1. Daily Return Behavior of the Insurance Industry: The Case of Contingent Commission Jiang Cheng Elyas Elyasiani Tzuting Lin Temple University 2007 ARIA Annual Meeting, Quebec City

2. 2 Motivation New York Attorney General Eliot Spitzer filed a civil suit in the State Supreme Court against Marsh & McLennan Cos. for “bid-rigging” and inappropriate use of “contingent commissions” on Oct. 14, 2004. We test the market reaction on insurance brokers and property-liability and life-health-accident insurers from the civil action suit using event study methodology within a GARCH framework. The bid-rigging event provides a good opportunity to test the effects of contingent commissions on the insurance industry.

3. 3 Findings The event generated negative effects both within the brokerage sector and for individual brokerage firms, suggesting that the contagion effect dominates the competitive effect. The inter-sectoral information spillover effects across the brokerage, property-liability, and life-health sub-sectors of the insurance industry are also significant and mostly negative. Our results support the information-based hypothesis against the pure-panic contagion effect as the size of the impact due to the event is highly correlated with firm characteristics. ARCH/GARCH effects are significant for both the sectoral portfolios and about half of individual brokers and property-liability insurers.

4. 4 Insurance Marketing systems and Contingent Commission Direct Marketing Insurers (DMIs) : direct writer + exclusive agents Insurers with Independent Intermediaries (IIIs) : independent agents + brokers Contingent Commission pros: alignment of interests between insurers and brokers cons: the potential conflict of interest for brokers and against the buyers

5. 5 Literature Event Study: the effects of California’s Proposition 103 (Fields et al., 1990; Szewczyk and Varma, 1990; Shelor and Cross, 1990; Grace et al., 1995; and Brockett et al., 1999), the 1989 California earthquake (Shelor et al., 1992), trouble in investment portfolio of First Executive and Travelers (Fenn and Cole, 1994), Hurricane Andrew (Lamb, 1995; Angbazo and Narayanan, 1996), property-liability insurance market pullout (McNamara et al., 1997), the terrorist attacks of September 11, 2001 (Cummins and Lewis, 2003), the European Union Insurance Directives (Campbell et al., 2003), and the impact of operational loss events in the U.S. banking and insurance industries (Cummins et al., 2006a, 2006b). Contingent commission (Cummins and Doherty, 2006; Kleffner and Regan, 2007). Stock return data often exhibit GARCH properties (Engle, 1982; Bollerslev, 1987; Akgiray, 1989; Lamoureux and Lastrapes 1990).

6. 6 List of HypothesesOutcome of the Test H1H1 Announcement of the “bid-rigging” event has no intra- sectoral effect; contagion and competitive effects offset one another exactly. Rejected. H2H2 Announcement of the “bid-rigging” event produces competitive effect which dominates the contagion effect. Rejected. H3H3 Announcement of the “bid-rigging” event has no effect on the insurers. Rejected. H4H4 The response of insurers’ stock prices to announcements of the “bid-rigging” event is independent of the insurers’ marketing system. Rejected. H5H5 Announcement of the “bid-rigging” event does not differentially affect stock prices of insurers with respect to their size. Rejected. H6H6 Announcement of the “bid-rigging” event does not differentially affect stock prices of insurers with respect to their payment of net contingent commission. Rejected. H7H7 Announcement of the “bid-rigging” event does not differentially affect stock prices of insurers with respect to business concentration. Rejected.

7. 7 Methodology GARCH (1, 1) model Determinants of Abnormal Returns

8. 8 Data The SIC codes used are: 6331 for property-liability, 6311 for life, and 6320-6321 for health and accident insurers, and 6411 for the broker companies. 74 property-liability insurers (excluding AIG, ACE, and Hartford), 40 life- health-accident insurers, and 10 insurance brokers (excluding MMC). The market return is measured using the CRSP equally weighted index. The property-liability insurers’ financial data is obtained from the Best’s Key Rating Guide and A.M. Best’s Aggregates and Averages.

9. 9 Table 1. Estimation of Stock Brokers and Insurers Portfolios Return Sensitivities to the Bid-rigging Event Stock PortfolioInterceptMarketD -1 D0D0 ARCH 0 ARCH 1 GARCH 1 Persistence Broker 0.000854 (3.00)** 0.7395 (21.37)*** -0.0185 (-9.27)*** -0.0366 (-21.90)*** 0.00003036 (3.30)*** 0.2490 (6.00)*** 0.3110 (2.07)** 0.5600 Property-Liability 0.0000288 (0.19) 0.7930 (36.25)*** 0.00131 (0.57) -0.0162 (-6.59)*** 0.00000825 (5.36)*** 0.0377 (1.34) 0.3275 (2.71)*** 0.3652 Life-Health- Accident 0.0000857 (0.38) 0.9382 (30.40)*** 0.000188 (0.05) -0.0164 (-4.47)*** 0.00000639 (1.46) 0.01180 (0.47) 0.6915 (3.21)*** 0.7033

10. 10 Table 2. Estimation of Individual Stock Brokers Return Sensitivities to the Bid-rigging Event StockInterceptMarketD -1 D0D0 ARCH 0 ARCH 1 GARCH 1 Persistence Aon Corp.0.0008520.8002***-0.0188***-0.1935***0.000039***0.363***0.411***0.7736 Brooke Corp.0.0047761.2433*** 0.0041 0.0029 Brown & Brown0.0012460.6906*** 0.0007-0.0719*** Gallagher Arthur0.0000650.4013*** 0.00571*-0.0261***0.000040***0.516***0.239*0.7545 Hilb Rogal0.0007380.9180***-0.00299-0.0817*** Hub Intl. Ltd.0.0002390.2611** 0.0021-0.0258* National Fin.0.0013851.0209*** 0.0173* 0.0162*0.000222***0.305***0.02240.3271 Quotssmith Com.0.0009640.4933**-0.0340 0.0041 U S I Holdings0.0005550.6417*** 0.0138*-0.0566***0.000064***0.305***0.466***0.7710 Willis Group0.0005970.5780***-0.0139-0.0676***

11. 11 Table 3. Brokers Ranks, Revenues, Market Share and Contingent Commissions as Percent of Revenues Brokerage Industry Rank 2004 Revenues ($Millions) Marker Share % Percentage of Contingent Commissions to Revenues % Stock Aon23105.916.60%2.00% Brooke Co3265.9070.40%3.10% Brown & Brown Inc7638.2673.40%6.00% Gallagher Arthur J & Co31192.686.40%3.00% Hilb Rogal & Hamilton Co8601.7343.20%NA Hub Intl Ltd12231.441.20%6.00% National Financial Partners CoNA Quotesmith Com IncNA U S I Holdings Co10405.822.20%5.00% Wollis Group Holdings Limited41036.355.50%4.00% Marsh15804.431.10%7.30%

12. 12 Table 4. Estimation of Individual Stock Property-Liability Insurers Return Sensitivities to the Bid-Rigging Event StockInterceptMarketD -1 D0D0 ARCH 0 ARCH 1 GARCH 1 21st Century Group-0.0005411.3005***-0.0233-0.0152 21st Century Holding-0.0003061.1955***-0.0025720.0160.000427***0.3506***0.4522*** ACE Ltd.-0.0003850.7916*** 0.007412-0.0678***0.0000745***0.1839**0.3421* AIG-0.0002890.7914*** 0.0107**-0.0786***0.0000321**0.2397**0.4230* ALFA-0.0001151.1232***-0.0115-0.004559 Alleghany 0.001177*0.3698*** 0.0123 0.03530.000004260.0907**0.8720*** Allianz-0.0008321.3843*** 0.00731-0.0079090.00002120.12620.7340*** Allmerica-0.00111.5249***-0.0040251-0.013 Allstate 0.0003910.6477***-0.000404-0.001638 American Financial Group 0.0002190.7863*** 0.001352-0.0114

13. 13 Appendix B. Descriptive Statistics for Property-Liability Insurers Variables and DefinitionsMean Std. Deviation Abnormal return on the event day, October 14, 2004 -0.01540.0180 Abnormal return on one day before the event day, October 13, 2004 0.00020.0124 Cumulative abnormal return of the event day and one day before -0.00750.0186 Marketing dummy variable equal to one if the insurer is an III, and zero if the insurer is a DMI 0.79730.4048 Size=Log of the total admitted assets for insurer14.43191.6016 Contingent= ratio of insurer’s total payment of Net Contingent Commission to its Net Premium Written 1.09601.67728 Commercial=ratio of insurer’s premium written in commercial lines to total premiums written from all lines 0.55250.35618 The interaction term of the above two ratio: (Commercial*Contingent) 0.51750.88578 Leverage= ratio of insurers’ premium written to surplus 1.49510.83178 Return is the insurer’s return on policyholders’ surplus 8.152515.46658 Multi-line dummy=1 if the insurer has business in Life-Health-Accident insurance lines, and zero otherwise 0.21620.4145 Regulation dummy=1 if the insurer regulatory location is New York, and zero otherwise. 0.06760.25275

14. 14 Table 5. Determinants of the size of abnormal returns (Cross-Sectional Analysis) VariablesCoefficientt-ratio Intercept 0.03002( 1.24) Marketing dummy variable equal to one if the insurer is an III, and zero if the insurer is a DMI-0.01450 **(-2.28) Size= Log of the total admitted assets for insurer-0.00245 *(-1.71) Contingent= ratio of insurer’s total payment of Net Contingent Commission to its Net Premium Written 0.00920 ***( 2.98) Commercial= ratio of insurer’s premium written in commercial lines to total premiums written from all lines 0.01299( 1.58) The interaction term of the above two ratio: (Commercial*Contingent)-0.01941 ***(-3.34) Leverage= ratio of insurers’ premium written to surplus -0.00219(-0.77) Return is the insurer’s return on policyholders’ surplus-0.00013(-0.77) Multi-line dummy=1 if the insurer has business in Life-Health-Accident insurance lines, and zero otherwise-0.00077(-0.14) Regulation dummy=1 if the insurer regulatory location is New York, and zero otherwise.-0.00030(-0.03) Number of observations74 Adi. R-square 0.1522 F-statistic 2.28 **

15. 15 Comments and Suggestions? Thank you!