Printer Friendly

The impact of medical malpractice insurance rate regulation.

The Impact of Medical Malpractice Insurance Rate Regulation

Abstract

This study examines the impact of rate regulation laws on direct writer market share and loss ratios for the medical malpractice insurance industry. Direct writer market share is significantly higher in states which have both non-competitive rating (NCR) laws and active joint underwriting associations (JUAs) than in states which have only one or neither of these characteristics.

NCR laws have had little independent effect on underwriting results in this industry. However, direct writers appear to fare better under NCR laws than they do under competitive rating (CR) laws. While direct writer market share is positively related to the loss ratio in CR states, in NCR states this relationship is far weaker.

Introduction

The impact of rate regulation on underwriting results has been the subject of a number of studies. Most of these have examined the auto insurance industry. Others have focused on homeowners insurance and general liability.

To date, no study has examined the impact of rate regulation laws on medical malpractice insurers. Such a study is of interest, however, for at least three reasons. First, the structure of the medical malpractice insurance industry differs substantially from that of other lines of insurance which have been examined in the context of rate regulation. Examining the impact of rate regulation on such a dramatically different line of insurance provides a test of the robustness of earlier findings.

Second, such a study would enhance understanding of some important economic factors which have served to shape today's medical malpractice environment. Previous studies in what may be broadly referred to as "the economics of medical malpractice" have focused on the determinants of malpractice claims frequency and severity, and the impact of malpractice litigation on physician behavior. However, little attention has been paid to the medical malpractice insurance industry.

Third, gaining insight into the determinants of underwriting results in the medical malpractice insurance industry is important from a practical standpoint because this line of insurance has suffered acutely in recent years and, as a result, the cost of malpractice insurance has increased while its availability has dwindled.

This study investigates the impact of rate regulation laws on state-level loss ratios for the medical malpractice insurance industry. A telephone survey of each state insurance commissioner's office was conducted to obtain rating laws pertaining to medical malpractice insurers. The impacts of other variables of interest, such as direct writer market share, are also examined.

This article first overviews the literature on the effects of rate regulation on underwriting results. Then the medical malpractice insurance market is described and contrasted to other lines which have been examined previously for rate regulation effects. The specification of the model and the empirical results are presented next. The results and their implications are then discussed.

Previous Multivariate Approaches

The literature on the impact of rate regulation laws is extensive. This section briefly reviews the results of large sample studies which have employed multivariate techniques to assess the impact of rate regulation laws. For a full survey of the literature on the impact of rate regulation laws, see Harrington [8].

Ippolito [12] used the multivariate approach to control for factors that may influence loss ratios independently of rating laws. Ippolito examined the impact of auto insurance rating laws on loss ratios, employing two estimation strategies. The first strategy used pooled cross-section time series data, but did not correct for either heteroskedasticity or autoregression. The model was estimated by ordinary least squares (OLS), and no effect of rate regulation on loss ratios was observed. Ippolito's second strategy computed average loss ratios by state for 1971 through 1973 and 1973 through 1975. Both of these equations were estimated by OLS. Once again, no impact of rate regulation was observed. A study by the General Accounting Office [6] which employed OLS to estimate the impact of auto insurance regulation also found no significant difference between non-competitive rating law (NCR) and competition rating law (CR) states.

Cummins and Harrington [2] used by-state by-line loss ratios for four lines of insurance (private passenger auto, commercial auto, homeowners, and general liability) in 1977. The impact of rate regulation laws was estimated using weighted least squares, to correct for heteroskedasticity. Their results indicate that loss ratios were significantly lower in CR states than in NCR states for all lines examined except general liability. They also found lower loss ratios for direct writers.

Harrington [9] estimated the impact of auto insurance rate regulation laws, controlling for the average expected incurred loss in a state and adjusting for heteroskedasticity. Harrington used average values of the variables from 1976 to 1981 for all states and the District of Columbia, and ran separate estimates for direct writers and independent agents. His results indicate that states having prior approval rating laws (the most common form of NCR laws) experienced significantly higher loss ratios. Pauly et al. [18] also found that NCR states exhibited higher loss ratios.

While earlier studies, such as Ippolito [12] and the GAO [6], found no relationship between rating laws and loss ratios, later studies, such as Cummins and Harrington [2], Harrington [9], and Pauly et al. [18], found that NCR states had higher loss ratios. Since these later studies employed more sophisticated estimation techniques (most notably correcting for heteroskedasticity), and generally employed a superior set of additional explanatory variables (i.e., Ippolito failed to include direct writer market share; most of the explanatory variables in the GAO study were not statistically significant), they would appear to be more credible.

The Medical Malpractice Insurance Market

In contrasting the medical malpractice insurance market to other insurance markets, one is impressed by the differences. As Danzon [5] notes:

Medical malpractice insurance is not like automobile liability insurance or homeowners' insurance, with which most people are familiar: the physician cannot simply call up his local insurance agent, or choose from twenty or more companies listed in the yellow pages. Typically, there would be at most two or three companies writing a significant volume of business in his area, of which one may be a physician-owned company (p. 89).

Indeed, perhaps the most striking difference between the medical malpractice insurance market and other lines of insurance which have been examined for rate regulation effects is the degree of market concentration. Table 1 shows the median market share of the leading writer for each state for medical malpractice, general liability, homeowners, and auto insurance in 1976 and 1984. As the Table indicates, the malpractice insurance market is much more concentrated than the others, with a median leading writer market share of 44.6 percent.

In addition to having more concentrated markets than the other three lines of insurance, loss ratios have risen more rapidly for medical malpractice insurers, and by 1984 were substantially higher than loss ratios for the other lines of insurance. Table 2 compares median state-level loss ratios in 1976 and 1984 for all four lines of insurance. As the Table indicates, in 1976 the median loss ratio for medical malpractice was less than the corresponding loss ratios for general liability, auto, and homeowners insurance. By 1984, the pattern had completely reversed, with medical malpractice loss ratios substantially higher than loss ratios for these other lines.

The dramatic loss ratio growth occurred in spite of malpractice insurers' increased predilection for writing claims-made, as opposed to occurrence-based, policies during this period. Under an occurrence policy, which had been the traditional form of policy for medical malpractice prior to the mid-1970s, the insurer covers all claims resulting from incidents which occurred during the policy year, regardless of when such claims are actually filed.

By contrast, the claims-made policy covers only claims filed during the policy year. As Danzon [5, p.110] notes, the claims-made policy is advantageous to the insurer in that it:

. . . shifts the risk associated with future claims from the insurer to the policyholder. Although the policy guarantees the availability of an endorsement to cover claims incurred but not filed during the policy year, the price of this future coverage is not fixed in advance. (1)

Although loss ratios have risen rapidly in medical malpractice insurance, it seems likely that the appearance of claims-made policies has prevented even more unfavorable underwriting results in this industry. In transferring the risk associated with future claims to the insureds, the claims-made policy makes it easier for the insurer to determine adequate levels of premiums and reserves.

A further difference between medical malpractice insurance and the other lines is the rapid growth of direct writer market share in medical malpractice insurance markets. Table 3 compares median state-level direct writer market shares in 1976 and 1984 for the four lines of insurance. As the Table indicates, median direct writer market share increased by 721.1 percent between 1976 and 1984 for medical malpractice. By comparison, growth in direct writer market share for general liability (9.4 percent) and homeowners (26.1 percent) was modest.

Finally, observing the ranges in leading writer market share, loss ratios, and direct writer market share, one notices substantially greater cross-sectional variability in medical malpractice insurance markets. For example, Table 3 indicates that state-level direct writer market share for medical malpractice in 1984 ranged from a low of 0.6 percent to high of 81.9 percent. The corresponding ranges in direct writer market shares for the other lines of insurance in 1984 are considerably more compact.

To sum up, the medical malpractice insurance market differs from the other three lines examined in that it exhibits: greater market concentration; higher loss ratios; relatively rapid growth in loss ratios and direct writer market share; and greater cross-sectional variability. Evidently, both across states and over time, the medical malpractice insurance market exhibits considerably less stability than do general liability, auto, and homeowners insurance markets.

Much of the variability in the medical malpractice insurance market may be traced to the unpredictable nature of medical malpractice losses. Medical malpractice losses are extremely difficult to predict for several reasons. First, as Danzon [5, p. 90] points out, many states face small sample problems:

In 1976 there were 17,683 claims closed, for which 7,262 closed with payment to the plaintiff. But claims are concentrated only in a few states. Of the 7,262 paid claims, 4,647, or almot two-thirds, occurred in the seven states that each had more than 300 claims. New Hampshire had only 3 claims, and thirty-five states had fewer than 100.

The enormous range in the size of awards contributes to the variability in losses resulting from small sample size. Another factor contributing to malpractice loss uncertainty is the "long tail" in the disposition of claims. It may take a number of years between the occurrence of an injury, the filing of a claim, and its final disposition. It is hard to predict exactly when, if at all, a payment will have to be made.

Furthermore, that malpractice claims may be filed a number of years after an injury has occurred implies that there is a stockpile of potential claims. A change in what Danzon terms "sociolegal" trends -- attitudes of juries, changes in malpractice torts -- may induce a rash of claims from this available stockpile, resulting in a malpractice "crisis." As Danzon [5, p. 91] notes:

Although sociolegal risk affects all lines of liability insurance to some degree, it contributes more to the medical malpractice risk as a result of the long tail.

The difficulty in predicting medical malpractice losses has led to substantial variability in loss ratios, especially in smaller states. Furthermore, loss ratios have been rising rapidly over time, as noted above. These factors have made providing medical malpractice unattractive to many insurers, inducing national companies to exit a number of malpractice insurance markets, and increasing the need for direct writers (many of which the physician-or hospital-sponsored programs) to take up the slack.

Empirical Specification and Results

The effect of state rate regulation laws on medical malpractice loss ratios is estimated using a pooled state-level cross-section time series data for the years 1976 through 1984. A two-stage procedure is employed. In the first stage, the determinants of direct writer market share are estimated. The predicted value of direct writer market share, together with rate regulation laws and other variables thought to influence loss ratios, are then included in a second equation which estimates state-level loss ratios.

An alternative approach is to estimate a simultaneous equations system, in which the loss ratio would appear as an explanatory variable in the direct writer market share equation. Hoever, the present approach does not estimate a structural equation for direct writer market share. Instead, reduced form estimates of direct writer market shares are obtained. (2)

Since the predicted direct writer market share must be a number between 0 and 1, a logit model is estimated. (3) More specifically, the first stage model to be estimated is:

1n ([DW.sub.ti])/1 - [DW.sub.ti]) = [a.sub.o + [a.sub.1 . [NOCOMP.sub.ti] + [a.sub.2] . [NOCOMP.sub.ti] . [JUA.sub.ti] + [a.sub.3] . [JUA.sub.ti] + [a.sub.4] . [CLAIMS.sub.i] + [a.sub.5] . [DOCS.sub.ti] + [a.sub.6] . TREND,

where:

1n(.) = natural logarithm operator; [DW.sub.ti] = direct writer market share in year t for state i; [NOCOMP.sub.ti] = 1 if state i has NCR law in year t, 0 otherwise; [JUA.sub.ti] = 1 if state i has an active joint underwriting association in year t, 0 otherwise; (4) [CLAIMS.sub.i] = average per capita claims incidents from 1975 to 1978 in state i; (5) [DOCS.sub.ti] = number of nonfederal patient care physicians in year t for state i; and [TREND] = time trend variable. (6)

The variable NOCOMP measures whether the state has a competitive rating law for medical malpractice insurers or a non-competitive rating law. Data for constructing the variable NOCOMP were obtained from a telephone survey of the 50 state insurance commissioners' offices and the insurance commissioner's office for the District of Columbia. Table 4 lists states having competitive rating laws.

The presence of NCR laws and/or joint underwriting associations (JUAs) should serve to decrease the attractiveness of writing insurance, other things being equal. NCR laws restrict insurer discretion in setting rates. JUAs typically require contributions from insurers in their state of operation. If national companies tend to avoid such states, a higher direct writer market share might be observed in NCR states and in states with active JUAs

An interaction term, NOCOMP . JUA, is included to test whether NCR laws have had a different impact in JUA states compared to non-JUA states. It may be that national companies are particuarly anxious to avoid states where both NCR laws and JUAs are present. If so, one might expect to observe a positive relationship between direct writer market share and this interaction term.

The variable CLAIMS is included to control for state-specific differences in the medical malpractice environment. The variable TREND controls for the rapid growth in direct writer market share that has occurred in recent years.

DOCS is included because national companies, which are typically quite large, may be unlikely to write in any state unless they achieve a certain volume of insureds. This "minimum volume requirement" translates into a higher "minimum market share" requirement for national companies when they operate in smaller states. Therefore, one might expect to observe a higher market share for national companies in smaller states, and a correspondingly lower market share for direct writers in such states, ceteris paribus.

Separate regressions were run for all states and for non-JUA states. (7) While a variable measuring the presence of an active JUA is included in the all-states sample, it is important to note that JUA data are typically not reported to AM Best, so that the loss ratios do not reflect JUA experience. Nevertheless, the presence of a JUA may proxy for an unfavorable malpractice environment generally, and in this way may be related to the loss ratio data reported to Best.

The estimation results are shown on Table 5. direct writer market shares are particularly high in NCR states having active JUAs. On the other hand, the impact of NCR laws in non-JUA states appears to be far more modest. While the coefficients on NOCOMP are positive for both regressions presented in Table 5, they are not statistically significant. Direct writer market share is directly related to per capita malpractice claims incidents, number of physicians within a state, and the time trend variable.

The loss ratio (8) is estimated using the predicted value of DV from (1) and other relevant variables:

1n ([LR.sub.ti]) = [b.sub.0] + [b.sub.1] . [NOCOMP.sub.ti] + [b.sub.2] . [NOCOMP.sub.ti] . [DW.sub.ti] + [b.sub.3] . [DW.sub.ti] + [b.sub.4] . [JUA.sub.ti] +[b.sub.5] . [CLAIMS.sub.i] + [b.sub.6] . TREND,

where:

[LR.sub.ti] = dividend-adjusted loss ratio in year t for state i; (9)

[DW.sub.ti] = predicted value of direct writer market share in year t for state i; (10) and other variables are as defined above. (11)

Preliminary analysis of the data indicated that heteroskedasticity is present and, in particular, that the variance of the disturbance term is inversely related to the size of the market. (12) Harrington [9] and Cummins and Harrington [2] also found an inverse relationship between the disturbance variance and state-wide premiums for the insurance lines they examined. To correct for heteroskedasticity, equation (2) was estimated by weighted least squares, using the square root of state-level malpractive premiums as the weighting factor.

Direct writer market share is included as an explanatory variable since previous research has indicated that loss ratios may differ for direct writers and non-direct writers. Most research, such as Joskow [13], Cummins and VanDerhei [3], and Witt and Miller [20], has found that direct writers tend to have lower expense ratios and higher loss ratios, though Cummins and Harrington [2] found evidence of significantly lower loss ratios for direct writers.

Prior analyses of the relationship between direct writer market share and loss ratios may be of limited guidance in predicting the effects of direct writer market share on medical malpractice insurance loss ratios, however. The direct writers in studies of other insurance lines are quite different from the typical direct writer for medical malpractice insurance. For example, in the Cummins and VanDerhei study, direct writers included Allstate, State Farm, and Nationwide. These are large companies that operate in a number of states. By contrast, direct writers of medical malpractice insurance are typically much smaller, and often operate only within a single state. In contrast to the findings for large direct writers such as State Farm (see Joskow, [13], medical malpractice direct writers may not have relatively low expense ratios, for example. As a result, their premiums may not be lower and their loss ratios may not be higher than those of non-direct writers.

Joskow [13, p. ji1] has argued that ". . . the strategy of the direct writers . . . should be to lower their prices below bureau rates just enough to 'differentiate' their products, and then fill up their premium 'quota' with the best risks available." Joskow's argument is appropriate for the insurance lines which he examined (fire and auto) and probably for most other property-liability lines as well. In medical malpractice insurance, however, the situation is different. Many medical malpractice direct writers emerged in response to the inability of some physicians to obtain coverage. If anything, these physicians were in higher risk categories, not lower.

As observed above, another difference between medical malpractice direct writers and direct writers in other property-liability lines concerns growth in market share. Joskow [13] predicts that growth in direct writer market share should be gradual, increasing as equity capital is accumulated over time. (13) As the discussion above indicates, the observe gradual increases in direct writer market share for general liability and homeowners insurance are consistent with Joskow's prediction. But the rapid increase in medical malpractice direct writers market share is not, suggesting that it may have increased due to other factors. Rapid growth of physician-and hospital-sponsored direct writers may have occurred to fill the void created by the exit of non-direct writers, for example.

Finally, in estimating the impact of direct writer market share on loss ratios, one must be aware of the potential for reverse causality. Since direct writer market shares may have increased in states where other insurers had exited due to high loss ratios, this suggests that one might observe a positive relationship between loss ratios and direct writer market share. Once the direct writer has entered the market, however, it will affect loss ratios in a direction which is ambiguous a-priori. Using the predicted value of direct writer market share obtained from the equation (1) estimates and controlling for differences in state-level malpractice incidents should serve to mitigate this problem.

An interaction term, NOCOMP . DW, is included to test for the possibility that non-competitive rating laws affect direct writers differently than non-direct writers. NCR laws may affect direct writers differently for at least two reasons. First, to the extent that many medical malpractice direct writers are physician- or hospital-sponsored, cooperation between the insurer and the insured in the setting of rates may be greater, which facilitates obtaining regulatory approval of rates. For example, Danzon [5, p. 95] has noted that:

. . . cooperation between medical societies and insurers reduces the costs of obtaining regulatory approval of rates. Although rate filings must still be supported by actuarial data, the consent of the insured is usually a sufficient condition for automatic approval.

Secondly, insurance regulators in NCR states having a large direct writer market share may simply be more lenient in allowing direct writers to set rates as they see fit. If these direct writers arose to fill the void created by the exit of national companies, insurance regulators might be wary of inducing direct writer exit by failing to approve their rates promptly.

If direct writers in NCR states can adjust rates more easily than can non-direct writers, this suggests that they may enjoy a competitive advantage in NCR states. This could improve underwriting results for direct writers in NCR states. This could improve underwriting results for direct writers in NCR states relative to direct writers in CR states. If so, the coefficient on the interaction term will be negative. Finally, CLAIMS and TREND are included to control for state- and time-specific effects on the loss ratio.

Table 6 shows the estimation results. Again, separate regressions were run for all states and non-JUA states. The coefficients on direct writer market share are positive in both regressions, and significant in the all-states sample. The interaction term, NOCOMP . DW, however, is negatively related to the loss ratio in both regressions. These results suggest that loss ratios are positively related to direct writer market share in CR states, but that this relationship appears to be much weaker in NCR states. Indeed, the coefficients on NOCOMP . DW and DW in the all-states sample suggest that the loss ratio is unrelated to direct writer market share in NCR states. The improved performance of direct writers in NCR states may stem from their greater ability to secure approval for rate changes in these states than can national companies, as discussed above. Other results indicate that the loss ratio is directly related to per capita claims incidents, and that it has risen dramatically over time.

Alternative Specifications

As a check on the robustness of these findings, alternative estimation strategies were employed. This section briefly summarizes these approaches.

First, the issue of claims-made versus occurrence-based policies was considered. Ideally, one would like to control for the mix of these policies within a state. While it was not possible to obtain data on the proportion of premiums written on a claims-made basis, states were identified which prohibited claims-made policies during the study period. Only 3 states prohibited such policies, however. With so few states denying claims-made policies, estimated relationships between the prohibition of claims-made policies and the loss ratio may reflect state-specific effects, not the effects of these prohibitions.

Nevertheless, if denial of claims-made policies is correlated with the present set of explanatory variables, the resulting estimates will be biased. While there are no academic studies on the relationship between denial of claims-made policies and the set of explanatory variables employed here, it is useful to speculate as to the nature of these potential relationships and their implications for these results.

Direct writers have tended to flourish in areas considered unattractive by national companies. States which deny laims-made policies may fall into this category. If a disproportionate share of direct writers write in states which deny claims-made policies, the present estimates may overstate the higher loss ratios of direct writers, since loss ratios are generally higher under occurrence-based policies.

To address this potential problem, equations (1) and (2) were reestimated, excluding states which denied claims-made policies. (A drawback to this approach is that it requires excluding the largest NCR state, New York, and one of the largest, Michigan. The importance of these omissions is magnified by the weighting procedure employed to correct for heteroskedasticity.) In the sample including JUA states, direct writer market share was directly and significantly related to the loss ratio, and the interaction term showed a negative and significant relationship. The coefficients on these variables were smaller than those reported in the text, however. In the non-JUA states sample, the estimated coefficients on these variables were not significant. Coefficients on other explanatory variables showed little change.

A second approach assumed that the loss ratio and direct writer market share were related in a purely recursive fashion. In particular, it was assumed that only past values of the loss ratio affect current values of direct writer. Market share which, in turn, affect current values of the loss ratio. In this case, OLS estimation of equation (2) (i.e., using the actual values of direct writer market share rather than predicted values) may provide consistent estimates, as Kmenta [11] has noted. (14)

This approach yielded a positive and significant relationship between direct writer market share and loss ratio in the all-states sample, and a negative and significant relationship on the NCR variable interacted with direct writer market share. Again, however, the coefficients were not as large as those reported in the text, and the estimates were insignificant in the smaller, non-JUA sample, althought the signs were in the expected direction. A positive and significant relationship was observed between the JUA dummy and the loss ratio. Other results were quite similar to those reported in the text.

Finally, equation (2) was reestimated including an interaction term between NOCOMP and the presence of a JUA. As mentioned earlier, JUA premiums are typically not reported to AM Best, so only the impact of the presence of a JUA on underwriting results in the non-JUA sector can be measured.

Thus, this interaction term tests whether the presence of a JUA affected non-JUA underwriting results differently in NCR and CR states. including this interaction term had only modest effects on the other explanatory variables in equation (2) with the exception of JUA status, which was directly and significantly related to the loss ratio. The interaction between rating law and JUA states was negative and significant, suggesting that underwriting results for direct writers and national companies are more favorable in NCR states with JUAs than in CR states with JUAs.

Interpretation of this result is confounded by the fact that multicollinearity may be present. As Harrington [8, p. 611] has noted:

The inclusion of even one variable interacting type of regulation and some other factor may lead to a multicollinearity problem and reduced efficiency. Including of additional interaction terms may cause severe multicollinearity. (15)

To examine whether the differential effects of direct writers and the presence of an active JUA in NCR versus CR states denote multicollinearity caused by the interaction terms, equation (2) was also estimated separately for NCR and CR states. Direct writer market share and the presence of a JUA were directly and significantly related to the loss ratio in CR states. No significantly relationships between these two variables and the loss ratio were observed in NCR states, however.

Nevertheless, the explanation for the differential impact of JUAs in NCR states versus CR states is unclear. The results in Table 5 suggest that national companies have tended to avoid NCR states with active JUAs. This could indicate that there is a shortage of coverage in the non-JUA sector in these states. As a result, national companies who write in these states may be able to do so selectively, providing coverage to lower risk medical specialists and hospitals, rather than to higher risk malpractice clients, who get relegated to the JUAs. This explanation is speculative, however, and these results should be viewed as preliminary.

Conclusion

With the growing frequency and severity of medical malpractice litigation, understanding the behavior of all players involved in this process is of considerable policy interest. To date, however, little attention has been paid to the medical malpractice insurance industry. This article has attempted to bridge this gap.

The two-stage estimation strategy employed gives insight into the determinants of direct writer market shares as well as loss ratios, both of which the grown substantially in this industry. Direct writers, which are typically small physician- or hospital-sponsored programs, have flourished in NCR states having active JUAs. Such states may be particularly unappealing to national companies. Direct writers may have grown in these states to fill the void left by the exit of national companies. In non-JUa states, however, NCR laws are not strongly related to direct writer market share. Direct writer market shares are relatively large in states characterized by high per capita claims incidents.

Evidence also suggests that NCR laws affect direct writers and national companies differently. In particular, underwriting results for direct writers improve relative to those of national companies in NCR states. In CR states, however, loss ratios appear to be significantly higher for direct writers.

While these findings on the impact of rating laws and direct writer market share on underwriting results are suggestive, and consistent with what one might expect, they must be viewed with caution. Data limitations did not allow controlling for the extent of premiums written on a claims-made basis within a state. This will be an important direction for future research when such data become available.

Even if NCR laws help lower the loss ratio of direct writers relative to national companies, and lead to substantial improvements in underwriting results when direct writer market share is large, it would be erroneous to conclude that implementing NCR laws is the best way to improve underwriting results in medical malpractice. The results of Table 6 could also be interpreted as implying that efforts to limit direct writer market share may be just as effective in terms of lowering the loss ratio in CR states. (Of course, if availability of insurance is a problem in these states, limiting direct writer market share may be impossible.) Since the emergence of direct writers as a major source of medical malpractice insurance is a relatively recent phenomenon, it would be prudent for CR states to continue to observe and evaluate this insurance alterntive before implementing any changes in rating laws.

Furthermore, implementing NCR laws may exacerbate problems of insurance availability. For example, the growth of direct writers in NCR states with active Juas may have occurred in part because national companies have tended to avoid such states. CR laws may be advantageous in that they attract the added competition and choice offered by national companies.

On the other hand, achieving competitive results might not be the main objective of rate regulation in a particular state. Harrington [8, p. 617] notes that:

if the underlying objective of regulation in a particular state is to achieve results that may differ substantially from those that would arise even if the market were perfectly competitive ... then an NCR law may facilitate attainment of the regulatory objective.

While NCR laws may lead to improved underwriting results for direct writers, the impact of such laws on premiums, the degree of competition, and the availability of malpractice insurance must be better understood if insurance regulators are to make informed decisions about the appropriate medical malpractice rating law in their states. Future research should focus on these issues.

(1) See also Danzon [5, p. 111] for a discussion of why the claims-made policy may be the optimal form of policy for insureds as well.

(2) The rationale behind the reduced form specification is that, while direct writer market share may depend upon current and lagged values of the loss ratio, these values of the loss ratio themselves depend upon exogenous variables employed in the text, such as malpractice claims incidents and the time trend.

(3) Disturbance terms are known to be heteroskedastic under the logit model. In this case, the variance of the disturbance term, e, will be given by:

VAR(e) = 1 / DW(1 - DW)

This problem is corrected by using weighted least squares. The weights are obtained by running an OLS regression on equation (1) and using the resulting estimates of direct writer marker share. While this procedure is frequently used for grouped date, it may also be applied in cases as in the text, where the dependent variable is a proportion and thus confined to the unit interval. See Maddala [16] and Greene [7] for further details on this estimation procedure.

While the logit method described above provides a convenient method for estimating market share variables, it does not allow for 0 percent or 100 percent market share. However, only one direct writer market share observation out of a possible 459 had an extremen value (zero).

(4) If one takes a sufficiently broad view, JUA status, and even rating law type, may be regarded as endogenous. However, Harrington [8] has argued that there is little evidence suggesting that rating laws need be specified as endogenous. Both rating law type and JUA status were quite stable over time for each state, suggesting that these variables had been largely determined prior to the period of the study. In view of this, as well as the difficulties posed by identifying a multiple equation system, it seems reasonable to treat rating laws and JUA status as predetermined.

(5) The years 1975 to 1978 are used in constructing the variable CLAIMS to obtain an adequate number of observations for claims incidents in some of the smaller states. CLAIMS is included in the analysis to control for the malpractice environment in each state. However, omitting this variable had every little effect on the results reported in the text.

(6) Data on direct writer market share (as well as loss ratio and market concentration data) were obtained from Best's Review: Property, Casualty Edition, for the years 1977 through 1985. The JUA data were obtained from the Alliance of American Insurers. The claims incidents variable was constructed from the National Association of Insurance Commissioner's medical malpractice closed claims study (1980). The variable DOCS was obtained from the American Medical Association's Profile of Medical Practice for the Years 1977 through 1985.

To facilitate comparisons of coefficients, continuous explanatory variables and the time trend were normalized to lie between 0 and 1.

(7) The insurance industry data used in this study are drawn from AM Best. Best receives insurance information from the voluntary portion of the insurance market; i.E., from direct writers andnational companies. However, JUAs sometimes contract with these private insurers. The extent of such contracting is uncertain, but, judging from the low levels of premiums reported to Best in states with large active JUAs like Massachusetts, it appears to be small. Nevertheless, some of the insurance data reported to Best for JUA states may reflect coverage for JUAs.

This ambiguity is, of course, avoided in the non-JUA states sample. Furthermore since the results for the all-states and non-JUA states samples are fairly consistent, the inability to control for JUA contracting with private insurers does not appear to be a serious problem.

(8) The loss ratio reflects claims paid and premiums earned for both physicians and hospitals. However, the strong majority of premiums are paid by physicians, not hospitals. In California, for example, data from a study by the General Accounting Office [6] indicate that hospitals paid $67.3 million in medical malpractice insurance premiums in 1983. Data from the AMA's Socioeconomic Monitoring System indicate that California's physicians paid $473.8 million in medical malpractice insurance premiums in 1983, accounting for over 87 percent of medical malpractice premiums in that state.

(9) The dependent variable is expressed in logarithmic form for two reasons. First, in some cases the actual loss ratio was a small positive number. Such outliers could lead to seriously downward-biased estimates and non-positive predicted loss ratios. The logarithmic specification precludes this possibility. (In three cases, states reported non-positive loss ratios. Omitting three out of a possible 459 observations was deemed the lesser evil than rising downward bias leading to non-positive predicted loss ratios.)

Second, the distribution of the loss ratio data suggests that it is log-normally distributed. As Maddala [15] notes, expressing the dependent variable in logarithmic form may help symmetrize its distribution.

(10) Identification of DW requires that at least one explanatory variable in equation (1) be omitted from (2). The variable DOCS was chosen for this purpose. While it might be argued that DOCS belongs in equation (2) as a separate regressor (perhaps as a measure of average scale of operations), prior statistical tests for direct relationships between DOCS and the loss ratio consistently found these relationships to by very weak.

In selecting DOCS to identify DW, the presumption is that DOCS provides a much better measure of the total size of the malpractice insurance market within a given state than it does of the average size of an insurer within that state (i.e., scale of operations). As argued in the text, in states where the total market size is small, national companies will tend to have higher market shares in order to obtain enough insureds to make entry worthwile. As a result, the state-level market shares of natonal companies should be inversely related to DOCS, and the market shares of direct writers should be directly related to DOCS. But if DOCS primarily measures the average size of an insurer within a state (i.e., the larger is DOCS, the larger the average scale of operations), one would not expect to observe such a strong positive relationship between direct writers market share and DOCS as is reported on Table 5. Indeed, if DOCS provided a good measures of the average size of an insurer within a state, one might expect direct writer market share to be inversely related to DOCS. Since direct writers are typically much smaller than national companies, one would expect that states having a large market share of direct writers will also be states where the average scale of operations is low.

The strong positive relationship observed between DOCS and direct writer market share is consistent with the presumption that DOCS provides a much better measure of the total size of the malpractice market within a state than the average scale of operations.

(11) Preliminary analyses included the market share of the three leading writers in each state as an explanatory variable measuring market concentration. This variable was never statistically significant. Although the results reported in the text are little affected by including three firm market share, this variable was excluded because it may be endogenous and because it was significantly correlated with some of the explanatory variables in the text. Excluding this variable lowered the standard errors of some of the estimates slightly.

(12) The data were stratified into two subsamples, one for observations from states having a larger volume of maloractice premiums, the other from states having a smaller volume of malpractice premiums. The Goldfeld-Quandt test for homoskedasticity (see Kmenta [11], for details) was preformed. The results strongly indicate that the disturbance variance is inversely related to total premiums written in a state. The estimated disturbance variance of the smaller premium sample was significantly larger than the estimated disturbance variance of the larger premium sample, at the I percent level of significance.

(13) The direct writer market shares in the insurance lines examined by Joskow were fairly concentrated. This allowed for the possibility of oligopolistic pricing and, hence, a rate of return in excess of the competitive opportunity cost of capital over the short run. Entry into the direct writer market, however, should gradually erode these profits and increase direct writer market share.

(14) This will be true provided that error terms in the direct writer market share equation and the loss ratio equation are independent and nonautoregressive (in which case these estimates will also be asymptotically efficient).

(15) Harrington [8] has also suggested that a parsimonious approach to model specification may be desirable in estimating the impact of NCR laws, as this will decrease the potential for problems of multicollinearity. This approach was used in estimating the impact of NCR laws.

References

[1.] Best's Review: Property/Casualty Edition, Oldwick, NJ, various issues, 1977-1985.

[2.] Cummins, J., and Harrington, S., "The Impact of Rate Regulation in U.S. Property-Liability Insurance Markets: a Cross-Sectional Analysis of Individual Firm Loss Ratios," The Geneva Papers on risk and Insurance, Vol. 12 (January 1987), pp. 50-62.

[3.] cummins, J. and VanDerhei, J., "A Note on the Relative Efficiency of Property-Liability Distribution Systems," Bell Journal of Economics, Vol. 10 (Autumn 1979), pp. 709-19.

[4.] Danzon, P., "The Frequency and Severity of Medical Malpractice Claims," Journal of Law and Economics Vol. 27 (April 1984), pp. 115-48.

[5.] Danzon, P., Medical Malpractice: Theory, Evidence and Public Policy, Cambridge: Harvard University Press, 1985.

[6.] General Accounting Office, Medical Malpractice: Case Study in California, Washington, DC, 1986.

[7.] Greene, W. "LIMDEP: An Econometric Modeling Program for the IBM PC." American Statistician Vol. 39 (August 1985), p. 210.

[8.] Harrington, S., "The Impact of Rate Regulation Laws on Prices and Underwriting Results in the Property-Liability Insurance Industry: A Survey," Journal of Risk and Insurance, Vol. 51 (December 1984), pp. 577-623.

[9.] Harrington, S., "A Note on the Impact of Auto Insurance Rate Regulation," Review of Economics and Statistics, Vol. 69 (February 1987), pp. 166-70.

[10.] Heidrich, G., and Kenney, R., "The Solvency Crisis in Medical Malpractice Insurance," The Examiner, Vol. 11 (Spring 1987), pp. 6-9.

[11.] Kmenta, J., Elements of Econometrics, New York: MacMillan Publishing Co., 1986.

[12.] Ippolito, R., "The Effects of Price Regulation in the Automobile Insurance Industry," Journal of Law and Economics, Vol. 22 (April 1979), pp. 55-89.

[13.] Joskow, P., "Cartels, Competition and Regulation in the Property-Liability Insurance Industry," Bell Journal of Economics and Management Science, Vol. 4 (Autumn 1973), pp. 375-427.

[14.] Maddala, G., "The Use of Variance Components Models in Pooling Cross-Section and Time Series Data," Econometrica Vol. 39 (March 1971), pp. 341-58.

[15.] Maddala, G., Econometrics, New York: McGraw Hill Book Co., 1977

[16.] Maddala, G., Limited-Dependent and Qualitative Models in Econometrics, New York; Cambridge University Press, 1983.

[17.] Medical Malpractice Closed Claims: 1975-1978, M. Patricia Sowka (ed.), Brookfield, Wisconsin: National Association of Insurance Commissioners, 1980.

[18.] Pauly, M., H. Kunreuther, and P. Kleindorfer, "Regulation and Quality Competition in the U.S. Insurance Industry," in: Economics of Insurance Regulation: A Cross-National Study, J. Finsinger and M. Pauly (eds.), New York: St. Martin's Press, 1986, pp. 65-107.

[19.] Profile of Medical Practice, Chicago, IL: American Medical Association, various issues, 1977-1985.

[20.] Witt, R., and Miller, H., "Price Competition, Regulation, and Systematic Underwriting Risk in Auto insurance Markets," CPCU Journal, Vol. 34 (December 1981), pp. 174-89.

John A. Rizzo is an economist with the National Center forHealth Services Research and Health Care Technology Assessment (NCHSR). This work was completed while Dr. Rizzo was at the Center for Health Policy Research, American Medical Association (AMA) in Chicago. The author would like to thank Scott Harrington, Greg Heidrich, A. Ronald Gallant, Daid Emmons, Richard Wilke, and participants in the session on medical malpractice insurance issues at the 1987 ARIA meeting for helpful comments and suggestions. He also thanks the editor, an associate editor, and two anonymous referees whose comments and suggestions led to substantial improvements over an earlier version of this pqper. The author accepts responsibility for any remaining errors. The views expressed in this paper are the author's own and no official endorsement by either NCHSR or AMA is intended or should be inferred.
COPYRIGHT 1989 American Risk and Insurance Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Rizzo, John A.
Publication:Journal of Risk and Insurance
Date:Sep 1, 1989
Words:7416
Previous Article:Aspects of optimal multiperiod life insurance.
Next Article:Asset/liability management for the life insurer: situation analysis and strategy formulation.
Topics:

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters