1 The Journal of Risk and Insurance, 1998, Volume 65, No. 3, Market Structure and Performance in Private Passenger Automobile Insurance Vickie L. Bajtelsmit Raja Bouzouita ABSTRACT The exemption of the insurance industry from federal antitrust law has generated some controversy, particularly in light of the well-publicized price and availability crises in certain lines of insurance. Antitrust restrictions in the United States are generally based on the assumption that high levels of concentration in an industry will make it more likely that firms will collude to raise prices and restrict supply, resulting in higher prices for consumers. This study examines the relationship between profitability and market structure in automobile insurance and tests for the existence of a positive relationship between concentration and performance. The results of the analysis show a significant positive impact of concentration on profitability for combined liability and physical damage lines in private passenger automobile insurance for the period 1984 through Differences in rate regulation across states are not found to impact profitability. INTRODUCTION Despite the proliferation of research on the relationship between concentration and profitability in many industries, 1 the existence of such a relationship in the property and liability insurance sector has not received much attention in the literature. Carroll (1993) examines this issue for the workers compensation market and is unable to find the positive relationship that is predicted by industrial organization theory. Since automobile insurance is the largest source of income for propertyliability insurers 2 and competition in consumer markets tends to be local, concentration may be a more significant factor in pricing for this line than it is for workers compensation. Although many companies sell automobile insurance, most are affiliated with larger groups of insurers and a fairly small number of companies control a large share of the business in the United States, with more concentration Vickie L. Bajtelsmit is in the Department of Finance and Real Estate at Colorado State University and Raja Bouzouita is in the Department of Economics and Finance at Central Missouri State University. The authors gratefully acknowledge advice on earlier drafts of this paper received from J. David Cummins, Lisa Posey, Linda Lee, session participants at the 1995 American Risk and Insurance Association Meeting, and two anonymous referees. 1 See, for example, Scherer and Ross (1990). 2 Direct premiums written in the auto line (commercial and personal combined) accounted for approximately forty-four percent of industry premiums in 1990 and forty-six percent in (A.M. Best Company 1991, 1993)
2 504 The Journal of Risk and Insurance in certain states. 3 Under these circumstances, market structure may play a greater role in the pricing of private passenger auto insurance. The auto insurance industry has been under fire in the last two decades and has been accused of making excessive profits by colluding to restrict supply and keep prices artificially high. The passage of Proposition 103 in the state of California demonstrates that such beliefs may be widely held and can have dramatic consequences. Legislative proposals at the federal level, intended to repeal the antitrust exemption for insurers under the McCarran Ferguson Act, can also be attributed to this type of consumer attitude. Although the last decade has been a period of rising auto insurance premiums, particularly in certain geographical areas, the industry has consistently argued that higher prices are the result of rising costs, and that profit margins are slim. 4 This argument apparently won judicial support in California where the voter-approved twenty percent rollback of premiums was halted based on evidence that such a reduction would imply negative profits for many insurers. The purpose of this study is to investigate the relationship between profitability and concentration in the United States private-passenger automobile insurance industry. Since antitrust regulation is based on the assumption that concentration promotes collusive behavior, the results of this study should prove valuable to policy-makers, industry professionals and, ultimately, to consumers as well. The next section of the paper details the theoretical background and reviews previous empirical studies. The third section presents the data and methodology followed by a summary of the empirical results. Conclusions and policy implications are provided in the final section of the paper. BACKGROUND The industrial organization literature provides two competing hypotheses for the relationship that exists between performance and concentration. The structureconduct-performance paradigm (SCP) suggests that market share concentration creates conditions for collusive anti-competitive behavior that can lead to monopoly profits. The SCP paradigm therefore predicts a positive relationship between profitability and market concentration level. The efficient structure (ES) hypothesis offers an alternative explanation for the positive relationship between profit and concentration. First proposed by Demsetz (1973), the ES hypothesis suggests that firms with superior efficiency (lower costs) will gain a larger market share. Thus, if larger firms have a comparative advantage in production or services provided, they may achieve higher profits without resorting to collusive measures such as raising prices or restricting supply. From a public policy standpoint, the SCP paradigm and the ES hypothesis have conflicting implications for industrial 3 For example, although there were approximately three thousand companies selling automobile insurance in the United States in 1992, there were half as many affiliated groups. Chidambaran et al. (1997) report that the largest four auto insurers control more than forty percent of the national market. In local markets, this percentage is often much higher. 4 See Cummins and Tennyson (1992) for a discussion of this issue. Their analysis shows that higher costs are a plausible explanation for recent increases in automobile premium levels.
3 Market Structure and Performance 505 regulation. The first provides justification for strict antitrust regulation and the latter would imply that an unrestricted marketplace would result in the lowest prices to consumers. Structurally, the insurance market has generally been characterized as monopolistically competitive since there are a large number of companies, low concentration, and low entry barriers (Joskow 1973; Cummins and Weiss 1992; Joskow and McLaughlin 1992). However, in markets with heterogeneous products 5 and individually determined prices, a large number of firms does not necessarily imply that prices are competitive. Although Joskow s (1973) earlier study concluded that the insurance market has all the characteristics of a competitive market, Cummins and Weiss (1992) suggest that overall market concentration measures may disguise higher levels of concentration in certain lines of business. This conclusion is supported by Chidambaran et al. (1997) who find that concentration levels in private passenger automobile insurance are much higher than in most other lines of business. 6 Similarly, the use of national as opposed to state premium data is likely to further disguise the concentration that exists in local markets, particularly for automobile insurance. Even though Carroll s (1993) empirical results do not support either of the competing explanations for insurer performance in the worker s compensation line, inherent differences between the worker s compensation and automobile insurance make it difficult to draw parallel inferences without further analysis. Thus, the purpose of this study is to examine the relationship between performance and market structure in private-passenger automobile insurance markets and to determine whether concentration within a state has a positive effect on insurer profitability. DATA AND METHODOLOGY State regulation of auto insurance and the local marketing of personal auto insurance policies to consumers, imply that any analysis of performance should be at the state level. 7 Relevant company data by state for operations in all fifty states 8 obtained from the National Association of Insurance Commissioners Although insurance policies are relatively homogeneous due to standardization of contracts, contract prices are individually determined and consumers are not well informed about price dispersion. Heterogeneity is also introduced by differences in claim settlement procedures by company and firm specific cost structures (Eastman 1994). 6 The mean four-firm concentration ratio for shows that 41.15% of all premiums are written by the four largest firms. Only fidelity, boiler and machinery lines have higher levels of concentration (Chidambaran et al. 1997). 7 Although Joskow and McLaughlin (1991) and Chidambaran et al. (1997) make the argument that the appropriate unit of measurement for insurance studies is a national rather than a state market, in private passenger auto insurance, this argument is less convincing. Consumers tend to search only locally and the degree of regulation differs significantly by state. In addition, there are large differences in concentration by state, with Herfindahl indices in personal auto liability insurance ranging from 0.04 to The District of Columbia and United States territories are excluded from this analysis due to incomplete data.
4 506 The Journal of Risk and Insurance annual reports for the private passenger physical damage and liability lines are used in this study. Consistent with Weiss (1974) and Carroll (1993), industry profitability in a given state s is hypothesized to be a function of state-specific and market variables in the following general form: Profit s = α + β 1 (Concentration s ) + Σ j=2 to n β j X sj, (1) where X s2...x sn are control variables for market and state characteristics. This general form allows us to determine the relative importance of concentration as a determinant of industry performance in state markets. Applying the general model in equation (1) to the private passenger automobile insurance lines, the profitability π in state s and year t is estimated in the following equation: π st = α + β 1 HERF st + β 2 MCR st + β 3 DWMS st + β 4 RESMS st + β 5 NF st + β 6 WAGE st + β 7 LAGINT st + β 8 DELAY st + β 9 GPOP st + β 10 RATEREG st + ε st, (2) where π st = profit margin in state s and year t, measured as the ratio of premiums earned (adjusted for policyholder dividends) minus losses incurred to premiums earned; HERF st = Herfindahl index of concentration measured as the sum of squared market share of premiums written for state s and year t; MCR st = minimum capital requirements for multiple line property-liability insurers; 9 DWMS st = direct writers market share of premiums written; RESMS st = residual market share, measured as the percentage of total premiums written in the shared market; NF st = no fault, a binary variable equal to 1 if the state has no-fault laws, 0 otherwise; WAGE st = average wage for employees in the industry in state s in year t ; 10 LAGINT st = one-period lagged yield on intermediate-term government bonds; DELAY st = the extent to which the firm is able to delay paying losses, measured as the ratio of losses unpaid to losses incurred; GPOP st = state population growth rate; 11 RATEREG st = rate regulation, a binary variable equal to 1 if competitive, 0 if non-competitive; 12 ε st = error term for state s in year t, assuming the error components model, i.e. µ s + υ t + ω st. The independent variables include state-specific and market-specific factors. The Herfindahl concentration index (HERF), which can be derived from market 9 Information on minimum capital requirements by state was obtained from the National Association of Insurance Commissioners. 10 The WAGE date is the reported average wage by state for SIC classification 6331 (fire, marine, and casualty insurance) from County Business Patterns for the years in question. 11 This data is taken from the National Economic, Social, and Environmental Data Bank, U.S. Department of Commerce (August 1995). 12 A competitive regulatory environment includes states with file-and-use, use-and-file, and no regulation. Non-competitive regulation includes prior approval, modified prior approval, and file-anduse with bureau adherence. This data was obtained from the National Association of Insurance Commissioners.
5 Market Structure and Performance 507 structure theory (Crowling and Waterson 1976), is included as a test of the SCP paradigm. Intuitively, it is superior to the alternative three-firm, four-firm, or fivefirm concentration ratios since it encompasses information on all the firms in the market rather than just information on a few of the largest firms. Due to the possibility of cross-subsidization between lines, the physical damage and liability data are aggregated for the analysis. 13 Similarly, concentration ratios are calculated, assuming groups of affiliated companies act as a unit. Since barriers to entry are often associated with higher concentration levels, 14 the minimum capital requirement in each state (MCR) is included as a control. The use of the profit margin (premiums earned, adjusted for dividends, minus losses incurred, divided by premiums earned) as a measure of state profitability, although common in insurance studies, may create some distortions since it does not include any accounting for expenses or investment income. 15 Although expenses cannot be measured directly by state, the direct writer market share, the existence of no fault laws, the size of the residual market in each state, and the average state wages are included as controls for these factors. The direct writer s share of the market (DWMS) is expected to be positively related to profit. Direct writers market their policies using a salaried sales force and generally have lower expense ratios than firms using an independent agency system. 16 The size of the residual market (RESMS) is expected to have a positive effect on profitability. If cross-subsidization occurs between the voluntary and involuntary markets as suggested by Harrington (1990), higher rates in the voluntary market will be used to offset the unprofitable business in the residual markets. 17 If rates are not competitively determined, insurers would be expected to tighten underwriting standards, which would push more drivers to the involuntary market and would also result in higher profitability in the voluntary market. No fault laws have been empirically shown to reduce operating expenses, particularly where the law restricts pain and suffering damages (Cummins and Weiss 1991; Johnson et al. 1992). It is therefore expected that the existence of a no fault law in the state (NF) will be negatively related to pre-expense profit. Operating expenses are also expected to differ due to state-by-state cost differentials. Since the insurance industry is labor intensive, the average wage for insurance company workers in the state (WAGE) is included to control for differences in marginal costs across states. The sign on the WAGE variable is a 13 The equations were also estimated separately for physical damage and liability. The results of the aggregated lines confirm that some cross-subsidization is occurring. If firms consider the two lines as part of the same business, they may under-price in the more competitive line and make up the difference in the other, a common practice in other types of product markets. 14 The profit-concentration relationship is more likely to exist where industries exhibit significant barriers to entry. Collusive behavior and supernormal profits are not sustainable if other firms can enter and charge lower prices to capture the market. Barriers to entry are thus any costs borne by entrants but not by existing firms in the market, including economies of scale, and other cost advantages such as experience with vendors, capital requirements or product differentiation advantages. See Tirole (1990) and Scherer and Ross (1990) for a general discussion. 15 See Carroll (1993) for a discussion. 16 For further discussion, see Cummins and VanDerhei (1979) and Barrese and Nelson (1992). 17 Residual plans generally operate at a loss (Insurance Information Institute 1990).
6 508 The Journal of Risk and Insurance priori indeterminate. If insurers are able to adjust rates accordingly, it would be expected that pre-expense profit would be higher in high-wage states. But wages may also be an indication of overall cost factors in the state, which might imply that losses and expenses would both be higher, thus having an ambiguous impact on profit. It is also arguable that higher wages are an indication of greater employee productivity, which might translate into higher profits. As in the case of expenses, investment income is not considered in the profit measure, due to the lack of availability of this information by state. In periods of high interest rates, the present value of expected losses is lower, resulting in lower premiums. In policy years following high interest rate environments, it is therefore more likely that underwriting profits (premiums minus losses and expenses) will be lower. Underwriting profit should therefore be negatively related to the lagged interest rate (LAGINT). The ratio of unpaid losses to incurred losses (DELAY) provides a measure of the insurers ability to delay payment of claims. This may be due to intentional delay by insurers or the result of an inefficient state judicial process. In either case, the delay should have a positive impact on profitability due to the lower present value of the expected loss. Growth in population (GPOP) is included as a control for information asymmetry in personal lines since insurers in states that are experiencing large population growth have more new policyholders. If these movers are a worsethan-average or better-than-average risk for their age, the insurers will initially experience underwriting losses or gains, which will impact profit levels. 18 Given that better-than average risks will be inclined to reveal such information, it is expected that the information asymmetry will be more likely to result in lower profits for insurers. An alternative explanation for the predicted impact of growth in population is offered by Pauly et al. (1986) who suggest that the increased search behavior of movers will result in greater price competition which should be reflected in lower equilibrium prices. There are two main theories regarding the effect of rate regulation (RATEREG) on the property-liability industry. Stigler s (1971) capture theory suggests that a powerful industry lobby will encourage regulators to set rates at levels that favor the industry. This implies that the existence of strict rate regulation in a state will result in higher profits to the insurance industry. Peltzman s (1976) political support theory generalizes capture theory to allow for competing interest groups and predicts that regulators will take action to maximize political support. Thus, political pressure from strong consumer groups might lead regulators to decrease insurance prices, resulting in reduced profits or, if insurance lobbies are more powerful, regulation might favor the industry as predicted by capture theory. Equation (2) is estimated for the pooled cross-section and time-series data set (fifty states over the period ) using an error components model where the error term for state s in year t is of the form ε st = µ s + υ t + ω st and is assumed to be 18 Cummins and Weiss (1992) use information asymmetry as an explanation for the aging effect : new insureds generate higher losses than insureds with similar characteristics that have been with the company for several years.
7 Market Structure and Performance 509 an independent and identically distributed random variable with mean zero and constant variance. 19 Since it has been shown that concentration in a given market is endogenous (Schmalensee, 1987), the potential endogeneity bias is controlled for by using an instrumental variables approach to obtain consistent estimates of the regression coefficients. 20 EMPIRICAL RESULTS Summary statistics for the variables used in the analysis are provided in Table 1. Concentration in private passenger automobile insurance had a mean value of.1779 over this time period but ranged from.0484 for the least concentrated state-year observation, to.8594 in the most concentrated. 21 The empirical results reported in Table 2 confirm the existence of a strong positive relationship between profitability and concentration by state, for both physical damage and liability lines in private passenger automobile insurance. Several of the explanatory variables have the expected effect on profitability. The lagged interest rate is negative and significant for both lines, a result that provides evidence that firms are pricing based on present value of expected losses, which is lower in a high interest rate environment. For similar reasons, the delayed payment of incurred liability losses, whether intentional or due to the state legal environment, has a significantly positive impact on profitability in the liability line. The growth in population has a significantly negative effect on profits for the liability line, indicating that underwriting profits are lower in states where there are more newer policyholders, possibly due to information asymmetry and adverse selection. The states with more efficient systems of delivery as measured by direct writers share do not exhibit significantly greater profitability in private passenger lines although higher state wages have a negative impact as would be expected. The results of this study do not provide support for either of the hypotheses that link pricing behavior to the form of rate regulation. Rate regulation is not found to be a significant factor in insurer profitability. However, the existence of no fault increases pre-expense profit for the liability line and has a negative effect on combined line profitability. 19 In preliminary runs, the data set was checked for a possible heteroskedasticity problem. White s test was performed and the chi-squared statistic failed to reject the hypothesis of homoskedasticity at the one- percent significance level. The Durbin-Watson statistics for the individual states over the time period show that only four states exhibit significant first order autocorrelation, which was deemed insufficient to warrant the use of an alternative empirical technique. 20 Instruments used for the Herfindahl variable include one-year lagged profit margin, minimum capital requirements, wage levels for the state, state motor vehicle registrations as a percent of national registrations, percent of the state population in metropolitan areas, population density, and state per capita income. 21 One way of interpreting this statistic is that the inverse of the ratio provides an estimate of the number of equal sized firms that would produce an equivalent ratio. The mean Herfindahl of.05 thus implies that a state with approximately twenty equal sized insurers doing business would be equivalently concentrated.
8 510 The Journal of Risk and Insurance Table 1 Descriptive Statistics ; Private Passenger Automobile Combined Lines Mean Std. Dev. Minimum Maximum Profit Herfindahl Index Direct Writer Share Delay Residual Market Share Regulation Liability Mean Std. Dev. Minimum Maximum Profit Herfindahl Index Direct Writer Share Delay Residual Market Share Physical Damage Mean Std. Dev. Minimum Maximum Profit Herfindahl Index Direct Writer Share Delay Residual Market Share
9 Market Structure and Performance 511 Table 2 Profit and Concentration Model for Private Passenger Automobile Lines Parameter Liability Parameter Estimate (t-statistic) Physical Damage Parameter Estimate (t-statistic) Combined Lines Parameter Estimate (t-statistic) Intercept (2.7124)*** Herfindahl Index (3.4249)*** Direct Writer Share (1.1273) Rate Regulation ( ) Residual Market Share ( ) Min. Capital Req. ( ) Delay (6.5291)*** Lagged Interest Rate ( )*** Growth in Population ( )** No Fault ( )* Wages ( ) *** ( ) (6.1953)*** (5.6636)*** (2.8510)*** (4.5292)*** ( ) (0.2352) (0.0907) ( ) ( ) (0.2828) (0.9395) (0.2199) na (1.4721) ( ) *** ( ) *** (0.7135) ( ) na ( ) ** ( ) *** MSE * Significant at the 0.10 level. ** *** Significant at the 0.05 level. Significant at the 0.01 level. CONCLUSIONS Unlike Carroll s (1993) results for the workers compensation market, this study provides evidence that supports the structure-conduct-performance paradigm in the context of private passenger automobile insurance. Estimation of a profitability equation by state demonstrates that concentration has a significantly positive affect on profitability, even after controlling for regulatory and cost differences by state and over time. These results differ from previous studies using state level data from the previous decade (Ippolito 1979; Cummins and Harrington 1987) but confirm Chidambaran et al. s (1997) findings regarding the concentrationperformance relationship. The use of state-level data, which is particularly important in studies of markets in which competition is primarily local in nature,
10 512 The Journal of Risk and Insurance allows the inclusion of several important control variables not present in the latter study. Although the profit-concentration relationship has been found to exist in other industries (Schmalensee 1986), the insurance industry has thus far argued successfully against claims of collusive pricing based on the relatively large number of firms in the United States. However, the large number of insurance firms disguises the fact that there are a much smaller number of affiliated groups of firms in this industry. When the market is examined by state, the number of firms that do business in particular lines of insurance is relatively small, as evidenced by concentration ratios by state and by line. Even though the empirical findings of this study may call into question the competitiveness of the private passenger automobile insurance business, an alternative explanation is that higher profitability in the more concentrated states is the result of greater efficiency of large firms as suggested by Demsetz (1973). Further study of the efficiency issue is necessary before any conclusions can be made regarding the underlying cause of the profit-concentration relationship. It should be noted, however, that the measure of profitability used in this study is preexpense. If the insurers in more concentrated states have lower expense ratios due to their size, the pre-expense profit measure used in this study should actually be equal to, or lower than, profit in less concentrated states in the absence of other factors. Although this study does not provide conclusive proof of anti-competitive behavior on the part of automobile insurers, neither does it rule out the possibility. Whether the profit-concentration relationship found in this study is due to cartellike pricing behavior, product differentiation (in terms of claims settlement procedures), lower expenses, or consumer search behavior, closer attention to size and concentration levels in this industry is clearly warranted. REFERENCES A.M. Best Company Best s Aggregates and Averages. Oldwick, NJ: A.M. Best. Barrese, James, and Jack M. Nelson Independent and Exclusive Agency Insurers: A Reexamination of the Cost Differential. The Journal of Risk and Insurance 59(3): Carroll, Anne Structure and Performance of the Private Workers Compensation Market. The Journal of Risk and Insurance 60(2): Chidambaran, N. K., Thomas A. Pugel, and Anthony Saunders An Investigation of the Performance of the U.S. Property-Liability Insurance Industry. The Journal of Risk and Insurance 64(2): Crowling, Keith, and Michael Waterson Price, Cost Margins and Market Structure. Economica 43(1): Cummins, J. David, and Sharon Tennyson Controlling Automobile Insurance Costs. Journal of Economic Perspectives 6(2):
11 Market Structure and Performance 513 Cummins, J. David, and Jack A. VanDerhei A Note on the Relative Efficiency of Property-Liability Insurance Distribution Systems. Bell Journal of Economics 10(Autumn): Cummins, J. David, and Mary A. Weiss The Structure, Conduct and Regulation of the Property-Liability Insurance Industry. In The Financial Condition and Regulation of Insurance Companies, edited by A.W. Kopke and R.E. Randall. Boston: Federal Reserve Bank of Boston. Cummins, J. David, and Mary A. Weiss The Effects of No Fault on Automobile Insurance Loss Costs. Geneva Papers on Risk and Insurance 16(1): Demsetz, Harold Industry Structure, Market Rivalry and Public Policy. Journal of Law and Economics 16(1): 1-9. Eastman, Kevin L The Effect of Rate Regulation on Price Dispersion in the U.S. Automobile Insurance Market. Journal of Insurance Regulation 13(2): Goldberg, Lawrence G., and Anoop Rai The Structure-Performance Relationship for European Banking. Journal of Banking & Finance 20(4): Harrington, Scott E The Relationship Between Voluntary and Involuntary Market Rates and Rate Regulation in Automobile Insurance. The Journal of Risk and Insurance 57(1): Insurance Information Institute Insurance Facts. New York, NY: Insurance Information Institute. Johnson, Joseph E., George B. Flanigan, and Daniel T. Winkler Cost Implications of No Fault Automobile Insurance. The Journal of Risk and Insurance 59(1): Joskow, Paul L Cartels, Competition and Regulation in the Property- Liability Insurance Industry. Bell Journal of Economics 4(2): Joskow, Paul L., and Linda Mclaughlin McCarran-Ferguson Act Reform: More Competition or More Regulation? Journal of Risk and Uncertainty 4(4): Pauly, Mark, Howard Kunreuther and Paul Kleindorfer Regulation and Quality Competition in the U.S. Insurance Industry. In The Economics of Insurance, edited by J. Finsinger and M Pauly. London: MacMillan. Peltzman, S Toward a More General Theory of Regulation. Journal of Law and Economics 19(August): Scherer, F., and David Ross Industrial Market Structure and Economic Performance. 3d ed. Boston: Houghton Mifflin Company. Schmalensee, R Inter-industry Studies of Structure and Performance. In Handbook of Industrial Organization, edited by R. Schmalensee and R. Willg. Amsterdam: North Holland.
12 514 The Journal of Risk and Insurance Smirlock, M Evidence on the (non) Relationship Between Concentration and Profitability in Banking. Journal of Money, Credit, and Banking 17(1): Stigler, George A Theory of Oligopoly. Journal of Political Economy (Feb.): Tirole, Jean The Theory of Industrial Organization. Cambridge, Mass.: MIT Press. Weiss, Mary A Efficiency in the Property-Liability Insurance Industry. The Journal of Risk and Insurance 58(September): Winter, Ralph The Dynamics of Competitive Insurance Markets. Journal of Financial Intermediation 3(4):