Regulator Performance, Regulatory Environment and

Outcomes An Examination of Insurance Regulator Career

Incentives on State Insurance Markets

by

2007 Annual Meeting of the American Risk and Insurance Association

Quebec City, Canada

Martin F. GraceGeorgia State University

mgrace@gsu.edu

Richard D. PhillipsGeorgia State University

rphillips@gsu.edu

Revolving Door

The door can swing in two directions

Industry Government

Government Industry

DAVID LAURISKI, chosen as the Labor Department’s Assistant Secretary of Mine Safety and Health, previously spent 30 years in the mining industry, during which time he advocated loosening of coal dust standards. Once in office, he issued controversial rules (later blockedby the Senate) that would have reduced coal-dust testing in mines. Lauriski resigned from his position in late 2004 and took a job with a mine-industry consulting company. The Charleston Gazette later reported that Lauriski had been negotiating for private-sector jobs as early as six months before leaving office.Source: A Matter of Trust (Revolving Door Working Group: Washington D.C., 2005)

Revolving Dooro Debate exists whether revolving doors between a regulatory

agency and the industry it oversees should be open or closed

o Traditional concern is about capture by regulated industry o Stigler (Bell Journal 1971)

o However, agency can be captured by other than industry interestso Peltzman (J. of Law and Economics 1976)

o More recent literature suggests o Regulators may pursue own private incentives

o Laffont (various papers with various authors)

o Revolving doors may provide incentives that actually increase overall efficiency of the regulated market place

o Che (RAND 1995)o Salant (RAND 1995)

Research Objectives of this Paper

o Who are the regulators of the insurance industry?o Where do they come from?o What professional backgrounds do they have?o How do they become insurance commissioner?o Where do they go when they are no longer the insurance

commissioner?

o Can we find evidence that the decisions of insurance regulators, in states that provide them a means, are influenced by their o Pre-agency employmento Post-agency career ambitions, or by o The manner through which they attained the office?

o Can we explain the rather surprising result that rate regulation, on average, appears to have little effect on insurance prices yet in some states and at some times the effect appears large?

Prior Literatureo Regulatory Incentives

o Peltzman (JLE 1976)o Laffont and Martimort (RAND 1999)o Che (RAND 1995)o Salant (RAND 1995)

o Selection Mechanismo Besley and Coate (J. of the European Econ. Assoc. 2003)

o Rate Regulation and Automobile Insuranceo Harrington (Brookings-AEI 2002)o Cummins, Phillips, and Tennyson (JIR 2001)o And a number of other papers not cited here…

Revolving Door Sample of the Empirical Literature

o Gormley (American Journal of Political Science 1979)o Revolving door between industry and government for FCC regulators

o Cohen (American Journal of Political Science 1986)o Revolving door between government and industry for FCC regulators

o Gely and Zardkoohi (American Law & Economic Review 2001) o Revolving door between U.S. cabinet positions and lobbying firms

o Boylan (American Law & Economic Review 2005)o Revolving door between U.S. attorneys and judgeships and big-firm

partnerships

Industrial Focus of Our Researcho Personal auto insurance is an ideal setting

o Some states provide authority for regulator to approve rates prior to their use (Prior Approval States)

o Some states provide limited or no authority to approve rates (Competitive States)

o Some states have elected commissioners – others are appointed

o One insurance commissioner per state

o Proxy for the price of automobile insurance is well established in the literature

Database of State Insurance Commissioners: 1985 - 2002

o Biographical information on all state insurance commissioners

o Information collected includedo Gendero Educational backgroundo Prior agency professional

backgroundo Post agency career choice

o Data Sourceso NCOIL’s The Insurance Legislative Fact Book and Almanac (various

years)o Trade press and national and local newspapers in Lexis-Nexis and

Factivao Google/Internet Searches

Total number of insurance comm’s 254- Number of interim comm’s 10- Number of comm’s w/missing data 23Number of comm’s in study 221 Comm’s from regulated states 111 Comm’s from competitive states 110

Anecdotal Comments about the Commissioners

o One state had three insurance commissioners, in a row, convicted of either o Lying to a federal agento Accepting illegal campaign contributions, and/or o Bribery

Most recent two seem to be clean!

o Six commissioners in our sample resigned for ethical reasons or were serving jail time. Since our sample two more have been forced from office over ethical issues.

Anecdotal Comments about the Commissioners (2)

o 52% of commissioners had no industry experience prior to their appointment or election

o Biographical statisticso Just over 75 % are maleo 48% are lawyerso 22%went to graduate schoolo Some of undergraduate majors represented in our sample*

o Businesso Educationo Psychologyo Mortuary Science!

o Two states have had only one commissioner during the entire time period of our study

* - Note - data on majors is incomplete

Empirical Test

where pit = unit price of personal automobile liability insurance in year t and state iXm = is a vector of explanatory market variables in year t and state iPA = is an indicator for prior approval statute in year t and state iXr = is a vector of indicator variables controlling for the professional background, the prior and post agency employment choices of the insurance regulators, and the manner by which the regulator attained the position in year t and state it = is the year specific error termηi= is the state specific error term eit= random error termm, r, = estimated coefficients

i

m rit it it i t itp PA PA m r

it itX X

Estimation Methodologyo Two sources of potential endogeneity

o Decision by the states to allow the regulator to intervene may be jointly determined with insurance prices

o Prior literature mixed whether this selectivity bias existso Cummins, Phillips and Tennyson (2001)o Harrington (2002)

o Econometric technique: Heckman’s (1978) two-stage sample selection model extended to a treatment effects model

o Estimate first-stage Probit regressiono Calculate inverse Mills terms for regulated and unregulated

state-year observations

Estimation Methodology (2)o Two source of potential endogeneity, continued

o Commissioner’s post employment choices may be jointly determined with the decisions she makes while in office

o Econometric technique: Similar to Heckman’s two-stage sample selection model except post-agency choice is modeled using multinomial logistic regression.

o See Lee (1983) for details.

o Test null hypothesis of exogoneity using F-test restrictions ono Regulation inverse Mills terms jointly equal zeroo Post-agency career choice inverse Mills terms jointly equal zero

o Regression methodologyo Two-way fixed effects estimated using weighted least squares

Data Sources and Sampleo Data Sources

o State insurance marketso AIPSOo NAIC

o Political Environmento Berry et al (1998) Political Ideology Index

o Demographic and Economic Environmento U.S. Bureau of the Census

o Include all state-year observations from 1985 – 2002o Exclude the District of Columbiao Eliminate state-year observations where we do not have

complete data on the insurance commissionero Final sample is 708 state-year observations

Personal Automobile Unit Price Ratio All Coverages by Regulatory Regime: 1985 - 2002

1.04

1.08

1.12

1.16

1.20

1.2419

8519

8619

8719

8819

8919

9019

9119

9219

9319

9419

9519

9619

9719

9819

9920

0020

0120

02

Year

Prem

ium

Rat

io

Competitive StatesRegulated States

Premium Ratio is defined as the ratio of direct premiums earned divided by direct loss and loss adjustment expenses incurred plus policyholder dividends paid. Source: NAIC

Summary StatisticsRegulated vs. Unregulated States: 1985-

2002Regulated CompetitiveVariable Mean MeanUnit Price Ratio 1.191 1.194 0.462Ave. Loss Per Car Year ($2004) $ 688.12 $ 607.41 6.317 ***Ave. Prem Per Car Year ($2004) $ 816.97 $ 721.29 6.605 ***Percent Car Years in Residual Market 6.12% 0.83% 7.880 ***Total Written Car Years (000's) 2,767 3,440 2.984 ***Percent Liability Losses Incurred in No-fault 8.86% 8.33% 0.597Percent Losses Incurred in Liability Coverages 71.39% 69.04% 4.629 ***Percent Premiums by Direct Writers 62.10% 63.93% 1.936 *Percent Private Sector Employees in Ins. Ind. 1.89% 1.87% 0.495Percent State GDP from Insurance 2.40% 2.37% 0.346Percent of National Insurance GDP from State 2.02% 2.45% 2.216 **Indicator = 1 if commissioner is elected 0.293 0.105 6.410 ***Liberal Ideology Index 54.064 45.466 4.647 ***Real Per Capita Income ($2004) $ 28,688 $ 27,840 2.549 **Population Density 256.78 120.74 7.368 ***Percent Miles Driven in Urban Areas 56.04% 50.05% 4.821 ***Poverty Rate 12.33% 12.14% 0.726Note - there are 363 unregulated and 345 regulated state-year observations.***, **, * denotes statistical significance at the 1, 5 and 10 percent levels, respectively.

T-Testmreg = mcomp

Biographical Statistics of Insurance Comm’s Regulated vs. Unregulated States: 1985 - 2002

Variable Regulated CompetitiveCharacteristics and Background

Indicator = 1 if commissioner is elected 22.52% 6.36% 3.51 ***Tenure in Office 4.64 3.95 1.46 *Consumer Advocate 3.60% 0.91% 1.36 *College Graduate 90.99% 90.91% 0.02Graduate from Law School 49.55% 38.18% 1.71 **

gradschool Degree from Graduate School 20.72% 28.18% 1.29 *Male 74.77% 78.18% 0.60

T-Testmreg = mcomp

***, **, * denotes statistical significance at the 1,5, or 10 percent levels, respectively

Pre-and-Post Agency Choices of Ins. Comm’s Regulated vs. Unregulated States: 1985-2002

Variable Regulated CompetitiveCareer Prior to Tenure as Insurance Commissioner

Career Insurance Bureaucrat 18.92% 18.18% 0.14Career Bureaucrat Not J ust Insurance 15.32% 24.55% 1.73 **Career Politician 13.51% 7.27% 1.53 *Private Sector Insurance Industry 17.12% 16.36% 0.15Private Sector Not Exclusively Insurance 12.61% 6.36% 1.60 *Mixture Gov't and Private Sector 22.52% 27.27% 0.82

Employment After Tenure as Insurance CommissionerReturned to Insurance Department 2.70% 5.45% 1.04Seek Higher Office 8.11% 6.36% 0.50Private Sector Insurance 45.05% 50.00% 0.74Private Sector Other Than Insurance 9.91% 15.45% 1.24Indicted, J ailed or Resign Amid Ethical Charges 3.60% 1.82% 0.82Died 0.90% 0.00% 1.00Retired 5.41% 6.36% 0.30Still in office in 2002 11.71% 1.82% 2.99 ***

T-Testmreg = mcomp

***, **, * denotes statistical significance at the 1,5, or 10 percent levels, respectively

State Insurance Commissioners Career Choice Transition Matrix

Career Prior to Ins. Comm. Num

Ins. Dep't

Lateral Move

in State Gov't

Seek Higher Office

Private Sector

Ins.

Private Sector Other Than Ins. Indicted Died Retired

Still in office

in 2002

Career Ins. Bureaucrat 41 7 5 0 15 3 1 0 4 6

Career Bureaucrat Not J ust Ins.

44 1 9 1 23 7 1 0 0 2

Career Politician 23 0 3 8 7 1 0 0 2 2

Private Sector Ins. 37 0 0 0 23 5 1 0 5 3

Private Sector Not J ust Ins. 21 0 3 1 9 5 2 0 1 0

Mixture Gov't & Private Ind. 55 1 8 6 28 7 1 1 1 2

Totals 221 9 28 16 105 28 6 1 13 15

Estimated Marginal Effects from Multinomial Logistical Regression of Post-Agency Career Choice

Post Employment Varialbe / ChoiceGraduate from Law School -0.103 * 0.038 0.218 *** -0.086

Degree from Graduate School -0.011 0.036 -0.036 0.052

Male -0.004 -0.036 0.072 -0.012

Mixture Gov't and Private Sector -0.072 0.098 0.016 -0.012

Career Politician -0.125 *** 0.622 *** -0.390 ** -0.117 **

Prior Industry Experience -0.257 *** 0.007 0.151 0.062

Percent State GDP from Insurance 0.501 -1.974 -2.104 3.332Percent of National Insurance GDP from State -1.784 -0.500 3.315 * -0.161

Ave. Size of the Residual Market Final Two Years in Office -0.732 -0.159 0.785 -0.019

Ave. Loss Per Car in State to Ave. Loss Per Car in Nation 0.265 ** 0.044 -0.345 * 0.077***, **, * denotes statistical significance at the 1, 5 and 10 percent levels, respectively.

Lateral Move into

Gov't

Seek Higher

Political

Private Sector -

Ins.

Private Sector - Not Ins.

Probit Regression Results of the Choice of Regulatory Regime: 1985 - 2002VariableIntercept 0.035Percent Private Sector Employees in Insurance Industry -41.069 ***Ave. Loss Per Car in State to Ave. Loss Per Car in Nation 0.351Lagged(Percent Car Years in Residual Market) 6.617 ***Lagged(Percentage Increase in Ave. Loss Per Car) 1.115Percent Premiums by Direct Writers -1.172 *Indicator = 1 if commissioner is elected 1.200 ***Percent Liability Losses Incurred in No-fault Coverage -0.757Liberal Ideology Index 0.007 ***Real Per Capita Income ($2004) 0.000Population Density 0.001 ***Percent Miles Driven in Urban Areas 1.327 ***Poverty Rate 0.009Log-Likelihood Function Value -448.82Pseudo R-Squared 20.84%***, **, * denotes statistical significance at the 1, 5 and 10 percent levels,respectively.

State Unit Price of Automobile Ins.Two Way Fixed-Effects Regression Results

Price Regulation EndogenousPost Agency Choice EndogenousVariableIntercept 2.298 *** 2.146 *** 2.147 *** 1.914 *** 1.964 ***% Liability Losses Incurred in No-fault -0.578 *** -0.598 *** -0.597 *** -0.581 *** -0.587 ***% Losses Incurred in Liability Coverages 0.839 *** 0.853 *** 0.854 *** 0.870 *** 0.872 ***% Premiums by Direct Writers -0.195 *** -0.153 *** -0.151 *** -0.182 *** -0.193 ***Ave. Loss Per Car in State to National Ave. -0.143 *** -0.134 *** -0.134 *** -0.125 *** -0.122 ***Log(Written Car Years) -0.100 *** -0.091 *** -0.092 *** -0.072 ** -0.076 **PA Indicator 0.017 * 0.001 0.004 0.008 0.004PA x Consumer Advocate -0.044 *** -0.051 *** -0.050 *** -0.057 *** -0.060 ***PA x Elect 0.032 0.053 ** 0.052 ** 0.044 ** 0.050 **PA x Prior Mix Gov't and Private Sector -0.005 -0.005 0.001 0.002PA x Prior Politico 0.022 ** 0.023 ** 0.015 0.014PA x Prior Insurance Industry -0.013 -0.013 -0.009 -0.009PA x Private Sector Other Industries -0.018 * -0.018 * -0.011 -0.011PA x Probability Post Lateral Move to Gov't -0.005 -0.005 0.006 0.007PA x Probability Post Higher Office 0.063 *** 0.063 *** 0.069 *** 0.072 ***PA x Probability Post Private Sector Ins. 0.017 0.017 0.010 0.011PA x Post Private Sector Other Than Ins. 0.008 0.008 0.005 0.005R-squared 0.668 0.698 0.699 0.706 0.706F-Statistics for Selectivity TestsRegulated = Unregulated IMR = 0 0.040 0.690Post Bureaucrat IMR = Post Higher Office IMR = Post Insurance Industry IMR = Post Other Industry IMR = 0 3.230 *** 3.310 ***

Model 5Model 1 Model 2 Model 3 Model 4

YesNo No No Yes YesNo No Yes No

***, **, * denotes statistical significance at the 1,5, or 10 percent levels, respectively

Estimated Effect of Regulator Career Incentives on the Price of Insurance in

Regulated States

Price Regulation EndogenousPost Agency Choice EndogenousVariableEstimated Effect of Prior Approval Regulation by Regulator TypePrior Approval + Prior Approval x Consumer Advocate = 0 -0.050 ** -0.047 * -0.049 * -0.055 **Prior Approval + Prior Approval x Elected Commissioner = 0 0.054 ** 0.056 ** 0.052 * 0.055 *Prior Approval + Prior Approval x Prior Mixed Public-Private = 0 -0.004 -0.001 0.009 0.006Prior Approval + Prior Approval x Prior Politico = 0 0.023 0.026 * 0.023 0.018Prior Approval + Prior Approval x Prior Insurance Industry = 0 -0.012 -0.010 -0.001 -0.005Prior Approval + Prior Approval x Prior Other Industry = 0 -0.017 -0.014 -0.003 -0.007Prior Approval + Prior Approval x Post Lateral Move to Gov't = 0 -0.004 -0.001 0.014 0.012Prior Approval + Prior Approval x Post Higher Office = 0 0.064 *** 0.066 *** 0.077 *** 0.076 ***Prior Approval + Prior Approval x Post Private Sector Insurance = 0 0.018 * 0.021 0.018 * 0.015Prior Approval + Prior Approval x Post Private Sector Other Than Ins. = 0 0.004 0.003 0.006 0.007***, **, * denotes statistical significance at the 1, 5 and 10 percent levels, respectively.

YesModel 1 Model 2 Model 3 Model 4 Model 5

No No No YesNo No Yes No Yes

Conclusionso The overall estimated impact of rate regulation appears small

o We find significant evidence of a revolving door from government to industryo Prices appear jointly determined with the post-agency career aspirations of

the insurance commissioners in regulated stateso Particularly strong for commissioners with aspirations to seek higher political officeo Weaker evidence for commissioners seeking post-industry employment

o We find no evidence the prior employment background of insurance regulators leads to price outcomes significantly different than what one would expect from a similarly situated state with a competitive rating law

o We find some evidence the selection mechanism motivates commissioners inconsistent with the results of Besley and Coate (2003)

o We find some evidence avowed consumer advocates are successful reducing the price of insurance