American Economic Association Controlling Automobile Insurance Costs Author(s): J. David Cummins and Sharon Tennyson Source: The Journal of Economic Perspectives, Vol. 6, No. 2 (Spring, 1992), pp. 95-115 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/2138410 Accessed: 06/04/2010 16:07 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aea. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact email@example.com. American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Economic Perspectives. http://www.jstor.org Perspectives- Volume Number Journal of Economic 6, 2-Spring 1992-Pages 95-115 Controlling Automobile Insurance Costs J. David Cummins and Sharon Tennyson A utomobile insurance prices rose rapidly throughout the 1980s, even as the general rate of inflation slowed and automobile accident rates declined. The auto insurance component of the consumer price index more than doubled over the decade, increasing at a rate almost twice as fast as the overall CPI, and faster than even medical care inflation. The discrepancy was even greater from 1984-1989, when the auto insurance CPI grew at 9 percent per year and the all-items CPI at only 3.5 percent; by contrast, the medical component of the CPI grew only 6.5 percent annually over this time period. Some states have been hit harder than others, and double-digit infla- tion rates are not uncommon. In 1989 the average premium per car ranged from $353 in Iowa to $1074 in New Jersey (National Association of Insurance Commissioners, 1990). Premiums for high risk groups such as young males can be several hundred percent higher than the average. These trends have made auto insurance harder to afford for many consumers, particularly in urban areas, and the number of uninsured motorists has grown. As a result, auto insurance has become a potent political issue. It has figured prominently in gubernatorial races from New Jersey to California, and has stimulated many legislative and consumer initiatives aimed at reducing premiums. The populist view is that high insurance costs are primarily attributable to inefficient and excessively profitable insurance companies. Insurers are accused of paying claims without adequate controls, padding their expenses, and passing the costs along to the consumer, all while being shielded by a federal antitrust exemption. Insurers counter that insurance markets are competitive * J. David Cummins is Professor of Insurance and Risk Management and Sharon Tennysonis Assistant Professorof Insurance, bothat the WhartonSchool, Universityof Pennsylvania, Philadelphia, Pennsylvania. 96 Journal of EconomicPerspectives and efficient. They attribute rising costs to the excesses of the tort liability system, which is largely beyond their control. Insurers argue that profits are declining, in part because rate regulation has prevented prices from fully adjusting to recognize claim cost inflation. Informed discussion of the auto insurance issue has been impeded by the complexity of the issue and the difficulty of obtaining objective information. Publicly available statistics generally are highly aggregated and presented in a form accessible only to the specialist. As a result of these limitations, much of the auto insurance debate has been emotional and uninformed and has focused on the wrong issues. The primary objective of this paper is to remedy this situation. We begin by providing an overview of the auto insurance system and the structure of the auto insurance market. We then turn to an analysis of the factors underlying the auto insurance price increases experienced in recent years. The paper concludes with recommendations for bringing costs under control. The Automobile Insurance Market Automobile Insurance Compensation Systems Private passenger auto insurance is actually a package of insurance cover- ages, some mandatory or quasi-mandatory and others optional, depending on the rules adopted by a given state. Optional coverages in all states include collision insurance, which provides compensation for accident damage to the insured's automobile, and comprehensive insurance, which compensates the insured for non-collision vehicle damages such as fire, theft and vandalism. Liability insurance is at least quasi-compulsory in all states.1 As with other types of liability insurance, auto liability insurance protects the insured against being held liable for others' losses, rather than his or her own direct losses (hence it is often called third-party coverage, whereas insurance for one's own losses is first-party coverage). Potential liability can arise from either personal injuries or property damage sustained by others in an automobile accident with the insured. The scope of liability for auto accident damages is the key difference between states' automobile insurance systems. Compensation regimes fall into two principal categories: tort and no-fault. Tort is the traditional mechanism for compensating victims of auto acci- dents. Under the tort regime, a driver can only be held liable for other parties' damages if the driver is found to be negligent in the context of the accident. The necessity of establishing blame implies that accident victims often must sue ILiability insurance is compulsory in 18 states. Most other states have so-called "financial responsi- bility" laws, requiring drivers to show proof of the ability to pay claims following an accident. Proof can consist of a liability insurance policy, a bond, or sufficient liquid assets. J. David Cummins and Sharon Tennyson 97 to collect damages resulting from an accident. This tendency is exacerbated in accidents involving personal injury because lawsuits can be brought not only for economic personal injury losses (medical bills and lost wages) and property losses, but also for general damages (so-called "pain and suffering" losses). Lawsuits are less common in cases involving only property damage losses, because the existence and extent of property damage is easier to ascertain and recovery is limited to economic loss. The tort system evolved primarily as a mechanism for deterring negligent behavior. This perspective has received strong criticism in the context of auto accidents. As the boom in automobile usage after World War II increased the frequency and severity of accidents, several influential studies criticized the tort compensation system as slow and inefficient (Conard, 1964; Keeton and O'Connell, 1965). It was argued that most auto accidents result from circum- stances in which greater care would make little difference. Thus, the deterrent effects of tort could not be sufficient to justify the transactions costs of assessing fault through the civil justice system. These studies also argued that the tort system was inequitable; they found that large claims tended to be undercom- pensated, probably due to defendants' greater incentives to fight such suits. No-fault automobile insurance is an alternative to the tort system that is intended to compensate injuries more efficiently, primarily by reducing trans- actions costs. No-fault insurance has two distinguishing features. First, some restriction is placed on the ability to sue others for accident damages. To be eligible for suit, the personal injury losses from an accident must exceed a threshold, defined either in verbal terms or as a dollar amount of economic loss. Among states which define tort eligibility in monetary terms the threshold varies considerably, from as low as $400 of medical expenses in Connecticut to $4,000 in Minnesota. Verbal thresholds are generally much more effective limits on lawsuits. For example, under Michigan's verbal threshold, an accident victim can sue for general damages only if the victim has suffered death, serious impairment of bodily function, or permanent serious disfigurement (American Insurance Association, 1990). With rare exceptions, no-fault addresses only personal injury losses rather than property losses, which are handled relatively efficiently under existing insurance regimes. The second defining characteristic of no-fault is that compulsory first-party coverage for the insured's own personal injury losses is added to the auto insurance policy. This coverage often provides generous medical benefits, although it limits recovery to economic losses, eliminating all compensation for "pain and suffering." The theory is that the savings in transactions costs and general damage payments from the elimination of relatively minor liability suits can be reallocated to pay most or all of the loss payments under this additional first-party coverage. No-fault was endorsed by the U.S. Department of Transportation (1971) and first implemented in Massachusetts in 1971. At the present time, there are 12 no-fault states and 25 states that have retained the tort system. There are 98 Perspectives Journal of Economic also 10 states with so-called "add-on" personal injury coverage; these states place no restrictions on the right to sue, but do require drivers to carry coverage for their own personal injuries. Finally, several states (Kentucky, New Jersey and Pennsylvania) operate "elective" no-fault systems, allowing each individual insured to choose between tort and no-fault insurance. The Industry Structure Using conventional measures, the private passenger auto insurance market seems to be competitively structured. There are numerous firms and concen- tration is relatively low. Countrywide, there are about 500 firms writing auto- mobile insurance.2 The largest firm, State Farm, wrote 21 percent of auto insurance premiums in 1989; the top four firms (add Allstate, Farmer's Group, and Nationwide) wrote 43 percent, and the top 20 firms account for two-thirds of the market. The countrywide Herfindahl index for the market was a relatively low 690 in 1989.3 Concentration at the state level is somewhat higher. The median number of firms per state is about 100. The median Herfindahl index by state was 1050 in 1989, and the median four-firm concentration ratio was 55 percent. Thirteen states have Herfindahl indices above 1250. Insurers tend to divide states into territories for purposes of pricing and marketing, and concentration may be higher at the territorial level in some areas. Auto insurance is marketed through three major distribution channels: independent sales agents, exclusive sales agents, and mail or telemarketing. Firms using the latter two categories are often referred to jointly as "direct writers" or "exclusive dealers." Direct writers are dominant in the private passenger auto market, accounting for about two-thirds of premium volume. Direct writers have lower expenses per dollar of premiums than firms using independent agents, and generally earn a higher rate of return on equity. One commonly accepted view is that exclusive dealing therefore represents a superior technology (Joskow, 1973; Cummins and VanDerhei, 1979). An alter- native view is that the expense difference is attributable to differences in services such as matching of policies with customer needs and assistance with claims (Pauly, Kleindorfer, and Kunreuther, 1986). However, no convincing evidence of a service differential has yet been presented. If no service differen- tial exists, the survival of independent agency firms may be attributable to the slow diffusion of information in insurance markets (Berger, Kleindorfer and Kunreuther, 1989), or to market segmentation. 2 An often-heard fallacy is that the market consists of more than 3000 firms. This is based on a simple tabulation of the number of licensed companies. Most of these companies, however, are clustered together in insurance groups under common ownership and management. Many of the companies not affiliated in this way are very small and write only specialty coverages, like fire insurance on farm properties. After adjusting for specialization and common ownership, an accurate count of auto insurance companies is about 500. 3The Herfindahl index is computed by taking the market shares of the firms in the market, treating them as whole numbers and squaring them, and then summing the squares. Thus, the Herfindahl measure of pure monopoly is 1002, or 10,000. Controlling InsuranceCosts 99 Automobile The McCarran-Ferguson Act of 1945 provides insurers with an exemption from federal antitrust laws. The principal rationale for the exemption is to allow insurers to pool loss data and thus to reduce sampling error. Data are pooled through rating bureaus and statistical organizations, such as the Insur- ance Services Office (ISO) and the National Association of Independent Insur- ers (NAII), and disseminated to member firms. A further provision of the McCarran-Ferguson Act is that insurers are regulated by the states rather than the federal government. About half of the states currently regulate automobile insurance rates. The typical system re- quires insurers to obtain approval from the state insurance commission prior to changing rates. Insurers must file rates and their statistical justification with the regulator. The criterion for rate approval is that rates should not be "excessive, inadequate, or unfairly discriminatory." In most states, the criterion for rate approval is a fair return on equity capital. However, this standard is often implemented using crude book value methods, and decisions have become increasingly politicized. The traditional focus of rate regulation was on solvency, preventing rates from being too low. Since the early 1970s, however, the emphasis has shifted towards controlling price inflation. Recent studies have found that the average markup of premiums over losses, a measure of the "price" of one dollar of insurance coverage, is lower in states which regulate insurance rates than in unregulated states (Harrington, 1987; Grabowski, Viscusi, and Evans, 1989). Explaining Automobile Insurance Price Increases Insurer Behavior as a Cause of Price Inflation Much of the controversy about the price of auto insurance focuses on the behavior of insurance companies. Many blame insurer collusion for the price inflation of the late 1980s. Others argue that insurers raised prices to current policy-holders to make up for losses experienced in the early to mid-1980s, periods of low profitability. Did insurers recover lost profits by raising insur- ance rates in the late 1980s? This section examines these allegations in the light of recent profit trends. Although entry on a small scale and entry into independent agency writing appear to be relatively easy, entry into the exclusive dealer segment of the market may be more costly. For example, Prudential Property and Casualty, which entered the market with a sales force comparable in size to that of any existing exclusive dealer, was highly unprofitable during its first few years of operation. The large numbers of high cost, fringe firms in the industry also suggest that exclusive dealers are protected by entry barriers, and that exclu- sive dealer prices may be above the competitive level. Moreover, insurance distribution channels are characterized by a type of resale-price maintenance. Under "anti-rebate" laws in effect in nearly all states, 100 Journal of Economic Perspectives agents are prohibited from discounting policies by "rebating" part of the sales commission back to the consumer. These statutes prevent the formation of wholesalers (super-agencies) and protect both agents and insurers from more vigorous price competition. Anti-rebate laws were enacted ostensibly to prevent "unfair" discrimination-that is, rate differentials not based on cost differen- tials. However, protection of commissions seems to have been the most impor- tant underlying rationale (Williams, 1969). Of course, anti-rebate laws may be beneficial in that important agent services are protected from cost-cutting (Katz, 1989); however, there is little evidence that the laws have actually had this effect. It has also been argued that the extensive information sharing allowed by the McCarran-Ferguson Act facilitates tacit price collusion. Traditionally, insur- ance rating bureaus like the ISO recommended rates and loading factors to its members and filed rates on their behalf with state insurance commissioners. Perhaps in response to growing sentiment against the industry's antitrust exemption, the ISO voluntarily ended these practices in 1989; it continues to pool data and recommend trend factors. Even in states with competitive rating, however, each firm's rates must be filed with the state insurance commission (there are a few exceptions, like Illinois). Rate filings are public information, and other insurers are the largest users of this information. In at least some states, it is common practice for small insurers to utilize the market leader's most recent rate filing in developing their own filing. Of course, parallelism in price changes is not necessarily evidence of collusion. Prices will tend to move together under perfect competition, and use of common loss data in rate development can lead to more efficient pricing. Given the number of firms in the industry and the relative ease of entry (at least into agency writing), it would seem that the potential for tacit collusion and price leadership in insurance is not great. Even in the absence of collusion, it has been argued that consumer search costs could raise prices above competitive levels. The importance of search costs in insurance markets has been empirically documented (Dahlby and West, 1986). Profitability patterns by insurance policy cohort also suggest that insur- ers gain information about the risk characteristics of policy-holders over time (D'Arcy and Doherty, 1990); this information asymmetry could effectively lock consumers into purchasing from a particular insurer. The evidence on consumer switching costs, coupled with the market dominance of the exclusive dealers, the high costs of entry into this segment of the market, and the presence of price leadership argue for some caution in drawing conclusions about the competitiveness of auto insurance markets. It is nonetheless difficult to argue that recent problems in the market are at- tributable to non-competitive practices of insurers. No significant changes in market structure or institutions have occurred, and regulatory emphasis and consumer pressure have been increasingly aimed at restricting price increases. It therefore seems unlikely that collusion is the cause of price inflation. J. David Cummins SharonTennyson 101 and Might the higher rates stem from insurer overpricing to recover past losses? On the surface, the possibility seems easy to dismiss. Insurance rate- making is prospective, so prices in a competitive market will reflect the present value of expected losses and expenses during the coverage period. Companies attempting to load prior losses into rates for future periods would lose market share to competitors and new entrants that did not use retroactive loadings. Even in the absence of competition, it is unclear that such a pricing strategy would be optimal, since past losses represent sunk costs. Despite this argument, certain cyclical features of insurance markets may result in a phenomenon which resembles retroactive loss loading. The existence of profit "cycles" in insurance markets, discussed by Ralph Winter in the Summer 1991 issue of this journal, has been extensively documented. One widely accepted view is that the cycle arises because insurance supply is limited by the stock of equity capital insurers hold, and capital is costly to obtain quickly due to information asymmetries in capital markets (Winter, 1991; Cummins and Danzon, 1991). According to Winter's theory, periods of low profitability will deplete equity capital and hence insurance capacity. Capital will not be immediately replaced due to the cost advantage of retained earnings, and hence existing capacity will earn higher than normal returns in the short run. This implies that relatively sudden, sizable premium increases are possible even in competi- tive insurance markets after periods of low profits. Insurance industry profits hit record lows in the mid-1980s, after falling steadily for several years. These "cyclical" effects may therefore have contributed to insurance premium in- creases in the late 1980s; even so, this would have resulted from competitive market forces rather than an insurer conspiracy. Both the argument for collusion and the argument for retroactive loss loading presume that industry profit levels were especially high during the late 1980s. However, the evidence on auto insurance profitability provides no indication of such excessive profits. Constructing an accurate measure of the insurance industry's profit mar- gin is difficult. The combinedratio is widely reported as a measure of profitabil- ity for a line of insurance. This ratio is the sum of what is known as the expense ratio (expenses/premiums) and the loss ratio (losses/premiums). The com- bined ratio is essentially the ratio of total underwriting costs to total premium revenues, with a value of 1 indicating that costs equal revenues. Hence, 1 minus the combined ratio is called the underwritingprofit ratio, a figure often used as a profit indicator, somewhat analogous to the return on sales. The underwriting profit ratio is usually negative, especially in liability insurance, and insurers sometimes use this fact to argue that underwriting operations are not profitable.4 This argument is fallacious because the 4Influential publications such as The New York Times have sometimes mistakenly drawn this conclusion. See, for example, Jill A. Fraser, "The Travelers Rides into the Storm," The New York TimesMagazine, December 2, 1990, p. 46. 102 Perspectives Journal of Economic underwriting profit measure ignores investment income. Keep in mind that automobile insurance is a discounted cash flow product: the company collects the premium at policy inception (or finances it during the policy period) and earns interest on unexpended premium balances until losses are paid. This would generally produce a negative underwriting profit, unless the profit rate is relatively high or interest rates are relatively low. An alternative profit measure developed by industry regulators is the overall operating ratio (OOR), which is a proxy for the ratio of the present value of losses and expenses to premiums. This ratio is equal to the combined ratio minus the ratio of investment income to premiums.5 Investment income is allocated by line on the basis of reserves, so that lines where losses are expected in the more distant future receive larger allocations. Both the combined ratio and the overall operating ratio have shown auto liability insurance to be unprofitable and auto physical damage insurance to be profitable during the 1980s. In auto liability, the combined ratio averaged 1.15 and the OOR 1.06 during the period 1980-1989. The corresponding averages for auto physical damage were .96 and .94, respectively. However, the OOR is far from an accurate measure of economic profits. It tends to be biased when the average yield on the insurer's existing investment portfolio differs from yield rates currently available in investment markets. Additional bias is present if the past growth rate in insurer liabilities exceeds current investment yields. Because it gives a return on sales rather than a return on equity, it is difficult to determine whether a given OOR is appropri- ate, too high, or too low. To correct at least some of the problems with the OOR, we use an alternative measure of profit that has recently been suggested in the literature (Winter, 1991). We call this measure the economicprofit margin. The economic profit is computed as the difference between premiums and the present value of losses, expenses, and policy-holder dividends.6 The economic profit divided by premiums is the economic profit margin. The economic profit margin is an improvement on the OOR but still does not relate returns to equity or adjust for federal taxation. Nevertheless, it should provide a reliable indication of profit trends.7 5Until 1990, the overall operating ratio recognized only investment income on policy-holder funds: that is, it did not include investment income on equity. The National Association of Insurance Commissioners (NAIC) changed the formula in 1990 to include the latter component of investment income. The present formula still does not recognize unrealized capital gains. 6The loss payout pattern was estimated from Schedules 0 and P of the industry-wide annual statement data presented in A.M. Best Company (1989, 1990). Losses were discounted using the U.S. Treasury yield curves presented in Coleman, Fisher, and Ibbotson (1989), for 1980-1988. For 1989, the intermediate term U.S. government bond yield was used for auto liability insurance and the 90-day treasury bill rate for physical damage insurance. Expenses and policy-holder dividends were not discounted. 7Ultimately, an internal rate of return calculation should be used to estimate insurance profits (see Cummins, 1990). However, a number of methodological issues need to be resolved before that method can be used with confidence for aggregate data. InsuranceCosts 103 Automobile Controlling Figure 1 Auto Liability Insurance Profits 0.1' 1-OOR E -\- ECON PROFIT MARGIN 0.05 - 0 0: 0 -0.05 80 81 82 83 84 85 86 87 88 89 The economic profit margin and the conventional underwriting profit measure for auto liability insurance, 1-OOR, are shown in Figure 1. The economic profit margin is higher than the OOR for all years shown in the figure, but both profit measures show that auto liability profits were negative for most of the decade. The economic profit margin implies that the largest decline occurred in 1983. Figure 2 presents the economic profit margins for auto liability, auto physical damage, and the weighted average of both lines combined. The figure shows that physical damage insurance was moderately profitable during the period 1980-85 and highly profitable from 1986 through 1989; liability insur- ance was increasingly unprofitable after 1983. Cross-line subsidization may be present, especially since auto liability and physical damage coverage are usually written as a package. The profits in physical damage coverage partially offset the losses in liability insurance so that overall auto insurance profits were Figure 2 Private Passenger Auto Profits 0 0 PHYSICAL DAMAGE ? -0.05 -4LIABILITY WEI(GHTED AVERA(GE -0.1 9 I I I I I I 1 1980 198 1 1982 1983 1984 1985 1986 1987 1988 1989 104 Perspectives Journal of Economic moderately positive from 1986-1988. Nevertheless, the weighted average profit generally trended downward during the 1980s. Cost Increases as a Cause of Price Inflation Auto insurance price increases of the late 1980s do not appear to originate in excessive insurer profits. Nor does it seem likely that cyclical effects alone caused the sustained price inflation observed. In this section we examine the growth in the major cost components of the auto insurance premium. Our findings point to cost inflation as the primary cause of premium inflation. Auto insurance policies typically provide coverage for a period of six months or one year. The premium is fixed at the policy issue date and cannot be increased during the policy period. The primary cost component of the insurance premium is expected loss payments under the policy. This portion of the price is known as the pure premium and is determined by two components: the expected number of claims per policy (claim frequency), and the expected cost of each claim (claim severity). Determining losses involves finding the expected value of the losses at current price levels, and then making adjustments for expected inflation. The proportion of losses expected to be paid in each time period-the "payout tail" -must also be estimated. In auto insurance, property losses are settled rela- tively quickly, usually within two years after the beginning of the policy period; however, bodily injury liability and personal injury protection (no-fault) loss settlements can cover substantial periods of time. This payout tail yields the company interest earnings on unexpended premium balances until losses are paid. Because premiums are collected in advance of loss payments, the insurer provides a discount to reflect projected interest earnings on premium balances. In a competitive market, losses will be discounted at a risk-adjusted discount rate, which consists of the rate of return on the company's asset portfolio less a risk-premium to compensate the firm's owners for bearing insurance risks (like the risk that losses will be greater than expected). The risk premium also compensates the insurer for the equity commitment, which ensures that losses will be paid even if they are larger than anticipated. Insurance premiums must also cover underwriting expenses, which in- clude marketing and administrative expenses as well as the state premium tax. The insurer is also subject to federal income taxes on its underwriting income (the difference between premiums and the sum of losses and expenses) and on investment income. In a competitive market, the price of insurance is thus determined as the present value of losses and expenses plus the present value of federal income taxes attributable to the insurance transaction (Myers and Cohn, 1987). To examine auto insurance costs within this pricing framework, we begin with the pure premium-expected losses per policy-the foundation of the insurance premium. In both tort and no-fault states, about one-third of the and J. David Cummins SharonTennyson 105 Figure 3 Claim Cost Inflation: 1984-89-Tort States 0.12 0.1I .0.08- 0.06- 0.04- 0.02- -0.02 - 0.04 BI LIABILITIY PD LIABILITFY COLLISION COMPREHENSIVE = PURE PREMIUM m FREQUENCY M SEVERIT1-Y Source:National Associationiof Independenit Insur-er-s pure premium goes for personal injury losses (medical bills, lost wages, and general damages). Another one-third of the total is attributable to collision losses, and the balance is split between comprehensive and property damage liability payments.8 As an approximation for trends in expected losses we examine trends in claim costs; that is, actual losses per insured vehicle. We focus on the period 1984-1989, when the inflation rate in auto insurance was highest. Loss infla- tion accounts for a large part of the growth in insurance premiums over this period: in tort states, losses per insured car grew at an annual rate of 7.2 percent, while in no-fault states losses grew 6.9 percent per year during this period. Loss inflation by insurance coverage for tort and no-fault states is shown in Figures 3 and 4, respectively. In interpreting the graphs, it is helpful to keep in mind that average losses are determined as the product of average claims frequency and severity. Also, changes in claims frequency reflect both the number of accidents and any changes in the proportion of accidents that lead to a claim being filed. Figure 3 shows that bodily injury liability was the primary source of loss inflation in tort states, growing at a rate of more than 11 percent per year over this period. Most of the loss growth in bodily injury liability was attributable to greater claim severity, although frequency also grew slightly. This growth in frequency is probably due to an increase in the propensity to file claims, since claims frequency for collision and property damage liability declined. The 8Countrywide, theft losses comprise about 30 percent of comprehensive losses or about 5 percent of the overall pure premium. 106 Journal of Economic Perspectives Figure 4 Claim Cost Inflation: 1984-89-No Fault States o. 12 i 0.1 0.08- 0.06- 0.04- 0.02- 0 -0.02- -0.04 BI LIABILITY COLLISION PIP PD LIABILITY COMPREHENSIVE z PURE PREMIUM E FREQUENCY M SEVERIT1-Y Source: National Association of Independent Insui-ei-s decreases in these latter two claims rates are more in line with trends in accident rates, which declined nearly 10 percent between 1984 and 1988.9 The property damage coverages also experienced loss inflation in tort states, due to increases in claims severity. This probably reflects the higher costs of repairing the more technologically advanced automobiles that have been introduced in recent years. Figure 4 demonstrates that bodily injury liability also experienced the highest rate of loss inflation in no-fault states. However, this cost component grew more slowly than in tort states (between 8 and 9 percent per year, compared with 11 percent in tort states). Contrary to the experience in tort states, the frequency of bodily injury claims declined in no-fault states during 1984-1989 period. The frequency of property damage liability, collision, and personal injury protection claims also declined in no-fault states, as expected given the down- ward trend in accident rates. Interestingly, the frequency of comprehensive claims increased in no-fault states. This is attributable primarily to higher auto theft rates in no-fault states, and reflects the fact that highly urbanized states are more heavily represented among the states with no-fault auto insurance laws. The higher theft rates are thus associated with the economic and demo- graphic characteristics of no-fault states and not with no-fault itself. Other factors external to insurance markets also furthered the growth in price inflation over the decade. Declining interest rates in the mid-1980s are one contributor. Other things held constant, a decline in interest rates will raise insurance premiums, since it raises the present discounted value of future 9In 1984, fatal accidents occurred at a rate of 2.31 per 100 million miles driven, and accidents leading to personal injury occurred at a rate of 125.24 per 100 million miles. The corresponding rates for 1988 were 2.08 and 113.72, decreases of 10 percent and 9.2 percent, respectively (Federal Highway Administration, 1988). InsuranceCosts 107 Automobile Controlling claims. In other words, when it is harder to earn investment income, insurance premiums will be higher. At mid-year 1984, the intermediate term government bond yield was about 13 percent; one year later, the yield had fallen to 9.8 percent; and two years later it was 7.4 percent. This decline in interest rates coincides with the sharp rise in auto insurance prices starting in 1984. Changes in effective corporate tax rates also account for part of the recent inflation in auto insurance prices. Prior to the Tax Reform Act of 1986, insurers used underwriting losses and investment income on tax-exempt bonds to avoid paying federal income taxes. The imposition of the alternative minimum tax (AMT) under the Tax Reform Act as well as provisions specifically directed at property-casualty insurers increased the effective federal tax rate to about 20 percent. We estimate that the higher tax rates accounted for an increase of about 3 percent in auto insurance prices, phased in over three years. Contrary to claims that inefficient insurance production caused premium inflation in the 1980s, growth in underwriting expenses did not contribute to auto insurance inflation during the period 1984-1989. Expenses as a percent of premiums actually declined during this period, from just over 24 percent to under 23 percent for auto insurance coverages. This decline indicates that auto insurance losses grew faster than expenses during the period.'0 It is clear that a number of factors have combined to push up auto insurance prices during the 1980s: high rates of pure premium inflation, especially increased severity of bodily injury liability insurance; the sharp decline in interest rates; higher federal income taxes imposed on insurers by the Tax Reform Act of 1986; and possibly cyclical factors. Some of these causes of insurance inflation-federal taxes, interest rates and profit cycles-are exogenous to the insurance sector. But others, like claims frequency and severity increases, are at least partly endogenous. Auto insurance reform must address these factors if inflation in premiums is to be reduced. Policies to Reduce Premium Inflation Controlling Liability Costs Through No-Fault Insurance The legal environment plays a critical role in driving up auto insurance costs. It is simply too easy and too profitable to file bodily injury liability claims on auto insurance. In tort states and low-threshold no-fault states, motorists experiencing minor accidents effectively receive a lottery ticket. The ticket gives them a high probability of winning some amount of money (like $5,000) and a low probability of no gain. The lottery winnings are the motorist's share of a general damage (pain and suffering) award. The price of the ticket is either zero or a small positive amount. 10Of course, even though the expense ratio did not increase, it could still be "too high" if insurers engage in non-price competition. Marketing costs alone accounted for 15.6 percent of premiums in 1989: that is, for 68 percent of the expenses included in the expense ratio. 108 Journal of Economic Perspectives To explore this lottery in more detail, consider the following sce- nario: Driver A is involved in a minor accident where it is reasonably clear that the other driver is at fault. Driver A sustains no injuries. However, he enters the lottery anyway by visiting a lawyer. The lawyer sends driver A to a cooperating physician, who conducts a battery of tests and administers some treatment. The process of compiling documentation to inflate a claim is known as the build-up. The lawyer then files a claim against the driver who hit driver A. The cost of the lottery ticket to driver A is the physician's fee plus the cost of the time required to visit the lawyer and the doctor. The lawyer charges no fee up front but is entitled to a share of the lottery winnings (perhaps a contin- gency fee of 33 percent). The insurer of the other driver now faces a dilemma. It suspects that driver A was not injured, but has received detailed documentation including a diagnosis and statement of treatment by a licensed physician. Furthermore, the claim is relatively small. The insurer calculates the expected costs of defending against the suit and finds that the costs exceed the expected gain (the amount of the suit times the probability of a successful defense). It settles out of court, and driver A has won the lottery. Several elements must be present to operate a successful personal injury lottery. First, the game must be played frequently to permit the lawyer to take advantage of the law of large numbers, because not all suits are successful. Second, there must be many physicians to choose from because of the low proportion of physicians willing to provide treatment for non-existent injuries. Third, driver A must be aware of the lottery and have easy access to a lottery attorney. Fourth, driver A must be able to play for a relatively low cost, which implies that the cost of the driver's time and the physician's up-front fee must be low. Fifth, the utility value to driver A of winning a few thousand dollars must be high. Finally, the court system must be sufficiently sympathetic to bodily injury claims so that the insurer's expected payoff from resisting the claim is lower than the cost of taking the case to court. These elements suggest that insurance lotteries are most likely to develop in areas where accident frequency is high, physicians and lawyers are easily accessible, information about playing the lottery can be widely disseminated, the utility value of winnings is high, and the cost of drivers' time is low. In short, insurance lotteries are most likely to develop in densely populated areas with economic problems and court systems that favor the plaintiff. Statistical evidence consistent with the insurance lottery hypothesis is provided by calculating the likelihood of filing a bodily injury claim in relation to the number of accidents. The frequency of property damage liability (PDL) claims is the best available estimate of the relative number of accidents in different geographical areas. While not every accident leads to a PDL claim, one would not expect the rate of PDL claims per accident to vary significantly by location. PDL claims are much less susceptible to fraud because it is relatively easy to determine whether a vehicle has been damaged and effective and J. David Cummins SharonTennyson 109 Table I Ratios of BIL to PDL Claims: 20 Largest Cities: 1985-1987 City StateAverage City/State Avg New York, NY 0.202 0.114 1.772 Los Angeles, CA 0.606 0.346 1.751 Chicago, IL 0.521 0.344 1.515 Houston, TX 0.290 0.240 1.208 Philadelphia, PA 0.750 0.209 3.589 San Diego, CA 0.314 0.346 0.908 Dallas, TX 0.208 0.240 0.867 SanAntonio, TX 0.288 0.240 1.200 Phoenix, AZ 0.411 0.385 1.068 Baltimore, MD 0.577 0.331 1.743 San Jose, CA 0.271 0.346 0.783 San Francisco, CA 0.374 0.346 1.081 Indianapolis, IN 0.248 0.218 1.138 Memphis, TN 0.320 0.237 1.350 Jacksonville, FL 0.110 0.171 0.643 Washington, DC 0.201 N.A. N.A. Milwaukee, WI 0.416 0.289 1.439 Boston, MA 0.213 0.185 1.151 Columbus, OH 0.240 0.247 0.972 New Orleans, LA 0.585 0.430 1.360 AVERAGES* 0.357 0.263 1.277 Source: Insurance Research Council (1990). * Unweighted. BIL = bodily injury liability; PDL = property damage liability. State average is a weighted average including the city. N.A. = not applicable. Detroit is not listed because comparable data is not available in Michigan. Insurers report territorial claims data to statistical agents such as the Insurance Services Office or the National Association of Independent Insurers. Because the agents use different definitions of territories within states, pooling of data from different agents was not feasible. The numbers reported in this table are from the statistical agent representing the largest proportion of the market in each state. This represents at least a majority market share in all states except California and Illinois, states in which agents are not required to report data to a statistical agent. mechanisms such as arbitration are in place to assign negligence without the necessity for a lawsuit. Thus, the first column of Table 1 presents the ratio of bodily injury liability (BIL) to property damage liability (PDL) claims for the nation's 20 largest cities. The second column presents the state average ratio for comparison, and the third column shows the ratio of the city ratio to the state average. In nearly every instance, the BIL/PDL claim ratio is higher in the large city than in the state as a whole. For example, in Los Angeles there were 60.6 BIL claims for every 100 PDL claims, whereas the statewide average for California was only 34.6. Chicago registered 52.1 BIL claims per 100 PDL claims, while the total for Illinois was only 34.4. In Philadelphia, there were 75 BIL claims per 100 110 Journal of Economic Perspectives PDL claims; while the statewide average for Pennsylvania was 20.9. The discrepancies were smaller for some of the other cities; however, it seems clear that insurance lotteries are in operation in many urban areas. This evidence is particularly convincing in view of the fact that accidents occurring in urban areas typically are less severe than those in less densely populated areas because they occur at lower speeds (Insurance Research Council, 1990). A second source of evidence on the insurance lottery is provided by a detailed closed-claim study conducted by the Automobile Insurers Rating Bureau of Massachusetts (Weisberg and Derrig, 1991). The Bureau investi- gated 597 bodily injury claims settled by member companies and estimated that over 30 percent of the claims involved explicit fraud or apparent overstatement of damages. The most successful insurance lottery uncovered was in Lawrence, a city long beset by high unemployment and crime rates. The BIL/PDL ratio in Lawrence is .472, while the average ratio in Massachusetts is only .185. A high proportion of the claims involved the same doctors and lawyers. Partly as a result of this study, Massachusetts recently instituted a joint industry-government effort to resist claims fraud. The legislation authorizes the creation of the Insurance Fraud Bureau of Massachusetts, a private organiza- tion funded by the insurance industry. Insurers are required to report sus- pected fraudulent activities to the Bureau, which has statutory power to investigate suspicious claims. If sufficient evidence exists to support prosecu- tion, cases are referred to the Massachusetts Attorney General. It is anticipated that by providing statutory authority to insurance investigators and a direct link with the Attorney General's office, the legislation will improve the cost- effectiveness of resisting fraud. Besides more vigorous enforcement, as in Massachusetts, probably the best way to stop insurance lotteries is to adopt no-fault insurance, since no-fault eliminates the right to sue for relatively minor injuries. This removes from the system the smaller claims, for which insurer resistance is not cost-effective. Insurers have a stronger economic incentive to resist larger claims, so a smaller proportion of questionable claims are likely to receive compensation under no-fault. It is also more difficult to develop buildups for large amounts if injuries are slight or nonexistent. In tort and add-on states, the number of bodily injury liability claims per 100 property damage liability claims grew from .235 in 1980 to .305 in 1989. Ninety percent of this increase took place from 1984 to 1989. The highest ratio of BIL to PDL claims in 1989 was in California (.557)."1 The growth in the relative number of BIL claims was considerably lower in no-fault states with verbal tort thresholds (where an injury must satisfy a verbal rather than a I IThis number is the statewide average over all insurers participating in the Fast Track Monitoring System. This system produces quarterly reports on the aggregate auto insurance claims experience of participating insurers. In most states this data source represents at least 60 percent of the market. Note that these data are not directly comparable to those reported in Table 1. InsuranceCosts 111 Automobile Controlling monetary definition of severity in order to qualify for tort) or dollar thresholds of an economically meaningful amount. In these states, the BIL/PDL claims ratio grew from .079 to .111 over the 1980-1989 period, an increase of only 3 percentage points. Thus, no-fault has potential for curbing inflation in bodily injury liability insurance premiums. In contrast, in states with a low-dollar tort threshold the ratio of BIL to PDL claims grew from .117 to .227 in the 1980s. This corroborates the findings of previous research (Cummins and Weiss, 1991a) that no-fault needs a strongly-worded dollar threshold, such as those presently in effect in Michigan and New York. For example, the Florida verbal threshold permits the insured to sue for permanent injuries, but does not specify that the injury be serious, as the Michigan and New York thresholds do. The BIL/PDL ratio in Florida grew from .095 in 1980 to .154 in 1989, versus remaining constant at .115 over this time period in New York.12 Additional problems must be overcome for no-fault to be a satisfactory solution to the problem of personal injury costs. Personal injury protection (PIP) coverage under no-fault is also susceptible to moral hazard, and PIP claims severity has been a major source of premium inflation in most no-fault states. PIP usually provides first dollar coverage, with no copayments or deductible, and very generous (sometimes unlimited) medical benefits. More attention should be paid to the incentive structure of the first-party medical coverage provided under no-fault. It tends to be more politically feasible to allow drivers to elect to use no-fault, rather than requiring them to do so. However, these laws are most effective if no-fault is the default option, rather than tort. For example, in New Jersey, where no-fault is the default, 83 percent of drivers are covered under no-fault; in Pennsylvania, where tort is the default, only about 24 percent of drivers have opted for no-fault. Since the laws are similar and demographics of the two states are not markedly different, this is likely due to status quo bias in decision-making (Samuelson and Zeckhauser, 1988). In this instance the re- sponse is consistent with rational behavior, however; Pennsylvanians faced significant costs to choosing no-fault, being required to read and sign eight separate documents in order to exercise all of their options under the insur- ance reform law. Of course, the adoption of no-fault laws involves some tradeoffs: for example, no-fault might weaken the deterrent effects of tort and lead to higher accident rates. But since all existing U.S. no-fault laws permit lawsuits for wrongful death and for serious injuries, any adverse effects of no-fault on incentives are likely to be small. Most recent research on the effects of no-fault on accident rates-for example, Zador and Lund (1986) and Cummins and Weiss (1991b)-supports this conclusion. 12Property damage liability data are not available in Michigan, because Michigan also has no-fault property damage coverage. 112 Journal of Economic Perspectives Controlling Prices Through Rate Regulation Rate regulation is often viewed as a promising solution to the auto insur- ance cost problem. Several states, notably California, have recently restored rate regulation and/or legislated mandatory insurance rate rollbacks. While some advocates of regulation take an anti-collusion stance, the more common justification for these policies is the belief that firms do not have the correct incentives to fight cost increases. The rationale is that by holding prices down, rate regulation will force firms to find ways to trim costs. As noted previously, insurers may not always choose to police fraudulent or escalated claims aggres- sively. If the cost of researching a claim's validity exceeds the expected payoff to the firm, then it will not pursue the investigation. If this behavior is efficient, there is no benefit to consumers from forcing firms to undertake more investigations. However, there are likely to be positive externalities from such investigations; the benefits to society include not only the potential reduction in the current claim payment, but the deterrent effect on future fraudulent claims to all insurers. In this case the individual firm, which will consider only current and future reductions in its own claims, will not have the socially correct incentives to resist claims. Rate regulation is not an appropriate device to address this problem. There is no reason to believe that restricting price increases will induce firms to resist claims inflation, because it doesn't affect the relative marginal costs and benefits from doing so. More likely, holding prices down will result in other distortions to the market. For example, there is some evidence that rate regulation results in decreased service expenditures in regulated states (Pauly, Kleindorfer and Kunreuther, 1986). Furthermore, the market share of exclu- sive dealers is lower in regulated states, leaving a larger proportion of risks to be underwritten by relatively inefficient firms. This is sometimes attributed to regulatory barriers to price competition (Joskow, 1973), but could also result from profit restrictions inducing the efficient firms to limit output in regulated states and concentrate in other, unregulated, markets. This latter point of view is reinforced by the frequent threats of insurers to withdraw from markets in response to restrictive regulation. But while withdrawal could be a rational response to the inability to earn a fair profit under the disputed law or regulation, it is often alleged that such threats represent a conspiracy to intimidate regulators, and thus to force the approval of higher rates or to impede auto insurance reform. Although little evidence exists that would shed any light on this question, there have been instances of insurers unilaterally withdrawing from states, which clearly does not make sense as part of an intimidation strategy since it leaves the fruits of the withdrawal to others. Also, regulators have some tools of intimidation, them- selves. Some states have forced insurers to continue to write automobile insurance as a condition of retaining their licenses to write other lines of business. Kemper actually paid "ransom" of $175 million to be permitted to J. David Cummins and Sharon Tennyson 113 withdraw from the automobile insurance market in Massachusetts, while retain- ing its licenses to write other lines.'3 The potential for market distortions under regulation is underscored by the experience in "residual" auto insurance markets. All states have residual market mechanisms to provide insurance for motorists who cannot obtain coverage in the regular market. The most common mechanism is the assigned risk plan, where drivers who cannot obtain coverage in the voluntary market are "assigned"' to insurers doing business in the state in proportion to the market shares of those insurers. This "involuntary" market tends to be larger in regulated than in unregu- lated states (Grabowski, Viscusi and Evans, 1989). States with extremely restric- tive regulation, like Massachusetts and New Jersey, have experienced the virtual collapse of the voluntary market. In both states, more than half of all drivers are insured through the involuntary market. In Massachusetts, the problem may have been exacerbated by the prohibition against the use of some types of information in classifying risks. Preventing insurers from using risk classification information is equivalent to requiring the use of pooled rates for drivers with different expected costs. This could lead to adverse selection and market failure even if rates were adequate on average, as lower risk drivers reduce insurance purchases (e.g., Rothschild and Stiglitz, 1976). Rate regulation might make more sense if auto insurance price increases had been caused by collusion or other anti-competitive practices of insurers. But as a method of holding down the growth in insurance costs, regulating prices is a very imperfect mechanism. A policy which affects insurers' claims settlement practices directly would be more advisable. This could be accom- plished either through mandatory measures such as the Massachusetts insur- ance fraud bureau, or by changing the relative costs and benefits of contesting claims, as through some form of no-fault insurance. Conclusion The auto insurance inflation of the 1980s was caused primarily by increases in cost factors, especially inflation in the severity of personal injury claims. There is no persuasive evidence that increasing profit rates or expense loadings contributed to inflation in premiums. Hence, policy measures to deal with insurance inflation must address the growth in the underlying cost factors. This means finding effective mechanisms for breaking up insurance lotteries and providing additional incentives for insurers to resist potentially fraudulent claims. Promising approaches along these lines would be for states to adopt 131t may not be feasible for insurers to withdraw from the market and then reenter in profitable lines through a subsidiary. For example, Fireman's Fund withdrew from Massachusetts and then attempted to reenter the commercial market through a subsidiary. This attempt was blocked by the insurance commissioner and the matter is now in court. 114 Journal of Economic Perspectives elective no-fault laws with strict verbal thresholds for pursuing liability suits, or anti-fraud efforts coordinated through organizations like the Massachusetts fraud bureau. Arbitrary rate rollbacks and additional rate regulation are unlikely to be effective in controlling costs because such measures merely address the symp- toms of the auto insurance problem, rather than the underlying causes. In- stead, state insurance departments should direct their attention towards ensuring that insurance markets are competitive. Anti-rebate laws should be repealed and more aggressive competition by insurance agents should be encouraged. Computerized insurance hotlines should be established, funded by the insurance industry, where buyers can obtain free information on the prices charged by competing insurers. Reducing excessive claims costs and ensuring competition in the insurance marketplace offer the best hope for controlling automobile insurance prices. However, even with a perfect auto insurance market, insurance coverage is likely to remain an expensive product, since the costs of medical care and auto repairs are likely to continue to rise more rapidly than the overall consumer price index. References A.M. Best Company, Best's Aggregates and Insurance." Working paper, Wharton School, Averages: Property-Casualty Edition. Oldwick: University of Pennsylvania, 1991. A.M. Best Company. 1989 and 1990 editions. Cummins, J. David and Mary A. Weiss, American Insurance Association, Summary "The Effects of No Fault Automobile Insur- of SelectedState Laws and Regulations Relating to ance Loss Costs," Geneva Papers on Risk and AutomobileInsurance. New York: American In- Insurance, January 1991a, 16, 20-38. surance Association, 1990. Cummins, J. David and Mary A. Weiss, Berger, Lawrence A., Paul R. Kleindorfer, "Incentive Effects of No Fault Automobile In- and Howard Kunreuther, "A Dynamic Model surance." Working paper, University of Penn- of the Transmission of Price Information in sylvania, 1991b. Auto Insurance Markets," Journal of Risk and Insurance, March 1989, 56, 17-33. Cummins, J. David and Jack VanDerhei, "A Coleman, T. S., L. Fisher, and R. Ibbotson, Note on the Relative Efficiency of Property- U.S. Treasury Yield Curves: 1926-1988. New Liability Insurance Distribution Systems," Bell York: Moody's Investors Service, 1989. Journal of Economics, Autumn 1979, 10, Conard, Alfred F., et al., AutomobileAccident 709-19. Costs and Payments. Ann Arbor: University of Dahlby, Bev and Douglas S. West, "Price Michigan Press, 1964. Dispersion in an Automobile Insurance Mar- Cummins, J. David, "Multi-Period Dis- ket," Journal of Political Economy,June 1986, counted Cash Flow Ratemaking Models in 94, 418-38. Property-Liability Insurance," Journal of Risk D'Arcy, Stephen P. and Neil A. Doherty, and Insurance, March 1990, 57, 79-109. "Adverse Selection, Private Information and Cummins, J. David and Patricia A. Danzon, Lowballing in Insurance Markets," Journal of "Price Shocks and Capital Flows in Liability Business, April 1990, 63, 145-64. Controlling Automobile Insurance Costs 115 Federal Highway Administration, Highway Commissioners, 1990. Loss Statistics. Washington, D.C.: Federal Pauly, Mark, Paul Kleindorfer and Howard Highway Administration, 1988. Kunreuther, "Regulation and Quality Compe- Grabowski, Henry, W. Kip Viscusi and tition in the U.S. Insurance Industry." In William S. Evans, "The Effects of Rate Regu- Finsanger, Jorg and Mark Pauly, eds., The lation on the Price and Availability of Automo- Economics of Insurance Regulation. New York: bile Insurance," Journal of Risk and Insurance, St. Martin's Press, 1986, 65-107. June 1989, 56, 275-99. Rothschild, Michael and Joseph E. Stiglitz, Harrington, Scott E., "A Random Coeffi- "Equilibrium In Competitive Insurance Mar- cient Model of Interstate Differences in the kets: An Essay on the Economics of Imperfect Impact of Rate Regulation on Auto Insurance Information," QuarterlyJournal of Economics, Prices," Review of Economics and Statistics, November 1976, 90:4, 629-49. February 1987, 69, 167-70. Samuelson, William and Richard Zeck- Insurance Research Council, Trendsin Auto hauser, "Status Quo Bias in Decision Making," Bodily Injury Claims. Oak Brook: Insurance Re- Journal of Risk and Uncertainty,March 1988, 1, search Council, 1990. 7-59. Joskow, Paul, "Cartels, Competition and U.S. Department of Transportation, "Mo- Regulation in the Property-Liability Insurance tor Vehicle Crash Losses and Their Compen- Industry," Bell Journal of Economicsand Man- sation in the U.S." Washington, D.C.: U.S. agement Science, Autumn 1973, 4, 375-427. Government Printing Office, 1971. Katz, Michael L., "Vertical Contractual Re- Weisberg, Herbert I. and Richard A. lations," In Schmalensee, Richard and Robert Derrig, "Fraud and Automobile Insurance: A Willig, eds., Handbook of Industrial Organiza- Report on the Baseline Study of Bodily Injury tion. Amsterdam: Elsevier Science Publishers, Claims in Massachusetts," Journal of Insurance 1989, 656-721. Regulation, March 1991, 9, 497-541. Keeton, Robert E. and Jeffrey O'Connell, Williams, C. Arthur, Jr., Price Discrimina- Basic Protection for the Traffic Victim. Boston: tion in Property and Liability Insurance." In Little, Brown and Co., 1965. Kimball, Spencer L. and Herbert S. Denen- Myers, Stewart and Richard Cohn, "In- berg, eds., Insurance, Government, and Social surance Rate Regulation and the Capital Asset Policy. Homewood: Richard D. Irwin, 1969, Pricing Model," In Cummins, J. David and 209-42. Scott E. Harrington, eds., Fair Rate of Return Winter, Ralph A., "The Liability Insurance in Property-LiabilityInsurance. Norwell: Kluwer Market," Journal of EconomicPerspectives,Sum- Academic Publishers, 1987, 55-78. mer 1991, 3, 115-36. National Association of Insurance Com- Zador, Paul and Adrian Lund, "Re-Analysis missioners, State Average Expendituresand Pre- of the Effects of No Fault Auto Insurance on miumsfor Personal AutomobileInsurance in 1989. Fatal Crashes," Journal of Risk and Insurance, Kansas City: National Association of Insurance June 1986, 53, 226-41.