Gibson, J.
In these proceedings, brought pursuant to article 78 of the Civil Practice Act and consolidated by our order, we review a determination of the Superintendent of Insurance disapproving petitioners’ proposed revision of the rates for automobile liability insurance in certain categories and classifications. No rate dispute of this nature has previously been litigated in this State.
Petitioners are rating organizations, licensed by the Superintendent under section 181 of the Insurance Law and existing ‘ ‘ for the purpose of making rates to be used by more than one authorized insurer ” (Insurance Law, § 180, subd. 4). Petitioner National Bureau of Casualty Underwriters acts for certain stock companies, 137 of which wrote automobile liability insurance in New York in 1956. Petitioner Mutual Insurance Bating Bureau represents 39 mutual companies writing such insurance in New York. The members of these two organizations and the nonmember subscribers to their automobile liability insurance rating services in New York write approximately 80% of the automobile liability insurance premiums written within the State. The remaining 20% of these premiums are written by 9 insurance companies which are not members or subscribers of either organization.
In October, 1957, petitioners filed with the Superintendent of Insurance, pursuant to subdivision 1 of section 184 of the Insurance Law, separate schedules of revised rates. The revisions proposed are identical and provide for changes which, calculated in terms of State-wide averages, would increase rates for bodily injury and property damages liability insurance, combined, by 9.5% for private passenger cars and by 5.9% for commercial vehicles, and for certain garage risks would result in an average decrease of 1.8%. As between the 58 rating-territories in the State, the proposed revisions are not uniform and, in fact, involve decreases for passenger vehicles, in one or more classifications, in 27 of these territories and general decreases in certain other categories as well. The rates which would result from the revision sought would also, as heretofore, vary greatly in different territories. For example, the lowest proposed rate is $39.40 in Jefferson County and the highest is $395.60 in Kings County, in each ease for coverage of $10,000 [76]*76per person and $20,000 per accident, for bodily injury, and of $5,000 per accident for property damage. Our general discussion will proceed, however, as does that of each party, on the basis of the State-wide average increases, amounting, as above stated, to 9.5% for passenger cars and 5.9% for commercial vehicles, and without regard to the many variations as between certain categories and classifications and as between rating territories as well.
The regulation of insurance rates is mandated by article 8 of the Insurance Law and that article prescribes and defines the functions of the superintendent and of insurers and their rating organizations in that process. Rates had long been subject to regulation in New York when, in 1944, the Supreme Court of the United States held in United States v. Underwriters Assn. (322 U. S. 533) that insurance companies conducting business transactions across State lines were subject to the regulatory power of Congress under the commerce clause of the United States Constitution (art. I, § 8, cl. 3) and hence, in cases of monopolistic practices, to the provisions of the Sherman Anti-Trust Act (U. S. Code, tit. 15, §§ 1, 2). Following that decision, Congress acted to provide that after June 30, 1948, the Federal anti-trust acts should be “ applicable to the business of insurance to the extent that such business is not regulated by State Law” (U. S. Code, tit. 15, § 1012). In New York, the Joint Legislative Committee on Insurance Rates and Regulation in its 1948 report recommended acceptance of ‘‘ the Federal authorities ‘ invitation ’ to deal affirmatively and effectively with those activities and practices which might otherwise be the subject of Federal regulation.” (N. Y. Legis. Doc., 1948, No. 46, p. 15.) . The Legislature thereupon enacted various revisions of article 8. (L. 1948, ch. 618.) These included the addition, as a part of section 180, of a statement of the purpose of the article as “to promote the public welfare by regulating insurance rates to the end that they shall not be excessive, inadequate, unfairly discriminatory or otherwise unreasonable and to authorize and regulate cooperative action among insurers in rate making and in others matters within the scope of this article.”
Stated in broad outline, the statutory procedure requires that every rating organization, and every insurer which makes and files it own rates, “ shall make rates ” (§ 183, subd. 1), which “ shall be reasonable and adequate for the class of risks to which they apply” (§ 183, subd. 1, par. [b]), giving consideration to, among other factors, past and prospective loss experience, past and prospective expenses, and a reasonable profit [77]*77(§ 183, subd. 1, par. [d]), and shall file with the superintendent schedules of such rates (§ 184, subd. 1). The superintendent shall review such filings ‘ ‘ to determine whether they meet the requirements of [the] article ”, and each such filing “ shall be deemed to meet the requirements of [the] article unless disapproved by the superintendent ’ ’ within a waiting period of 15 days, which may be extended for an additional period not exceeding 15 days. (§ 184, subd. 4.) If, however, the superintendent finds that a filing does not meet the requirements of the article, he shall give ‘ ‘ written notice of disapproval of such filing specifying therein in what respects he finds such filing fails to meet the requirements of [the] article and stating that such filing shall not become effective.” (§ 184, subd. 5.) “Any insurer or rating organization aggrieved by any order or decision of the superintendent made without a hearing” may request a hearing thereon (§ 186-b) and the “ findings, determinations and orders of the superintendent made after notice and hearing * * * shall be subject to judicial review” (§ 187, subd. 3).
Certain conclusions are immediately apparent upon consideration of the statute in the light of the Federal legislation which occasioned its revision in 1948 and in the light, also, of the decision in the Underwriters case (322 U. S. 533, supra) which impelled that Congressional action. Obviously, regulation that would be effective and not merely perfunctory was contemplated, whatever the form of the rate-making process. The statute preserved the concept of co-operative regulatory action by continuing the function of the rating organization, long since recognized as “ a development of insurance arising out of the difficulties of rate making based on the experience of a single company” (Matter of Importers & Exporters Ins. Co. v. Rhoades, 239 N. Y. 420, 423). Proposed revisions would continue to originate with the rating organizations and insurers on the basis of statistical and other data collated by them. There remained absent ‘“an exact statutory formula for rate-making ’ ” as had been proposed on occasion in the past. (2 Benjamin on Administrative Adjudication, pp. 26, 29, quoting Insurance Law Revision, Tentative Draft of 1937, Comment on § 73.) Instead, the prescribed test is, primarily, that the rates be “ reasonable and adequate ’’. (Insurance Law, § 183, subd. 1, par. [b].)
The application of the tests of reasonableness and adequacy necessarily requires the exercise of judgment on the part of the superintendent in analyzing and treating the factual data before him and in reaching his ultimate determination. He is [78]*78not restricted to a rigid formula of his own or another’s devising nor, of course, is he bound to hold rules of past practice immutable, but his determination must, like any other administrative decision, be grounded on proof found by him to be preponderant and which, upon review, will meet the judicial test of substantial evidence. Beyond this restatement of familiar principles, we need not discuss the conflicts in the various theories advanced by the parties as to the Superintendent’s powers and the procedural questions involved in their exercise.
The proposed revisions were considered by the Superintendent, without a hearing, and were disapproved on November 12, 1957 upon three grounds, which he later summarized as follows:
‘‘1. The proposed rate level changes in automobile liability insurance covering private passenger cars are based upon loss experience for policy years 1955 and 1956. This period is too limited and does not constitute a sufficiently credible base.
“2. The loading for general administration expense is predicated upon an unsound basis.
“3. The Compulsory Automobile Insurance Law having been in force only since February 1, 1957, there has not been sufficient opportunity to evaluate the experience thereunder.”
Following a hearing, requested and granted pursuant to the statute, the Superintendent on January 21, 1958 again disapproved the filings, reiterating two of the three conclusions upon which his initial disapproval was predicated. Thus, the period of loss experience for the policy years 1955 and 1956 was again found too limited to constitute a sufficiently reliable or “ credible ” base, for the reasons that only 40% of the total incurred bodily injury losses shown for the policy year 1955 represented actual payments, 60% being in reserve and representing'estimates; and but 6% of such losses for the policy year 1956 represented actual payments, 94% being still in reserve and thus representing estimates. The Superintendent considered that the loss experience for the policy years 1952 through 1956 would constitute a ‘‘ more realistic and credible base”, as reflecting “a more mature development and a far lesser degree of judgment factors ”, and found that on the basis of that five-year average experience the present rates were entirely adequate.
The additional ground of disapproval was a finding of inequity in the method employed for allocation or loading of general administration expenses, whereby the percentage of premium allowed therefor would produce $4.57 per car in New York for general administration expenses while the correspond[79]*79ing allowance country-wide, applied to the lower average country-wide premium, provided $2.94 per car.
Respondent states in his brief that the further ground stated in his initial denial of approval — the absence of experience as to the effect of the compulsory insurance statute (Motor Vehicle Financial Security Act, Vehicle and Traffic Law, art. 6-A, eff. Feb. 1, 1957)—was not a ground of his final determination, although referred to and discussed therein. The record does not, in any event, support any inference that loss experience would be improved as a result of the act.
The basic and difficult problem of rate making is, of course, that of accurately predicting actual losses, not merely for the future period to which the rates proposed will be applicable but, also, for the last two available policy years which, in the past, have been used as the measure of loss experience to be applied to future rates. The usual term of an automobile policy being one year, experience data is collated for all policies written in a calendar year, with the result that a very large proportion of the losses attributable to policies written in a particular calendar year will not be incurred until some time in the next year. The latest policy year considered in connection with these filings was 1956 and, of course, some policies issued in 1956 expired as late as December 31,1957, subsequent to the rate filings made in October, 1957. Further, of course, a large proportion of even the known claims remained unpaid at the end of the calendar year following that in which the policies applicable to those claims were written. Finally, some proportion of the claims (especially those for bodily injury) attributable to a particular policy year will, because of litigation or for other reasons, customarily remain undisposed of for several years. As has been noted, the superintendent found that at the time of these filings but 40% of the total incurred bodily injury losses for the policy year 1955 represented actual payments as did but 6% of those for the policy year 1956. Necessarily, therefore, the amounts of the losses incurred but not paid and of the losses to be incurred during the unexpired portion of the later of the two policy years were estimated and this was done on the basis of past experience and adjusted by the application of certain factors customarily used.
The identical 1957 rate filings made by petitioners were prepared upon the bases customarily employed for many years past. The process requires provision, first, for expenses of operation such as commissions and like production costs, general administration expenses, taxes and fees, and audit and [80]*80inspection costs, plus a 3.5% margin for underwriting profit and contingencies, and, second, for losses. The provision for expenses was less than that made in the last prior filings (by-reason of a decreased allowance for production costs) so that in the case of bodily injury the expense provision was reduced from 45.48% to 40.97% of the premium, leaving 59.03% of future gross premiums available for payment of future losses. By a similar reduction of the expense item, the portion of the property damage premium allocated to losses was increased to 56.53%. Petitioners then computed, again by customary methods, the “ experience loss ratio ”, which is the percentage of the present premium currently being expended for losses. Division of the figure found as the experience loss ratio by the percentage of the future rate that will be available for losses yields the ratio of the required future rate level to the present rate level, and deduction therefrom of unity (the present rate level) furnishes the percentage of increase required.
The accuracy of the statistical data underlying the computations is not questioned, insofar as it is derived from actual experience. As respects the provision for future losses, the question is, then, as to the correctness of the experience loss ratio computed on the basis of the mean experience of the last two available policy years. This is the crux of the controversy as to the first ground of the Superintendent’s disapproval.
As has been noted, the determination stresses the substantial portion of losses necessarily estimated. Petitioners adduced evidence, however, that past experience had shown their computation of the experience loss ratio substantially accurate upon application of the “ loss development factor ” derived by comparison of the losses actually paid, in a year in which substantially all claims had been paid, with the reserve which had actually been established. There was testimony that over a 10-year period the average loss cost was within five cents of the estimate for bodily injury, or $30.03 as against $30.08, and within 13 cents of that for property damage, or $11 as against $11.13.
This result was obtained upon examination of the yearly loss costs reported for each year of the 10-year period and upon comparison, in each case, of the loss costs reported as of 12 months with those reported as of 36 months, in the case of [81]*81bodily injuries, and as of 24 months in the ease of property damage. The respondent attacks these computations as not developed for longer periods. He contends, on the basis of a compilation published by the Department of Insurance and in evidence, that development beyond three years and to the end of five-year and six-year periods demonstrates that the loss ratios at the end of the longer periods are from 3 to 4% lower than the corresponding loss ratios at the end of three years. Stated in simpler terms, the contention is that when all claims incurred in a particular year have been finally settled, their actual cost is shown to be 3 to 4% less than their assumed cost at the end of three years.
The compilation upon which this argument is predicated appears on one page of a 143-page booklet, other portions of which were referred to in the testimony. The particular compilation to which respondent adverts in his brief was not discussed on the hearing nor was it specifically referred to when counsel for respondent offered the booklet in evidence. It was received, without objection, under a limitation expressed in terms not entirely clear. In offering it, counsel for the Insurance Department said, in part: “ Mr. Donovan referred to certain pages and I would like to incorporate in the record all references made thereto referring to automobile bodily injury and property damage liability insurance.” The record continues: “Mr. Friendly: As so limited, I have no objection. Hearing Officer: There is no point in putting the entire exhibit in. Mr. Gross: Only as to the references I made to automobile bodily injury. Mr. Donovan: No objection.”
Petitioners advance reasons of apparent cogency, which might well have been subjects of testimony, for according the exhibit less than full weight. It is therefore unfortunate that it was not analyzed on the hearing, since there was little else in the evidence to impeach petitioners’ proof as to the accuracy of the loss development factors underlying their ratings. We may, however, properly take cognizance of one of the objections now urged by petitioners,— that while the rate filings here in controversy involve only so-called statutory limits losses ($10,000/ $20,000 bodily injuries and $5,000 property damage) and petitioners’ calculation of the loss development factor was concerned only with losses within those limits, the questioned compilation covers total limits losses. It is argued that with respect to losses in the latter category there is greater possibility of error in estimating reserves than in eases of statutory limits losses, apparently on the theory that claims for injuries [82]*82so serious as to involve damages in excess of statutory limits are more difficult to evaluate and their ultimate disposition, after verdict. or adjustment, more likely to entail disproportionately large variations from initial estimates.
The decision does not indicate the weight accorded the disputed exhibit by the superintendent in finding that ‘ ‘ in forecasting future expectations reliance upon the loss experience of the companies for the [two] most recent policy years is too limited and is unsound.” Indeed, the supposed compelling significance of the exhibit is entirely a matter of argument on this appeal. The decision itself refers specifically to a statement prepared by the bureaus, at the Superintendent’s request, showing experience for private passenger cars for the five-year period 1952 through 1956, which statement reflects, in some part, the loss development procedure now attacked but also gives effect, as to bodily injury, to loss development to 60 months by use of a loss development factor of 0.984. If it be assumed that the disputed compilation was properly received within the limitations imposed by counsel upon the exhibit of which it is a part, its weight would be for the Superintendent’s determination. It would follow that the credibility, otherwise unimpeached, of petitioners’ proof of the accuracy of their loss development procedures would also be within an area of decision into which we may not intrude. Here, however, mathematical factors are concerned and it is apparent that the percentage of error in the loss ratios asserted by the Attorney-General upon his analysis of the exhibit is insufficient, alone or in combination with the supposed excessive loading for general administration expense, to discount all of the proposed increase and to establish the adequacy of the present rates.
The particular factual dispute which we have discussed was not resolved by specific findings. The determination proceeded instead on the broad basis of administrative judgment and policy. We thus reach the question whether the Superintendent properly adopted and applied a new standard or measure of computation—the five-year experience base—whereby it was determined that the current rates were adequate.
We may not question the Superintendent’s general authority, as a matter of administrative judgment, to reject the basis of two-year loss experience, however long accepted, and to require the use of such other standard as should be reasonable. It is clear, however, that if the application of any standard, however reasonable in general concept it may seem, will cause an illegal and unreasonable result in a particular case, the principle of [83]*83administrative discretion in the formulation of standards does not render that result invulnerable to attack. The crux of this case is not the propriety, as an abstract proposition, of one rate-making basis or another, but is, rather, the question whether the rates which remain in force by reason of the disapproval of the revisions are in fact adequate and those proposed unreasonable. (Insurance Law, § 184, subd. 5; § 183, subd. 1, par. [b].)
In our view, there is no substantial evidence to support a finding that no increase is necessary to render the rates adequate. If the insurers are entitled to some increase in the current rates, it follows that those rates are no longer adequate.
It must be presumed that the current rates were reasonable and adequate when approved by the Superintendent. (People ex rel. Wellington Apts. v. Miller, 288 N. Y. 31, 33; Anderson v. Erie R. R. Co., 223 N. Y. 277, 283.) The purely factual or statistical data submitted in support of the claim of subsequently increased and increasing loss costs is not questioned. The dispute is as to the accuracy of the method and result of petitioners’ loss development. As we have stated, modification of the proposed new rate to give effect to respondent’s contentions as respects both (1) the margin of error in the “ incurred-earned loss ratios ” compiled in his exhibit and (2) the supposed excessive loading for general administration expenses would not have the effect of absorbing the entire amount of the proposed increase, which the otherwise uncontroverted and unimpeached evidence seems clearly to warrant.
Our conclusion that loss experience has worsened since promulgation of the current rates appears to be strongly and additionally fortified by all the evidence of trends which the record contains. Because of delay incidental to the controversy, opportunity was afforded to develop additional proof of this nature after the rate filings had been made.
During the 15-month period ending March 31, 1957, petitioners’ members’ and subscribers’ average paid claim costs in New York increased by 7.9% for bodily injury claims and by 7.2% for property damage claims. As of June 30, 1957, such average paid claim costs for bodily injury were 8.2% higher, and those for property damage 9.9% higher, than for the year ending December 31, 1955.
A frequency chart of motor vehicle accidents in New York, involving bodily injuries, charted per 100 cars registered, shows a gradual upward curve from December 31, 1954 to December 3.1, 1956 and a sharp uptrend of 11.1% from the latter date to [84]*84September 30, 1957. During the latter period, accidents involving property damage increased by 7.5% after a previous rise and fall.
Similarly, the number of persons injured in automobile accidents in New York, per 10,000 cars registered, increased during these periods. Again there is a gradual upward trend followed in the period from December 31, 1956 to September 30, 1957 by a sharper increase, computed as 12.6%.
The charted upward trend of plaintiffs’ verdicts in the Supreme Court may be of less significance, since the data, based on Judicial Conference Reports, combines all types of cases.
The compilation of underwriting results for New York State of petitioner National Bureau’s members and subscribers showed gains in 1953 and 1954 and losses in 1955 and 1956, the loss in 1955 being approximately $5,900,000 and that in 1956 rising to approximately $24,000,000, the latter figure constituting a loss of 10.1% on 1956 premiums of approximately $238,500,000. The results of the Mutual Bureau’s members’ and subscribers’ operations do not appear.
Petitioners offered evidence of the combined loss experience of their members and subscribers for bodily injury and property damage, combined, in the form of average loss cost per passenger car at statutory limits ($10,000/$20,000/$5,000) for -each of the years 1952 through 1956. Even if effect be given to respondent’s criticism of petitioners’ methods of loss development, the data shown is impressive and may not be ignored. In 1951, 1952 and 1953 the average losses per car were relatively stable, being $48.03, $47.60 and $49.34, respectively. Thereafter there was marked acceleration in the loss rate, losses rising to $53.25, $59.80 and $62.64 for 1954, 1955 and 1956. These figures compare with a loss provision of $56.02 in the present rates, and a loss provision of $61.34 proposed by the rate filings in dispute and, according to the computations, would imply a loss provision of $54.06 under the Superintendent’s basis of a five-year loss experience. If the latter basis were sound, then, despite all the evidence to the contrary, even the present rates would have to be appraised as excessive.
The effect of the application of the five-year standard is, of course, to weight the result with the relatively lower losses in the three earlier years and to give less effect to the substantially higher losses in the last two years and, also, to that extent to ignore the marked acceleration which commenced in 1954, continued through 1955 and 1956 and, according to all the evidence of trend, would continue in 1957.
[85]*85Again giving effect to the presumption that the current rates were lawfully established and, therefore, neither unreasonable nor more than adequate, we find in the record no substantial evidence supportive of the determination that those rates remain adequate. That determination would be sound only if a very sharp downward loss trend could have been predicted for 1957 but, as has been shown, every indication apparent from the evidence offered in January, 1958 was directly to the contrary.
We are constrained, therefore, to annul the determination and, necessarily under the circumstances, to remit the cases for further proceedings.
Our conclusion renders unnecessary any extended discussion of the second ground of the Superintendent’s disapproval—that the loading for general administration expenses, by allocation of a percentage of the premium derived from country-wide experience, is predicated upon an unsound basis. The decision states that the use of this basis would produce an average of $4.57 per car in New York as against a country-wide average of $2.94 per car. “ This inequity ”, the decision continues, “ calls for corrective action.” These observations do not seem to us to partake of the nature of a finding. The test is not whether inequality exists as between the allowance for expense in New York and like allowances elsewhere but whether the proposed expense item is accurate and reasonable. It may well be that unit cost rather than a percentage figure might more fairly be applied but this record neither sustains that conclusion nor affords statistical evidence upon which a unit cost could be determined. On the contrary, the computation of the administration cost factor in the proposed rates was upon statistical data which petitioners were required by the Superintendent and his predecessors to report for the purposes of such computation and the absence of data especially applicable to New York was due to the Superintendent’s failure to require the submission of such evidence. (See Insurance Law, § 183, subds. 6, 8.) We need not determine whether this second ground of disapproval constituted arbitrary action on the part of the Superintendent since we find it, in any event, insufficient to constitute a factual finding of excessiveness. Insofar as it may have been intended to foreshadow a change of policy, it is not before us.
The nature and content of the decision in these cases, based on broad expressions of judgment and administrative policy, serve to point up very sharply indeed the clear necessity for specific findings upon the factual issues, made in terms of the [86]*86statistical and monetary factors involved. Such findings are necessary in aid of intelligent judicial review and that necessity is emphasized when, as here, the factors involved are numerous and complex, but disapproval may properly rest on consideration of but some of them and may not require a determination of all of the issues presented. (Matter of Elite Dairy Prods. v. Ten Eyck, 271 N. Y. 488, 498; 1 Benjamin on Administrative Adjudication, p. 251 et seq.)
The statute requires that in an initial disapproval of a rate filing the superintendent shall specify “ in what respects he finds such filing fails to meet the requirements of this article [8] ”. (Insurance Law, § 184, subd. 5.) In essence, the stated ground of the initial disapproval was “ that the proposed rates are deemed to be excessive.” As has been noted, petitioners thereupon demanded and received a hearing, pursuant to the statute, and the superintendent’s determination followed.
The statute further provides: ‘‘ The findings, determinations and orders of the superintendent made after notice and hearing, pursuant to the preceding sections of this article shall be subject to judicial review.” (Insurance Law, § 187, subd. 3.) The determination was clearly deficient in omitting “ findings ” such as were contemplated by the statute or such as are necessary to judicial review, regardless of statutory provision therefor. (See 1 Benjamin on Administrative Adjudication, p. 252.) The broad and general conclusions of the determination which have been discussed were followed merely by the statement 'that ‘ ‘ nothing of probative force has been added to warrant a modification of the conclusion reached by the Superintendent of Insurance in his letters of disapproval dated November 12, 1957.” The testimony and the great and complex mass of evidentiary data submitted to the superintendent required precise professional analysis and, beyond that, findings in such detail and of such particularity as to furnish an adequate basis for review.
The difficulties inherent in the statutory scheme of regulation, whereby the superintendent may not directly fix rates but must approve or disapprove proposed rates in toto, furnish an additional ground for requiring findings in terms of the statistical and monetary factors involved. In many cases the area of dispute might thus be narrowed and the treatment of new filings expedited, after a decision adverse in part, and without the necessity of judicial review.
The determination should be annulled and the cases remitted to the Superintendent of Insurance for further proceedings not inconsistent with this opinion, with costs to petitioners.
[87]*87Foster, P. J., Coon, Herlihy and Reynolds, JJ., concur.
Determination annulled and cases remitted to the Superintendent of Insurance for further proceedings not inconsistent with the opinion herein, with costs to petitioners.
For example, the experience loss ratio with respect to the premium for bodily injury coverage was computed at 64.92 and divided by 59.03 (the percentage of the future premium dollar available for losses), the quotient of 1.10 indicating a future rate of 110% of the present premium, or an increase of 10%.