Massachusetts Automobile Rating & Accident Prevention Bureau v. Commissioner of Insurance

389 Mass. 824 (1983) 453 N.E.2d 381

MASSACHUSETTS AUTOMOBILE RATING AND ACCIDENT PREVENTION BUREAU & others
vs.
COMMISSIONER OF INSURANCE (and a companion case).

Supreme Judicial Court of Massachusetts, Suffolk.

May 4, 1983. August 4, 1983.

Present: HENNESSEY, C.J., LIACOS, NOLAN, LYNCH, & O'CONNOR, JJ.

Acheson H. Callaghan, Jr. (Scott P. Lewis & Douglas H. Wilkins with him) for Massachusetts Automobile Rating and Accident Prevention Bureau & others.

Francis H. Fox for Liberty Mutual Insurance Company & another.

Stephen S. Ostrach, Assistant Attorney General (Scott A. Smith & John P. Graceffa, Assistant Attorneys General, with him) for the defendant.

Stephen J. Paris, for Independent Insurance Agents of Massachusetts, Inc., amicus curiae, submitted a brief.

LYNCH, J.

This is a reservation and report of two cases by a single justice of this court arising out of the decision by the Commissioner of Insurance (Commissioner) fixing and establishing the 1983 automobile insurance rates. The plaintiffs in the two cases are the Massachusetts Automobile Rating and Accident Prevention Bureau (bureau), together with several named insurance companies in the first case, and Liberty Mutual Insurance Company and Liberty Mutual Fire Insurance Company in the second. The plaintiffs brought complaints under G.L.c. 175, § 113B, and alleged both statutory and constitutional violations by the Commissioner in establishing the 1983 automobile rates. The plaintiffs first contend that the rates are not adequate, just, reasonable, and nondiscriminatory, as required by G.L.c. 175, § 113B. Additionally, the plaintiffs allege that the rates are confiscatory and, therefore, unconstitutional.

The Commissioner established the 1983 automobile insurance rates after a hearing in which the bureau, the State Rating Bureau, and the office of the Attorney General, were the primary participants. At the hearing, the bureau recommended a rate increase of 18.5%. The 1983 rates that were established provide for a much smaller increase. The Commissioner determined the total premium for each insurance coverage based on the sum of three separately calculated factors: (1) an allowance for losses, which primarily involves the payment of claims; (2) an allowance for company expenses; and (3) an allowance for underwriting profits. The determination of the allowance for underwriting profits is directly influenced by the Commissioner's calculation of the investment income earned by the industry through its use of policyholder premiums and its own capital.

The central disputes between the parties involve the Commissioner's determination of the allowance for losses and the allowance for underwriting profits. The plaintiffs argue generally that the evidence presented at the hearings was insufficient to support the Commissioner's findings for these provisions and that the Commissioner's findings are inadequate, inconsistent, and erroneous as a matter of law. More specifically, the plaintiffs allege two substantial errors by the Commissioner as to the profits provisions and several errors with regard to the determination of the loss provisions. We outline briefly these allegations of error. First, as to the profit provisions, the plaintiffs allege that the Commissioner has committed legal error in his choice of the surplus allocation model to be used in determining the underwriting profit provisions for the various lines of insurance coverage. Second, the plaintiffs attack the Commissioner's choice of the appropriate tax rate to be used with the hypothetical statutory/regulatory model company, by which the industry's net investment profits are calculated. The plaintiffs state that the tax rate used by the Commissioner is inconsistent with the model's basic assumption.

As to the loss provisions, the issues raised by the plaintiffs relate almost exclusively to projected changes in the claim frequency of various insurance line coverages. The plaintiffs argue first that the Commissioner has failed to follow established rate making methodology in determining the property damage liability claim frequency trend. The plaintiffs allege that in projecting this claim frequency trend the Commissioner relied primarily on internal claim data, rather than on an analysis of specific external phenomena, as the Commissioner has done in the past. Where the Commissioner did analyze external phenomena, the plaintiffs challenge the Commissioner's analysis of the likely effect of these external factors. The plaintiffs particularly disagree with the Commissioner's conclusions that the stability in vehicle-miles travelled per automobile, the stiffening of the drunk driving laws in Massachusetts, and the increase in merit rating surcharges for drunk driving offenses, will reduce claim frequency. Finally, the plaintiffs have challenged the Commissioner's decision on a single cost claim issue in the loss provision, the Commissioner's projection that the Hospital Cost Containment Act, inserted by St. 1982, c. 372, which is designed to control hospital costs, will lead to a slight reduction in the hospital costs covered by bodily injury insurance. Based on these alleged errors by the Commissioner, the plaintiffs contend that the rates the Commissioner established for 1983 automobile insurance do not meet statutory or constitutional requirements.

We uphold the Commissioner's decision with respect to the allowance for profits, but remand for further consideration of the allowance for losses with respect to the Commissioner's decision on the projected effect of the Hospital Cost Containment Act on bodily injury insurance costs.

1. Standard of Review.

The bureau devotes a large portion of its brief positing its theory of the proper standard of our review that we should apply to the issues presented. However, the standard of review for this case has been fully developed and established by our prior decisions in this area. Our statutory review "of the Commissioner's decision is limited to a determination of whether the rates are `adequate, just, reasonable and non-discriminatory.' G.L.c. 175, § 113B" (footnote omitted). Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. 333, 336 (1981).^[1] We have repeatedly stated that under this standard of review, "[j]udicial inquiry is limited to whether the rates set by [the Commissioner] have reasonable support in evidence." Id. at 337. See Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. 592, 596 (1980); Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 795 n. 4 (1976). Since fixing the rates is not a judicial function, id., we do not substitute our judgment for that made by the Commissioner as to the adequacy or reasonableness of the premium charges ultimately set. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 336. The Commissioner must make findings that indicate the over-all basis for his decision. Id. at 337. See G.L.c. 175, § 113B; Insurance Rating Bd. v. Commissioner of Ins., 359 Mass. 111, 118 (1971). However, we review the evidence and the findings based on the established standards and give due weight to the Commissioner's experience, technical competence, and specialized knowledge, as well as the discretionary authority vested in the Commissioner by the Legislature. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 337. Attorney Gen. v. Commissioner of Ins., supra.

2. Allowance for Profit.

a. The statutory/regulatory company investment model and the appropriate tax rate. In establishing the companies' underwriting profit allowance, the Commissioner's objective is "to fix an allowance for underwriting profit which, when added to investment income, would yield a return on shareholder capital comparable to that of unregulated industries of comparable risk." Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 604-605. An important component of this calculation is the Commissioner's determination of the potential investment income to be attributed to the industry. The importance of this figure, as we have noted, results from, "[t]he financial facts of life — not unrecognized in the industry ... that money is made on investments, not on underwriting." Id. Typically, the premiums which are collected by the industry are combined with a percentage of capital contributed by shareholders, and used for investment.

Beginning with the decision on the 1976 automobile insurance rates, the Commissioner has employed a hypothetical insurance company, the statutory/regulatory company model (model), as the vehicle for estimating these potential investment returns. The basic theory and purpose of this model have been extensively discussed in our prior decisions. See generally Attorney Gen. v. Commissioner of Ins., supra at 813-815; Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 604-606. Briefly, the Commissioner has assumed that this simplified model company invests all its available funds in Treasury securities with maturity dates selected to meet the model company's actual payment patterns for losses and expenses. This basic, conservative investment assumption was designed for rate making purposes to establish a minimum, reasonable investment yield standard that companies can meet in actual practice.^[2]

In past years, the Commissioner has discounted the resulting investment income by the nominal Federal tax rates on income from such securities, presently 46%, to produce the net investment profit figure. The use of this nominal tax rate has been predicated on the implicit assumption that this model company performs no functions other than earning investment income and paying Federal taxes. However, in this year's decision, the Commissioner expressed dissatisfaction with the model's simplicity and decided to take a first step in making it more sophisticated. Accordingly, he modified the basic model by rejecting this unrealistic tax assumption and using an effective tax rate in the model which more closely approximates actual company experience. The Commissioner concluded that using a 28% effective tax rate in the model would be appropriate and he applied that tax rate to the model's projected investment income.

The plaintiffs assert that the Commissioner's decision to modify the model and his choice of the 28% effective tax rate are the most serious errors in his decision and, consequently, present the most important issues raised in their appeal.^[3] They attack this part of the Commissioner's decision on several grounds. The plaintiffs argue, first, that the Commissioner has failed both to explain adequately his reasons for departing from prior practice and to make clear findings to support the change. They further contend that the Commissioner, by using an effective tax rate for the model company, has created a hybrid model combining hypothetical investment practices with "real-world" tax data. This resulting hybrid model, they assert, is internally inconsistent with the statutory/regulatory assumption and inconsistent with the Commissioner's prior decisions in this area. They claim that if the model company is investing only in Treasury securities, it must be paying taxes at the applicable tax rate of 46% on such securities. If, however, the Commissioner chooses to use effective tax rate figures drawn from actual companies' experience in the model, the plaintiffs state that he is obligated to use actual portfolio compositions, investment returns, and underwriting results in order to be consistent.

Finally, the plaintiffs argue that even if what they denominate "real-world" tax rates should be used, the Commissioner's choice of 28% as this effective tax rate is not reasonably supported in the evidence as required by our prior decisions. Moreover, they claim that the Commissioner's choice of a 28% tax rate is unrealistic and violates the principle that any simplifying assumptions used in the model should not be "fanciful or impossible to match in the real world." Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 608, quoting Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 822 (1976).

We have reviewed each of the plaintiffs' contentions and we conclude that the Commissioner's decision on this issue is proper and supported by the record. We disagree fundamentally with the plaintiffs' analysis and characterization of the Commissioner's decision on these issues. Contrary to the plaintiffs' assertion, the Commissioner has provided adequate reasons for his decision to modify and rationalize the model. The Commissioner stated that he has been dissatisfied with the model's lack of sophistication and he noted the testimony of Dr. William Fairley, a recognized expert in the field, that numerous simplifications existed in the original model. Dr. Fairley further testified that it was intended that the model would be made more sophisticated with experience. Thus the statutory/regulatory assumption was to be only the first step in the development of a profit model that would reflect the actual investment experience of the insurance companies. Using an effective tax rate in the model, the Commissioner found, would achieve further refinement of the model.

The Commissioner's primary criticism of the model is that in its recent projections the model has consistently underestimated the actual investment profits earned by the industry. The Commissioner identified the continued use of the nominal 46% tax rate as a principal factor in this underestimation. When the model was originally formulated the use of this nominal rate was appropriate. However, the financial environment in which this tax assumption was proper has altered significantly, vitiating this assumption. These changed circumstances result from the interrelationship between the model's projected investment profits and the underwriting profit allowance that is determined by the Commissioner. In theory, this underwriting profit allowance reflects profits earned by the insurance companies from the collection of premiums. In practice, whether these "profits" are actually positive or negative depends on which of the following circumstances are met. Where the projected investment income is relatively low, the premium rates of the various insurance coverages are set at a level at which the companies earn a profit on their underwriting business. Assuming that the expenses of the model company are not so high that a net loss results when they are combined with the underwriting profits, the use of the nominal 46% tax rate for investment income may be appropriate. However, where the projected investment income is relatively high, the Commissioner will set premium rates at a level where an underwriting loss will result. This situation presently exists. The recent substantial rise in interest rates has driven the underwriting profits set by the Commissioner from the small positive value fixed in 1976 to a negative value in this year's decision on the major insurance coverages.^[4]

The Commissioner's choice of a new tax rate merely recognizes this previously ignored internal factor, that the model company would, as do real companies, combine its underwriting losses with its expenses to reduce the effective tax rate on its investment income. The Commissioner's use of this tax rate merely involves a rationalization and extension of the theories behind the original model's assumptions, not a rejection or adulteration of the model as the plaintiffs assert. Nor does the adoption of this new tax rate create internal inconsistencies in the model. The Commissioner has not combined "real-world" tax rates with hypothetical investment strategies, but rather has extrapolated from his observation of actual companies' experience to obtain a more detailed understanding of the model company's internal actions. This, in turn, provides a more precise calculation of the companies' investment profits.

The plaintiffs' analysis attempts to treat the model company as if its investment actions occurred in a vacuum. However, it is clear that in the present rate setting circumstances the model company would pay taxes at an effective rate considerably less than the 46% rate because of the projected underwriting losses. Without the rationalization of the model to incorporate this tax effect, the projected investment profits would likely be "a very misleading indication of the actual profit made by the insurer[s]." Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 812 (1976). The use of this more realistic tax rate for the model company has the effect of more closely "corresponding to the investment policy [that the Commissioner has] assumed." Id. at 815 n. 28. We conclude that the Commissioner adequately explained his reasons for abandoning the existing assumption about the applicable tax rate and we defer to his experience, expertise, and discretion in this decision. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. 333, 343 (1981).

The plaintiffs also contend that the actual choice of the 28% figure as the proper tax rate to be used in the model does not have reasonable support in the evidence. We disagree. In his testimony at the hearing, Dr. Fairley stated that 28% is a reasonable estimate of the actual effective tax rate. In reaching this conclusion, Dr. Fairley relied specifically upon the calculations and tax rate arguments presented by Dr. Hill and Dr. Modigliani at the 1982 hearing.^[5] This estimate was consistent with the actual tax rates of ten insurance companies appearing in the Value Line Investment Survey. Dr. Fairley testified that the effective tax rates for all these companies were 20% or less. The Commissioner expressly relied on the testimony of this expert witness in choosing the 28% tax rate as the highest tax rate the model company would pay. This conclusion has reasonable support in the evidence. Moreover the evidence of the Value Line Investment Survey completely rebuts the plaintiffs' assertion that this assumed tax rate is "fanciful or impossible to match in the real world." Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 608, quoting Attorney Gen. v. Commissioner of Ins., supra at 822.^[6]

b. The surplus allocation model. As part of the calculation of projected investment income discussed above, the Commissioner must determine the amount of surplus capital which the model company will allocate to investment purposes. This year, the Commissioner has chosen to use the surplus allocation model proposed by the State Rating Bureau (SRB) at the hearing. This model is basically the same model that has been employed since the 1976 automobile insurance rate decision. This model allocates the surplus equally between the different insurance coverage lines, a process referred to by the parties as the surplus flow "block assumption." This results in different rates of investment return being assigned to the different insurance coverage lines. The bureau offered an alternative model which assigned different amounts of surplus to the different insurance lines. This approach, conversely, results in equal rates of return being assigned to the different lines. The Commission rejected both the bureau's model which altered the traditional block assumption and a third model offered by the Attorney General. The Commissioner concluded that neither of these alternative models presented an adequate theoretical or empirical basis for its assumptions and noted that this court has in the past approved this basic surplus allocation model. See Attorney Gen. v. Commissioner of Ins., supra at 819. The plaintiffs now argue that this choice constituted clear legal error.

The plaintiffs' novel argument begins with the uncontroverted principle that an investor in a regulated industry is entitled to receive a return on his investment commensurate with returns on investments in other enterprises having commensurate risks. FPC v. Hope Natural Gas Co., 320 U.S. 591, 603 (1944). See Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 339; Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 604-605; New England Tel. & Tel. Co. v. Department of Pub. Utils., 360 Mass. 443, 472 (1971). The plaintiffs then assert that the Commissioner, by using this surplus allocation model which uses a single average beta of liabilities for underwriting risk, found as a fact that the risk of underwriting insurance was the same for all insurance line coverages, an assertion that the Commissioner vigorously disputes. From this, the plaintiffs state, it follows that if the "risk" is the same for all coverages then the rate of return must also be the same in order to satisfy the Hope principle. They contend that since the surplus allocation model used by the Commissioner produces different rates of return for the different coverages it constitutes legal error for the Commissioner to use it, and they conclude the Commissioner must adopt the model they proposed since it is the only model that provides for equal rates of return for the different coverages.

In his brief for the Commissioner, the Attorney General responds that the bureau's argument is predicated on the mischaracterization of a theoretical working assumption used by the Commissioner in his calculations as a finding of fact by the Commissioner. The focus of the dispute between the parties centers on the Commissioner's use of a single average beta of liabilities in his calculations for the profits projected for each line. The beta of liabilities is used as a measure of how investors view the specific riskiness of underwriting automobile insurance as compared to investing in the average unregulated company. See Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 343. No direct method exists for calculating the beta of liabilities for a particular line of insurance. The indirect methods all rely primarily on computations based on the stock market performance of companies which write a variety of insurance lines. Since the actual beta for a specific line of insurance cannot be determined, an average beta is used. The use of the average beta of liabilities is based itself on an assumption, concededly not true, that the risk of underwriting is constant across all lines of coverage. Dr. Fairley testified that the assumption that "the risk of underwriting is a constant across lines," is a method for obtaining an estimate of average underwriting risk which can be used to calculate profits in the model.^[7] That the beta of liabilities is an assumed average of the true betas for each line causes no over-all error in the actual calculations.

The Attorney General argues that the use of this concededly imperfect working assumption about the beta of liabilities for the model does not amount to a finding that the assumption is true. Rather, the Attorney General states, the average beta of liabilities is just that, an average, not a finding that all individual betas of liabilities are equal. If the different betas of liabilities were in fact equal, it would not be necessary to calculate an average. The Attorney General concludes that since the plaintiffs' characterization of what the Commissioner actually determined is incorrect, the necessary premise for their invocation of the Hope principle, even assuming that this principle is applicable to this case, does not exist.

We agree with the Attorney General that the Commissioner made no finding that the risk is equal for all insurance coverage lines. Moreover, theoretical assumptions aside, the record is replete with uncontroverted testimony that the risk of underwriting varies among the different insurance lines. Dr. Richard Derrig, the industry's expert witness on the surplus flow issue, admitted that the bodily injury coverage is a high risk line while, by comparison, the risk for automobile property damage coverage is less than average. Richard Cohn, who helped develop the Myers-Cohn present value model used in the decision to calculate underwriting profits, also testified that the risk over time for underwriting bodily injury coverage is greater than the risk for underwriting automobile property damage. These conclusions, which are generally accepted in the industry, were supported by other witnesses at the hearing. In its own filing with the Commissioner, the bureau stated that: "Conventional wisdom offers the precept that longer-tailed insurance lines are riskier than shorter-tailed lines." This court has also recognized that different lines of insurance have greater risk than other lines. See Attorney Gen. v. Commissioner of Ins., supra at 817-818 ("Among these elements [of investment risk] is the risk inherent in the line of insurance itself; some lines will have greater unpredictability and fluctuation of losses than others and an investor in a company which wrote such lines would demand a greater expected return than he would in a company in which the return was more certain"). Assuming, arguendo, that the principle expounded in Hope Natural Gas is applicable to a comparison of the rates of return between insurance lines within the same regulated industry, a proposition not without doubt, this principle is not controlling here where the record amply demonstrates that the actual riskiness of investing in the various lines of insurance is not comparable. Therefore, the Commissioner's use of the surplus allocation model which results in a different rate of return was not legally erroneous and we defer to the Commissioner's experience and expertise in his choice of methodology. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 342.

3. Allowance for Losses.

The second major area of the Commissioner's decision challenged by the plaintiffs involves the Commissioner's provision in the rates for losses. The loss provisions established by the Commissioner are based primarily on the Commissioner's projection of the claim frequency changes and the claim cost changes from 1981 to 1983. A settled and direct methodology for determining claim cost trends has been developed in recent years. As a result, the parties have been in general agreement in this area. This year, the plaintiffs raise only a single cost claim issue: They dispute the validity of the Commissioner's assessment of the effect of the Hospital Cost Containment Act.

The Commissioner's estimation of claim frequency trends, on the other hand, continues to generate considerable controversy. This year the plaintiffs raise three challenges to this part of the Commissioner's decision. First, the plaintiffs object to the methodology used by the Commissioner for determining property damage liability claim frequency. Next, the plaintiffs dispute the Commissioner's conclusion that the putative "energy effect" on driving will lead to a continuing decline in claim frequency. Finally, the plaintiffs similarly challenge the Commissioner's projection of further claim frequency decline due to the new Massachusetts law focusing on the problem of drunk driving, in combination with changes in the merit rating surcharges for repeat offenders of the law. We turn first to these claim frequency issues.

a. Property damage liability claim frequency. In attempting to estimate claim frequency changes, the Commissioner traditionally has relied on analysis of the effect of specified external phenomena that are likely to exert a measurable impact on claim frequency, rather than simply extrapolating from reported internal claim frequency data. We expressed general approval of this practice in our decision reviewing the 1976 rates. See Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 801 (1976) ("It seems to us significant that the Commissioner did not merely attempt to project the data, as did the parties; rather he reached for the underlying causal phenomena and considered what their continuing influence would be"). The basic assumption for this methodology has been that there is no underlying trend in claim frequency.

However, in determining the change in property damage liability claim frequency, the Commissioner concluded that exclusive use of this approach would be inappropriate. The Commissioner observed contrary to that basic assumption that the underlying trend for property damage liability claim frequency has been one of consistent decline for the past four years. Since, contrary to the methodology's assumption, property damage liability claim frequency has demonstrated an underlying trend, the Commissioner decided to reject the plaintiffs' proposal that he base any adjustments in property damage liability claim frequency exclusively on observable external phenomena. Instead, he accepted the SRB's recommendation that he consider both internal statistical data and external indications for determining this frequency. He also adopted the projected change in claim frequency calculated by the SRB.

In projecting the claim frequency change, the SRB incorporated the observed decline in property damage liability claim frequency shown by internal industry data. The SRB attributed this decline principally to two external factors: the change in vehicle-miles travelled, and the increase to $1,000 in merit rating surcharges combined with the shortening of the period in which these payments must be made from three years to one year. The SRB projected a 2% decrease in claim frequency due to a change in vehicle-miles travelled and a 3% decline due to the effect of the increased merit rating surcharges. The Commissioner noted that the suggested percentage decline in claim frequency resulting from changes in the merit rating surcharge is consistent with the decrease he has attributed to the program in recent years.

The plaintiffs raise several objections to this procedure and the Commissioner's conclusion. The plaintiffs first contend that in projecting the decrease in property damage liability claim frequency, the Commissioner simply extrapolated from internal claim frequency data and did not provide the necessary causal analysis. However, as noted above, it is clear that the Commissioner, in addition to considering the unique internal trend data on property damage liability claim frequency, focused on two significant external causes for this trend, the change in vehicle-miles travelled and the changes in the merit rating program.

The plaintiffs also argue that this method of predicting future property damage liability increase is inconsistent with the methodology the Commissioner has used in past years. While the Attorney General, in his brief for the Commissioner, concedes that this linking of internal data with analysis of external cause reflects a modification of past methodology, he asserts that this modification is proper for forecasting property damage liability claim frequency. As the Commissioner explained, he focused initially on the internal claim frequency due to the unique trend displayed by these data. Although we have intimated that simple extrapolation of internal data is not preferred, see Attorney Gen. v. Commissioner of Ins., supra at 801, the Commissioner's analysis included consideration of external causes. Moreover, as we have previously recognized, in matters such as choice of methodology, we give due weight to the Commissioner's expertise, experience, and legislatively conferred discretion. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 342.

Finally, the plaintiffs dispute the Commissioner's assumption that the increases in the merit surcharges will have an effect on driving behavior and they argue that the evidence presented by the SRB supporting a merit rating effect was insubstantial. However, we do not weigh the evidence de novo. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 337. The Commissioner considered the evidence presented of a deterrent effect resulting from the merit rating surcharges and concluded that a 3% decline for property damage liability claim frequency was reasonable and was consistent with the decrease in claim frequency previously attributed to the merit rating surcharges. There was the requisite substantial support in the evidence for the Commissioner to accept the frequency adjustments suggested by the SRB. See id. at 338-339; Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. 592, 600-601 (1980).

b. Energy effect. Since the 1980 rate hearings the Commissioner has included an analysis of the effect of gasoline availability and price on insurance claim frequency trends. This impact is referred to by the parties as the "energy effect" on claim frequency. In measuring the energy effect in this year's automobile rate hearings, the Commissioner adopted a model proposed by the SRB and rejected a model offered by the bureau. Using this methodology, the Commissioner projected a 2% decrease in claim frequency for 1983. The plaintiffs argue that the Commissioner erred in not accepting their model and that the Commissioner's projection of claim frequency trends is not supported by the evidence.

The model proposed by the plaintiffs focused primarily on projections of increased automobile fleet fuel efficiency. Using historical data of miles per gallon fuel efficiency achieved in Massachusetts in combination with a national miles per gallon fuel efficiency forecast by a United States Department of Energy model, the plaintiffs' model estimated that fuel efficiency would increase from between 10% and 12% from 1981 to 1983. From these calculations, the plaintiffs suggested that 11% would be the best estimate of the improvement in fuel efficiency for the period and that this would result in an 11% increase in vehicle-miles travelled. The plaintiffs estimated from this projected figure that claim frequency would increase by 1%.

The SRB and the plaintiffs agreed that fuel consumption would remain stable, that the exposure change (i.e., the number of vehicles insured) would increase 5%, and that the elasticity of accidents on vehicle-miles travelled was .55, but they disagreed on the projection of the change in the total vehicle-miles travelled, and, implicitly, the increase in fuel efficiency. The SRB used a regression line to analyze the relationship between gasoline consumption and vehiclemiles travelled and to project the total vehicle-miles travelled in 1983. This regression line incorporates actual Massachusetts experience during the past six years. Unlike the model used by the bureau, the model used by the SRB does not require an explicit estimate of fleet fuel efficiency, a problematic estimate which the Commissioner found has resulted in serious errors in the recent projections by the bureau of total vehicle-miles travelled. By contrast, the Commissioner noted that the model employed by the SRB has proved remarkably accurate, predicting the change in vehicle-miles travelled within .1% in 1981.

Additionally, the Commissioner noted that in calculating gasoline consumption, the model proposed by the bureau uses numerous independent variables such as gasoline price, personal and family income, inflation, population density, and the number of households. None of these values is known to a certainty. The uncertainty in the estimates of these variables reduces the reliability of the bureau's results. Based on this record, we cannot say that the Commissioner's choice of methodology was erroneous. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. 333, 342 (1981). There is reasonable support in the evidence for the Commissioner's projection of the energy effect on claim frequency.

c. Changes in sanctions for driving under the influence of alcohol. Two significant events occurred in Massachusetts in 1982 designed to increase the deterrent effect of sanctions against drivers who operate automobiles under the influence of alcohol. First, the Legislature enacted St. 1982, c. 373, entitled, "An Act increasing the penalties for operating a motor vehicle while driving under the influence of intoxicating liquors." This law, in essential part, provides for increased penal sanctions for those convicted under the statute and reduces judicial discretion in sentencing. Second, the Commissioner increased to $1,000 the merit rating insurance surcharges imposed on individuals convicted for the second or third time of the offense of driving under the influence of alcohol. Additionally, this payment must be made in the first year, rather than being paid over a three-year period as previously had been allowed.

The SRB introduced a methodology at the hearing which attempted to predict the impact these changes would exert on bodily injury and property damage liability claim frequency. As a means of determining the likely effect of these changes, the SRB reviewed the results of two analogous programs designed to decrease the incidence of drunk driving. The SRB's first source was the experience of the Alcohol Safety Action Programs (ASAP) conducted in several locations in the United States during the past decade. The data on the effect of these programs were varied, but many programs showed a decrease in offenses ranging from 15% to 25%. From these results, the SRB concluded that an effectively implemented program directed at deterring drunk driving could produce a similar decrease in such offenses.

The SRB also studied the effects of the recent enactment by the State of Maine of a drunk driving law which requires a mandatory jail sentence of one year for first offenders convicted of driving under the influence of alcohol. The study revealed that during approximately the first year, the number of fatal alcohol related accidents in Maine declined by 35%. The SRB noted that, unlike the Maine law, the Massachusetts law does not require mandatory jail sentencing for first offenders. However, based on the results of the Maine law in its first year, and the evidence of the ASAP studies, the SRB estimated that the deterrent effect of the new Massachusetts law and increased merit rating surcharges would lead to a minimum 10% reduction in driving under the influence offenses in Massachusetts. Using statistics reported by the National Highway Traffic Safety Administration on the frequency of each type of accident due to the effects of driving under the influence of alcohol, the SRB projected a claim frequency decrease in bodily injury coverage of 1.1% and property damage liability coverage of .5% based on the anticipated 10% decline in drunk driving offenses during 1983. The Commissioner adopted these SRB projections and rejected criticism by the bureau of the SRB's analysis. The plaintiffs renew their arguments here.

The plaintiffs disagree with the major premise of the SRB's analysis that the new law and increased merit rating surcharges will have a deterrent effect. They assert that the Massachusetts law is not sufficiently similar to the law in Maine or the ASAP's results to justify extrapolating future Massachusetts results from this evidence. The plaintiffs claim that, unlike these programs, the Massachusetts law does not provide for increased enforcement of the law, and they assert that this fact lessens any supposed deterrent impact. In sum, the plaintiffs conclude that the anticipated change in claim frequency due to these measures is too speculative to warrant use in this year's rates.

After reviewing the evidence and the arguments of both the plaintiffs and the SRB, the Commissioner concluded that the increased sanctions against drunk driving will have some impact on the problem. Although the use of evidence gleaned from other jurisdictions' programs, which are admittedly different in degree and kind from the changes made in Massachusetts, is not completely satisfactory, we cannot say that the Commissioner's assumption does not have reasonable support in the evidence. The plaintiffs' criticism, renewed here, of the SRB's methodology and evidence were considered by the Commissioner and rejected by him. As we noted above, we do not substitute our judgment for that of the Commissioner. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 337.

d. The effect of the Hospital Cost Containment Act. The Commissioner accepted the SRB's estimate that the Hospital Cost Containment Act (Act), would result in a 2% reduction in hospital costs over a six-year period. Recognizing that 2% would not be a flat reduction in cost but would start at a lower level, the SRB then reduced its estimate of first-year savings to 1% and reasoned that the effect on automobile coverage for 1983 would be a .7% reduction to the medical care components. Both the SRB and the Commissioner premised their estimates on the estimate of an unidentified sponsor of the Act that it would result in an overall savings in hospital costs of $500,000,000 over six years. Nowhere in the record do we find an explanation by the Commissioner, the SRB, or the "legislative sponsor" of the basis for this estimate. At the hearings, the Commissioner characterized the estimate as the "best guesstimate" of the author or of the committee as it was prepared in the last moments of the legislative process to expedite the passage of the bill. The actuary who prepared this portion of the study for the SRB testified that he did not know how the sponsor of the legislation derived this estimate, that it was really too early to be able to tell what the full impact of the bill would be, and that he based his opinion exclusively on the bill itself and the summary of the bill apparently prepared by the legislative sponsors. The exhibit he identified as the summary he relied upon does not state that the Act will reduce hospital costs $500,000,000 over the six-year period. It does state the Act will limit the growth of hospital costs by 7.5% over six years. The Commissioner in his opinion recognized that the effect of the Act would not immediately become known and in order for him to make this determination it was necessary for him to exercise "some judgment." It is one thing to exercise some judgment, and another to speculate as to the existence of some future fact with no explanation of how the exercise of judgment leads to the conclusion that cost reduction will occur in the future. A mere speculation concerning a matter of fact cannot form an adequate foundation for the Commissioner's decision. Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 805 (1976). On this record we conclude that reasonable support is lacking for the premise that the Act will result in a 1% reduction of hospital costs in the year 1983.

The record before us does not enable us to say what reduction, if any, will result from the Act. If after remand the Commissioner's decision contains any factor for reduction in rates arising from the Act it must contain an adequate foundation for that decision.

4. Confiscation.

The plaintiffs' final argument is that rates set by the Commissioner are unconstitutionally confiscatory. When a claim that the rates fixed are unconstitutionally confiscatory is made, our decisions direct us to make an independent review of the facts as well as our traditional review of legal issues. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 346. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 596. Attorney Gen. v. Commissioner of Ins., supra at 795 n. 4. Accord Boston Gas Co. v. Department of Pub. Utils., 368 Mass. 51, 54 (1975). However, such independent review of factual issues does not include the Commissioner's subsidiary findings, but extends only to the Commissioner's ultimate findings and conclusions. Consistent with traditional principles of administrative law, the Commissioner's decisions will be accorded the "presumptive validity" given to actions taken by an agency. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 596. The essentially legislative process of rate making must not "be usurped in the guise of independent review." Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. at 346. Moreover, the plaintiffs bear the burden of demonstrating that the rates fixed deprive them of the opportunity to achieve a fair return and that these insufficient returns are directly attributable to the inadequacy of the rates set and not the result of other factors. See Fitchburg Gas & Elec. Light Co. v. Department of Pub. Utils., 371 Mass. 881, 884 (1977).

On this record, the plaintiffs have failed to meet their burden of proving the rates are unconstitutionally confiscatory. Their arguments on this point consist mainly of a restatement of the specific individual contentions that they have previously advanced and which we have already considered and rejected in previous portions of this opinion. In resolving the plaintiffs' statutory challenges against them, we have necessarily determined that the rates established by the Commissioner satisfy the statutory requirements that they be "adequate, just, reasonable and nondiscriminatory." G.L.c. 175, § 113B. Perforce, such rates which meet this statutory standard cannot be unconstitutionally confiscatory. The plaintiffs' constitutional challenge, therefore, is without merit.

5. Conclusion.

These cases are remanded to the single justice who is to enter judgment reversing the decision of the Commissioner and directing him to fix and establish the 1983 automobile insurance rates, redetermining the allowance for losses in accordance with this opinion.

So ordered.

NOTES

[1] We shall discuss the appropriate standard of review for the plaintiffs' constitutional claim of confiscation when we consider that specific challenge, infra.

[2] We have noted in past decisions that the rate of investment returns to actual insurers has been greater than the income attributed to them by the model. Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. 333, 340 (1981). Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. 592, 608 (1980).

[3] The plaintiffs assert that this modification of the model results in a change of 5.3% in the over-all rate level which will cost the companies approximately $61,000,000.

[4] The parties stipulated that the underwriting profit provisions established for the various coverages are: (1) negative 22.3% for bodily injury coverages, (2) negative 3.6% for property damage liability, and (3) positive .4% for physical damage coverages.

[5] The Commissioner agreed to take official notice of the 1982 hearing record to the extent it was referred to by witnesses in this year's hearings.

[6] We also note that in the 1980 rate hearing the bureau projected an average tax rate of 22% for its companies, Massachusetts Auto. Rating & Accident Prevention Bureau v. Commissioner of Ins., 381 Mass. at 607-608, and that in the 1981 rate hearing the bureau offered 19.28% as an estimate of the actual investment tax rate for the company in its proposed profit model.

[7] The plaintiffs concede in their brief that reliance on this working assumption "is admittedly a result of the inability of any party to suggest a way to measure beta by coverage."