Kaplan, J.

These cases question the legality of indus-trywide automobile insurance rates for the policy year 1976 set by the Commissioner of Insurance (Commissioner) under G. L. c. 175, § 113B. 2 The Commissioner commenced hearings on the bodily-injury insurance coverages on September 2, 1975, and issued ids decision establishing those rates on November 17, 1975. The Massachusetts Automobile Rating and Accident Prevention Bureau (Bureau), an unincorporated association of insurance companies writing motor vehicle insurance in the Commonwealth, together with various of the insurance companies, filed a petition for review of that decision in the Supreme Judicial Court for Suffolk County within the twenty-day period prescribed by G. L. c. 175, § 113B. The Attorney General also sought review. In the meantime, the Commissioner, who had recessed the hearings on other coverages to await action by the Legislature on changes in laws governing automobile insurance, on November 19, 1975, *795 resumed hearings on the property-damage rates. His decision issued on December 30, 1975, and again the Bureau, various companies, and the Attorney General sought review. In addition, the Bureau and the companies filed a complaint seeking a declaration as to the effect on the setting of the 1976 rates of St. 1975, c. 707, § 1A, which repealed a certain “second look” provision of G. L. c. 175, § 113B, inserted by St. 1971, c. 977, § 1A, described below. 3 The matters were reserved and reported by the single justice for decision by the full court upon the pleadings and the record of the hearings. 4

The rates set represent a forecast based on analysis of past experience. The several components of a rate are isolated and the contribution of each to the total premium is analyzed separately. Crudely divided, the total premium breaks up into components as follows: an allowance for losses (the portion of the premium uséd mostly for the payment of claims), an allowance for expenses, an allow- *796 anee for profit, and, at least with regard to certain of the 1976 rates, a possible allowance for the inadequacy of those rates for the previous year.

I. Allowance for Losses

A. The Formation of the Rate Level Pure Premium. Whereas the Commissioner innovated in his methods for establishing the allowances for expenses and profits, he used the conventional method for determining the loss allowance. The rate setter uses as his basis for prediction the most recent year for which reasonably complete claims data are available; in setting the 1976 rates, he used the claims experience for the 1974 policy year. The rate setter takes the total claims arising from 1974 policies plus certain expenses allocated to those claims (1974 “losses” in insurance terminology) and divides by the total insurance exposure, i.e., the number of car years for which policies were written. Thereby he derives the “1974 Raw Pure Premium.” This figure is the observed 1974 loss cost per vehicle. From it, the rate setter builds a parallel predicted figure for 1976 just adequate to cover the expected 1976 loss — the “1976 Rate Level Pure Premium.” He does this by making several adjustments in the 1974 Raw Pure Premium to take account of the fact that the data arising out of 1974 policies are not final, and the further fact that the 1976 experience will differ in predictable ways from that observed for the 1974 policy year. Each of the adjustments is expressed as a multiplicative factor.

(1) The development factor. The 1974 data are not final at the time of the rate setting process. Some claims have not yet been filed, and the loss adjuster’s estimate of the size of claims filed but not paid may prove to.be erroneous. The adjustment of the 1974 Raw Pure Premium to reflect expected final loss experience is made by means of a “development factor” which is estimated from the observed change during the course of a year of the losses attributed to other policy years. See Massachusetts Bonding & Ins. Co. v. Commissioner of Ins., 329 Mass. 265, 274 (1952).

*797 (2) Trend and projection factors. The losses on the 1976 policies will be paid, on average, later than those on the 1974 policies, and it can be expected that cost levels for the losses in the two years will differ, necessitating an adjustment for the change in those levels. The 1974 losses reflect experience for policies centered on July 1,1974, and the period for which rates are set centers on July 1,1976. The “trend” factor adjusts for observed changes in price levels from July, 1974, to a later point for which data are available, and the “projection” factor adjusts for any expected continued change in prices to July, 1976. 5

(3) Frequency adjustment. If the probability that an insured vehicle will be involved in a claim is expected to be different in 1976 than in 1974, a further adjustment is needed to reflect the difference.

Before getting into our analysis of the factors just mentioned, we note that the data for 1974 which form the basis for the estimation of the 1976 Rate Level Pure Premium were collected and summarized by the Bureau, and are not directly challenged here. The Bureau’s contention is that the Commissioner committed certain errors in fixing the adjustment factors in the light of these data and others. As the type of insurance coverage has a bearing on the estimation of the adjustment factors, we have to consider separately the bodily-injury and property-damage coverages.

B. The Bodily-Injury Coverages. (1) The development factor. The Bureau recommended a development factor of 1.085, based on an examination of the change in the course of one year (1974) of the losses attributed to a series of recent policy years. However, the actuary for the Division of Insurance (Division) observed that the development factor for changes in the losses in the previous year (1973) was much less. He considered the factor to *798 be a “predictor” which should not vary radically from one rate revision to the next; and he attributed the difference in the factors for the two years to “some form of variation in the data rather than to [the] intrinsic property of the losses themselves.” Accordingly, he recommended, as the factor for 1976, a “more stable” predictor, the average for the two years, 1.038. This recommendation the Commissioner adopted.

The Bureau argues, first, that the industry is entitled to consistency of approach, and hence the traditional procedure, using a factor derived from a single year’s observations, must be followed. See Boston Gas Co. v. Department of Pub. Util., 367 Mass. 92, 104 (1975) (condemning “decision according to the whim or caprice of the Department”). But the Commissioner gave a convincing reason for his departure from the tradition in citing the need for a less volatile predictor. Cf. id. at 104. Indeed the Bureau’s own actuary testified to the desirability of averaging in order to make the development factor more stable and accurate. 6

Second, the Bureau argues that if an averaging technique is to be used, the average should be extended over a longer period, and the development factor observed for 1973 should be rejected as “aberrant.” A longer-term average, dropping the low year 1973, would produce a larger factor. 7 But the Bureau presents no reasoned argument demonstrating error on the Commissioner’s part.

The 1973 data, which yield a development factor of 0.990, are viewed as “aberrant” merely because the figure is lower than the other numbers of the series. There is no sugges- *799 tian that the 1973 data are unreliable, and it would seem odd to reject the data from the one year in which the insurers’ estimates of their losses at the beginning of the year were nearly exactly correct as measured by the changes at the end of the year. As to the desirability of a longer-term average, the Bureau points to testimony of the Division’s actuary suggesting a five-year period. But the suggestion was quite tentative and was accompanied by the thought that the problem should be studied more fully before the next rate revision with an attempt to reach agreement on a stable development factor. We may note that if an average over a longer period would 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.

On the whole we cannot conclude that the choice of a two-year average was wrong. Our view is fortified somewhat by the fact that, in portions of the hearing dealing with other coverages, the Bureau itself proposed development factors based on two-year averages. See note 6 supra. Compare Massachusetts Bonding & Ins. Co. v. Commissioner of Ins., 329 Mass. 265, 274-277 (1952) (upholding Commissioner’s determination of development factor despite “impressive” evidence in support of alternative procedure) .

(2) Frequency adjustment. In its original presentation the Bureau argued that 1974, the basis year, was aberrant in respect to frequency because of the effects of the fuel shortage in reducing the use of cars and so the number of claims. Therefore the Bureau recommended a frequency adjustment of 1.05. It refused to acknowledge, at least initially, that there were any long-term trends in the accident-frequency data. The Division’s actuary, on the other hand, presented a study, using Registry of Motor Vehicles accident information from 1971 to 1975, which revealed a long-term downward trend in accident frequency after an adjustment for the effects of the energy shortage. His conclusion from the composite study was that the required upward adjustment of the 1974 data for *800 the energy shortage was more than offset by the downward trend, and he therefore recommended a frequency factor of 0.950.

Thereupon the Bureau, adopting the Division’s approach, attempted to deal with trend, but it used somewhat different data taken from the registry: it calculated the trend and the fuel-shortage adjustment from a set of registry subcategories of data 8 supposed to parallel the insurance claims. Weighing each category by a factor which it considered appropriate, the Bureau derived an overall frequently factor of 1.100.

The Commissioner concluded that no frequency adjustment could be justified. He thought the two competing presentations “have merit on the assumption that a trending method is to be used at all. At this point, I am not convinced that any adjustment or trend is called for.” He found there was “no compelling evidence that 1976 frequency rates will either be higher or lower than 1974 rates.” The Commissioner’s conclusion was consistent with the approach taken in previous rate decisions, in which no frequency adjustment was made.

In criticism of the Commissioner’s decision, the Bureau says that while “trending” might be unjustified (the position first taken by the Bureau), the Commissioner erred in seemingly disregarding the need for an adjustment of the 1974 data to correct for the energy shortage. Here, we *801 think, the Bureau reads the Commissioner’s opinion too finely. The Division and Bureau studies were both com-posities embracing not only a search for trends in accident frequencies, but adjustment for the gasoline shortage; the Commissioner’s opinion does not ignore the latter consideration (it refers to the effects on the accident rate of “the increased prices of fuel,” “the energy crisis,” and “external, changing forces,... with relatively short life spans”); the opinion simply reflects the Commissioner’s judgment that neither study was strong enough to warrant a departure from the 1974 data.

As to the Bureau’s argument that its study must be given credence, it may be observed that the Bureau acknowledged the poor correlation between the registry’s statistics and the claims against insurers 9 and that the Division was concerned that “additional random variation” was introduced in the Bureau’s study by using registry data on the number of persons injured, rather than on the number of accidents, in calculating frequency; and there was further concern by the Division that the Bureau might be giving undue weight to certain subcategories of the registry data. See note 8 supra. The Commissioner could likewise find the Division’s presentation, finding a downward trend, less than persuasive. 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. He ended by finding that there were a “host of... influencing variables,” and that “[wjhether their impacts will diminish or rise next year is an extraordinarily tough question.” His conclusion that “[n]one of the data exist in stable, easily trended series,” and hence that a frequency adjustment was inappropriate, is not shown to be erroneous.

C. The Property-Damage Coverages. (1) Development factor for property-damage liability losses. As in the *802 bodily-injury portion of the hearing, the Bureau advocated here the use of a development factor derived from the observation of a single year (1974). Its recommended factor to be applied in calculating the 1976 Rate Level Pure Premium was 1.157%. To prevent undue fluctuation from year to year, the Commissioner, following the Division’s recommendation, adopted a two-year average — a development factor of 1.1155.

To the general criticism already noted, the Bureau adds the point that there was no evidence in the record of the development factors for prior years. The arguments then follow that the decision was based on material outside the record, requiring reversal (see American Employers’ Ins. Co. v. Commissioner of Ins., 298 Mass. 161, 166-167 [1937]), or is equally vulnerable because of the absence of a showing that the development factor was unstable in fact from year to year.

As to the first contention, we need only note that the two-year average was admitted in evidence and the Bureau has not claimed that it is inaccurate. And, since the development factor for 1974 is in the record, the factor for 1973 may be derived from the average by a trivial calculation.

As to the second argument, since the average is different from the development factor for 1973 there must be some instability in the factor. And if the factor was not unstable, no prejudice could be shown from the use of an average instead of a single-year result.

(2) Trend and projection factor for comprehensive coverages. It was only with respect to the “comprehensive” coverages that any significant question arose in finding a trend and projection factor appropriate to the losses. Comprehensive coverage provides insurance against damage of the insured’s own vehicle due to theft, vandalism, fire, or other hazards. The Division recommended that the trend and projection factor be drawn from the Consumer Price Index (CPI) for Automobile Repairs and Maintenance, and suggested a combined factor of 1.248. The Bureau was content to use the CPI for the projection factor, but recommended that the trend factor be drawn from internal *803 data: it would calculate the trend from the change observed in the cost of the average comprehensive claim in Massachusetts from a certain period in 1974 to an identical period in 1975. This comparison suggested that the cost of a comprehensive claim had increased by 31.9% and the Bureau requested recognition of a 25% change in the trend factor; its recommended combined trend and projection factor would then be 1.405. 10

There was testimony that the theft frequency had increased disproportionately to other covered losses 11 and that a theft claim was considerably more expensive on average than claims for the other losses. As the CPI cannot reveal the increased cost of an average comprehensive claim arising from the change in the mix of covered losses, the Commissioner properly rejected the Division’s suggestion that the CPI be used to derive a trend factor.

But the Commissioner also found it “not obvious” that the Bureau’s figures were cogent. He noted a claim by the Division’s actuary that two Bureau exhibits were mutually inconsistent and both conflicted with certain industry “fast-track” data for Massachusetts. (He also acknowledged that there was no conclusive proof of inconsistency.) He remarked that “[i]ntemal data is flawed to the extent that it allows for [sic: reflects] overpayment of losses through poor carrier policing of damage appraisers (many of whom have business relationships with the body shops *804 at which the work is done).” The internal data also appeared to depend heavily on the sample data base and time period chosen for comparison. Nevertheless, the Commissioner thought the “best answer” lay closer to the Bureau’s figure than the Division’s. Yet he felt he could not accept the precise Bureau figure because of the questions raised and because the Bureau had failed to prepare an external index (based on data other than claim experience) justifying the increase. Mediating between the Bureau and Division figures (1.40 and 1.25) on the basis of their relative persuasiveness, he selected a combined trend and projection factor of 1.35, thus giving greater weight to the Bureau figure.

The Bureau argues that the Commissioner’s judgment is itself flawed, and in substantial part we agree. Any apparent discrepancy between the Bureau’s exhibits and between those exhibits and the “fast-track” data is readily explained. 12 The Commissioner’s references to the “body-shop overpayment problem” and to the failure of the Bureau to prepare an external index also suffer on analysis. Both subjects are related to the Commissioner’s effort to introduce “normative” considerations into ratemaking: his view (with which we agree in principle) is that regulation should provide cost control as well as cost observation. To introduce such control, the Commissioner would prepare an index based on data external to the insurance industry and his “best choice” between that index and one based on internal data would be the lower of the two: if the external *805 indicator is lower, it should be chosen so that the regulated industry is made subject “to the same discipline as the outside world”; if the internal indicator is lower, then it should be chosen because “there is no excuse for the regulator to pass out free money.” 13 An external index may be useful in helping to assess the credence to be given internal data, but we doubt that the Commissioner’s rule of preference would always be appropriate. For if the cost changes reflected in the internal data are not subject to control by the insurer, then recognizing only the lower costs shown in an external index would go beyond subjecting a regulated industry to the rigors of simulated competition; it would prevent legitimate price changes based on real and uncontrollable changes in costs.

As the “body-shop overpayment problem” presumably is subject to control by the insurers, there may be justification for turning here to an external index. But it seems the “problem” was first raised in the decision itself; no party points to any evidence concerning body-shop practices. And even if such a problem were shown to exist, overpayments must be proved to have become more prevalent in 1975 than in 1974 in order to justify a change in the trend factor rather than a change of the underlying loss base. In the absence of evidence on the record, the Commissioner’s comment becomes speculation concerning a matter of fact and it cannot form an adequate foundation for decision. See American Employers’ Ins. Co. v. Commissioner of Ins., 335 Mass. 748, 756 (1957); American Employers’ Ins. Co. v. Commissioner of Ins., 298 Mass. 161, 167-168 (1937).

As to the failure of the Bureau to prepare an index based on external data, we need only say that it was somewhat unclear to the parties how to create such an index *806 and that the Bureau was not aware of the consequences of the failure to prepare such a display.

The rationale for decision which we find to have merit is the Commissioner’s observation that the trend factor drawn from internal data was suspect because it seemed to depend heavily on the data base and time periods chosen for comparison. (The Bureau compared both the first six months and the first eight months of 1975 with like periods in 1974.) Although the Division’s actuary felt that a trend shown from less than full-year data would be “comparatively close” to full-year experience, the Bureau’s exhibits put this in doubt. The eight-month data, on which the Bureau relied, would show a cost increase of 31.9% whereas a similar calculation with the six-month data would show a cost increase of approximately 36.2%. The sensitivity of the trend factor to the data base is not directly revealed but is suggested by the different percentage increase in claims (from 1974 to 1975) derived in calculations founded on different bases. 14 Indeed, the Bureau’s request for a trend factor of 1.25, rather than the factor of 1.319 derived from the eight-month data, indicates a certain skepticism about its own evidence. In view of the “softness” of the Bureau’s requested figure, we cannot find that it was error for the Commissioner to discount the historical display on which the figure was based slightly more than did the Bureau. 15

(3) Special discount for anti-theft devices. Statute 1975, c. 707, § 2, amending G. L. c. 175, § 113B, provides: “In fixing and establishing classifications of risks for comprehensive fire and theft coverage so-called to motor vehicles, the commissioner shall provide for appropriate reduc- *807 tians in the premium charges covering such vehicles if such vehicle is equipped with an anti-theft mechanism or device approved by the commissioner.” 16 During the course of the hearing the Bureau filed an exhibit suggesting a 10% discount on comprehensive coverages for vehicles having an alarm that could be heard at a distance of 300 feet and that was set off “by entry through any door or opening of the hood.” In response, the Commissioner ruled: “Rather than accepting the Bureau’s arbitrary standard, I will leave the definition of an effective anti-theft device to the individual companies for 1976, ordering only that a 10% discount be offered for devices the insurer determines to be effective.” Further, the Commissioner directed the Bureau to prepare for the next year’s hearing a definition of suitable devices and documentation for setting the discount.

In challenging this disposition the Attorney General does not claim that 10% was inadequate as a discount; he argues that the Commissioner was unfaithful to the statute in allowing each company to determine what constitutes a suitable device. The Bureau, however, was not aware of experience anywhere in the country concerning such devices; and, in the absence of pertinent data, the Commissioner asserts that his action was a “pragmatic and sensible” solution to the problem for 1976.

The better course might well have been for the Commissioner to attempt to define and approve suitable devices and set a discount. These would have been somewhat arbitrary, but industry-wide experience on the loss rates of equipped cars could then have been analyzed systematically for use in future years. Under the Commissioner’s ruling the data for 1976 may prove less valuable. Our review of the Commissioner’s decision is perforce undertaken late in the policy year and there would be little benefit in inviting the Commissioner to run a different course for *808 1976. We acknowledge, too, that had the Commissioner adopted a tentative definition, consumers who purchase devices in reliance on it might have found that these did not meet the standard finally adopted. It was indicated that a definition would be forthcoming that might apply to the next year’s rates.

II. Allowance for Expenses

A. The Method of Determining the Allowance for Expenses. Before the decision on 1975 rates by the Commissioner’s predecessor, it had been customary to calculate the expense allowance by multiplying the Rate Level Pure Premium by an expense multiplier. This multiplier was defined as the ratio of the expenses attributed to the insurance fine in the basis year (1974 in the instant hearings) to the developed losses for that line for the year. The defect of the method is that, as anticipated losses increase, the dollar amount allocated to expenses grows proportionately even in the absence of evidence that any such increase in expenses is expected. Although the Bureau advocated a return to the multiplier method in this year’s hearing, the Bureau’s actuary acknowledged that it could result in “extra money” for the insurers. 17

A new approach was introduced for 1975 and in the decision under review the Commissioner continued and sought to refine his predecessor’s work. The portion of company expense attributed to claims adjustment was considered to follow losses closely so that the multiplier method could continue to be used for those expenses. 18 However, the *809 manner of determining the expense allowance for acquisition (agents’ and brokers’ commissions), field administration, and general administration costs was to be similar to that for establishing the 1976 Rate Level Pure Premium: the total of these expenses for the 1974 policy year would be divided by the 1974 exposure (yielding the average expense per insured vehicle) and adjusted to the 1976 policy year by the application of an appropriate trend and projection factor. The component of the total premium thus calculated may be termed the “Expense Pure Premium.” Finally, the allowance for incidental taxes, licenses, and fees, which traditionally had been part of the multiplicative expense factor, was considered to be related to the total premium, rather than the 1976 Rate Level Pure Premium or Expense Pure Premium, and so was to be separately calculated. 19

B. Alleged Errors in Application of the Method. The Bureau states in its brief that the method used by the Commissioner in determining the expense allowance “may in the abstract be preferable” to the older multiplier method. It argues, however, that the trend and projection factor (to adjust certain of the 1974 expenses to 1976 cost levels) was too low, and also that including the acquisition expenses in the Expense Pure Premium, rather than compensating for them by the multiplier method, was error.

(1) Trend and projection factor. The factor as approved by the Commissioner was admittedly a compromise. The Division developed an “internal” estimate of the factor based on the change in insurance company expenses *810 from 1973 to 1974 (as revealed by statistics maintained by the Bureau), and recommended a factor of 1.0929. The Bureau — assuming the multiplier procedure was to be discarded — proposed an “external” measure of the factor, derived from statistics maintained by the United States Department of Labor concerning wages for fire, marine, and casualty insurance carriers. It recommended a projection factor of 1.198.

The “best choice” view of the Commissioner has been described above. It would apply here, but the Commissioner found that neither party had prepared a fully convincing analysis. The Division’s estimate did not deal with the probability that the rate of growth of inflation had increased since 1974. The Bureau’s factor disregarded the apparently contradictory actual experience of the insurers 20 and was also out of line with the over-all trend of wages in private industry. 21 As a “best choice” rule could not be safely applied, the prudent alternative, according to the Commissioner, was to average the “probably understated” factor of the Division and the “probably overstated” fac *811 tor of the Bureau. He adopted a trend and projection factor of 1.145. We cannot say he was wrong. Cf. Boston Gas Co. v. Department of Pub. Util., 368 Mass. 780,804 (1975).

It should be noted for future reference that the Bureau attacked the Division’s estimate, founded on internal data, as unreliable because the insurers estimate some expenses by setting them equal to a fraction of premium volume and therefore the expenses reported do not coincide with actual expenses, but are more directly affected by the rates set for the year. This is a curious commentary on the manner of reporting expense information to the Division: if reported expenses are inaccurate for projection, the question is raised whether they may not be inaccurate in absolute amount. 22

(2) Acquisition expenses. The commissions to agents and brokers for selling policies have been traditionally determined as percentages of the premiums on the policies. In the past this practice has conformed to the ratemaking process in which the expenses have been determined by the multiplier method. 23 The conclusion that acquisition expenses should not be so treated in ratemaking is hardly startling: if costs of medical care were expected to skyrocket in a given year, thereby affecting the projected losses, but salaries were not expected to rise at all, the old method would nonetheless raise the allowance for acquisition expenses in proportion to the expected losses, despite the close connection of commissions to wages.

*812 In the end the Bureau did not really attack the Commissioner’s fundamental approach to acquisition expenses, but argued instead that the companies had not had a chance to adjust their contractual commitments to it because by statute (G. L. c. 175, § 163) insurers must give agents and brokers ninety days’ notice of changes in their contracts and the Commissioner’s decision was not issued until mid-November. 24 The answer is that the predecessor Commissioner treated acquisition expenses in the same fashion as the current Commissioner who indicated his intentions early in the hearings. 25

III. Allowance for Profit

A. The Method of Determining the Profit Allowance. Traditionally, the allowance for profit was 1% of the total premium for compulsory coverages and 5% of the total premium for other coverages. But these figures were likely to be a very misleading indication of the actual profit made by the insurer on a given line of insurance. Because the premium dollars are paid to the insurers early in the policy year and the payment of insurance claims stretches out over several years, the cash flow produces funds that the insurer can invest. See generally Birkinsha, Investment Income and Underwriting Profit: “And Never the Twain Shall Meet”?, 8 B.C. Ind. & Com. L. Rev. 713 (1967); Comment, Insurance Ratemaking Problems; Administrative Discretion, Investment Income, and Prepaid Expenses, 16 Wayne St. L. Rev. 95 (1969). This income from invest *813 ment is not directly adverted to in formulating the profit allowance by the traditional method. The Commissioner concluded that the method was the “shoddiest component” of ratemaking and that the 1% and 5% figures “hang solely on threads of imagination unsupported as measures of reasonable profit in either hearing evidence or actuarial literature.”

The Commissioner proposed a radically different procedure — altogether novel in automobile insurance rate regulation in the Commonwealth — designed to set the profit allowance for a given insurance line in a more realistic fashion. His goal was to adjust the allowance for profit so that it, together with the investment income, would “allow the average regulatee a rate of return on [shareholders’] invested capital roughly equal to that earned by businesses with similar riskiness in the unregulated, competitive sectors of the economy.” The several stages of the Commissioner’s procedure follow.

(1) A “minimum reasonable investment yield” is determined for the funds provided by the policyholders and shareholders and available for investment by the insurer. One approach to this end would be to examine and analyze the investment income actually realized by insurers. But the Commissioner, quoting from an earlier decision of his own, noted the “multitudinous problems” involved in trying to measure actual investment income: insurers pool the cash from several lines of business, so that income cannot easily be attributed to a particular line; and the investment results among real-world insurance companies fluctuate widely from year to year, making difficult the achievement of stable results. Moreover, if a real investment program is examined, complex problems arise in dealing with realized and unrealized gains and with income taxes. Characterizing the problems as a “Gordian Knot,” the Commissioner chose to slash through it by setting up a hypothetical investment program. He selected as the minimum reasonable investment yield the average yield on Treasury securities of various maturities, each weighted in the averaging in accordance with the cash flow pattern *814 of the line of insurance under consideration. The choice of such a “risk-free” yield was fair and neutral, according to the Commissioner, for if an insurer pursued an investment policy offering an opportunity for more profit, it did so at the risk of having to absorb losses.

(2) A cash flow exhibit is prepared for the line of insurance under examination which discloses how funds become available for investment, and which calculates the amount obtained from investing these funds at the minimum reasonable investment yield. If the premium for the line of insurance is set properly, as the model assumes, then the total premium receipts will exactly equal the payout of the profit allowance, losses, and expenses. 26 The pattern of rapid buildup and gradual release of cash generates in-vestable funds, which are assumed to be invested at the minimum reasonable investment yield. The yield from that investment is divided by the total premium receipts (premium volume) to give what is termed the underwriting investment income as a percentage of premium volume.

(3) The return on shareholder-contributed capital is calculated. To the underwriting investment income as a percentage of premium volume (derived in Step 2) is added the profit allowance 27 (which is also expressed as a percentage of premium volume), giving the total return from underwriting as a percentage of premium volume. Since capital serves as a guaranty fund for payment of claims in the automotive insurance industry, the insurer will monitor and maintain a ratio between premium volume, a mea *815 sure of his loss exposure, and capital. The return from underwriting as a percentage of premium volume is multipled by this ratio to give the return from underwriting as a percentage of capital. Next, added to this figure is the return from investing the capital itself at the minimum reasonable investment yield (the return is expressed as a percentage of capital). The total, which represents the pre-tax return on capital, is then multiplied by a tax conversion factor to produce the estimated after-tax return on capital.

(4) The rates of return on unregulated industries are gathered and a target rate of return for the insurer is determined. (One of the issues before us is how this determination should be made.)

(5) The profit allowance is adjusted so that the total return on capital determined in Step 3 equals the target rate of return derived in Step 4.

Essentially the procedure involves an analysis of a hypothetical or model insurance company which writes only the line of insurance under consideration. The various elements of the analysis were considered at both the bodily-injury and property-damage hearings; some were challenged and are discussed below. 28 As it turned out, on the Commissioner’s calculations a profit allowance of negative 4% for bodily-injury coverages was found to meet the target rate of return. Because of the investment income made possible by the comparatively slow payout of bodily-injury claims, the amount collected in premiums can be 4% less *816 than the amount absorbed by expenses or paid in claims and still provide those who invest in the insurance company with a reasonable return. For the property-damage coverages, with a much faster payout of claims and hence less underwriting investment income, a positive 5% profit allowance was found necessary to produce the target return. By coincidence a 5% allowance has been traditional in property-damage ratemaking.

B. Alleged Errors in the Method. None of the parties to the hearing challenged the Commissioner’s general method of setting the profit allowance. 29 Not surprisingly, however, the figures used in applying the method have provoked disagreement.

(1) Comparisons with other industries. In the bodily-injury decision the Commissioner took for comparison the rates of return on total capital as reported by Forbes magazine (January 1, 1975) for a recent twelve-month period for 850 of the largest domestic corporations, and, after certain adjustments, he derived a 10% average return on total capital. 30 Because the bodily-injury insurance is a “slow pay” line, the Commissioner felt that an insurance company engaged in selling such insurance is subject to somewhat greater risk in an inflationary period than the average domestic company, and he therefore adjusted the target rate of return to be used in the profit calculation *817 upward to 11.5%. In the property-damage decision the Commissioner chose the 10% figure as the target, making no such upward adjustment, because the faster payout of property-damage losses made the return less vulnerable to inflation.

Did the Commissioner make an appropriate reference to the unregulated sector? The Bureau argues that the proper comparison is not with the return on total capital (debt plus equity), but rather the return on equity alone. And the average return on equity, as derived from Forbes, is in the vicinity of 14%.

The Commissioner argued briefly (in the bodily-injury decision) that the Bureau’s view would result in a higher total return for a firm with a capital structure consisting only of equity than for a firm with identical total capital but with both debt and equity in the structure, and that this must be unsound because the “capital at risk” in the two cases is the same. Because insurers seldom issue debt, the Commissioner’s view was that setting the target rate by comparison with the return on equity in the unregulated sector would give the insurers an “extra return” on their capital bases.

The Commissioner’s reasoning is open to question. The rate of return to an investor in the model insurer should equal what he would expect on an investment of comparable riskiness in the competitive market. See, e.g., New England Tel. & Tel. Co. v. Department of Pub. Util., 360 Mass. 443, 472 (1971), quoting Federal Power Comm’n v. Hope Natural Gas Co., 320 U.S. 591, 603 (1944). See generally V. Brudney & M.A. Chirelstein, Corporate Finance 56-81, 378-380 (1972). And the Commissioner’s approach seems suspect because it fails to confront and to consider all the elements of risk of an investment in such an insurer. Among these elements is the risk inherent in the line of insurance itself; some lines will have greater unpredictability and fluctuation of losses than others 31 and an investor *818 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. The risk of an investment in an insurer is also affected by the risk that is characteristic of the insurer’s internal investment policy. Finally there is risk to the investor associated with the ratio which the insurer adopts between premium volume and capital: if the insurer maintains a high ratio, then adverse loss or expense experience will consume a greater portion of the capital than it will in the case of an insurer with a low ratio; conversely, favorable experience will result in greater return to capital than in a low-ratio insurer. See In re Application of Ins. Rating Bd., 63 N.J. 413, 417 (1973). Hence, other things being equal, the high-ratio insurer is more risky than a low-ratio insurer and the reasonable investor would demand a greater average return from it than from the low-ratio insurer. An assessment of these risks, and perhaps others, that characterize the model insurer and a comparison with the risk-return relationship of unregulated enterprises may be thought necessary if the target return of the model insurer is to be properly determined. 32

*819 But even if the Commissioner’s view, just criticized, is taken to be weakly supported, we believe a remand to find an appropriate target rate of return is not required. In the course of the property-damage hearings, experts on both sides, testifying on the present subject, all agreed that the target for the model company should be set equal to that of an equity investment of comparable riskiness. The problem, as they saw it, lay in assessing the riskiness of the model corporation. 33 Although the experts felt that prospective shareholders would demand about 15% from an equity investment of average riskiness and that they would demand that much or more from an investment in a real-world liability insurer, they agreed that the model insurer might appropriately pay less. The hypothetical model assumed a premium volume to capital ratio of two to one (see note 28 supra), whereas real insurers are currently writing at a more risky ratio of three to one or higher. And the model insurer was to follow a risk-free investment policy, whereas the real insurers typically invest with assumption of risk. Because of the diminished risk of investment in the model insurer, a target return below that of the market generally was justified. In fact, two of the three experts called indicated that 10% — the target chosen, coincidentally, 34 by the Commissioner — was reasonable for the model property-damage insurer. Although the Commissioner characterized the experts’ figures as “loose estimates” and the experts hedged their estimates by saying that more careful analysis would be needed to reach a precise figure, we feel their testimony, in its cumulative effect, *820 provides adequate support for the Commissioner’s figure. 35 In reaching this result we are mindful that the Commissioner’s approach to the profit allowance is not only novel but complicated, and that somewhat greater imprecision must be tolerated in its initial application than might be acceptable in later years.

Turning to the target for the bodily-injury line of insurance (and deferring until (2) below the separate question of the adjustment of 1.5% for the consequences of longer delay in the payment of those claims), we think the basic 10% can be justified here even though there was no expert testimony on bodily injury comparable to that received in respect to property damage. The assumptions concerning the model insurer were identical for both bodily injury and property damage, and so the adjustment of the target return in light of the risk of the internal investment policy and ratio of premium volume to capital would likely be identical for the two coverages. In fact, in the course of the property-damage hearing the experts were asked to comment on bodily-injury decision, and their remarks provide the necessary support we would otherwise expect on remand.

(2) Adjustment for risk in the bodily-injury decision. As already suggested, with payment of claims for bodily injury extending over a much longer period than payments for property damage, inflation is a more serious problem increasing the riskiness of the bodily-injury line. The Attorney General argues that the Commissioner’s upward adjustment of 1.5% was not adequately supported by the evidence. 36 To the Attorney General’s suggestion that there *821 was inadequate record evidence that inflation affects the risk, it is a fair answer that that premise is obvious and may be taken to have been acknowledged by all. As to the amount of the adjustment, we agree with the Commissioner that, while disallowance of any adjustment might be “procedurally fair” because the Bureau did not come forward on the matter, it would be “less than logically fair.” Here the judgmental estimate of the Commissioner of the proper adjustment is of the type that can be expected on the initial application of a novel methodology. 37 Crude estimation could not be tolerated in normal circumstances, and we observe that the Commissioner called on future participants in the hearing process to present “more formal data and analysis.”

(3) Deduction of certain expenses from the minimum reasonable investment yield. In the bodily-injury decision the Commissioner allowed a deduction of 0.6% from the weighted average of Treasury securities constituting the minimum reasonable investment yield so as to compensate the insurer for investment expenses and for unin-vested cash. In the property-damage decision, on the other hand, the Commissioner found that an insurer following the hypothetical investment policy would be able to invest nearly all its funds and would have minimum investment expenses; thus he concluded that a deduction would be inconsistent with the model and he refused to allow it. The Attorney General argues that the 0.6% deduction in the bodily-injury decision should be disallowed and the Bureau argues that the deduction should have been allowed in the property-damage decision. The results are fully supported by the respective records on which they *822 stand; all witnesses agreed with the need for an adjustment in the bodily-injury hearings and the experts split at the property-damage hearings. We think no correction of either decision is warranted.

(4) Claimed deviation of model from reality. Although the Bureau states early in its brief that, at least for purposes of this appeal, it is not challenging the Commissioner’s method of calculating the profit allowance (see note 29 supra), it nonetheless later criticizes the method as a “blackboard” exercise lacking realism. It points, by way of example, to the fact that the billing and receipt of premiums in 1976 were delayed by tardiness in the issuance of the property-damage decision, making the cash flow exhibit somewhat erroneous, and also to the fact that the investment yields on government securities at the time premium dollars became available for investment have proved to be less than those used in the model. And it says that the omission of the property-damage decision to take account of investment expenses and uninvested cash is á sign of failure to face facts. But the first two problems do not really grow from the model; they rather are the products of disparities that inevitably will exist between a projection of the future and the future as it proves to be. And although, as to the third point, the model may depart from the behavior of real-world companies, it is of the essence of the use of a model that it makes some simplifying assumptions. The model is unrealistic only in the sense that no actual insurer happens to conform its behavior to the assumptions, not in the sense that its assumptions are fanciful or impossible to match in the real world. If under the model there is neither investment expense nor uninvested cash, then the failure of real-world insurers to follow the model must arise from the fact that they find these added expenses are less than the extra profits derived from a more aggressive investment policy. The same observation may be made about other departures by a real insurer from features of the model that are subject to the insurer’s control.

*823 IV. Questions About the “Second Look”

A. As to the Bodily-Injury Decision. Applicable to the Commissioner’s consideration in 1975 of bodily-injury rates for 1976 was an unusual “second look” statute providing that if the premium charges for any current calendar year (here 1975) were found to be “excessive, unjust, unreasonable, discriminatory or inadequate,” the Commissioner was to “take these facts into account” and fix the charges for the ensuing year (1976) at levels which he determined would produce “adequate, just, reasonable, nondiscriminatory or not excessive charges” for the two years taken together. G. L. c. 175, § 113B, as amended through St. 1971, c. 977, § 1A; see Employers’ Commercial Union Ins. Co. v. Commissioner of Ins., 362 Mass. 34, 37 (1972). In the bodily-injury decision the Commissioner decided against adjusting the rates, otherwise to be set for 1976, by reason of a claimed inadequacy of the 1975 rates.

The Bureau’s rather strenuous argument in its brief that the Commissioner erred is not matched by much that is substantial in the transcript of the hearing, and the Commissioner’s observation that “[n]o party at the hearings explored this matter in great depth, either factually or legally,” seems quite accurate. A Bureau exhibit and counsel for the Bureau in an opening statement made passing reference to the inadequacy of 1975 rates, and the only focus on the matter during the hearing was in a few general questions put to the Bureau’s actuary. 38 Indeed most of the attention given to rates for prior years was on the question whether the 1976 rates should be adjusted downward, pursuant to special legislation to recapture any excess profit made by insurers in 1974. 39 St. 1973, c. 1113. *824 The failure of the Bureau to press a case for adjustment at the hearing may well be taken to bar their raising the issue on appeal.

But if the merits are to be examined as far as is possible, we think the Commissioner had adequate pragmatic reasons to justify his action. Rates frequently miss their target for the given year but prove more accurate over the longer term; the public would be disserved by constant tinkering with the rates to meet a tight standard of accuracy in each year. So also yearly adjustments to compensate for adverse experience might blunt efforts to encourage efficient and economical operation by the insurers. Such considerations led to the conclusion that the second look provision should not require rate adjustment where, as in 1975, the current rates could not be said to be dramatically less accurate than rates in other years. This view is consistent with the statutory language. “Inadequate” and “excessive” do not define precise boundaries (cf. Massachusetts Bonding & Ins. Co. v. Commissioner of Ins., 329 Mass. 265, 270 [1952]), and the sense of the statute is that adjustment is not contemplated except where experience shows an egregious failing either way. No such showing was made here. 40

B. As to Property-Damage Decision. On November 26, 1975, while the hearing on property-damage rates was under way, St. 1975, c. 707, §§ 1A and 9, was approved repealing the second look provision. The Commissioner read the statute as ending his obligation to consider adjustment *825 of 1976 rates to take account of any 1975 shortfall, which according to the Bureau’s allegations was very considerable. 41

The Commissioner’s reading of the statute appears to us to be correct. While the repealer does not state expressly that it is to affect the setting of 1976 rates, and could possibly be thought to deal only with 1977 rates so far as adjustment for 1976 experience was concerned, the fact is that § 9 says the repealer provision of § 1A “shall take effect upon the passage of this act,” and the entire act has an emergency preamble which says that deferred operation “would tend to defeat... [the act’s] purpose which is to provide immediate relief to owners of motor vehicles from the high costs of compulsory property protection insurance.” The milieu in which the repealer was enacted suggested strongly to the Commissioner, as it does to us, that the Legislature wanted truly immediate effect. 42 Decisions interpreting quite different enactments with different backgrounds, such as City Council of Waltham v. Vinciullo, 364 Mass. 624, 628-629 (1974), cited by the Bureau, are not helpful here. This brings us to the Bureau’s constitutional claim which, however, we think not so weighty as to influence interpretation of the statute even if there was real room for doubt as to its intended meaning. See Mile Road Corp. v. Boston, 345 Mass. 379, 383, appeal dismissed, 373 U.S. 541 (1963).

The situation during the period of the second look statute was that, in addition to a remedy for correction of rates by regular appeal, there was an unusual administra *826 tive-judicial remedy in the form of a second look. An appeal raised questions about the legality of the rates. The second look did not concern itself with legal issues but with practical experience under the rates. The second look procedure was made available without any constitutional compulsion, for it is clear that the rates for a given year cannot be argued to be confiscatory because of industry losses in previous years. Public Util. Comm’rs v. New York Tel. Co., 271 U.S. 23, 31 (1926).

We think the State could decide to abandon the second look procedure and retain only the appeal. Abolition of the second look in fixing 1976 rates would raise no question if legislated before 1975 rates were set; and we think the same would hold for abolition of the second look legislated during the period for taking an appeal, because the right to appeal would not have been lost.

It can be argued that the present case is legally different from the one last mentioned because abolition of the second look occurred after the time for taking an appeal had run and the insurers had failed to appeal though entitled to do so. Debate in terms of “vested rights” and “retro-activity” actually comes down to the proposition that the insurers relied, in failing to take an appeal, on the then existence of the remedy of the second look, and that this reliance is entitled to constitutional protection to the extent of preserving the second look despite the legislative repeal. 43

Nothing in the record suggests that in fact the insurers relied on the possibility of a second look in forgoing direct appeal. Nothing in the theory or design of the second look statute suggests that it was enacted procedurally as an alternative to a direct appeal, or substantively to give the same relief as a direct appeal; its function was rather to *827 adjust in year two for the unfolding of events in year one in a manner which differed from a legally correct prediction or estimate made by the Commissioner in setting rates for year one. If legal error was committed by the Commissioner, direct appeal was the only way to correct it, notwithstanding the possibility of some equitable offset in year two through the second look.

Nor, in the circumstances, could the insurers reasonably have relied on the second look as a basis for forgoing an appeal. As members of an industry whose rates had long been regulated in the public interest, the companies must be charged with knowledge that changes in regulatory legislation were likely and that the possibility of cumulative relief embodied in the second look statute was not an eternal expedient or verity. They were bound to know that much more solid interests or expectations than a second look have been held subject to the evolving exercise of regulatory powers. See, e.g., Mile Rd. Corp. v. Boston, 345 Mass. 379, 383-384, appeal dismissed, 373 U.S. 541 (1963); Paquette v. Fall River, 338 Mass. 368, 375-376 (1959); Boston Real Estate Bd. v. Department of Pub. Util., 334 Mass. 477, 488-491 (1956); Federal Housing Administration v. Darlington, Inc., 358 U.S. 84, 90-92 (1958); Queen-side Hills Realty Co. v. Saxl, 328 U.S. 80, 83 (1946); Western Union Tel. Co. v. Louisville & Nashville R.R., 258 U.S. 13, 20, 22 (1922); Union Dry Goods Co. v. Georgia Pub. Serv. Corp., 248 U.S. 372, 377 (1919); Louisville & Nashville R.R. v. Mottley, 219 U.S. 467, 482-486 (1911) ; South Terminal Corp. v. EPA, 504 F.2d 646, 678 (1st Cir. 1974). Their forgoing an appeal could not assure the companies of a second look which the Legislature could otherwise constitutionally repeal. Cf. Hughes, C.J., in Sproles v. Binford, 286 U.S. 374, 390-391 (1932).

V. Disposition

The Bureau makes the general criticism that even if the Commissioner’s decisions are correct in their particulars, the end results are wrong because, at places where the *828 Commissioner had room for choice, he ruled so consistently against the Bureau as to suggest an unfair predisposition. 44 The Bureau is complaining that in a sense the whole in each case has turned out to be less than the sum of its parts.

On a general review of the decisions we do not think the criticism is well founded, and we are of opinion that the rates set are not shown to infringe the standards of the statute or to be confiscatory. 45

The cases are remanded to the county court where judgments are to be entered denying the petitions in Nos. 75-406, 75-403, 76-22 and 76-21 and dismissing the action for declaratory relief, No. 76-24.

So ordered.

Notes

Statute 1976, c. 266, approved by the Governor on August 4, 1976, makes important changes for the future in the laws of the Commonwealth regarding automobile insurance. Under the new arrangements, competition among insurers is chiefly relied on in the first instance to bring about fair rates, but the Commissioner reserves certain controls and he may finally resort to ratemaking. See St. 1976, c. 266, §§ 12 and 13; § 19 repealing present c. 175E and substituting a new c. 175E. And see especially §§ 5,7, 8 of new c. 175E.

Because the issues concerning the “second look” provision are presented for review in the § 113B petition, we need not decide whether the Bureau and companies were entitled to seek review of that matter by way of an action for declaratory relief.

To recall what is our function in rate review: We are obliged to determine whether the rates approved by the Commissioner meet the statutory standard as “adequate, just, reasonable and nondiscriminatory” premium charges. G. L. c. 175, § 113B. The actual fixing of rates is not a judicial function (see American Employers’ Ins. Co. v. Commissioner of Ins., 335 Mass, 748, 750 [1957]); thus we are not to substitute our judgment for the Commissioner’s, but rather to decide whether the rates set by him have “ ‘reasonable support in the evidence.’ ” See Insurance Rating Bd. v. Commissioner of Ins., 359 Mass. 111, 117-118 (1971); American Employers’ Ins. Co. v. Commissioner of Ins., 335 Mass. at 750; Massachusetts Bonding & Ins. Co. v. Commissioner of Ins., 329 Mass. 265, 273 (1952). We are to give due respect to the Commissioner’s expertness (which in the present cases, as the parties acknowledge, is not merely presumed, but real). See Atlantic Coast Line R.R. v. Florida, 295 U.S. 301, 317 (1935); 4 K.C. Davis, Administrative Law § 30.09 (1958). The Bureau asks us to perform a further duty which involves the Constitution: to determine whether the rates have not been set so low as to be “confiscatory.” With regard to this latter determination our decisions say that we are to make an independent review of the facts as well as the law. See Boston Gas Co. v. Department of Pub. Util., 368 Mass. 51, 54 (1975); American Employers’ Ins. Co. v. Commissioner of Ins., supra at 750-751.

In fact, of course, the losses on both 1974 and 1976 policies will be paid out over several years and thus the adjustment for the change in cost levels is greatly simplified. It may nonetheless be accurate if the pattern of the cash flow for 1976 policies parallels that for 1974 policies and if the rate of change in cost levels remains constant.

The Bureau advocated a development factor derived from a two-year average in setting the rates for increased limits coverage so as to achieve accuracy and stability. And the Bureau presentation concerning optional medical payment coverages included development factors calculated by two- and three-year averages. No issue about either of these coverages has been raised for review.

The development factors for recent years are: 1.085 (1974), .990 (1973), 1.083 (1972), 1.097 (1971), 1.040 (1970). The parties used different terminology, referring, for example, to the data derived from 1974 experience as the 1975 factor since that is the year in which the review was made.

The Registry of Motor Vehicles tabulates the number of persons injured in accidents in four categories: fatal, serious visible injury, minor visible injury, and nonvisible injury. The Bureau argued that the reports of nonvisible injuries should not be included in calculating the frequency factor because such injuries are unlikely to generate insurance claims and because reports of such injuries are not a reliable measure of claims by reason of the impact of the no-fault law. The Bureau, using the Division’s method of analysis, found that the accident frequency for the categories of fatal and serious visible injury, viewed together, was increasing, whereas the frequency in the category of minor visible injury was decreasing; and it argued that the first figure should be given approximately seven times the weight of the second.

The Division’s actuary used registry data on the number of accidents in which personal injury resulted, rather than data concerning the number and nature of the injuries.

Because of certain errors in the way insurers had been reporting their claims experience to the Bureau, many of the statistics maintained by the Bureau were unusable for frequency determination.

One Bureau witness recommended that the full 1.319 trend factor be recognized, yielding a combined trend and projection factor of 1.483.

Although the exhibits indicate that the theft rate in Massachusetts had increased dramatically (21.7% increase from 1974 to 1975 as revealed by registry statistics), the Commissioner concluded that the frequency of all comprehensive claims had moved comparably (20%). Thus the disproportionate increase of theft losses to other covered losses was not sharply revealed by the exhibits. All the witnesses who testified on the subject agreed, however, that such a shift had been occurring.

There was testimony that the increasing theft rate is inadequate to explain all the burgeoning increase in comprehensive losses. Some testimony suggested that the recent experience with comprehensive coverages might be related as well to the general state of the economy.

One exhibit (Ex. 96) gives a “claim count” for accidents in the first six months of 1975 that is more than the “number of claims” revealed on another exhibit (Ex. 87) for the first eight months of 1975'. The Bureau’s manager testified that the exhibits were prepared at different times and each used a different data base; each exhibit was internally consistent and was valid. Although no party raises the point, we observe that one exhibit is based on “closed loss experience,” apparently data on claims made and settled in the period, whereas the other is based on only a “claim count,” apparently a measure only of claims made. Thus the exhibits would not seem inconsistent.

No evidence was admitted concerning “fast-track” data. An exhibit introduced for identification by the Division reveals, however, that “fast-track” data on comprehensive coverages were not available on a State-by-State basis.

The rule may not diminish the forces that would cause the internal index to drop below the external index: because the rates are set industrywide, each insurer has the incentive to maximize his profit by beating the industry average, thus driving the internal index down. The Commissioner’s rule assumes, however, that this incentive to the individual insurers may often prove inadequate to keep the industry average within bounds.

Compare Ex. 87 (29.1% increase) with Ex. 96 (34.5% increase).

The Bureau’s trend and projection factor of 1.405 is a composite of a 1.25 trend factor and a 1.124 projection factor. Since the projection factor is not subject to dispute, the Commissioner’s overall factor of 1.35 may be seen to be equivalent to a trend' factor of 1.201. The difference between the trend factor requested and that given is thus comparable to the difference between the result drawn from the eight-month data and that drawn from the six-month data.

Although the section was not to be effective until January 1, 1976, it was apparently viewed by the parties as applicable to the setting of 1976 rates because of the provision that “all things necessary to be done prior to said effective date may he done.” St. 1975, c. 707, § 9.

If the Commissioner, using a multiplier method, had recognized the adjustments suggested by the Bureau in the bodily-injury hearings to predict the 1976 Rate Level Pure Premium from the 1974 experience, the allowance for 1976 expenses would have been 40.5% more than was spent in 1974.

The expenses associated with the investigation and adjustment of claims are divided into two portions: the allocated and the unallocated claims adjustment expense. The allocated adjustment expense is included as a component of losses and therefore compensation for this expense is part of the Rate Level Pure Premium. The allowance for unallocated adjustment expense is determined from the Rate Level *809 Pure Premium by multiplying the premium by the ratio of the 1974 unallocated adjustment expenses to the 1974 developed losses. The acceptance by the Commissioner, in this calculation, of the Division’s suggested development factor, resulted in an expense multiplier of 0.149 for the bodily-injury coverages, rather than the 0.143 which would have followed from acceptance of the Bureau’s suggested development factor.

The taxes, licenses and fees that will actually become due are calculated from the total premium rather them from one of its components. The Commissioner determined that an allowance of 2.3% of the total premium would be adequate to compensate for these expenses.

The Division’s factor, which displays the insurer’s actual experience, is not directly comparable with the Bureau’s because each factor is drawn from data of a different time period: the Bureau’s factor is derived from the change in the expenses under analysis between 1973 and 1974; the Division’s is based on the data from 1974 through part of 1975. Nevertheless, the discrepancy between the two factors is suggestive that the internal factor would differ from the external factor in an identical time period.

Statistics maintained by the United States Department of Labor of the average private weekly wage would suggest a trend and projection factor of 1.107, rather than the factor of 1.198 requested by the Bureau. The reason for the greater increase of wages in the fire, marine, and casualty insurance sector over that in the private sector as a whole was not revealed at the hearings.

It is true, as the Bureau points out, that in last year’s decision the Commissioner’s predecessor suggested that the trend and projection factor for expenses be derived from statistics from the Department of Labor concerning wages in the insurance sector. But the alleged inconsistency is adequately explained: the Commissioner’s rule of comparing external and internal factors and selecting the lower had not been formulated then, nor had an appropriate internal index been developed. And we add that the Bureau objected to the use of an external index at last year’s hearings.

If the rates are set by examining reported expenses, and the reported expenses are in turn determined by the rates, the process is entirely circular.

Although the Bureau frames its argument in terms of the continuation of the traditional method for allocating expenses, that method, which determines the expense allowance as a certain proportion of the loss allowance, will no longer yield an allowance for acquisition expenses that maintains a fixed relationship to the total premium. The reason is that the profit allowance now forms a separate part of the total premium which may vary independently of the loss allowance. In seeking to maintain a fixed relationship of acquisition expenses to total premium, the Bureau is in fact arguing that the acquisition expenses be accorded the novel treatment that is given to the allowance for taxes, licenses, and fees. See note 19 supra.

It is a premise of the Bureau’s argument that contracts with agents and brokers expire and are renewed as of January 1, 1976. The Commissioner questions the accuracy of the premise.

The Bureau raises the same objections with regard to the Commissioner’s treatment of expenses in both the bodily-injury and property-damage decisions. It argues, however, that “the impact of the Commissioner’s error is considerably greater in the property-damage context, because all the parties agree those rates require a substantially greater increase. If expenses were increased proportionately, in the traditional manner, they would reflect that greater increase.” But of course the whole point of the Commissioner’s approach to expenses is to avoid such an artificial justification for increase of the expense allowance.

The Commissioner’s exposition is not completely clear as to what is included in the cash flow. If, as we suggest in the text, the cash flow includes receipt of the total premium and the payout of losses, expenses, and the profit allowance (not including income from investment), then calculating the correct profit allowance will involve the determination of a self-consistent solution since the cash flow exhibit both includes the profit allowance and is used in determining the allowance.

Since the profit allowance is the figure that the method seeks to derive, the appropriate figure to be included in Step 3 must be determined either by trial and error (see Step 5) or by algebraic manipulation. See also note 26 supra.

In both the bodily-injury hearing and the property-damage hearing the Commissioner relied on cash flow exhibits prepared by the Bureau. In both he assumed a two to one ratio of premium volume to capital (see R. Kenney, Fundamentals of Fire and Casualty Insurance Strength 97-102 [4th ed. 1967]), and a tax rate of 48% corresponding to the investment policy he assumed. (Real insurers are observed to have a higher ratio and may benefit from a lower effective tax rate.) In the bodily-injury decision the Commissioner set a minimum reasonable investment yield at 7.42% and in the property-damage decision a yield of 6.42%, the difference being due chiefly to the different weights given to securities of various maturities, as demanded by the different cash flows. The target rate of return was 11.5% in the bodily-injury decision and 10% in the property-damage decision.

The Bureau is careful in its brief to limit its acceptance of the Commissioner’s procedure. It claims that its acquiescence in the new procedure was the result of a “pragmatic” decision to allow a “full discussion of the method.” The Bureau does not attempt to defend the traditional method nor does it propose any alternative.

Of the 850 corporations listed in Forbes, forty-three had deficit returns in the period examined but the negative return on capital was not tabulated. The Commissioner, in calculating the average, made allowance for the shortcoming in the data as best he could by setting at zero the. return for those companies showing deficits.

The Attorney General claims that the failure of the Commissioner to collect the missing data means that the 10% is not based on adequate evidence. We believe the alteration in result would be negligible, considering especially the other elements of judgment involved in the profit calculation. But in view of the fact that we affirm the Commissioner’s decision on the basis of testimony not founded on the Forbes presentation, we need not discuss the issue further.

The Commissioner’s adjustment for the risk of inflation in the bodily-injury decision might be seen as a compensation for one aspect of the risk inherent in the line of insurance.

Returning to the Commissioner’s justification for setting the target rate equal to the average return on capital, we acknowledge the correctness of his observation that the return on equity should not ordinarily be equal for firms that differ only in their capital structure. Because of the first claim of debt to income, the risk of an equity investment in a firm with both debt and equity in its capital structure is greater than the risk of an investment in a firm without debt. Thus, because of the difference in risk, the expected return must differ. The possible difficulty with his justification lies in his apparent assumption that the formal capital structure of an insurer largely controls its risk vis á vis the corporate average. Although, as the Commissioner notes, insurers seldom issue debt — an obligation to repay borrowed funds would conflict with the role of capital as a guaranty fund in an insurance company (see In re Application of Ins. Rating Bd., 63 N.J. 413, 415-416 [1973]) — an investment in an insurer can be subject to varying amounts of risk, as the text indicates. Indeed, because of the first claim of policyholders to money, it would seem that the premium volume to capital relationship in an insurer has an effect on the risk of an equity investment comparable to that of the debt to equity relationship in an ordinary corporation.

The simplifying assumption made in the determination of the minimum reasonable investment yield introduces some complications in comparing the model insurer with real companies in the unregulated sector — real-world insurers do not follow such an investment policy.

If the premium volume to capital ratio or the internal investment policy of the model were different, the risk of an equity investment would vary and, accordingly, the target rate of return, under the experts’ procedure, would change. Since the variation of these assumptions would not introduce debt into the model insurer’s capital structure, the Commissioner’s result, on the other hand, presumably would not be altered.

The Attorney General argues that since the model insurer is less risky than the average firm in the Forbes survey, the target return should he less than the average return on capital. The argument is dubious since it assumes that the return on capital sets the target. We note, however, that under the experts’ approach the return on equity is in fact adjusted to be less than the appropriate Forbes average.

The Attorney General also argues that the adjustment is inconsistent with the model and inconsistent with the property-damage decision. But, as the text above indicates, the assessment- of the risks inherent in the line of insurance and the adjustment of target return *821 accordingly are essential to the proper use of the model. And, although the adjustment is not made in the property-damage decision, the results are not inconsistent since they are justified by the lesser vulnerability of the property-damage insurer to inflation.

In setting the target rate of return the experts would advocate that all aspects of the risk of the model insurer be considered together (in setting the so called “Beta” factor), rather than by the serial process described here. The final results would be expected to be identical.

The Commissioner noted that “[p]ast hearing transcripts indicate that the pattern of light treatment is more the rule than the exception.” So far as the record reveals, the only adjustment to the rates made under the second look was a 26% rebate in 1972 to bring the profit margin within ten percentage points or so of target. See Employers’ Commercial Union Ins. Co. v. Commissioner of Ins., 362 Mass. 34 (1972).

A Bureau witness indicated that automobile insurers made a profit on bodily-injury coverage of approximately 10.9% of premium volume, *824 without including investment income. There were offsetting losses, again neglecting investment income, on property-damage coverages. The failure of the Commissioner to recapture any profit is not attacked in these appeals.

The Bureau argues that the Commissioner’s decision with regard to the second look must be reversed and remanded because of the failure to “state the facts relied upon in making his findings” as required by the statute. G. L. c. 175, § 113B, as amended through St. 1971, c. 977, § 1A. We note, however, that the Commissioner did set out his assessment of the adequacy of the 1975 rates for bodily injury. And remand for further findings would be pointless since even the Bureau does not appear to assert that the bodily-injury experience was egregious.

The claimed inadequacy in the 1975 property-damage rates ranged from 20.5% of the premium for some coverages to 50% for others. The total amount of the 1975 inadequacy was claimed to exceed $200,000,000.

Other provisions of the act selected by the Legislature for immediate effect along with the repealer dealt with the possible consequences of the withdrawal of insurance companies from writing automobile liability insurance in the Commonwealth — a possibility apparently feared imminent at the time of enactment. And the fact that the effective date of the repealer coincided with the resumption of the property damage hearings is suggestive.

Cases regarding the reduction of the period of a statute of limitations (e.g., Mulvey v. Boston, 197 Mass. 178, 182-184 [1908]; Sanford v. Hampden Paint & Chem. Co., 179 Mass. 10, 14 [1901]) generally come down to the same consideration of reliance; however, they do not generally arise in an area of intensive statutory and administrative regulation, as does the present controversy.

The criticism is directed at one point to the collation of particulars resulting in the allowance for profit.

See note 4 supra.