.04924

(0.40)

(0.22)

(-0.05)

(-0.20)

(-1.05)

(-0.97)

Log(2003N

etPremiumWritten)

0.00179

0.00677

0.00402

0.00988

0.00205

0.00181

(0.41)

(1.06)

(0.93)

(1.58)

(0.76)

(0.47)

Log(Market

Capitalization)

-0.00379

-0.00935

-0.00423

-0.01021

(-1.06)

(-1.80)*

(-1.15)

(-1.93)*

Log(2003ContingentCommissionsPaid)

0.001

0.00116

(0.89)

(0.71)

2003Contin

gentCommissions/NWP

0.00147

0.00251

0.00209

0.00320

(0.88)

(1.04)

(1.30)

(1.39)

Broker

0.00355

0.00909

0.00253

0.0073

0.00319

0.00787

(0.42)

(0.74)

(0.29)

(0.59)

(0.35)

(0.61)

IndependentAgent

-0.01686

-0.02468

-0.0176

-0.02616

-0.01757

-0.02175

(-1.75)*

(-1.77)*

(-1.82)*

(-1.87)*

(-1.76)*

(-1.53)

ExclusiveA

gent

-0.0257

-0.0361

-0.02925

-0.03967

-0.02316

-0.02953

(-1.72)*

(-1.67)

(-2.07)**

(-1.95)*

(-1.67)*

(1.50)

DirectMarketing

0.00597

0.0154

0.00425

0.1343

0.00526

0.1210

(0.68)

(1.22)

(0.50)

(1.09)

(0.61)

(0.98)

Adj.R-square

0.0379

0.0760

0.0375

0.0846

0.0344

0.0415

F-statistic

1.38

1.79

1.37

1.89*

1.42

1.51

***Significantatthe1%level

**Significantatthe5%level

*Significant

atthe10%level

23


24/28

will be limited to firms that use independent agents and brokers as their pri-

mary means of distribution, we conduct cross-sectional regression with theevent-period abnormal returns as the dependent variable and distributionmethod with controls as the independent variables. We only include firmsnot named in the lawsuit in these regressions. Results are reported in table5. Our results show that when controlling for firm size (measured in bothnatural log of NPW and natural log of market capitalization) and ratio ofcontingent commissions to NPW, firms that use independent agents and ex-clusive agents as distribution methods have significantly negative stock pricereturns in the (-1,1) and (-1,3) windows, partially supporting hypothesis 2.While we do not observe a significant result for firms that use brokers as adistribution method, we do observe a significant coefficient on agents, consis-

tent with our hypothesis. Our observation of negative returns for exclusiveagents does not fully support our theoretical prediction, though, and sug-gests that some investors may have information about commission practiceswith exclusive agents as well as independent agents. With very low ad-justed R-square values and insignificant F-statistics in some specifications,the power of some of our models is questionable.

There are several potential explanations for the low power of some ofour results. First, some companies in the sample also have non-insurancebusinesses that may mitigate the results. For example, Berkshire Hathaway,which had a positive excess return for the (-1,1) window, operates manybusinesses in addition to its insurance businesses. Second, the power ofindicator variables to capture effects is particularly limited in cases where

one firm may employ several distribution methods. With limited data avail-able regarding the proportion of business distribution methods employedby each firm, we are unable to identify a more robust variable. Finally,our variable for ratio of contingent commissions to NPW is not broken outbetween profit-based contingent commissions and volume-based contingentcommissions. Since the NYAG lawsuit was primarily focused on volume-based contingent commissions, we may be observing confounded effects.

Also, we include two size-related variables in our return-generating vari-ables. Wilder (2004) shows that in his sample, the larger insurers weremore likely to use volume-based contingent commissions. If volume-basedcontingent commissions are, in fact, correlated with firm size, then we will

confound results by including firm size as a return-generating variable. Fi-nally, we might consider that managerial effectiveness was a factor in excessreturns during the event period. However, by including a book-to-marketfactor as a return-generating variable, we also eliminate this as an explana-tion for excess return. We conducted unreported cross-sectional regressions

24


25/28

using A.M. Best ratings as a proxy for managerial effectiveness, but did not

observe significance on any of the variables.

Table 6: Cumulative Average Abnormal ReturnsResults of a seemingly unrelated regression (SUR) model analyzing returns over

a 255-day estimation period and two event windows. Window (-1,1) captures the

effect of New York Attorney General Elliot Spitzers announcement of a lawsuit

against three insurance companies and one insurance broker. Window (-1,3) cap-

tures the effect of a feature New York Timesarticle outlining the lawsuit and its

potential to affect other firms. The benchmark is a four-factor return generating

model including the excess market return, the difference in returns between a set

a of portfolios of small asset value firms and a set of portfolios of large asset value

firms, the difference in returns between a set of portfolios of portfolios of highbook-to-market firms and a set of portfolios of low book-to-market firms and the

difference in returns between a portfolio of high net premium written (NPW) insur-

ance companies and a portfolio of low NPW insurance companies. The cumulative

average abnormal return (CAAR) for each window is reported, the F-statistic for

the significance of the joint generalized least squares regression and the p-value for

the hypothesis that the CAAR=0 in the event period.




(-1,16) 3 -8.37% 6.564 0.01061,759




(-1,16) 72 0.31% 0.016 0.8996

(1,17204)

To test the third hypothesis, that markets overreact to news and nega-tive returns are reversed shortly after the announcement of negative news,

we conducted an event study for both the group of firms named in the law-suit and the other property-casualty firms not named over a (-1,16) eventwindow. The results are presented in table 6. In that window, negativereturns for named firms were less negative than in the shorter event win-dows (-8.37%), but other firms actually essentially recovered over that event

25


26/28

window (0.31%). This supports our third hypothesis.

7 Conclusion

In this paper, we have shown that the prevailing compensation methods forindependent insurance distributors may create perverse incentives for agentsand brokers. In the face of these perverse incentives, agents and brokersmay gain advantage from collusion with insurers, expropriating wealth fromcustomers. Upon announcement of a lawsuit for such alleged behavior, theabnormal stock returns for most property and casualty insurance companiesfell significantly. We showed that this drop was not unique to firms namedin the lawsuit, but was spread across the insurance industry. While we were

able to demonstrate that firms using independent agents as a distributionmethod had a significant drop in share price, we were not able to identifyother factors that had a significant impact in the drop.

Finally, we showed that negative returns persisted for the firms named inthe lawsuit, but were less negative than in the period immediately after theannouncement. However, firms not named in the suit essentially recoveredall losses over a 17-day event window. This supports the hypothesis thatmarkets may overreact to news.

26


27/28

References

Akerlof, George A., 1970, The market for lemons: Quality uncertainty and themarket mechanism, The Quarterly Journal of Economics84, 488.

Antweiler, Werner, and Murray Z. Frank, 2005, Do us stock markets typicallyoverreact to corporate news stories?,working paper.

Baucus, Melissa S., and David A. Baucus, 1997, Paying the piper: An empiricalexamination of longer-term financial consequences of illegal corporatebehavior,Academy of Management Journal40, 129.

Bhagat, Sanjai, and Roberta Romana, 2001a, Event Studies and the Law: PartI Technique and Corporate Litigation, Yale Law School John M.Olin Center for Studies in Law, Economics and Public Policy WorkingPapers Series.

, 2001b, Event Studies and the Law: Part II Empirical Studies andCorporate Law, Yale Law School John M. Olin Center for Studies inLaw, Economics and Public Policy Working Papers Series.

Binder, John J., 1985a, Measuring the effects of regulation with stock price data,The RAND Journal of Economics16, 167.

, 1985b, On the use of the multivariate regression model in event studies,Journal of Accounting Research23, 370.

Crawford, Vincent P., and Joel Sobel, 1982, Strategic information transmission,Econometrica50, 1431.

Cummins, David J., and Neil A. Doherty, 2005, The economics of insuranceintermediaries,working paper.

DArcy, Stephen P., and Neil A. Doherty, 1990, Adverse selection, private infor-mation, and lowballing in insurance markets,The Journal of Business63, 145.

Griffen, Paul A., Joseph A. Grundfest, and Michael A. Perino, 2000, Stock priceresponse to news of securities fraud litigation: Market efficiency andthe slow diffusion of costly information, NBER Working Paper.

Hay, Bruce L., 1996, Contingent fees and agency costs, The Journal of LegalStudies25, 503.

27


28/28

Henderson, Glenn V., Jr., 1990, Problems and solutions in conducting event

studies,The Journal of Risk and Insurance57, 282.

Karafiath, Imre, 1988, Using dummy variables in the event methodology, TheFinancial Review23, 351.

Lengwiler, Yvan, and Elmar Wolfstetter, 2005, Bid rigging: An analysis of cor-ruption in auctions,CESInfo Working Paper.

Levitt, Steven D., and Chad Syverson, 2005, Market distortions when agentsare better informed: The value of information in real estate, NBERWorking Paper.

Malatesta, Paul H., 1986, Measuring abnormal performance: The event param-

eter approach using joint generalized least squares, The Journal ofFinancial and Quantitative Analysis21, 27.

Miceli, Thomas J., Katherine Pancak, and C.F. Sirmans, 2006, Is the residentalreal estate brokerage compensation model broken?, working paper.

Prince, David W., and Paul H. Rubin, 2002, The effects of product liabilitylitigation on the value of firms,American Law and Economics Review4, 44.

Regan, Laureen, and Sharon Tennyson, 1996, Agent discretion and the choice ofinsurance marketing system, Journal of Law and Economics39, 637.

Wilder, Jeffrey, 2004, Competing for the effort of a common agent: Contingencyfees in commercial lines insurance, working paper.

28

Contingent Commissions - Conflict of Interests.pdf

Documents