Opinion
May 9, 1990.
August 16, 1990.
Present: LIACOS, C.J., ABRAMS, NOLAN, O'CONNOR, GREANEY, JJ.
Insurance, Motor vehicle insurance. Commissioner of Insurance, Rating, Practice, Civil, Review of action of Commissioner of Insurance. Administrative Law, Hearing, Standard of proof, Findings, Substantial evidence, Judicial review. Evidence. Administrative proceeding.
In establishing private passenger automobile insurance rates for 1990, the Commissioner of Insurance was not required by G.L.c. 175, § 113B, as amended by St. 1988, c. 273, § 35, to abandon certain methodologies adopted for use in setting 1989 rates unless he made new findings based on substantial evidence justifying deviation from a methodology based on internal data generated by the insurance industry. [371-374] In proceedings before the Commissioner of Insurance to establish private passenger automobile rates for 1990, substantial evidence supported the commissioner's decisions to adopt methodologies which did not rely exclusively on internal data generated by the insurance industry, to utilize the same methodologies for setting 1990 rates as used for 1989 rates, and not to utilize the methodology proposed by a certain representative of the insurance industry. [374-376] In establishing private passenger automobile insurance rates for 1990, the Commissioner of Insurance was not required by G.L.c. 175, § 113B, as amended by St. 1988. c. 273, § 35, to justify any deviation between the trend and projection factors produced by the methodologies he adopted and the factors produced by the separate methodology proposed by the insurance industry's representative; furthermore, apart from § 35, the evidence recited provided reasonable support for the commissioner's decision to utilize only two years of internal data generated by the insurance industry rather than six years of such data as provided by the industry's representative.[376-379] On the record of a proceeding before the Commissioner of Insurance to establish private passenger automobile insurance rates for 1990, there was no basis for this court to disturb the commissioner's assessment of either the probative worth of the predictive accuracy data provided by the insurance industry's representative, or of the likelihood that his new methodologies would perpetuate inadequate rates. [379-380]
CIVIL ACTION commenced in the Supreme Judicial Court for the county of Suffolk on January 3, 1990.
The case was reported by Greaney, J.
Stephen B. Deutsch ( Karen G. Wisbaum with him) for the plaintiff.
Thomas A. Barnico, Assistant Attorney General, for the Commissioner of Insurance.
On joint motion of the parties, a single justice of this court reserved and reported without decision the petition by Aetna Casualty and Surety Company (Aetna) for judicial review of the decision by the Commissioner of Insurance (commissioner) which fixed and established private passenger automobile insurance rates for 1990. G.L.c. 175, § 113B (1988 ed.).
The principal issue raised by Aetna is whether the commissioner, in establishing rates reflecting an average increase of 7.9%, complied with an amendment to G.L.c. 175, § 113B, inserted by St. 1988, c. 273, § 35. In setting rates for 1990, the commissioner adopted claims frequency and cost trend and projection factors which differed from those factors that he found would be derived from exclusive reliance on recent internal data generated by the insurance industry.
Automobile insurance "rates set by the commissioner are comprised of three basic components: (1) the loss allowance (expenses from payment of claims); (2) the expense allowance (nonclaim expenses); and (3) the underwriting profit allowance (an estimate of profit solely from the business of selling insurance)." Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 284 (1987). Aetna challenges only the claims frequency and claims cost projection factors adopted by the commissioner to arrive at the loss allowance component of the 1990 rates.
Aetna contends that the commissioner violated St. 1988, c. 273, § 35, by (a) improperly placing a burden on the insurance industry to justify the use of internal data; (b) failing to make findings and rulings based on substantial evidence justifying his decision not to rely exclusively on internal data; (c) rejecting the methodology proposed by the industry, which relied on six years of internal data; (d) allegedly "choosing a methodology which incorporated the same downward bias that plagued his former methodology and led to a decade of inadequate rates"; and (e) failing to give adequate consideration to predictive accuracy data presented by the industry. We reject these contentions and uphold the decision of the commissioner.
In the rate hearings, the insurance industry was represented by the Automobile Insurers Bureau of Massachusetts (AIB), an unincorporated voluntary association of insurance companies (including Aetna) licensed to write automobile insurance in Massachusetts. AIB formerly was known as the Massachusetts Automobile Rating and Accident Prevention Bureau. For convenience, we shall refer to this association as the AIB throughout this opinion.
1. Background of § 35. Section 35 was enacted following the decision by the commissioner in March, 1988, setting automobile insurance rates for 1987 and 1988. That decision came about by virtue of orders of this court and by a single justice directing that the commissioner reconsider his earlier decisions on 1987 and 1988 rates in light of our decision in Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282 (1987). See Attorney Gen. v. Commissioner of Ins., 403 Mass. 370, 371-372 (1988) (upholding commissioner's March, 1988, decision on remand). In his March, 1988, decision, the commissioner concluded that the rates previously set for 1985-1988 contained inadequate loss allowances. In initially setting such rates, the commissioner had not considered, or had excluded evidence of, the discrepancy between prior predicted loss severity and frequency for a given year and the industry's actual experienced loss severity and frequency for that year. See Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 293-297 (1987). To remedy the inadequate loss allowances for 1987 and 1988, the commissioner granted rate increases of 8.3% for 1987 and 7.7% for 1988 (in addition to increases already established in the original 1987 and 1988 rate proceedings). Attorney Gen. v. Commissioner of Ins., supra at 372. In his March, 1988, decision the commissioner announced that "[a] full-scale examination into our loss trending methodology will be undertaken in proceedings to set 1989 rates."
On November 5, 1988, while proceedings on the 1989 rates were under way, the Governor signed St. 1988, c. 273. Section 1 stated that "the purpose of this act is to restructure the automobile insurance system so as to provide residents of the commonwealth with effective automobile insurance protection at rates which are adequate, just, reasonable and nondiscriminatory." One aspect of this restructuring was contained in § 35 of the act, set forth in full in the margin. The principal effect of § 35 is to require the commissioner explicitly to consider recent internal data, to make findings as to the claim severity and frequency trend and projection factors indicated by such data, and to justify any material deviation between those factors and the factors ultimately adopted.
Statute 1988, c. 273, § 35, reads: "In setting the claims frequency and cost trend and projection factors used to fix and establish classifications of risks and premium charges, the commissioner shall explicitly consider recent actual Massachusetts claims frequency, cost trend and loss data, and shall make express findings as to the claims frequency and cost trend and projection factors which such data would indicate for the respective coverages. He shall also consider such other evidence, argument and considerations as he finds credible and relevant. He shall justify the claims frequency and cost trend and projection factors which he uses with specific findings of fact and conclusions of law regarding all disputed material issues, and if the claims frequency, cost trend and projection factors which he uses materially deviate from the claims frequency, cost trend and projection factors derived from such recent Massachusetts data, he shall explicitly set forth the reasons therefor by making specific findings of fact and rulings, which shall justify such deviation and which shall be based on substantial evidence. For the purpose of evaluating any methodology proposed by a party to the rate hearing to be used for trending or projecting claims frequency or costs in setting the premium charges for the rate year in issue, if the commissioner makes specific findings that the same methodology was used in fixing and establishing premium charges in prior years in Massachusetts, a party may introduce into evidence the actual results caused by the use of that methodology in prior years."
Section 77 of c. 273 provided that § 35 would take effect on January 1, 1989. Other sections of c. 273 raised the tort threshold to $2,000 (§ 55), raised personal injury protection benefits to $8,000 (§ 15), made underinsurance coverage optional (§ 46), changed the "trigger" for compulsory and noncompulsory underinsurance coverages (§ 46), and prohibited "stacking" underinsurance coverages in separate policies (§ 47). See generally St. 1988, c. 273, § 1.
In December, 1988, the commissioner issued his decision on 1989 automobile insurance rates. The commissioner noted that one of the purposes of the hearings on 1989 rates was to select a method to replace the traditional loss methodology that had been utilized in setting rates for 1985 and subsequent years. The commissioner stated that he viewed his task as determining which of the methodologies proposed by the parties would provide the most reasonable and accurate prediction of 1989 loss costs. The commissioner evaluated each proposed methodology on a coverage-by-coverage basis for predictive accuracy with respect to both claim cost and claim frequency. He made numerous findings bearing on which aspects of the competing proposals were likely to produce accurate loss cost projections. As a result of this evaluation, the commissioner adopted methodologies for determining trend and projection factors for claim cost and claim frequency which, with respect to many coverages, differed from the methodologies previously utilized. The commissioner set forth his reasons for his methodological choices in considerable detail. The commissioner rejected the methodologies proposed by the AIB. No one sought judicial review of the commissioner's decision setting rates for 1989.
The commissioner described this traditional loss methodology as follows: "In setting the rates over the years, the Division has consistently projected loss costs by separately examining claim cost data and claim frequency data and setting a claim cost trend and projection factor as well as a claim frequency trend and projection factor for each coverage. To determine the claim cost factors for the major coverages — BI, collision, and PDL — the Division has for the past eight years relied on a combination of internal data — that is, Massachusetts automobile insurance claims experience — and external data consisting of a set of inflation indices selected to collectively represent the components of insurance claim costs. For the past five years, the Division has attributed 60-percent weight to the external indices and 40 percent to internal data. For the A-2, D and Comprehensive coverages, only external data have been used. Trends in the external indices and internal data have consistently been projected using a linear regression methodology.
"Characterizing the determination of the claim frequency factors over the last decade has been a heavy reliance on the null hypothesis — that is, the principle that no change will be forecast for claim frequencies unless there is compelling evidence that claim frequencies in the policy year will be higher or lower than in the base year. Certain external influences which were expected to affect frequency during the trend and projection period were recognized. In addition, various parties have recommended, and the Division has on occasion agreed, that internal historical changes in claim frequency which could be expected to continue into the trend and projection period should be taken into account, notwithstanding the presumption of the null hypothesis."
During 1989, hearings were held for the purpose of setting automobile insurance rates for 1990. Three parties participated in the portion of the main rate hearing involving loss trending and projection factors: the AIB, the State Rating Bureau (SRB) and the Attorney General. The AIB proposed the same loss cost and frequency trending methodology that it had proposed at the conclusion of the hearing on 1989 rates. For present purposes, the AIB's proposed methodology may be characterized as relying exclusively on six years of internal data, weighting the five percentage changes from year to year, using 30-25-20-15-10 weights, with the most recent data weighted most heavily. The SRB proposed use of the methodologies adopted by the commissioner in his decision on 1989 rates. The Attorney General proposed methodologies which, with minor modifications, were the same as those he proposed at the hearing to set 1989 rates.
The parties, including the AIB, treated the hearing as calling for the commissioner to choose among the parties' competing proposals. Indeed, they stipulated that that was the only issue concerning loss trending which the commissioner needed to address in his decision. No issue was identified under § 35 or regarding the use of claim cost and frequency trend and projection factors indicated by "recent" internal data.
There is no indication that Aetna independently sought to intervene in the hearings. Compare Attorney Gen. v. Commissioner of Ins., 403 Mass. 370, 379-380 (1988) (leaving open the question whether an individual insurance company has a right to intervene in such rate hearings). Cf. 211 Code Mass. Regs. § 77.04 (2) (petitions to participate as "an interested party"); § 77.05 (1) (a) (insurers' advisory filings); § 77.05 (2) ("any insurer" may file supplement to advisory filing).
The hearings culminated in the commissioner's December 15, 1989, decision fixing and establishing private passenger motor vehicle insurance rates for policy year 1990. The commissioner adopted the same loss cost and frequency trending methodologies he had used to set 1989 rates.
The commissioner described the "1989 Decision claim cost projection methodology" as follows:
"For the major coverages — A-1 and B, Collision, and PDL — the Division relied on a combination of internal data, that is, Massachusetts automobile insurance claims experience, and external data, consisting of a set of inflation indices selected to collectively present the components of insurance claim costs. The Division attributed 60 percent weight to the external indices and 40 percent weight to the internal indices, based on eight quarters of data and projected by linear regressions. For the Limited Collision coverage, the Division used the Collision severity trend factor. For Coverage U, the Division used the A-1 and B severity trend factor. For the A-2 and D coverages, the Division attributed 100 percent weight to the external indices. For Comprehensive, the Division attributed 60 percent weight to the external severity data and 40 percent weight to the internal pure premium data."
He described the "1989 Decision claim frequency projection methodology" as follows:
"For the major coverages — A-1 and B, collision and PDL — the Division assigned one-half weight to the projected claim frequency for internal data based on a linear trend of the most recent eight quarters of frequency data and one-half weight to the current base-year frequencies. For the Limited Collision coverage, the Division used the Collision frequency trend factor. For Coverages A-2, D and U, the Division used the A-1 and B Frequency trend factor. For Comprehensive, since the Division attributed 60 percent weight to the external severity data and 40 percent weight to the internal pure premium data for determining a claim cost trend and projection factor, it attributed a unity factor for the claim frequency trend and projection factor."
The commissioner noted that the AIB had not presented any new arguments or evidence (except for one additional data point for 1988) in support of its proposed methodology and had failed to address any of the reasons that its methodology was rejected in his decision on 1989 rates. He noted that the AIB's proposed methodology had been examined and evaluated thoroughly the year before and that the decision on 1989 rates explained in detail why that methodology had not been accepted. He concluded that "[o]n this record, there is nothing to support replacing the methodology adopted last year with AIB's recommended methodology this year." The commissioner then briefly reviewed the evidence offered by the AIB and restated the reasons the methodology had been rejected in the decision on 1989 rates. In general, except as noted below, the commissioner did not make new factual findings.
With respect to the use of external data in setting claim cost and frequency trending and projection factors, the commissioner noted that, in the decision on 1989 rates, he had "concluded that the external data provide important information about expected changes in the components of claim payments which cannot be ignored." He noted that the "reasons for relying on a blend of external and internal data included: [a] the external data used for trending are more recent; [b] the reliability of the internal data is questionable; [c] the internal data are subject to more distortion than the external data due to such things as database coding errors, changes in reserving practices, reserving estimation errors, and adjustments to place the Collision and Comprehensive data on a constant deductible basis; [d] the internal data are subject to a wide range of random fluctuation; [e] the accuracy of the internal trends is affected by any variations over time in the percentage of average claim costs attributable to excessive claim payments; [f] the internal data are affected by errors in estimating the impact of certain phenomena affecting claim costs; and [g] the internal trends are affected by unidentified transitory phenomena." The commissioner noted that the external indices used had been selected and relied upon in the past "because they are closely related to automobile insurance claim costs." The commissioner also cited testimony by two witness that use of the methodologies adopted for 1989 should produce unbiased rates for policy year 1990.
In his decision on 1990 rates, the commissioner stated that "external" data consist "of a set of inflation indices selected to collectively represent the components of insurance claim costs." Examples included consumer price indices produced by the bureau of labor statistics for all items, for medical care costs in Boston and for automobile maintenance and repair costs, and average weekly wage data compiled by the division of employment security. See Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 295 (1987) ("external indicators rely on hypothetical models and data from outside the insurance industry").
In his decision, the commissioner stated that "[t]his year, for the first time I must make findings under Section 35 of Chapter 273 of the Acts of 1988." Setting forth the claim cost and frequency projection factors which he found were indicated by the most recent eight quarters of internal data available, he compared those factors with the factors resulting from the methodologies he had decided to adopt. He observed that the claim cost factors he adopted "deviate very little from the internal claim cost factors derived from the recent Massachusetts data [and that the] claim frequency factors [he adopted] show a greater deviation from the internal factors derived from the recent Massachusetts data." To the extent that such deviation is material within the meaning of § 35, he stated that the deviation was necessary because of factors that result in distortions in the internal trends and because of the failure of any party to explain why unidentified historical phenomena affecting claim frequencies should be projected into the future.
2. Burden of proof and reliance on prior decision. Aetna argues (a) that the commissioner violated § 35 by placing a burden on the AIB to demonstrate that a methodology relying exclusively on internal data is superior to any other methodology; (b) that § 35 mandates that claim cost and frequency trend and projection factors be based exclusively on internal data unless the commissioner explicitly finds, based on substantial evidence, that the use of something other than internal data is justified; and (c) that the commissioner properly could not rely on his prior decision on 1989 rates — made before § 35 took effect and based on an "entirely separate evidentiary record" — to satisfy § 35's requirement that he make findings based on substantial evidence. In effect, Aetna argues that § 35 required the commissioner to abandon the methodologies adopted for use in setting 1989 rates unless he made new findings based on substantial evidence justifying deviation from an internal data methodology. We disagree.
Aetna concedes, however, that "much of the underlying data in the two records may well be the same." Aetna does not attempt to identify any significant differences in the two records. The record of the hearings to set rates for 1989 has not been made part to the record before this court.
Contrary to Aetna's contention, § 35 does not create a presumption mandating use of any methodology proposed by a party which relies exclusively on internal data, unless the commissioner justifies using a different methodology. Nowhere where in the language of the statute is there found any phrase giving presumptive effect to such data. The statute merely requires the commissioner to consider recent internal data. To the extent that § 35 creates anything akin to a "presumption," it does so only with respect to the use of trend and projection factors which the commissioner finds would be indicated by "recent actual Massachusetts . . . data."
Section 35 does not put a time frame on the phrase "recent actual Massachusetts . . . data." By failing to specify a time frame, the Legislature left the selection of an appropriate time frame to the judgment and discretion of the commissioner.
Here, as noted above, the commissioner made findings as to the factors that would be indicated by the most recent two years of internal data available. Aetna's complaint is that the commissioner did not utilize the AIB's proposed higher projection factors based on six years of internal data rather than two years. To the extent that the AIB's proposed projection factors differed from the factors which the commissioner found were indicated by recent internal data, the commissioner did not contravene § 35 by rejecting the AIB's proposed methodology based on the absence of evidence that that methodology was superior to the methodologies the commissioner decided to adopt.
The commissioner cannot be faulted for failing to make anew the findings he made and relied on one year earlier in developing new methodologies for calculating claim cost and frequency trend and projection factors. General Laws c. 175, § 113B, provided, both before and after amendment by St. 1988, c. 273, that the "commissioner may make . . . reasonable rules and regulations to facilitate the operation of this section . . . and to govern hearings . . . under this section." Pursuant to this authorization, the commissioner adopted regulations governing automobile insurance rate setting hearings. 211 Code Mass. Regs. § 77.00 (1987). One of those regulations provides that the presiding officer may direct the parties to confer to consider a simplification of the issues. 211 Code Mass. Regs. § 77.04 (4). In this case, at the request of the hearing officer, the parties stipulated to the issues requiring resolution by the commissioner in his decision. (See note 5, supra.) The parties did not identify any issue in dispute under § 35.
Moreover, it was not improper for the commissioner to rely on his decision on 1989 rates as precedent. Here, as the commissioner found, the methodologies proposed by the parties were substantially identical to the methodologies proposed or adopted one year earlier, and the evidence and arguments offered by the parties were also nearly identical to those offered the prior year. Where, as here, no contention is made that the same parties were not afforded an adequate opportunity to present evidence and arguments in the earlier proceeding or in the proceeding under review, we cannot say that the commissioner's implicit adoption of the findings he made in his decision on 1989 rates was impermissible. Cf. Zachs v. Department of Pub. Utils., 406 Mass. 217, 220-223 (1989) (department properly relied on nonadjudicative factual findings from a prior decision); Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 287 (1987) ("A party is entitled to 'reasoned consistency' in agency decisionmaking. Boston Gas Co. v. Department of Pub. Utils., 367 Mass. 92, 104 [1975]").
More importantly, Aetna makes no attempt to demonstrate that any significant new data or evidence was introduced in the hearings on 1990 rates that should have resulted in adoption of different methodologies for setting. 1990 rates. In the circumstances, no point would be served by remanding the case to the commissioner for an express restatement of the findings he implicitly adopted from his decision on 1989 rates. Cf. Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 824 n. 40 (1976).
Additionally, as we conclude in part 3 of this opinion, the commissioner's choice of methodology was supported by substantial evidence.
3. Substantial evidence. Aetna argues that, even if it was proper for the commissioner to rely on findings made in his decision on 1989 rates, those findings and the additional findings in his decision setting 1990 rates were not based on substantial evidence. Aetna concedes that the internal data prepared by insurance companies contain errors and random variations and appear to reflect unidentified transitory phenomena. Aetna contends, however, that the commissioner failed to demonstrate that any resulting distortion or lack of reliability of the internal data would have a material effect on the accuracy of the trend and projection factors that would be derived from those data. Here, too, Aetna relies largely on its contention that § 35 creates a mandatory preference for the use of any methodology relying exclusively on internal data. As noted above, this contention misconstrues § 35. In addition, the unreliability of the internal data was only one of the reasons articulated by the commissioner for rejecting exclusive reliance on internal data. Further, we have stated that (apart from § 35) the commissioner need not make findings on every controverted issue of fact or law so long as his findings indicate the over-all basis of his decision and permit effective appellate review. Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 292 (1987).
We conclude that the testimony and evidence recounted in the commissioner's decision on 1989 rates provided more than reasonable support for the commissioner's decision to adopt methodologies which did not rely exclusively on internal data. See id. at 296-297 (upholding commissioner's reliance on external indicators). In addition, the record of the hearing on 1990 rates contains testimony which independently provides reasonable evidentiary support for his decision to utilize the same methodologies for 1990 rates as used for 1989 rates. It lies within the discretion of the commissioner to determine the weight to be given to certain evidence supporting or detracting from reliance on particular methodologies or data. E.g., Attorney Gen. v. Commissioner of Ins., 403 Mass. 370, 377-378 (1988).
In his decision on 1989 rates, the commissioner cited testimony regarding various types of errors in the internal data, some of which had not been detected by the "supposedly rigorous edit and audit procedures used by [Commonwealth Automobile Reinsurers] and [AIB]." He found that internal data were subject to a wide range of random fluctuation, and that external data are in general more stable and are distorted much less as a result of random variation than internal data. He noted testimony that external data were more recent and were more accurate and responsive for purposes of trend projection than internal data. He noted that, even if internal data were perfect, "accurate prediction of future loss costs depends on much more than simply trending forward past loss costs." He concluded that the "external data provide important information about expected changes in the components of claim payments which cannot be ignored." He also addressed whether reliance on external data would "perpetuate the systematic mispredictions of loss costs caused by the traditional loss methodology as a whole over the years." After citing and analyzing the evidence, the commissioner concluded that it would not.
With respect to claim frequency trend and projection factors, the commissioner similarly discussed the evidence and evaluated each party's proposed methodology. He noted testimony that distorted internal claim frequency data were distorted by the same factors that distorted internal claim severity data. He discussed historical data indicating that 100% reliance on internal claim frequency data would produce a frequency trend factor that would be expected to be too high for 1989. The commissioner found that giving only half-weight to internal data produced claim trend factors that were reasonable when evaluated against actual experience.
For example, the record of the hearings on 1990 rates contains testimony regarding the unreliability of internal data, the continuous process of correcting such data, and the fact that such data can be affected by changes in a company's reserving practices and "randomness." There was testimony that, in view of such unreliability, it would be "foolhardy" to place 100% reliance on such internal data, and that such internal data contain "transitory type effects" of a magnitude not found in external indices. There was also testimony that the AIB's so-called predictive accuracy data were essentially the same as submitted the prior year, did not directly test the accuracy of the methodology adopted by the commissioner for 1989, and did not shed light on what elements of the prior methodologies contributed to prior prediction errors.
Last, Aetna argues that the commissioner's decision not to utilize the AIB's proposed methodology was not based on substantial evidence. We observe that the AIB proposed its methodology on the theory that, "if used consistently over time," it would result in rates that were inherently unbiased. The AIB did not claim, or seek to show, that its methodology would be accurate for any given year. Rather, as its witness testified during the hearing on 1990 rates, if used consistently over time, "eventually" the inevitable positive and negative errors produced by use of its methodology for specific years would cancel each other out.
As the commissioner noted, however, his obligation in setting rates is, as stated in G.L.c. 175, § 113B, to set "adequate, just, reasonable and nondiscriminatory premium charges to be used . . . in connection with the issue . . . of . . . policies . . . for the ensuring calendar year" (emphasis added). See Medical Malpractice Joint Underwriting Ass'n v. Commissioner of Ins., 395 Mass. 43, 52 (1985) (under G.L.c. 175, § 113B, "rates must be adequate for each rating period"). Thus, even if over a number of years rates set by used of the AIB's proposed methodology would average out to be accurate, the rate thereby set for a given policy year would nevertheless be subject to challenge if it could be demonstrated that it did not meet statutory standards for that specific policy year. In the circumstances, the commissioner properly could conclude that, by ignoring the principle that the rates set by the commissioner must be "adequate, just, reasonable and nondiscriminatory" for each policy year, the theoretical foundation of the AIB's proposed methodology was flawed. Cf. Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 296-297 (1987) (upholding commissioner's choice of methodology based on considerations of predictive accuracy).
4. Failure to justify deviations. Aetna also argues that § 35 required the commissioner to justify any deviation between the trend and projection factors produced by the methodologies he adopted and the factors produced by the separate methodology proposed by the AIB. Here, too, Aetna misreads § 35. The obligation to explain and justify deviations imposed by § 35 applies only to material deviations between the factors the commissioner uses and the factors he finds to be indicated by recent Massachusetts data.
Aetna notes that, with respect to claim frequently, the factors produced by the AIB's proposed methodology were roughly 10% higher, on average, than either the factors the commissioner found to be indicated by recent Massachusetts data or the factors produced by the methodologies adopted by the commissioner. Relying solely on § 35, Aetna faults the commissioner for failing to "explain why two years of internal data provides [ sic] a more appropriate basis for deriving internal claim frequency factors than would a longer period," such as the six years of internal data relied on by the AIB.
In his decision on 1990 rates, the commissioner referred to the explanations in his decision on 1989 rates for rejecting the AIB's proposed methodology. That decision reveals the basis for the commissioner's determination that two years of internal data provide a more appropriate basis for deriving internal claim frequency factors than would the longer period proposed by the AIB. In his decision on 1989 rates, the commissioner cited testimony that the six years of internal data the AIB utilized in its methodology embraced the interval 1983-1985, during which the industry experienced the largest two-year increase in pure premiums for most coverages since 1977. The commissioner characterized the 23.5% increase in pure premium from 1984 to 1985 as aberrant and observed that its use to arrive at trend factors could be expected to result in overprediction of future experience. The commissioner found valid a witness's criticism that, as a result, the AIB's methodology "produced trends . . . that are higher than recent experience — in some cases, much higher — and that more closely resemble the . . . trends observed in the more distant past." The commissioner cited testimony that data from the period 1982-1987 were "very stale" and that the AIB's choice of time period gave weight to older trends resulting in projections that were biased upward. He noted that the traditional methodology had used the most recent eight quarters of data for calculating severity trends. He also noted that the AIB had stressed the importance of using the same experience period for severity and frequency trends due to their inherently interdependent nature and that the SRB had noted that use of the same experience period for severity and frequency trends would eliminate certain distortions in the frequency trends.
The commissioner observed that for 1989 rates the AIB's proposed methodology would give 20% weight to the probability that such a large increase would repeat itself throughout the projection period.
Section 35, by its terms, does not impose on the commissioner any obligation to make findings and rulings on this particular question unless it was a "disputed material issue," which it was not. We observe, however, that, apart from § 35, the commissioner's decisions regarding choice of methodology are subject to judicial review under G.L.c. 175, § 113B, under standards which have been articulated in numerous decisions. Under such standards, our inquiry, "is limited to 'whether the rates have reasonable support in evidence.' Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. 333, 337 (1981).[] The commissioner is required to make findings which indicate the over-all basis for his decision. Id. We will not perform a de novo weighing of the evidence, and we give due weight to the commissioner's experience, technical competence, specialized knowledge, and discretionary authority. Id. See also Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 389 Mass. 824, 828 (1983)." Attorney Gen. v. Commissioner of Ins., 403 Mass. 370, 376 (1988). We have indicated that such judicial deference to the commissioner's experience, expertise, and discretion applies particularly to his choice of methodology. Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 389 Mass. 824, 837 (1983). Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 384 Mass. 333, 341-343 (1981). Cf. Attorney Gen. v. Department of Pub. Utils., 392 Mass. 262, 268 (1984). The evidence recited provides reasonable support for his decision to utilize only two years of internal data rather than six years of such data as proposed by the AIB. Cf. Attorney Gen. v. Commissioner of Ins., 370 Mass. 791, 799 (1976) (upholding choice of two-year average noting that, although an average over a longer period might "produce a result less subject to fluctuation, it might do so at the expense of using data that could be regarded as obsolescent or stale").
Section 35 requires that the commissioner's findings and rulings justifying a material deviation between the factors be uses and the factors indicated by recent actual Massachusetts data be based on "substantial evidence." We have noted that the substantial evidence standard is indistinguishable from the supported by reasonable evidence standard employed in reviewing the commissioner's rate setting decisions under G.L.c. 175, § 113B. Medical Malpractice Joint Underwriting Ass'n v. Commissioner of Ins., 395 Mass. 43, 54 (1985).
5. Bias and predictive accuracy. Aetna contends that the commissioner failed to demonstrate "that in choosing his new methodology he has avoided the downward bias which infected the former methodology and led to the catastrophic ten year record of losses which caused the current insurance crisis." Aetna also argues that the commissioner failed adequately to take into account certain predictive accuracy data indicating the relative track records of the methodologies proposed by the AIB, the Attorney General and the SRB, and the methodologies used by the commissioner for policy years 1978-1988.
Aetna concedes, as it must, that the methodologies adopted by the commissioner for 1989 and 1990 were not the same as those used in prior years. We have noted that evidence of predictive accuracy of the type offered by the AIB is "relevan[t] only if it [can] be shown that the same factors which led to inadequacy in the previous years' predictions were present in the current calculation." Massachusetts Auto. Rating Accident Prevention Bureau v. Commissioner of Ins., 401 Mass. 282, 294 (1987). The mere fact that the traditional methodologies employed before policy year 1989 and the new methodologies employed in setting 1989 and 1990 rates relied to some extent on external data (in setting claim cost factors) and on the "null hypothesis" (in setting claim frequency factors) and did not rely exclusively on internal data is not enough to show that the same factors which led to the inadequacy in the previous years' predictions were present in the current calculations. As the commissioner noted, the AIB's predictive accuracy data did not consider, or shed light on, the accuracy of any specific components of the methodologies used in the past, some but not all of which were used in setting 1989 and 1990 rates. In the absence of a showing that the same factors which led to the inadequacy in the previous years' predictions were present in the current calculation, the commissioner was not required to place any particular weight on, or even to consider, the predictive accuracy data offered by the AIB. Id. at 294, 297-298. Cf. § 35, last sentence. In addition, as noted above, the commissioner cited testimony from the hearing on 1990 rates that the claim cost and claim frequency methodologies adopted to set 1989 rates should produce unbiased rates for 1990. We find no basis for disturbing the commissioner's assessment of the probative worth of the AIB's predictive accuracy data or of the likelihood that his new methodologies would perpetuate inadequate rates.
6. Disposition. We remand this case to the county court for entry of a judgment affirming the decision of the commissioner fixing and establishing private passenger automobile insurance rates for 1990.
So ordered.