Regulatory competition and forbearance: Evidence from the life insurance industry

Journal of Banking & Finance 34 (2010) 522–532

Contents lists available at ScienceDirect

Journal of Banking & Finance journal homepage: www.elsevier.com/locate/jbf

Regulatory competition and forbearance: Evidence from the life insurance industry Michael K. McShane a,*, Larry A. Cox b, Richard J. Butler c a

Old Dominion University, Department of Finance, 2124 Constant Hall, Norfolk, VA 23529, USA The University of Mississippi, School of Business Administration, P.O. Box 1848, MS 38677-1848, USA c Brigham Young University, Department of Economics, 183 Faculty Office Building, Provo, UT 84602-2363, USA b

a r t i c l e

i n f o

Article history: Received 7 January 2009 Accepted 20 August 2009 Available online 26 August 2009 JEL classification: G28 G22

a b s t r a c t Regulatory separation theory indicates that a system with multiple regulators leads to less forbearance and limits producer gains while a model of banking regulation developed by Dell’Ariccia and Marquez (2006) predicts the opposite. Fragmented regulation of the US life insurance industry provides an especially rich environment for testing the effects of regulatory competition. We find positive relations between regulatory competition and profitability measures for this industry, which is consistent with the Dell’Ariccia and Marquez model. Our results have practical implications for the debate over federal versus state regulation of insurance and financial services in the US. Ó 2009 Elsevier B.V. All rights reserved.

Keywords: Regulatory competition Regulatory forbearance Insurance regulation Financial services

1. Introduction The economic theory of regulation (ETR) has evolved over the past four decades to explain the relations of producers in an industry, their consumers, and their regulators. Stigler (1971) often is credited with the seminal work in describing how producers can influence regulators of their industry to win benefits that otherwise might accrue to others, although these benefits are constrained by information and organization costs (see, e.g., Peltzman et al., 1989). Subsequent researchers expand the theory to better explain the limits of influence held by producers with regulators (Peltzman, 1976; Becker, 1983) and to generate empirical evidence in support of the expanded ETR (Danzon and Harrington, 2001; Muth et al., 2003). While many studies of the traditional ETR focus on industries overseen by a single regulator, researchers have begun to explore the effects of multiple regulators, especially in the financial services industry (see, e.g., Merton, 1995; Kane, 1999; Huizinga and Nicodème, 2006). Franks et al. (1998) specifically estimate both the direct and indirect costs of regulation for various sectors of the financial services industry in the UK, US, and France. A portion of their results imply that the cost of life insurance regulation is * Corresponding author. Tel.: +1 757 683 3602; fax: +1 757 683 5639. E-mail addresses: [email protected], [email protected] (M.K. McShane), [email protected] (L.A. Cox), [email protected] (R.J. Butler). 0378-4266/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jbankfin.2009.08.016

comparatively high in the US, which they attribute to the ‘‘extra layer” of state regulation. Franks et al. focus only on regulatory costs and not possible benefits, which they admit to being a generally serious omission in the finance literature. Merton (1995) and Kane (1999) suggest that one source of such benefits can be regulatory competition. Laffont and Martimort (1999) provide a theoretical framework showing that competition between multiple regulators raises the transaction costs of collusion between regulators and regulated firms. A major implication of their model is that separation of regulatory responsibilities among multiple regulators reduces regulatory forbearance and, therefore, producer gains attributable to their influence with regulators. This has become known as regulatory separation theory. Dell’Ariccia and Marquez (2006) develop an alternative regulatory model showing that, whether in financially integrated economies or within single economies, multiple regulators normally will generate a ‘‘competition in laxity” in setting and maintaining regulatory standards. In this study, we empirically test the very different implications of these alternative models of regulatory competition. In line with the theoretical work of Acharya (2003) and Dell’Ariccia and Marquez (2006), we focus on empirical examination of returns, i.e. profitability, of US life-health insurers during the period 1999 through 2003. The US insurance industry is particularly well-suited for testing because insurers are regulated by designated bodies in each state and territory in which they issue

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

policies. An insurer writing business in the US can be subject to as many as 55 regulators.1 While state insurance regulators attempt to coordinate many standards via model legislation developed through the National Association of Insurance Commissioners (NAIC), state legislatures independently pass their own regulations pertaining to capital adequacy, policy forms and rates, market conduct, and agent licensing, so regulations often are not uniform (Grace and Phillips, 2007; Klein and Wang, 2009). Further, regulatory compliance requirements and actions can vary across state regulators (Willenborg, 2000). Most state insurance regulators also have an economic development role within their states, which can only serve to further promote competition between state regulators. Insurers are net income generators for virtually all states in the US. In 2007, insurers paid premium, franchise, and income taxes and fees to states that exceeded the cost of state insurance regulation by approximately $15.5 billion (Lehmann, 2008). Fortunately for researchers, coordination of accounting standards via the NAIC allows us to collect and test comparable financial data for insurers domiciled in the US and also facilitates tracking of the number of regulators to which each insurer is subject. This means that US data are more homogeneous and tractable for comparative analysis than international regulatory data, which often are subject to heterogeneous legal systems, economic structures, and accounting practices. Our results generally support the implications of Dell’Ariccia and Marquez’s regulatory competition model. In particular, greater regulatory competition across the states in the US leads to higher profitability for regulated insurers, which suggests greater forbearance by state regulators and a ‘‘competition in laxity”. Our test results for capital adequacy measures also are consistent with those for profitability. These findings should be of interest to researchers, but also to political leaders, regulators, investors, and, of course, the insurance industry in the US and elsewhere.

2. Related literature The extant research on regulatory competition in financial markets can only be characterized as diverse. Kane (1997, 1999) contends that regulatory competition results in more efficient regulatory services. White (1994) argues that cartel instability results in cooperation problems among independent regulators, however. He specifically implies that regulators will cheat instead of cooperating on uniform regulation, resulting in a race to the bottom to attract firms. Vives (2001) argues that the current system of home and host country regulation of financial services in the European Union leads to negative cross-border externalities. While several recent papers have empirically investigated bank and insurance regulation (see, e.g., Gatzert and Schmeiser, 2008; Zhao et al., 2009), none have considered regulatory competition. As noted previously, Laffont and Martimort (1999) provide a theoretical framework in which regulation of a single set of agents is divided among multiple, self-interested regulators. Their model indicates that the separation of information among regulators results in Bayes–Nash behavior, which increases the costs of collusion and reduces the likelihood of regulatory capture.2 In the Laffont and Martimort model, producers subject to more regulators are less likely to be able to gain influence with, and therefore subsidies from, their multiple regulators. 1 Insurance regulatory bodies exist for each of the 50 states, four territories, and the District of Columbia. 2 An example of information separation in the US insurance industry is that the domestic state regulator of an insurer has more information about an insurer’s overall operations while a regulator in another state in which the insurer does business has more information about the insurer’s operations within that particular state.

523

Empirical tests of the Laffont and Martimort implications have been conducted using data from the US insurance industry. Virtually all of these studies apply a binary variable for single-state regulation to proxy for regulatory forbearance. Willenborg (2000) finds that as they approach insolvency, larger property–liability insurers are less likely to be subject to regulatory action when they are regulated by a single-state regulator. He finds no such evidence when larger insurers are regulated in multiple states and considers this evidence to be consistent with Laffont and Martimort’s regulatory separation hypothesis. Grace et al. (2003) investigate the costs of insolvency resolution and find these costs to be higher for insurers regulated by a single-state, which also supports greater forbearance by single-state regulators. In a study of insolvent insurer liquidations, Leverty and Grace (2004) do not find singlestate regulation to be a significant factor, however. Acharya (2003) proposes an alternative model specific to international banking systems in which central banks coordinate bank capital requirements across borders, but do not coordinate in other policy areas such as bank closures in the event of insolvency. He shows that this asymmetric coordination leads to competition between central banks, which results in ‘‘regression toward the worst forbearance” with respect to bank closure policies. Following the earlier work of White (1994) and Acharya (2003), Dell’Ariccia and Marquez (2006) develop an elegant model of competing, independent bank regulators in which the authors specifically consider the incentives for these regulators to form a regulatory union to set uniform solvency standards. While they initially model a two-nation banking system with two regulators, they then generalize their analysis to address the cases of multiple countries and of home and host country regulation. They specifically note that the implications of their model also should apply in a single economy where multiple regulators compete, such as the US banking industry, in which banks have choices of state or national regulation and several oversight agencies. While USdomiciled insurers do not have a federal option, they do have the ability to choose among 50 state regulators as their primary regulator because they can switch domiciles at relatively low cost. We consequently expect the implications of the Dell’Ariccia and Marquez model to apply to the US insurance market as well as, if not better than, the US banking market. The essence of the Dell’Ariccia and Marquez model is that, first, domestic regulators cannot internalize the spillover effects generated by setting stricter standards in their home markets and, second, if they are concerned about the owners of domestic financial institutions, regulators are likely to relax standards to increase the profitability of the domestic institutions. This ultimately can lead to a race to the bottom between domestic regulators and their external competitors as they implement increasingly lax standards. Dell’Ariccia and Marquez suggest that this race will be particularly strong when producers can change their domiciles simply by moving their headquarters, a condition quite applicable to the US insurance industry. We note that whether the relaxation of standards by local regulators truly benefits domestic producers and owners is not as relevant as the collective effect of regulatory competition on all producers and their owners that results in the ultimate ‘‘competition in laxity”. When this happens, producers subject to greater regulatory competition should reap benefits, such as profits, that may otherwise accrue to other claimants in the economic system, such as depositors and policyholders.

3. The US life-health insurance regulatory system Regulation of the US insurance industry by individual state regulators has been a contentious issue even prior to 1868 when the

524

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

US Supreme Court ruled in Paul v. Virginia that insurance is not interstate commerce, thereby allowing states to serve as industry regulators. State regulators often extol the monitoring efficacy of multiple state regulators and their local accessibility to consumers (Pickens, 2003). Once supportive of state regulation, many major insurers, especially those in the life insurance sector, now frequently decry the high cost of alleged inefficiencies and redundancies in the current, multi-jurisdictional system (see, e.g., Postal and Brady, 2007). In a related vein, researchers and other industry observers point out the disadvantageous position of insurers within the financial services industry as they compete with banks that hold the option of a federal charter (Carow, 2001). An insurer writing business in the US faces a different state regulator in each state in which the insurer is licensed. From a state regulator’s viewpoint, insurers doing business in the state are categorized as either domiciled in the state or foreign-licensed. This division has implications for a state’s regulatory responsibility. Each insurer is incorporated, that is domiciled, in one state and additionally must be licensed in each state in which it does business. If the insurer is also licensed in a state outside its state of domicile, the insurer is considered ‘‘foreignlicensed” in that state. To reduce the duplication of regulatory activities, the insurer’s state of domicile has primary regulatory responsibility for the insurer’s overall business. More specifically, a state is principally responsible for the overall financial regulation of its domiciled insurers, but also regulates the market practices of all insurers operating in the state. Klein and Wang (2009) suggest that states have become fairly standardized in financial regulation but exhibit more diversity when it comes to market regulation. We focus on the life-health, rather than property–casualty, insurance industry because we expect less noise in financial performance estimates for a number of reasons. Life-health insurance products and regulations generally are less subject to regional variation. For example, while life insurance contracts primarily encompass mortality and investment risks that are relatively uniform across the US, property–liability contracts often reflect catastrophe risks such as windstorm, hail, earth movement, and wildfire that can vary widely across states. The life-health industry generally does not face the same magnitude of shocks posed by catastrophic natural losses and sudden shifts in the legal liability system. Empirical evidence confirms that the life-health insurance industry is less subject to volatility in insurance underwriting results compared to the property–liability sector (Lamm-Tennant and Weiss, 1997). Life insurers also are much less subject to rate regulation, which can vary substantially by state for property–liability insurers (Grace and Phillips, 2008; Weiss and Choi, 2008). Finally, earnings management via accounting practices is less likely for life-health insurers because well-established actuarial tables limit managerial discretion in setting reserves (Petroni, 1992).

4.1. Profitability measures We apply various book profitability measures to assess insurer operating performance. Market, rather than book, returns would be theoretically preferable, but they are generally available only for a relatively small number of holding companies that control multiple, individual insurers and other types of firms, so those data are both quite limited and noisy. Tests of individual insurer data, rather than group or holding company data, make sense for a study of regulatory competition because insurance regulation is applied at the individual firm level.3 Further, while detailed accounting data are available for individual insurers that are part of publicly-held groups, accurate aggregation is virtually impossible for most variables, not the least of which is the number of regulators. Finally, approximately 23% of life insurers are not stock organizations and they account for nearly 25% of US premiums written. Only book data are available for these insurers in addition to the many stock insurers that are closely held. Following prior research, we use operating return on equity (OROE), operating return on assets (OROA), and operating return on sales (OROS) as alternative profitability measures (see, e.g., Mitton, 2006). We apply operating income in the numerator because Barber and Lyon (1996) find that it produces more powerful results. They reason that leverage, extraordinary items, and other discretionary items do not affect operating income and also suggest the application of multiple measures to estimate profitability.4 We first test OROE because state regulators generally focus on income and book equity to assess insurer viability. In addition, Petroni et al. (2000) argue that among profitability measures, return on equity is a more precise indicator of the effect of regulation on insurer profitability. Our profitability measures capture returns from both insurance underwriting and investment portfolio activities. We specifically follow Cummins and Nini (2002) in estimating operating income as net income before federal income taxes and policyholder dividends. The denominators in our OROE, OROA, and OROS measures are capital and surplus, total admitted assets, and net premiums written, respectively. 4.2. Regulatory factors The impact of regulation on profitability is the primary focus of our empirical investigation. We subsequently discuss two regulatory measures that can affect individual insurer profitability. 4.2.1. Regulatory competition The multi-jurisdictional regulatory system in the US allows us to observe the number of regulators faced by the insurer. We therefore use a simple count of the number of states in which the insurer is licensed to write insurance premiums as our primary proxy for regulatory competition.5 Such a count provides a richer measure than the single-state binary variable applied by previous researchers attempting to test regulatory forbearance. As discussed

4. Profitability and related factors Consistent with the theoretical model of Dell’Ariccia and Marquez (2006), we initially test the impact of regulatory competition on insurer profitability. If greater competition among regulators negatively affects profitability, the results are supportive of Laffont and Martimort’s regulatory separation theory, which suggests that competition reduces regulatory forbearance and, therefore, the ability of producers and their shareholders to extract benefits from other claimants. On the other hand, a positive effect of regulatory competition on insurer profits is consistent with Dell’Ariccia and Marquez’s hypothesis that regulatory competition leads to greater forbearance, which should improve producer profits.

3 Individual insurers, not the holding company or group, are required to submit annual accounting data to state insurance regulators. In addition, individual insurers are also subject to market conduct examinations, solvency examinations, and insolvency resolution by state regulators (Grace and Phillips, 2007). 4 Barber and Lyon (1996) note that the appropriateness of a particular book performance measure depends upon the research purpose. For example, OROA would not be a good measure for firms that have recently issued securities because the book value of assets will increase immediately without an immediately corresponding increase in operating income. In this case, a measure independent of the asset base, such as OROS, would be more accurate. OROS is far from a perfect measure, however, because it does not reflect the productivity of the asset base. 5 We count only the number of state regulators faced by the insurer and do not include regulators in US territories or the District of Columbia, so the maximum value is 50.

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

previously, the Dell’Ariccia and Marquez model of a system with multiple regulators overseeing different sets of producers is based upon assumptions that are particularly applicable to the US life insurance industry. A positive effect of regulatory competition would be supportive of the greater regulatory forbearance implied in Dell’Ariccia and Marquez’s model while a negative effect would indicate support for Laffont and Martimort’s regulatory separation theory. 4.2.2. State regulatory budgets State insurance departments vary by size, budget, and sophistication (see, e.g., Klein, 1995). To control for differences in state insurance departments, we therefore include 50 state regulatory budget variables for each insurer. The value of this proxy for each state is zero if the insurer is not licensed to write business in the state. If the insurer is licensed in the state, we apply the ratio of that state’s budget for the insurance department to total premiums written by all insurers doing business in the state. 4.3. Firm-specific control factors We necessarily employ control variables that can reflect differences in risk-bearing between insurers, which in turn should affect profitability. Our discussion of these variables follows. 4.3.1. Size Regan (1999) and Grace and Klein (2000) observe that expense ratios generally decrease with insurer size, which can be attributable to larger firms being more capable of capturing economies of scale in underwriting insurance contracts. Cummins and Nini (2002) and Liebenberg and Sommer (2008) find that larger property–liability insurers generate higher returns on equity, which they attribute to the greater market power and economies of scale and lower insolvency risk of larger insurers. Greene and Segal (2004) document a similar, positive relation between life-health insurer size and return on equity. In light of such consistent results, we expect size to be positively related to insurer profitability. We apply the natural logarithm of total admitted assets as our size proxy. 4.3.2. Geographic concentration Both policyholders and investors are concerned with the default risk of the insurer. Purchasers of insurance should view these contracts as risky debt and consequently pay lower premiums if the insurer issuing the contracts has higher default risk (Sommer, 1996). All else equal, a more geographically concentrated insurer should have higher default risk, which implies that policyholders will pay less and profits will be lower. On the other hand Cummins and Nini (2002) note that investors should expect higher returns for greater risk and they generate evidence of a positive relation between returns and geographic concentration in the property–liability insurance industry. Liebenberg and Sommer (2008) find similar results although they argue for an alternative rationale, i.e. strategically focused firms generate greater returns because the costs of diversification outweigh the benefits. Given the conflicting theory and explanations of empirical results, we cannot predict the impact of geographic concentration on insurer performance. For empirical testing purposes, we apply a geographic Herfindahl index, which is the sum of squared percentages of premiums written in each state. 4.3.3. Line of business (LOB) concentration The same arguments related to geographic concentration, risk, and return should apply to LOB concentration. In addition, the managerial discretion hypothesis of Mayers and Smith (1988) suggests that insurers operating in fewer lines of business should have

525

lower monitoring costs for owners, resulting in higher returns. Cummins and Nini (2002) and Liebenberg and Sommer (2008) provide empirical evidence of positive relations between property–liability insurer returns and LOB concentrations. We consequently expect insurers in more concentrated lines to generally exhibit higher returns. Our LOB concentration proxy is the LOB Herfindahl index, which is the sum of squared percentages of premiums written in each line of business. We adopt the same LOBs as Pottier and Sommer (1997a) to calculate this index, i.e. industrial life, ordinary life, individual annuities, credit life, group life, group annuities, group accident and health, credit accident and health, and other.

4.3.4. Stock organizational form Stock insurers should have lower owner–manager agency costs while mutual insurers should have lower owner–policyholder agency costs (Mayers and Smith, 1988). As a consequence, stock managers should invest in higher growth opportunities than mutual managers, who tend to invest in lower-return projects for which well-established actuarial information is available. Cummins and Nini (2002) and Liebenberg and Sommer (2008) offer empirical support for stock insurers generating higher returns than mutual insurers in the property–liability insurance industry. We similarly expect the stock insurers in our sample to exhibit greater profitability. Our stock variable is defined as one if the insurer is a stock organization and zero otherwise.

4.3.5. Group affiliation Some insurers are affiliated with groups of insurers while others are unaffiliated and independent. Grace and Klein (2000) argue that economies of scope can enable insurance groups to provide services at lower costs than unaffiliated insurers. Their empirical results and a study of property–casualty insurers by Regan (1999) suggest that group members report relatively higher expenses than unaffiliated insurers, however. One explanation is that more complex group structures can generate higher organizational costs that offset any group economies (Colquitt and Sommer, 2003). Because an insurance group can allow an affiliated insurer to fail while protecting its other assets, insurance purchasers may view group-affiliated insurers as riskier and thus insist on lower premiums (Cummins and Sommer, 1996), which would directly reduce profitability. Liebenberg and Sommer’s (2008) results for a property–casualty insurer sample show a negative relation between group affiliation and insurer returns. Reflecting the preponderance of empirical research, we expect group-affiliated insurers to generate relatively low profits. Our group affiliation proxy is assigned a value of one if the insurer is a member of a group and zero otherwise.

4.3.6. Financial strength Insurance rating firms supply important monitoring services (Cummins, 1988). Research on the property–liability insurance industry reveals that firms with greater financial strength, as measured by insurance rating firms, command higher premiums (Sommer, 1996). Cummins and Nini (2002) suggest that insurers with higher ratings are perceived as safer, which should result in higher returns, and their empirical tests of A.M. Best’s ratings and insurer returns are consistent with this expectation. Pottier and Sommer (2002) test a variety of financial strength measures and conclude that Best’s ratings better reflect insurers’ abilities to meet policyholder obligations. We therefore use Best’s ratings as our financial strength measure and expect a positive relation with insurer profitability. Following Pottier and Sommer (2002), we estimate financial strength by classifying Best’s ratings as follows: A++, A+ = 5; A, A = 4; B++, B+ = 3; B, B = 2; C++, C+ = 1; and other = 0.

526

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

4.3.7. Independent broker distribution Insurers distribute their products in a variety of ways, including directly through exclusive agencies and through independent agencies or brokers. We henceforth reference the former as direct distribution and the latter as broker distribution. Cummins and Weiss (1992) note that most of the relevant research indicates that direct distribution costs less and produces a higher return on equity than broker distribution. Gardner and Grace (1993) suggest that direct distribution generates lower agency costs than broker distribution, but their empirical tests are inconclusive. Regan (1999) argues that direct distribution systems produce comparative advantages in more standardized, less complex lines, but that broker distribution can generate advantages in more complicated lines. Her empirical results indicate that broker distribution leads to higher expense ratios in general, but she finds no differences in more heavily regulated insurance lines. Given the preponderance of theory and evidence we expect insurers that use broker distribution to incur greater expenses and, hence, lower profitability. We apply a simple binary variable with a value of one if the insurer uses independent broker distribution and zero otherwise. We follow Baranoff and Sager (2003) by classifying an insurer as using independent brokers if A.M. Best assigns the insurer a marketing code that contains a B (e.g., B, BA, AB, BD, and DB). Under the Best Key Rating scheme B, A, and D indicate broker, agent, and direct marketer, respectively. 4.3.8. Underwriting risk Different lines of business often are subject to varied regulations and expose insurers to different insurance underwriting risks (Cummins et al., 1999). We therefore control for variation in line of business (LOB) exposure by including a variable for each major LOB. We apply premiums written in a particular LOB, scaled by the total premiums written by the insurer, as our underwriting risk measure for all lines. The six LOBs include three lines of insurance issued to individuals (life, annuities, and accident and health) and three lines issued to groups (life, annuities, and accident and health). 4.3.9. Investment risk While underwriting risk arguably represents the primary source of insurer default risk (Downs and Sommer, 1999), investment risk is another substantial source for insurers (Cummins and Nini, 2002). We control for risk in insurer investment portfolios by including the percentage of total assets invested in each of the major asset categories in our model. These categories are cash and short-term investments, bonds, stocks, mortgages, and real estate. 4.3.10. Time effects We include year dummy variables to control for differences over time that could affect insurer profitability. 5. Research design 5.1. Data Our primary data source is the NAIC InfoPro Database for the period 1999–2003, which encompasses virtually the universe of life-health insurers licensed to do business in the US during this period. We omit firms that do not have positive assets, premiums, or surplus. We also follow Cummins et al. (1999) in excluding insurers that have average returns on equity greater than +100% or less than 100% over the years in our sample. We focus only on insurance regulation in the 50 states, so do not factor in insurance written in the District of Columbia or the four US territories.

Firms that are pure reinsurers, i.e. only report reinsurance assumptions and no cessions, are excluded. We use the A.M. Best Key Rating database for the financial strength and broker distribution variables. Data for state insurance department budgets is extracted from the NAICs Annual Insurance Department Resource Report. After merging the datasets and retaining only insurers with complete information for all five years, our final sample consists of 554 insurers, which allow a total of 2770 observations over the five years. We initially test a balanced subpanel because doing so generally reduces bias in comparison with tests of unbalanced panel data (Verbeek and Nijman, 1996). The primary disadvantage of using a balanced subpanel is the reduction in sample size, which may result in inferior estimates of variance components (Baltagi and Li, 1990). We therefore also generate estimates using the full, unbalanced panel for robustness purposes. 5.2. Model We combine all five years of data into a balanced panel and apply the following model:

Profitability ¼ b0 þ b1 Regulatory Competition þ b2 Size þ b3 LOB Conc: þ b4 Geo Conc: þ b5 Stock Form þ b6 Group Aff: þ b7 Fin: Strength þ b8 Ind: Broker þ b914 Underwriting Risk þ b1519 Investment Risk þ b2069 State Regulatory Budgets þ b7073 Time Effects þ 

ð1Þ

where LOB indicates insurance underwriting line of business. Specific proxies for all independent variables are defined in the previous section. 5.3. Estimation method In our regression analysis, we use pooled OLS, random effects, and fixed effects estimation, testing throughout to determine the most appropriate method. We apply a LaGrange multiplier (LM) test to determine whether pooled OLS is more suitable than the random or fixed effects methods. The Hausman test (1978) is used, as needed, to determine whether random effects or fixed effects estimation is better for our data. We also investigate the potential endogeneity of the regulatory competition variable because insurers with greater profitability may be more able to expand into more states. We test for this endogeneity via another test suggested by Hausman (1978) by first regressing the number of regulators on all other independent variables in Eq. (1) and instrumental variables to obtain the residuals (û). Then û is added to Eq. (1) and tested for significance via OLS regression. If û is not significant, then OLS results are appropriate. Otherwise, OLS estimates are considered inconsistent and 2SLS regression should be applied. As instrumental variables, we use the relative contiguous-state insurance department budget variables to represent the monitoring of a state’s regulator by the neighboring state regulators. For each insurer domiciled in a particular state, this measure is the mean of the relative budgets (budget/total premiums) of the contiguous states. For example, if the insurer is domiciled in Indiana, the measure is the sum of the relative state budgets of Ohio, Michigan, Illinois, and Kentucky divided by four. If the insurer is not domiciled in the state, the variable is set at zero. Butler and Appel (1990) show that contiguous states’ legislative activity affect a state legislature’s propensity to pass maximum benefit levels with

527

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

respect to workers’ compensation. We extrapolate on this logic by using contiguous-state budgets to discern possible effects of competitive legislative pressure with respect to regulatory monitoring activity.

6. Results 6.1. Summary data and statistics Table 1 shows the number of licensed and domiciled life-health insurers by state. Texas has the most licensed insurers operating within its boundaries while New York has the least, but these two states rank first and second, respectively, in the number of domiciled insurers. New York insurance regulation is perceived as particularly stringent and is applied extraterritorially (Pottier and Sommer, 1998). Insurance groups therefore frequently set up single-state insurers to do business in New York, which is a reason that New York has both a low number of licensed insurers and a high number of domiciled insurers. Based upon the total premiums written in each state, New York represents the second largest lifehealth insurance market after California. The information in Table 2 disaggregates our 554 sample insurers by the number of states in which insurers are licensed and, hence, the number of regulators faced. Substantial variation in the number of regulators overseeing our sample insurers is apparent. For example, 86 insurers face only one state regulator, while 49 face all 50 state regulators. Such variation is important if our regulatory competition proxy is to be effective. Table 3 provides financial and line of business information in four classifications of insurers based upon the number of regulators faced. On average, insurers facing 26–50 regulators have higher returns on equity than insurers facing fewer than 26 regulators, a pattern suggestive of competition in laxity. The results for the ROA measure do not show any such relation and, if anything, the ROS estimates run counter to the ROE results. Other notable patterns are that total assets, group affiliations, and Best’s ratings in-

crease with the number of states in which insurers are licensed. The vast majority of insurers in all categories have a stock organizational form although, as noted previously, we cannot discern the proportion that is closely held. The more widely licensed insurers also tend to underwrite relatively more individual annuities. Conversely, the less dispersed insurers underwrite a higher percentage of group accident and health insurance. Regarding asset holdings, the more dispersed insurers tend to hold more bonds while those operating in fewer states tend to hold more stocks and cash or cash equivalents. 6.2. Regression analysis and results Correlation analysis of the proxies for independent variables in our model, for which results are not fully reported here, reveals a few Pearson correlation coefficients exceeding ±0.50. We observe positive relations of this magnitude between size and both our regulatory competition and financial strength estimates. Similar magnitudes of negative correlation apply between geographic concentration and both regulatory competition and size. We therefore compute the variance inflation factors (VIFs) developed by Belsley et al. (1980), and find that all fall below 5.0. Hence, collinearity is unlikely to be a problem in our regressions. We also alternately omit these variables without substantial changes in regression results. The LaGrange multiplier (LM) test results show that fixed and random effects regressions are more suitable than the pooled OLS method, while a subsequent Hausman test indicates that fixed effects regression is more appropriate than random effects estimation. In the first stage of the test for the potential endogeneity of the regulatory competition variable described earlier, an F-test of our instrumental variables, the relative contiguous-state insurance department budget variables, indicates a jointly significant effect (p-value: 0.0087) on regulatory competition. In the second stage, the residuals from the first stage have a p-value of 0.5544, which does not does not allow us to reject exogeneity at any reasonable significance level. We therefore do not perform 2SLS regressions.

Table 1 Number of licensed and domiciled life-health insurers by state. State

Number of licensed insurers

Number of domiciled insurers

Total premiums written in state ($mil)

State

Number of licensed insurers

Number of domiciled insurers

Total premiums written in state ($mil)

AK AL AR AZ CA CO CT DE FL GA HI IA ID IL IN KS KY LA MA MD ME MI MN MO MS

236 313 340 328 282 322 239 291 325 311 246 291 301 346 336 322 314 348 239 298 205 283 262 339 331

0 7 16 14 11 5 13 15 9 9 2 15 4 43 22 5 3 19 11 5 2 12 9 22 11

670 4,900 2,900 8,600 46,000 7,700 12,000 3,400 26,000 10,000 1,700 6,400 2,000 27,000 8,800 5,200 3,800 5,600 19,000 9,900 1,500 16,000 8,400 9,000 3,300

MT NC ND NE NH NJ NM NV NY OH OK OR PA RI SC SD TN TX UT VA VT WA WI WV WY

298 292 301 318 188 232 322 305 101 318 332 305 298 226 324 304 335 363 323 300 193 291 283 286 278

0 3 3 15 2 4 1 0 58 20 14 2 21 1 9 2 6 62 11 8 2 9 15 0 0

890 12,000 890 3,900 2,400 19,000 1,700 2,200 36,000 19,000 3,700 4,300 23,000 1,700 4,600 1,300 7,800 31,000 2,800 12,000 820 8,000 9,500 1,800 570

The table lists the number of licensed and domiciled insurers in each state and the total life-health insurance premiums written in each state in 2001. Sample size: 554 insurers.

528

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

Table 2 Number of regulators faced by life-health insurers. No. of regulators faced

No. of insurers

No. of regulators faced

No. of insurers

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

86 27 12 6 9 8 11 2 4 9 6 6 6 3 7 3 4 5 3 3 3 2 5 2 7

26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

5 2 7 4 4 8 1 2 4 3 2 5 7 7 7 8 4 8 5 7 21 18 20 107 49

The table categorizes insurers by the number of regulators faced, which is determined by the number of states in which the insurer is licensed in 2001. Sample size: 554 insurers.

Table 4 contains results for fixed effects regressions for our balanced panel.6 Coefficient estimates are provided for the eight variables for which we previously made predictions, but, for the sake of space, are omitted for the variables controlling for underwriting risk, investment risk, and regulatory budgets. Regulatory competition is positively related to all three profitability measures with significance at the 1% level in the OROE results and at the 5% level for OROS. As explained by Ziliak and McCloskey (2004), researchers should consider the economic, as well as statistical, significance of the independent variables that they analyze. Looking at the product of the coefficient for OROE and the standard deviation of that variable, we find that an increase of one standard deviation in the number of regulators is associated with an increase of operating return on equity from the mean of 5.1% to a mean of 7.9%. A similar calculation for OROS estimates reveals an even more substantial rise. We conclude that regulatory competition positively affects profitability and is economically significant, consistent with Dell’Ariccia and Marquez’s (2006) model. In other words, greater regulatory competition leads to higher profitability, which suggests greater regulatory forbearance. Size is positively related to profitability in all cases and significant at the 10% level for return on equity and at the 5% level for return on assets and sales, consistent with prior research. The product of the coefficient and standard deviation of our size variable reveals that an increase of one standard deviation in the size of the firm is associated with an increase in return on equity of only 5.5%, that is, from a mean of 5.1% to a mean of 5.4%. The economic significance of the size factor therefore is questionable at best. The line of business (LOB) concentration variable is positive 6 A possible concern for fixed effect regressions is that some variables may not have ‘‘within” variation. For our main variable of interest, regulatory competition, about 30% of the insurers in our sample have variation in the number of regulators they face over the five year period. The variables that have the least variation over time are group affiliation (6%), stock organization form (2%), and independent broker distribution (1%).

in all cases and significant for the OROE regression at the 1% level. We further find that a one standard deviation increase in LOB concentration is associated with an 18% rise in return on equity, which would be 6.0% compared to the mean OROE of 5.1%. In such a lowmargin business, this impresses us as economically significant. The LOB results are consistent with both the managerial discretion and strategic focus hypotheses. In contrast to previous findings in studies of property–liability insurers, we do not find significant relations between geographic concentration and profitability. A possible explanation is that geographic concentration is not as strong a risk factor for life-health insurers as for property–liability insurers because the latter often are exposed to catastrophic losses that are regional in nature. The stock organizational form variable is not significant, which suggests that mutual and stock insurers are sorted into lines of business that allow them comparative advantages in minimizing agency costs, so that differences in efficiency will not be apparent (see, e.g., Cummins and Zi, 1998). Group affiliation also is not a significant factor for our sample. These results run counter to previous empirical findings for the property–liability insurance industry, but are in line with the hypothesis that group economies of scope may offset the higher costs of the more complex group structure.7 The financial strength variable is positive in all cases and significant at the 5% level in the OROE and OROA regressions and at the 10% level for OROS. In terms of economic significance, an increase by one level in our scheme of Best’s ratings relates to a 40% increase in OROE, which represents a rise to 7.1% from the mean of 5.1%. Although this increase is economically significant, we note that a one-unit change in our proxy actually can represent as much as a three-level rise or fall in Best’s ratings, so such a strong effect is to be expected. Nevertheless, our findings generally support the hypotheses that financial rating firms provide insurer monitoring services that keep agency costs in check and that consumers pay more for the products of insurers that they perceive as safer. The independent broker variable is insignificant for all regressions, which sustains the argument that neither distribution system has an absolute advantage, although either could have a comparative advantage depending upon the complexity of the product being distributed. 6.3. Robustness tests If some of the regressors are affected by insurer profitability, endogeneity could cause bias and inconsistency problems for our results. Such results for our main variable of interest, regulatory competition, would be of particular concern. To investigate this issue we run separate regressions using lagged values for the size, LOB concentration, geographic concentration, financial strength, underwriting risk, and investment risk variables, then compare the results for the regulatory competition variables in these regressions with our original results. Lags of suspected endogenous regressors are considered to be predetermined variables, which removes some of the endogeneity in large samples. The p-values for

7 To address the possibility that decisions on the number of states in which to operate are made at the group rather than the individual insurer level, a regression with data aggregated by group would be informative. Our regulatory competition and independent broker variables are impossible to aggregate for a group, however. For example, our independent broker proxy is a binary variable that cannot be aggregated if a group contains insurers that use different types of distribution systems. We do investigate whether the other independent variables in the main regression are sensitive to group affiliation by splitting insurers into group and non-group subsamples and running separate regressions for the subsamples. We find that most of the independent variables are not sensitive to group affiliation. The only difference between the two subsample results is for the size variable. Size is positive and significant for the all regressions of the non-group subsample but is significant only for the OROE regression of the group subsample.

529

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532 Table 3 Selected data for insurers by aggregate number of regulators faced. Aggregate number of regulators faced (measured by number of states in which insurers are licensed) (1–5) (6–25) (26–45)

(46–50)

Percent of insurers (%) Means for each category Profitability (OROE) Profitability (OROA) Profitability (OROS)

25

39

Total assets ($mil) LOB concentration Stock form (%) Group affiliation (%) Independent broker (%) Best’s rating Life (individual) (%) Annuities (individual) (%) A & H (individual) (%) Life (group) (%) Annuities (group) (%) A & H (group) (%) Bond (%) Stock (%) Real estate (%) Cash (%)

219 0.739 90.10 64.10 19.30 1.81 26.90 8.60 10.00 5.50 1.10 34.00 60.00 10.50 2.00 21.90

18

0.065 0.0220 0.1264

18

0.058 0.0172 0.1060

0.077 0.0280 0.0433

324 0.688 91.00 69.10 32.00 2.4 32.70 9.20 13.90 10.00 0.70 22.80 69.70 6.90 1.80 13.50

0.078 0.0134 0.0953

1,630 0.693 90.00 79.70 26.00 3.14 30.60 15.60 18.60 7.70 1.20 23.30 75.80 6.00 0.80 10.20

10,000 0.618 86.10 95.90 38.50 4.07 38.80 17.40 9.40 8.50 5.20 17.40 74.70 5.70 0.80 4.90

Each insurer is classified into one of four categories depending on the number of regulators faced by the insurer. The table shows the means of selected data for each category in 2001. Sample size: 554 insurers. OROE is operational return on equity (ratio of operating income to capital and surplus). Operating income is net income before federal income taxes and policyholder dividends. Total assets is total admitted assets. Line of business (LOB) concentration = LOB Herfindahl index, which is the sum of squared percentages of premiums written in each line of business. Stock form % means the percentage of insurers that have a stock organizational form. Group affiliation % means the percent of insurers that are affiliated with a group. Independent broker % means the percent of insurers that use independent broker distribution. Best’s rating means A.M. Best’s financial strength rating (A++, A+ = 5; A, A = 4; etc.). Life (individual) % means the percent of the insurer’s total premiums that are individual life premiums. Annuities (individual) % means the percent of the insurer’s total premiums that are individual annuity premiums. A & H (individual) % means the percent of the insurer’s total premiums that are individual accident and health premiums. Life (group) % means the percent of the insurer’s total premiums that are group life premiums. Annuities (group) % means the percent of the insurer’s total premiums that are group annuity premiums. A & H (group) % means the percent of the insurer’s total premiums that are group accident and health premiums. Bond % means the percent of the insurer’s investment assets that are in bonds. Stock % means the percent of the insurer’s investment assets that are in stocks. Real Estate % means the percent of the insurer’s investments assets that are in real estate. Cash % means the percent of the insurer’s total investment assets that are in cash.

Table 4 Fixed effects regressions of regulatory competition on various measures of firm performance. Balanced panel consists of 554 insurers with complete data for all years 1999–2003. OROE

Intercept Regulatory competition Size LOB concentration Geographic concentration Stock organization form Group affiliation Financial strength Independent broker R2 No. of firms = 554 No. of observations = 2770

OROA

OROS

Coeff.

p-Value

Coeff.

p-Value

Coeff.

p-Value

0.100 0.027 0.021 0.271 0.008 0.029 0.070 0.397 0.418 0.487

0.875 0.004*** 0.084* 0.005*** 0.901 0.688 0.100 0.030** 0.657

0.242 0.004 0.015 0.039 0.001 0.001 0.005 0.104 0.195 0.645

0.148 0.160 0.021** 0.117 0.971 0.975 0.686 0.031** 0.432

5.401 0.100 0.252 0.370 0.248 0.086 0.037 1.535 0.724 0.265

0.090 0.036** 0.037** 0.238 0.605 0.815 0.861 0.095* 0.878

Dependent variables: OROE is operational return on equity (ratio of operating income to capital and surplus). OROA is operational return on assets (ratio of operating income to total admitted assets). OROS is operational return on sales (ratio of operating income to net premiums written). Operating income is net income before federal income taxes and policyholder dividends. Independent variables shown: Regulatory competition = number of states in which the insurer is licensed. Size = natural logarithm of total admitted assets. Line of business (LOB) concentration = LOB Herfindahl index, which is the sum of squared percentages of premiums written in each line of business. Geographic concentration = geographic Herfindahl index, which is the sum of squared percentages of premiums written in each state. Stock organizational form = 1 if the insurer is a stock organization, 0 otherwise. Group affiliation = 1 if the insurer is a member of a group, 0 otherwise. Financial strength = Best’s rating (A++, A+ = 5; A, A = 4; etc.). Independent broker = 1 if the insurer uses independent broker distribution, 0 otherwise. Independent variables not shown: The following independent variables are included as control variables in the regression (see Eq. (1)), but results for these variables are not shown in this table. Underwriting risk consists of six variables: XLOB/(total premiums written), where XLOB are the premiums written in a specific line of business. The six lines of business used are three individual lines (life, annuities, and accident and health) and three group lines (life, annuities, and accident and health). Investment risk consists of five variables: XA/(total investment assets), where XA is a specific asset category. The five asset categories used are cash and short-term investments, bonds, stocks, mortgages, and real estate. For each state, a relative budget variable is defined as the ratio of the state’s insurance department budget to the total premiums written by all insurers in the state if the insurer is licensed in the state, or 0 if the insurer is not licensed in the state. Time effects are controlled for by including year dummy variables. * Indicate significance at the 10%, level. ** Indicate significance at the 5%, level. *** Indicate significance at the 1%, level.

530

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

Table 5 Fixed effects regressions of regulatory competition on various measures of firm performance. Unbalanced panel consists of 737 insurers with complete data during any year from 1999 to 2003. OROE

Intercept Regulatory competition Size LOB concentration Geographic concentration Stock organization form Group affiliation Financial strength Independent broker R2 No. of firms = 737 No. of observations = 3350

OROA

OROS

Coeff.

p-Value

Coeff.

p-Value

Coeff.

p-Value

0.180 0.030 0.081 0.094 0.123 0.040 0.061 0.573 1.065 0.4961

0.511 0.012** 0.002*** 0.021** 0.255 0.692 0.238 0.015** 0.355

0.131 0.009 0.056 0.007 0.056 0.009 0.015 0.121 0.456 0.5093

0.125 0.140 <0.001*** 0.347 0.287 0.852 0.540 0.093* 0.418

1.781 0.078 0.260 0.262 0.593 0.605 0.029 0.354 2.391 0.5859

0.592 0.077* <0.001*** 0.325 0.726 0.698 0.971 0.337 0.491

Dependent variables: OROE is operational return on equity (ratio of operating income to capital and surplus). OROA is operational return on assets (ratio of operating income to total admitted assets). OROS is operational return on sales (ratio of operating income to net premiums written). Operating income is net income before federal income taxes and policyholder dividends. Independent variables shown: Regulatory competition = number of states in which the insurer is licensed. Size = natural logarithm of total admitted assets. Line of business (LOB) concentration = LOB Herfindahl index, which is the sum of squared percentages of premiums written in each line of business. Geographic concentration = geographic Herfindahl index, which is the sum of squared percentages of premiums written in each state. Stock organizational form = 1 if the insurer is a stock organization, 0 otherwise. Group affiliation = 1 if the insurer is a member of a group, 0 otherwise. Financial strength = Best’s rating (A++, A+ = 5; A, A = 4; etc.). Independent broker = 1 if the insurer uses independent broker distribution, 0 otherwise. Independent variables not shown: The following independent variables are included as control variables in the regression (see Eq. (1)), but results for these variables are not shown in this table. Underwriting risk consists of six variables: XLOB/(total premiums written), where XLOB are the premiums written in a specific line of business. The six lines of business used are three individual lines (life, annuities, and accident and health) and three group lines (life, annuities, and accident and health). Investment risk consists of five variables: XA/(total investment assets), where XA is a specific asset category. The five asset categories used are cash and short-term investments, bonds, stocks, mortgages, and real estate. For each state, a relative budget variable is defined as the ratio of the state’s insurance department budget to the total premiums written by all insurers in the state if the insurer is licensed in the state, or 0 if the insurer is not licensed in the state. Time effects are controlled for by including year dummy variables. * Indicate significance at the 10% level. ** Indicate significance at the 5% level. *** Indicate significance at the 1% level.

the regulatory competition coefficients are 0.002, 0.177, 0.028 for the ROE, ROA, and ROS measures, respectively. The magnitude of the coefficients and these p-values are very similar to those for the original regressions that are shown in Table 4. Our initial tests used a balanced panel consisting of 554 insurers for which complete data are available for all five years. To ensure that we have not discarded valuable information that may affect variance estimates we run the same regression on an unbalanced panel of the 737 insurers reporting complete data for any of the five years. This increases the total number of observations from 2770 to 3350. The results for the unbalanced panel, shown in Table 5, generally are consistent with our prior results, except for a few, relatively minor exceptions. The regulatory competition measure remains positive for all three profitability measures and only slightly less significant. Size results become even more significant. Dell’Ariccia and Marquez (2006) contend that regulators prefer that entities they regulate maximize profits given the regulatory monitoring level and they model this relation, hence our primary focus on profitability tests. Just as Dell’Ariccia and Marquez iterate the ways in which national regulators can differentially regulate international banks, we previously described the variety of methods that state regulators use to control incentives for managers of US life insurers. Not all of these methods are quantifiable using publicly available data, but some applications of financial measures definitely are. Klein and Wang (2009) explain that state regulators of US insurers primarily focus on minimum capital requirements and risk-based capital (RBC) measures. RBC estimates are designed to incorporate a variety of risk charges for credit, market, and insurance underwriting risk. Because minimum capital requirements are very low and generally non-binding (see, e.g., Klein, 1995), the primary measure of capital adequacy published and monitored by regulators is RBC. We consequently apply our model to determine

if the ‘‘competition in laxity” we observed for profitability measures also applies for the primary capital adequacy measure monitored by state regulators. To control for outliers, we omit insurers that have an RBC ratio in excess of six standard deviations from the mean, which reduces our sample size by about 1.5%. The results in Table 6 show that our measure of regulatory competition is statistically significant at the 10% level and negatively related to capital adequacy. Specifically, greater regulatory competition correlates to lower capital adequacy, on average, among US life insurers. This evidence suggests that insurers facing more regulators hold less capital, and offers additional support for a competition in laxity. Similarly, size is significantly and negatively related to capital adequacy, which is consistent with the findings of Pottier and Sommer (1997b).

7. Conclusion Even though financial services firms—such as banks, investment companies, and insurers—face multiple regulators, empirical work related to multi-jurisdictional regulation of these essential industries remains relatively rare. Such research has been limited because theoretical models involving multiple regulators have not been well-developed until recently. Prominent models of regulatory competition were introduced by Laffont and Martimort (1999) and Dell’Ariccia and Marquez (2006), but they produce very different predictions about the effect of competing regulators on the firms that they regulate. Laffont and Martimort’s regulatory separation model predicts that firms facing more regulators are less likely to capture the regulators, which implies less regulatory forbearance. Alternatively, the Dell’Ariccia and Marquez (2006) model predicts that regulatory competition will lead to greater regulatory forbearance. The US life insurance industry, with over 50 regulators and relatively uniform

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532 Table 6 Fixed effects regressions of regulatory competition on risk-based capital ratio (RBCR). Balanced panel consists of 554 insurers with complete data for all years 1999–2003. Unbalanced panel consists of 737 insurers with complete data during any year from 1999 to 2003. Dependent variable: RBCR

Intercept Regulatory competition Size LOB concentration Geographic concentration Stock organization form Group affiliation Financial strength Independent broker R2 Number of firms Number of observations

Balanced panel

Unbalanced panel

Coeff.

p-Value

Coeff.

p-Value

5.559 1.536 3.169 1.564 0.131 0.459 0.677 5.705 6.517 0.310 545 2725

0.015** 0.070* 0.053* 0.714 0.840 0.927 0.815 0.648 0.919

4.077 1.634 1.738 2.190 0.447 0.659 0.794 6.938 3.057 0.352 721 3243

0.050* 0.073* 0.015** 0.224 0.949 0.862 0.943 0.710 0.865

Dependent variable: Risk-based capital ratio (RBCR) is the ratio of the insurer’s actual level of capital to its risk-based capital. Independent variables shown: Regulatory competition = number of states in which the insurer is licensed. Size = natural logarithm of total admitted assets. Line of business (LOB) concentration = LOB Herfindahl index, which is the sum of squared percentages of premiums written in each line of business. Geographic concentration = geographic Herfindahl index, which is the sum of squared percentages of premiums written in each state. Stock Organizational form = 1 if the insurer is a stock organization, 0 otherwise. Group affiliation = 1 if the insurer is a member of a group, 0 otherwise. Financial strength = Best’s rating (A++, A+ = 5; A, A = 4; etc.). Independent broker = 1 if the insurer uses independent broker distribution, 0 otherwise. Independent variables not shown: The following independent variables are included as control variables in the regression (see Eq. (1)), but results for these variables are not shown in this table. Underwriting risk consists of six variables: XLOB/(total premiums written), where XLOB are the premiums written in a specific line of business. The six lines of business used are three individual lines (life, annuities, and accident and health) and three group lines (life, annuities, and accident and health). Investment risk consists of five variables: XA/(total investment assets), where XA is a specific asset category. The five asset categories used are cash and short-term investments, bonds, stocks, mortgages, and real estate. For each state, a relative budget variable is defined as the ratio of the state’s insurance department budget to the total premiums written by all insurers in the state if the insurer is licensed in the state, or 0 if the insurer is not licensed in the state. Time effects are controlled for by including year dummy variables. * Indicate significance at the 10% level. ** Indicate significance at the 5% level.

and comprehensive financial data, represents an excellent arena for empirically testing the implications of these theories. Our results generally support the Dell’Ariccia and Marquez model in that greater regulatory competition is positively and significantly related to various measures of insurer profitability, especially operating return on equity. The economic significance of regulatory competition on profitability is particularly notable. Further confirmation is provided by the significantly negative relation that we observed between regulatory competition and risk-based capital. Overall, our results are consistent with regulatory competition leading to greater forbearance and, hence, stronger financial performance among the regulated firms. For years, state insurance regulators and their supporters in the US have contended that the regulation of insurers at the state level, which implies some redundancies and inefficiencies, actually produces economic benefits including better protection of consumer interests, in comparison with a single-regulator system (see, e.g., Pickens, 2003). The regulatory separation theory of Laffont and Martimort (1999) and supporting empirical literature provides a possible theoretical rationale for a system of multiple, competing regulators that can reduce the costs of regulatory capture by the regulated industry. Our results run counter to these arguments

531

that regulatory competition essentially is good, however. These findings should be very relevant during a time when various initiatives that advocate at least some degree of federal regulation to supplant multi-state regulation are being considered by the US Congress. Recent changes in the balance of power in the executive and legislative branches, coupled with both direct and indirect emergency financial support by the US government to American International Group and several large, publicly held life insurers via their banking subsidiaries, only add impetus to this issue in the US. The implications of our findings for the regulation of financial entities across national borders are much tougher to judge. The subject of our empirical work, US insurance regulation, is a unique amalgam of competition and cooperation among state-based regulators who face the long-term risk of losing their sinecures to federal entities. We certainly cannot maintain that our findings can be projected to other nations. What we can say is our research suggests that the implications of the regulatory competition model developed by Dell’Ariccia and Marquez, apparently with the international banking system in Europe in mind, apply to a different financial industry and very different regulatory system ‘‘across the pond”. Further research on why US insurers cluster at the two extremes of the spectrum with regard to number of states licensed impresses us as warranted. The determinants of this separation and how managers of insurers exploit niches in the diverse regulatory environment provided by state regulation should be of particular interest. If the US moves toward a single-regulator system, will this render some firms’ strategies obsolete? Will it also affect the mix of firms in the industry? The US insurance industry data is well-suited for testing such regulatory issues. The applicability to and implications for international regulation also impress us as ripe areas for further exploration.

Acknowledgments The authors thank Marty Grace, Ty Leverty, Walt Mayer, Rich Phillips, participants in the 2007 Risk Theory Society Seminar, Ike Mathur (the editor), and an anonymous reviewer for their helpful input on prior versions of this paper. We also thank Rich Phillips for providing state budgetary data.

References Acharya, V., 2003. Is the international convergence of capital adequacy regulation desirable? Journal of Finance 58, 2745–2781. Baltagi, B., Li, Q., 1990. A comparison of variance components estimators using balanced versus unbalanced data. Econometric Theory 6, 283–285. Baranoff, E., Sager, T., 2003. The relations among organizational and distribution forms and capital and asset risk structures in the life insurance industry. Journal of Risk and Insurance 70, 375–400. Barber, B., Lyon, J., 1996. Detecting abnormal operating performance: The empirical power and specification of test statistics. Journal of Financial Economics 41, 359–399. Becker, G., 1983. A theory of competition among pressure groups for political influence. Quarterly Journal of Economics 98, 371–400. Belsley, D., Kuh, E., Welsch, R., 1980. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. John Wiley, New York. Butler, R., Appel, D., 1990. Benefit increases in workers compensation. Southern Economic Journal 56, 594–606. Carow, K., 2001. The wealth effects of allowing bank entry into the insurance market. Journal of Risk and Insurance 68, 129–150. Colquitt, L., Sommer, D., 2003. An exploratory analysis of insurer groups. Risk Management and Insurance Review 2, 83–96. Cummins, D., 1988. Risk-based premiums for insurance guaranty funds. Journal of Finance 43, 823–839. Cummins, D., Nini, G., 2002. Optimal capital utilization by financial firms: Evidence from the property–casualty insurance industry. Journal of Financial Services Research 21, 15–53. Cummins, D., Sommer, D., 1996. Capital and risk in property–liability insurance markets. Journal of Banking and Finance 20, 1069–1092.

532

M.K. McShane et al. / Journal of Banking & Finance 34 (2010) 522–532

Cummins, D., Tennyson, S., Weiss, M., 1999. Consolidation and efficiency in the life insurance industry. Journal of Banking and Finance 23, 325–357. Cummins, D., Weiss, M., 1992. Controlling automobile insurance costs. Journal of Economic Perspectives 6, 95–115. Cummins, D., Zi, H., 1998. Comparison of frontier efficiency methods: An application to the U.S. life insurance industry. Journal of Productivity Analysis 19, 131–152. Danzon, P., Harrington, S., 2001. Workers’ compensation rate regulation: How price controls increase costs. Journal of Law and Economics 44, 1–36. Dell’Ariccia, G., Marquez, R., 2006. Competition among regulators and credit market integration. Journal of Financial Economics 79, 401–430. Downs, D., Sommer, D., 1999. Monitoring, ownership and risk-taking: The impact of guaranty funds. Journal of Risk and Insurance 66, 477–497. Franks, J.R., Schaefer, S.M., Staunton, M.D., 1998. The direct and compliance costs of financial regulation. Journal of Banking and Finance 21, 1547–1572. Gardner, L., Grace, M., 1993. X-Efficiency in the US life insurance industry. Journal of Banking and Finance 17, 410–497. Gatzert, N., Schmeiser, H., 2008. Combining fair pricing and capital requirements for non-life insurance companies. Journal of Banking and Finance 32, 2589–2596. Grace, M., Klein, R., 2000. Efficiency implications of alternative regulatory structures for insurance. In: Wallison, P. (Ed.), Optional Federal Chartering and Regulation of Insurance Companies. American Enterprise Institute, Washington, DC, pp. 79–131. Grace, M., Klein, R., Phillips, R., 2003. Insurance company failures: Why do they cost so much? Working paper, Georgia State University. Grace, M., Phillips, R., 2007. The allocation of governmental regulatory authority: Federalism and the case of insurance regulation. Journal of Risk and Insurance 74, 207–238. Grace, M., Phillips, R., 2008. Regulator performance, regulatory environment and outcomes: An examination of insurance regulator career incentives on state insurance markets. Journal of Banking and Finance 32, 116–133. Greene, W., Segal, D., 2004. Profitability and efficiency in the U.S. life insurance industry. Journal of Productivity Analysis 21, 229–247. Hausman, J., 1978. Specification tests in econometrics. Econometrica 46, 1251– 1271. Huizinga, H., Nicodème, G., 2006. Deposit insurance and international bank liabilities. Journal of Banking and Finance 30, 965–987. Kane, E., 1997. Ethical foundations of financial regulation. Journal of Financial Services Research 12, 51–74. Kane, E., 1999. Implications of superhero metaphors for the issue of banking powers. Journal of Banking and Finance 23, 663–673. Klein, R., 1995. Insurance regulation in transition. The Journal of Risk and Insurance 62, 363–404. Klein, R., Wang, S., 2009. Catastrophe risk financing in the US and the EU: A comparative analysis of alternative regulatory approaches. Journal of Risk and Insurance 76, 607–637. Laffont, J., Martimort, D., 1999. Separation of regulators against collusive behavior. RAND Journal of Economics 30, 232–262. Lamm-Tennant, J., Weiss, M., 1997. International insurance cycles: Rational expectations/institutional intervention. Journal of Risk and Insurance 64, 415– 439. Lehmann, R.J., 2008. The cost of regulation. Best’s Review 109, 32–38. Leverty, T., Grace, M., 2004. Dupes or incompetents? An examination of management’s impact on property–liability insurer distress. Working paper, Georgia State University. Liebenberg, A., Sommer, D., 2008. Effects of corporate diversification: Evidence from the property–liability insurance industry. Journal of Risk and Insurance 75, 893–919.

Mayers, D., Smith, C., 1988. Ownership structure across lines of property casualty insurance. Journal of Law and Economics 31, 351–378. Merton, R.C., 1995. Financial innovation and the management and regulation of financial institutions. Journal of Banking and Finance 19, 461–481. Mitton, T., 2006. Stock market liberalization and operating performance at the firm level. Journal of Financial Economics 81, 625–647. Muth, M., Rucker, R., Thurman, W., Chuang, C., 2003. The fable of the bees revisited: Causes and consequences of the U.S. honey program. Journal of Law and Economics 46, 479–516. Peltzman, S., 1976. Toward a more general theory of regulation. Journal of Law and Economics 19, 211–240. Peltzman, S., Levine, M., Noll, R., 1989. The economic theory of regulation after a decade of deregulation. Brookings Papers on Economic Activity. Microeconomics, pp. 1–59. Petroni, K., 1992. Optmistic reporting in the property–casualty insurance industry. Journal of Accounting and Economics 15, 485–508. Petroni, K., Ryan, S., Wahlen, J., 2000. Discretionary and non-discretionary revisions of loss reserves by property–casualty insurers: Differential implications for future profitability, risk, and market value. Review of Accounting Studies 5, 95– 125. Pickens, M., 2003. The NAIC’s 2003 agenda modernizing state insurance regulation: A legacy of consistency, efficiency and trust. Journal of Insurance Regulation 21, 117–125. Postal, A.D., Brady, M., 2007. Industry comments to treasury support of optional federal charter. National Underwriter (Life & Health ed.) 111, 6–40. Pottier, S., Sommer, D., 1997a. Agency theory and life insurer ownership structure. Journal of Risk and Insurance 64, 529–543. Pottier, S., Sommer, D., 1997b. Life insurer risk-based capital measures. Journal of Insurance Regulation 16, 179–196. Pottier, S., Sommer, D., 1998. Regulatory stringency and New York licensed life insurers. Journal of Risk and Insurance 65, 485–502. Pottier, S., Sommer, D., 2002. The effectiveness of public and private sector summary risk measures in predicting insurer insolvencies. Journal of Financial Services Research 21, 101–116. Regan, L., 1999. Expense ratios across insurance distribution systems: An analysis by line of business. Risk Management and Insurance Review 2, 44–59. Sommer, D., 1996. The impact of firm risk on property–liability insurance prices. Journal of Risk and Insurance 63, 501–514. Stigler, G., 1971. The theory of economic regulation. Bell Journal of Economics and Management Science 2, 3–21. Verbeek, M., Nijman, T., 1996. Incomplete panels and selection bias. In: Matyas, L., Sevestre, P. (Eds.), The Econometrics of Panel Data. Kluwer Academic Publishers, Amsterdam, pp. 449–490. Vives, X., 2001. Restructuring financial regulation in the European monetary union. Journal of Financial Services Research 19, 57–82. Weiss, M., Choi, B.P., 2008. State regulation and the structure, conduct, efficiency and performance of US auto insurers. Journal of Banking and Finance 32, 134– 156. White, L., 1994. On the international harmonization of bank regulation. Oxford Review of Economic Policy 10, 94–105. Willenborg, M., 2000. Regulatory separation as a mechanism to curb capture: A study of the decision to act against distressed insurers. Journal of Risk and Insurance 67, 593–616. Zhao, T., Casu, B., Ferrari, A., 2009. The impact of regulatory reforms on cost structure, ownership and competition in Indian banking. Journal of Banking and Finance, doi:10.1016/j.jbankfin.2009.07.022. Ziliak, S., McCloskey, D., 2004. Size matters: The standard error of regressions in the American economic review. The Journal of Socio-Economics 33, 527–546.