ROBERT P. SMITH, Jr., Chief Judge,
dissenting.
In my opinion Rule 4-43.03 permissibly exercises the responsibility placed in the *915Department of Insurance by chapters 626 and 627, Florida Statutes (1979), to interpret and then implement the “unfair discrimination” prohibition expressed in section 626.9541(15)(h). The economic impact statement accompanying the rule is also adequate, and any defect in it is formal only, not impairing either “the fairness of the proceedings or the correctness of the action.” Secs. 120.54(2)(a), 120.68(8), Fla. Stat. (1979). I therefore would reverse the hearing officer’s order invalidating this rule.
We have said many times that a prime goal of the Administrative Procedure Act of 1974 is “to encourage agencies of the executive branch to interpret statutes in their regulatory care deliberately, decisively, prospectively, and after consideration of comments from the general public and affected parties — that is, to interpret their statutes by rulemaking.” State Department of Health and Rehabilitative Services v. Framat Realty, Inc., 407 So.2d 288, 241 (Fla. 1st DCA 1981). The incentives for rulemaking were described at length in McDonald v. Department of Banking and Finance, 346 So.2d 569 (Fla. 1st DCA 1977) and are illustrated repeatedly by decisions since.
Rulemaking is therefore particularly favored as agency action. Correspondingly, our decisions firmly require a high degree of deference toward that kind of agency action, both in its function of selecting a permissible interpretation of the empowering statute and in its function of deciding how the statute so interpreted shall be implemented.
Though adjudication has other functions, notably determining particular facts, adjudication and rulemaking share those two essentially policymaking functions: first to determine which of several permissible conceptual interpretations of a statute is most consonant with the statutory purpose; then to determine and announce within the range afforded by the statute so interpreted, how and to what extent the statute shall be carried out in practical affairs.
Sometimes our analysis of agency action tends to blur together or telescope these two functions, so that they appear as one. Especially is that likely when the subject is a rule and the occasion is a challenge. For a rule speaks generally, briefly and prospectively, in style much like a statute. Though it of course is not a statute, and is dependent for its life upon a statute, the very appearance of a rule may provoke a certain skeptical eye, as though viewing an expression that pretends to have, though it cannot have, the stature and independence of legislation.
When the rule is thus suspect, we are apt to telescope its two distinct functions— choosing a permissible conceptual interpretation of the statute, then determining how far that practically shall be implemented— in concluding that the rule is beyond the statute. This error commonly attributes to the rulemaker a claim his rule does not make, that the final implementing command of the rule was inexorably intended by the legislature. Had the legislature so intended, the mispremised rebuttal goes, the legislature would have said so. What the rebuttal overlooks is the delegatory nature of much regulatory legislation and the progressively deductive method of rulemak-ing.
Here, Rule 4-43.08 represents, though characteristically of a rule it does not articulate as would an agency order, a permissible choice from among permissible conceptual interpretations of section 626.-9541(15)(h).
The permissible interpretations of § 626.9S41(15)(h)
By chapter 77-468, the Florida Insurance and Tort Reform Act of 1977, the legislature comprehensively enacted, reenacted, revised and reorganized numerous statutes pertaining to those subjects. Section 19 of the act added to section 626.9541(15), Florida Statutes (1976 Supp.) the new subsection (h) on which this controversy is centered:
(h) No insurer shall, with respect to premiums charged for automobile insurance, unfairly discriminate solely on the basis of age, sex, marital status or scholastic achievement.
*916Section 626.9541(15)(h), Fla.Stat. (1977) through (1981).
The words of subsection (h), like those of so many regulatory statutes, are susceptible of being interpreted, systematically and true to the statutory words, in several permissible ways. Because the key words— “unfairly discriminate solely” — are designedly general, the choice of interpretation is neither obvious nor self-executing. Refinement and exposition of the meaning^) within the words, by someone, is required. This case largely has to do with whom that someone is, whether the Insurance Commissioner and the Department of Insurance, or one of the hearing officers of the Division of Administrative Hearings, or three of the judges of this court.
The permissible interpretations of subsection (h) appear to fall into two major groups, depending on the effect to be given the word “unfairly.” The first major group would read “unfairly” as descriptive not restrictive, that is, as declaring “unfair” per se every discrimination solely based on “age, sex, marital status or scholastic achievement.” Nothing in the words themselves prevents that choice of meaning for “unfairly,” and disreputable connotations in the word “discriminate” add legitimacy to this choice of meaning for “unfairly”: any discrimination based solely on those characteristics “unfairly discriminate[s].” Within this major category are at least four alternative applications of subsection (h) as a whole:
First, reading “unfairly” as descriptive of the prohibited discrimination, not restrictive of its definition, one may legitimately read the remaining words in subsection (h) as emphasizing “solely” in a literal sense, so that unfair discrimination consists only of discrimination based “solely” on one of the characteristics that are listed with a final disjunctive “or.” Thus, discrimination based solely on “age” or any other single characteristic listed would be deemed unfair and forbidden, but discrimination based collectively on two or more or all of those factors would not be unfair. In combination, one might argue, two or more of the listed characteristics ameliorate the unfair discrimination that is based “solely” on one of them.
Second, again reading “unfairly” as descriptive not restrictive, one may choose a different course, interpreting subsection (h) as prohibiting discrimination based “solely” on one or more of the four listed characteristics, singly or in any combination.
Third and Fourth, still reading “unfairly” as before, one might regard the phrase “unfairly discriminates solely on the basis of” as addressing, in practical terms, insurance rates and premiums that are fixed “solely on the basis of” the listed characteristics, either [one alone, as in First] or [one or more, as in Second]. This view would regard an insurance premium as “unfairly discriminatory” only if it were wholly determined by [one] [one or more] of the listed characteristics, not if it were determined also by other unlisted factors known to ratemaking, such as miles traveled [by the principal insured automobile] [by all owned automobiles] [by the principal driver] or the geographic region where the automobile is [garaged] [principally driven].
The other major category of alternative interpretations would regard the word “unfairly” as restricting or further defining the prohibited discrimination, not merely descriptive of discrimination otherwise defined in subsection (h). By these views the term “unfairly” invokes other considerations, outside subsection (h) itself, that contribute to identifying the discrimination proscribed. In this category, adding to the range of choices, are:
Fifth, the statute may be read as proscribing only those discriminations that “unfairly,” in some technical sense peculiar to insurance protocols, employ the listed characteristics as described First above.
Sixth, again reading the statute as proscribing only those discriminations that are “unfair” by technical insurance protocols, one may interpret the statute as aimed against only those “unfair” discriminations that employ the listed factors as described Second above.
*917Seventh and Eighth, still reading “unfairly” as referring only to technical insurance protocols, one may interpret subsection (h) as addressing only those discriminatory rates and premiums that are fixed as described Third or Fourth above.
Ninth, reading “unfairly” as restricting or further defining the prohibited discrimination, not merely descriptive of it, one may treat the term “unfairly” as calling upon general normative standards [independent of] [in addition to] technical insurance protocols. So elaborating “unfairly discriminate,” one again might restrictively interpret “solely” as in First above.
Tenth, reading “unfairly” as in Ninth, one may read “solely” as in Second.
Eleventh and Twelfth, reading “unfairly” as in Ninth, one may interpret the balance of subsection (h) as in Third or Fourth.
These alternative interpretations of subsection (h), and the gradations between them, may be arrayed as follows:
California’s Justice Traynor said it well: “Rare are the statutes that rest in peace beyond the range of controversy. Large problems of interpretation inevitably arise. Plain words, like plain people, are not always so plain as they seem.”
The Department’s interpretation of subsection (h)
The Department’s conceptual interpretation of subsection (h), evidenced by the rule itself, is as stated Tenth above. Again, the words of the statute:
(h) No insurer shall, with respect to premiums charged for automobile insurance, unfairly discriminate solely on the basis of age, sex, marital status or scholastic achievement.
The rule reflects the Department’s view that “unfairly” further restricts or defines the discrimination aimed at, and is not merely descriptive of discrimination otherwise defined in subsection (h). Evidence of this, the rule does not list “age” among the classifications to be avoided in ratemaking. Had the Department considered “unfairly” merely a descriptive term, the Department could not have considered itself free to omit “age” from the rule’s list of proscribed classifications. Indeed, had the Department so interpreted “unfairly,” it is not clear why a rule would have been thought necessary at all, for the other interpretative problems are not nearly so substantial, and could easily have been resolved through adjudication.
The rule further reflects an interpretation of “solely” in subsection (h) as referring to discrimination based “solely” on any one or more of the four listed characteristics, as stated Second and subsumed in Tenth.
Finally, as stated Tenth, the Department’s interpretation of “fairly” invokes both technical insurance protocols, relating to actuarial and predictive accuracy, and general normative standards of what is fair. In its rulemaking, the Department used these two fairness standards harmoniously and conservatively, as calling on the Department to scrutinize more closely, by “fairness” standards normatively conceived, the actuarial or predictive accuracy of classifications based on age, gender, marital status and scholarship. Thus, in a noncon-stitutional but nonetheless very real sense, subsection (h) was interpreted as having designated the specified classifications as suspect. The legislative omission of, say, “miles driven,” was not accidental, though we can imagine that discrimination based on that factor can also be done unfairly. Thus, as in constitutional equal protection analysis that is familiar to courts, which *918also is essentially a “fairness” inquiry, subsection (h) was interpreted as requiring the Department to be assured that discrimination based solely on the stated characteristics is “fair” in the sense of closely predicting the accident-proneness of those prejudiced by falling into one of those classifications.
Again the Department’s omission of “age” from the rule’s list of prohibited classifications speaks worlds: the rulemaking process simply convinced the Department that discrimination based on age closely predicts, satisfactorily at least for now, the accident-proneness of those so classified.
Though such elucidation is unnecessary in rulemaking, the Department’s promulgating document confirms at some length this analysis of the statute and the rule. Excerpts from Florida’s Automobile Insurance Rate Classification. System: Report to the Insurance Commissioner and The Commissioner’s Orders and Findings, dated October 1, 1979, are set out in the Appendix to this dissenting opinion.
In my opinion, therefore, the Department predicates its rule on an entirely permissible interpretation of subsection (h). On that subject, that should end the inquiry. The controlling decisions are quite clear:
In Jax Liquors, Inc. v. Division of Alcoholic Beverages and Tobacco, 388 So.2d 1306 (Fla. 1st DCA 1980), this court rejected a rule-challenge asserting, among other things, that the rule “exercises regulatory powers not granted the Division by the Tied House Evil statute,” saying that “[t]he evidence submitted by appellant to the hearing officer does not show that Rule 7A-4.13 is without a rational relationship to the Tied House Evil statute.” 388 So.2d at 1307, 1308. The court declared that “appellant has assumed a heavy burden of proof in its attack upon the Rule,” noting that it was not a case of “policymaking by case-by-case adjudication,” nor an appeal from section 120.54 rulemaking proceedings “in which the agency’s duty is to supply a full and fair opportunity for affected persons to attack or comment on the proposed rule by debate and appropriate evidence, and our duty is to assure that the agency has done so.” 388 So.2d at 1307.
In ABC Liquors, Inc. v. Department of Business Regulation, 397 So.2d 696, 697 (Fla. 1st DCA 1981), this court stated that “APA processes thus lend added strength to the principle stated by the Supreme Court in King v. Seamon, 59 So.2d 859, 861 (Fla.1952)”:
The contemporaneous construction placed upon a statute by the officials charged with the duty of executing it should not be disregarded or overturned by this court except for the most cogent reasons, and unless clearly erroneous.
In State Department of Health and Rehabilitative Services v. Framat Realty, Inc., 407 So.2d 238, 241-2 (Fla. 1st DCA 1981), the court reversed a DOAH hearing officer’s order invalidating a rule as beyond the agency’s statutory authority, saying:
Whether the Department’s interpretation of section 381.272(7) is the only possible interpretation of the statute, or the most desirable one, we need not say. It is within the range of permissible interpretations of the statute, and that interpretation has acquired legitimacy through rulemaking processes in which those challenging the rule fully participated or had an opportunity to participate.... We have repeatedly found that other APA processes press the executive inexorably toward rulemaking....
When as here an agency has responded to rulemaking incentives and has allowed affected parties to help shape the rules they know will regulate them in the future, the judiciary must not, and we shall not, overly restrict the range of an agency’s interpretative powers. Permissible interpretations of a statute must and will *919be sustained, though other interpretations are possible and may even seem preferable according to some views. If the rule binds too tightly to suit them, the appellee developers have their proper remedy in the representative and politically responsive branches, the legislative or executive, but not in the judiciary, nor in Section 120.56 rule challenge proceedings before a hearing officer.
In Department of Administration, et a1. v. Nelson, 424 So.2d 852, 858 (Fla. 1st DCA 1982), this court reversed a hearing officer’s order invalidating an agency rule as being beyond its legislative authority, and reiterated the principle “that when the agency committed with statutory authority to implement a statute has construed the statute in a permissible way under APA disciplines, that interpretation will be sustained though another interpretation may be possible. When the agency so interprets the statute through rulemaking, the presumption of correctness is stronger.”
The Department’s implementation of subsection (h) as so interpreted.
The second function of rulemaking, beyond interpreting the empowering statute conceptually, is to implement that statutory concept giving it practical and predictable effect. Obviously, the most sensitive decision to be made by the Department concerning the implementation of subsection (h), construed as Tenth above, was that of determining whether suspect classifications based on age, gender, marital status or scholarship, as distinguished from non-suspect classifications based directly on driver experience and propensities, pass close scrutiny for actuarial or predictive accuracy.
Determining where the line is to be drawn between fair and unfair actuarial predictions based on the enumerated factors is essentially a matter of agency policy, carrying into effect the legislature’s “fairness” standard against the suspect classes named in subsection (h). In APA terms the Department was entitled to choose between rulemaking or adjudication of particular cases in formulating its policy within the bounds of subsection (h) permissibly interpreted. McDonald, supra. Obviously, the Department decided that adjudication on individual and successive rate applications would be time-consuming for all, expensive for the state and for the industry, repetitive, and wanting in predictability. Obviously, too, if the controlling term “fairness” was to be left unimplemented until a number of adjudications fleshed out a definitive meaning, the statute already spoke sufficiently for that, and there was no need for rulemaking.
Thus, responsively to rulemaking incentives generally, and to practical incentives in the making and approving of automobile insurance rates, the Department embarked upon rulemaking proceedings characterized by Insurance Commissioner Gunter as “one of the most exhaustive studies of the automobile insurance rate classification system undertaken in recent years, certainly in this state.” App., 930. The promulgating document, Florida’s Automobile Insurance Rate Classification System, App. infra, abundantly documents the rulemaking study and its conclusions.
We are not informed, and it is unnecessary that we be satisfied, whether it might have been desirable for the Department to prescribe by rule some quantitatively verifiable threshold of accident-prone predictability — if there is such a thing — that would differentiate “fair” from “unfair” discrimination based on gender, marital status, and scholarship. See App., 934. It is plain *920enough from the rule and from the promulgating document that the Department found, as the legislature suspected, that all four of the specified classifications are “proxy” (App., 937) or “surrogate” (App., 937) classifications at best, and that three of the four, excluding “age,” too readily attribute accident-proneness to many persons caught in those classifications. App., 937-8. The overbroad classifications are used, the Department found, mainly because they are convenient to an industry in whose collective interest it is to continue their accustomed ways. App., 935.
Therefore, as a matter of policy within the constraints of subsection (h), permissibly interpreted, the rule characterizes all discriminations based on the three proxy factors as unfair actuarially. This policy judgment was required by subsection (h), was verified by rulemaking proceedings, was announced by the rule, and is now settled beyond legitimate dispute by the hearing officer or by this court. It is not our concern whether the Department might have adopted an intelligible rule permitting a little bit of discrimination based on those factors, or a moderate amount, or a whole lot, rather than none at all. The choice of implementing policy under subsection (h), permissibly interpreted, was wholly the Department’s.
This the decisions make clear:
In State Department of Agriculture and Consumer Services v. Denmark, 366 So.2d 469, 470 (Fla. 4th DCA 1979), the district court noted that “it would be an endless, and nigh impossible, task for the legislature itself to see to each and every necessary regulation which is formulated to implement and administer the laws passed. Therein lies the purpose and authority of the legislature and the executive.” The court sustained a Department of Agriculture rule quarantining horses for a certain disease that legislation had found a threat to Florida’s equine industry. It was contended that the rule-administered test “does not show the presence of [the disease] but merely the presence of antibodies in the bloodstream.” 366 So.2d at 469. Finding no “clear showing ... that the rules and regulations promulgated under the statutes were arbitrary or unreasonable or not consistent with the mandate of the legislature,” the court reversed a circuit court order invalidating the rule. 366 So.2d at 471.
In Florida Canners Association v. State of Florida, Department of Citrus, 371 So.2d 503 (Fla. 2d DCA 1979), the court considered and rejected a challenge to a Department of Citrus rule requiring the printed word “Florida” or the equivalent certification mark on processed grapefruit products packed in retail containers in Florida. The court characterized the challengers’ argument “that there is not competent substantial evidence in the record” supporting the rule as “amount[ing] to an attack on the wisdom of the rule.” 371 So.2d at 519:
They argue that there is no proof that the rule will accomplish the legislative objective. Further, they point to testimony of their advertising expert that designation of Florida origin on grapefruit products is undesirable and may have an adverse effect on sales.
But, obviously, there can be no guarantee of results in any decision on a matter of advertising. An exercise of police power will not be struck down because a law or regulation is unwise, improper, or out of harmony with a particular school of thought. [Citation omitted]. We feel that the evidence in the record as to the anticipated increase in competition from non-Florida grapefruit is sufficient to justify Respondent’s decision to require that Florida grapefruit products be identified in the marketplace. The wisdom of that decision is not a matter of judicial concern.
In Polk v. School Board of Polk County, 373 So.2d 960, 962 (Fla. 2d DCA 1979), the court rejected a rule-challenge to rulemak-ing action by a county school board restructuring high school attendance zones. The court stated:
Appellant argues first that there was no competent substantial evidence to support the school board’s adoption of the *921superintendent’s plan. Actually, there was no evidence at all in a judicial sense, because the determination of school boundaries was a quasi-legislative decision and those who presented their views at the public hearing did so in the context of a town meeting. Judicial review of quasi-legislative action is more limited than that of quasi-judicial action. Broward County v. Administration Commission, 321 So.2d 605 (Fla. 1st DCA 1975). The agency rule-making function involves the exercise of discretion, and absent a flagrant abuse of that discretion a court may not substitute its judgment for that of the agency. Section 120.68(12), Florida Statutes (Supp.1978); Citizens of Florida v. Mayo, 357 So.2d 731 (Fla.1978).
In Bowling v. Department of Insurance, 394 So.2d 165, 174 (Fla. 1st DCA 1981), this court compared the presumption of validity attending an agency rule with the deference to be accorded agency adjudication:
If the matter in issue were one of Department policy, a Department rule would be well-nigh conclusive, and a poli-cymaking order based on an adequate record foundation in this case would command great deference.
In Florida Commission on Human Relations v. Human Development Center, 413 So.2d 1251, 1253-54 (Fla. 1st DCA 1982), this court reversed a hearing officer’s order invalidating a rule as beyond the agency’s statutory powers. Referring to the “long road” traveled in that controversy, the court stated:
Along the way, the issue sub judice was before the Commission which concluded that it had the power to issue investigative subpoenas. Its conclusion is persuasive and entitled to great weight by this Court. Specifically, the statutory implementing proposals having made their way through the rule-making process, in which those challenging the rule fully participated or had an opportunity to participate, strengthens the case for judicial deference.. . .
... Where lawful rule-making authority is clearly conferred or fairly implied, and is consistent with the general statutory duties of the agency, a wide discretion is accorded it in the exercise of such authority. [citations omitted.]
In Moncrief v. State Commissioner of Insurance, 415 So.2d 785, 790 (Fla. 1st DCA 1982), this court stated, when reviewing agency adjudication:
[T]his is a case in which policy-making by rules is preferable to orders. This is so because the Department seeks to establish an industry-wide policy requiring bondsmen to maintain offices open to the public during certain specified hours. As the matter in issue is one of Department policy, a Department rule would be well-nigh conclusive.
In Cortese v. The School Board of Palm Beach County, 425 So.2d 554, 557-558 (Fla. 4th DCA 1982], the court upheld rulemaking by the board and explicitly endorsed the standard for review stated by the second district court of appeal in Polk, supra, that “[t]he agency rule-making function involves the exercise of discretion, and absent a flagrant abuse of that discretion a court may not substitute its judgment for that of the agency.”
The court wrongly preempts the Department’s rule and substitutes its own: the status quo ante
These proceedings seem to me flawed by rather serious misconceptions of the roles of hearing officer and court in rule-challenge proceedings. These were not proceedings for an agency order expounding policy satisfactorily on a record foundation of evidence. The rule itself and its antecedent proceedings are fully equivalent to that. This, therefore, should have been a simple matter of determining whether subsection (h) may permissibly be conceived as calling for the Department’s closer scrutiny on “fairness” grounds of the specified classifications, to verify their claimed accuracy in predicting accident-proneness. If the statute bears that as a permissible interpretation, then a rule proscribing further re*922liance on those classifications must be said to have concluded that they do not stand that closer scrutiny and that, being over-broad, they unfairly discriminate in the way forbidden by the statute. A policy decision of that sort, being reasonably related to the statute as permissibly construed, should readily be sustained under the authorities cited above.
Instead, we have here something else entirely, which gives no presumptive effect either to the rule as an act of delegated discretion, nor to the antecedent rulemak-ing proceedings in which the Department informed its discretion. What we have here is a seven-day evidentiary trial, presided over by a DOAH hearing officer and evaluated later by a court, on an assumed factual issue of whether discriminations based on gender, marital status and scholarship operate satisfactorily to predict accident-proneness. The hearing officer’s order, expounding his conclusion contrary to the Department’s rule, looks for all the world like a recommended order in section 120.57 proceedings. The only difference I can discern is that, because this rule-challenge was not an adjudicatory proceeding ending in an agency order, the Department had no opportunity by a final order to accept or reject, for stated policy considerations, the hearing officer’s recommended order. §§ 120.56(5), 120.59, Fla.Stat. (1979). McDonald,, supra.
So attuned are we both, hearing officer and court, to the adjudicatory mode of deci-sionmaking characteristic of the judiciary, that we seem not to acknowledge that rule-making is the preferred mode of decision-making by the executive. Apparently finding our prior decisions on the subject unpersuasive, the court joins the hearing officer in treating the issue essentially as one of pure fact or as one of pure law.
Thus the Department determines by its rule that all classifications based on gender, marital status and scholarship unfairly discriminate by standards fairly implied in subsection (h). To this the hearing officer replies, as though the burden were upon the Department to prove this factually to the hearing officer:
The facts proven in this cause do not support such an argument. It is not necessary to determine the validity of the present rate classification plans of Petitioners to determine from the facts presented in this ease and set forth above, that partial use of sex, marital status, and scholastic achievement as criteria in the formulation of private passenger automobile insurance premium rates can enhance the predictive accuracy of a rate classification plan and thus enhance the equitability of premiums. Thus it has been established that the use of the criteria prohibited by the Rule does not necessarily result in unfair discrimination. (Emphasis added.)
The court similarly deals with the issue as though the ultimate issue were one of fact to be resolved by the hearing officer and, as so resolved, to be honored by the reviewing court. The court goes so far as to escalate the hearing officer’s finding, “partial use of [the designated classifications] ... can enhance the predictive accuracy,” to something more decisive:
The hearing officer further found that the classification factors of sex, marital status and scholastic achievement, in light of the present state of the art in the industry, enhanced the actuarial soundness of a rate classification .... (supra, pp. 912,913, emphasis added.)
The court also joins the hearing officer in concluding that the Department failed to carry the burden of proof cast upon it: “Thus, as the hearing officer concluded, the Department has not established that the use of the criteria prohibited by Rule 4~43.-03 necessarily results in unfair discrimination.” Supra, p. 913, emphasis added.
This transformation of a rule-challenge proceeding into a conventional evidentiary trial runs precisely opposite to healthy forces in chapter 120 that press agencies toward rulemaking and away from case-by-case adjudication. With respect to “ultimate facts ... infused by policy considerations for which the agency has special re*923sponsibility,” we have repeatedly said that requiring proof of policy in 120.57 proceedings for an agency order “creates agency incentives for rulemaking, which as far as it goes displaces proof and debate of policy in 120.57 proceedings .... ” McDonald, 346 So.2d at 583; Anheuser-Busch, 393 So.2d at 1183 (emphasis added). In McDonald the ultimate factual issue, which was “infused by policy considerations for which the agency has special responsibility,” pertained to the bank’s “reasonable promise of success” and the skill of its proposed management, as described in the statute; in Anheuser-Busch that policy-ladened factual issue was whether manufacturers’ bar spending tended, in the same way as do gifts and rebates, to build favoritism among retailers, which was the statutory target. Many other examples could be given from decided cases. Here, plainly, the policy-ladened issue of ultimate fact is whether all prevailing dis-criminations based on gender, marital status and scholarship are fair, or many of them are, or some of them are, or none of them are. Choosing the degree of “unfairness” in predictive accuracy to be tolerated under the statute is ladened with policy considerations, and on that question the rule is well-nigh conclusive.
The inappropriateness of this evidentiary trial, and more so of the findings that resulted from it, is further illustrated by section 120.56(5), which regarding rule-challenges states that “[fjailure to proceed” in the manner provided by section 120.57 for rule challenges “shall not constitute failure to exhaust administrative remedies.” In other words, this court and others are authorized to entertain appeals directly from rulemaking, challenging the rule as beyond the delegated power of the statute, without need for evidentiary hearings or for a hearing officer’s order. E.g., Polk v. School Board of Polk County, supra; Florida Canners Association v. State of Florida, Department of Citrus, supra. See also Postal Colony v. Askew, 348 So.2d 338, 339 (Fla. 1st DCA 1977); Willis, supra, 344 So.2d at 592.
It seems inconceivable to me that the scope and method of rulemaking review should depend on the challenger’s strategic choice to appeal from rulemaking or to challenge first before a hearing officer. Doubtless there will be cases in which evidence illumines the issues, e.g., Postal Colony, supra, 348 So.2d at 339, but in no case should the hearing officer and court be seduced into attempting to separate and verify, through evidence, factual predicates in the policy choice represented by the rule.
Because the decisive question in this controversy is one of “ultimate facts ... infused by policy considerations for which the agency has special responsibility,” McDonald, 346 So.2d at 579, neither may the decisive question be put as one of pure law. The court does put it so, however, alternatively to its factual inquiry; and by various interpretative devices common to adjudicative method the court undertakes to show, as a matter of law, that the legislature intended by subsection (h) to preserve in insurance ratemaking some kind or some degree of discrimination based on the stated classifications. This purpose, the court considers, is contradicted by Rule 4-43.03.
I shall argue that the interpretative devices serve very poorly in this instance to isolate the exclusive meaning attributed by my colleagues, as a matter of law, to subsection (h). But preliminarily it must be said that the decisive issue is not one of pure law. For if it were, surely the same insight that declares the question one of law would also permit the court to instruct the Department on the subsidiary principles that, as a matter of law, differentiate fair and unfair discrimination. Obviously, invalidating this rule does not end the Department’s day-to-day responsibility to enforce subsection (h). That the court does not instruct upon the hidden but decisive subsidiary principles of subsection (h), differentiating fair from unfair discrimination, I take as proof that the statute does not as a matter of law yield those subsidiary principles, that more refined test of what is fair and unfair. The decisive question, therefore, remains a question of judgment; and unless (as I fear) the court now judges all such discriminations to be fair “in light of the present state of the art in the industry,” supra, p. 912, then the court evidently has reserved to itself the no less intrusive, but *924surely more burdensome, responsibility to make that judgment call on individual rate applications. But that, no more obviously than implementing the statutory standard through rulemaking, is the Department’s responsibility.
Our Framat and Nelson decisions teach that the only pertinent question of pure law in such a case as this, aside from procedural questions not here present, is whether the rule is predicated on a permissible interpretation of the statutory words. I have suggested that the range of permissible interpretations includes First through Twelfth above, and certainly includes Tenth, the interpretation by which the Department employed normative standards of “fairness” to sharpen its scrutiny into the claimed predictive accuracy of the suspect classifications.
Without particularly addressing this possible interpretation of subsection (h), which preserves the rule, the hearing officer and the court find the rule based on another interpretation, an illegitimate one beyond the range of the statute fairly read. In this, I respectfully suggest, two oversimplifications are indulged. The first is that the legislature expressed in subsection (h) but one possible conceptual scheme, that some kind or degree of the referenced discrimination is disapproved, but some also is approved and must necessarily survive. The second oversimplification addresses not the statute but the rule, and attributes to the rule, quite erroneously, an authenticating claim that its “blanket prohibition” of certain discrimination is but a paraphrase of the statute.
Thus the hearing officer states:
The Florida Insurance Code does not prohibit all rate classifications or premiums based upon sex, marital status or scholastic achievement, but rather only those that are solely based on those factors that unfairly discriminate.
Comparing the statute ... to the Rule in question, they clearly conflict. Rule 4-43.03 ... is a blanket prohibition....
And the court echoes:
In enacting [subsection (h)] the legislature obviously intended to permit discrimination based on sex, marital status, and scholastic achievement so long as this discrimination is not unfair or based solely on these factors. Yet the Department, by promulgating Rule 4-43.03, imposed a total prohibition.... Supra, p. 911.
Neither of these suppositions is correct. Given the range of permissible interpretations arising from the words themselves, there is no evidence in the words that all 160 legislators held a common purpose to approve and perpetuate some kind or degree of the referenced discrimination. It is also incorrect to say, by contrast, that the Department interpreted this same general language as a “total prohibition.” The Department’s interpretation is Tenth above, not Second.
Even if one were to concede, as I do not, that the one correct interpretation of subsection (h) permits some but disallows other discriminations based on the stated factors, the rule would appear to satisfy that interpretation. For through rulemaking processes the Department found that dis-criminations based on age do not unfairly discriminate, and so omitted age from the rule’s list of prohibited classifications. That is not an embarrassment to the rule, as the court seems to imply, supra p. 910, n. 2, but is verification that the rule lets stand some discriminations described in subsection (h) while forbidding others, as required by the court’s reading of the statute.
But of course the court is not satisfied with vertical differentiation, on fairness grounds, between the suspect classifications; the court requires a horizontal differentiation as well, saying that subsection (h) on fairness grounds permits some discrimi-nations based solely on gender, and so on, while disallowing others made on the same ground. This seems to me a permissible interpretation of the statute, but surely, due to the finer distinctions required, not the only permissible one.
The Department’s and the court’s conceptual interpretations of subsection (h) differ principally in the Department’s invocation *925of higher standards, assuring fairness nor-matively conceived, for judging the predictive accuracy of classifications by age, gender, marital status and scholarship. While the court is by no means unconcerned with the predictive accuracy of those classifications, the court obviously is satisfied with their actuarial soundness, whereas the Department is not. See the court’s approving reference to “the present state of the art in the industry,” supra p. 913, text at n. 8. These relatively higher and lower standards stem from two different interpretations of subsection (h): the Department’s is as stated Tenth, and the court’s is as stated Sixth, invoking only technical protocols for “fairness,” no heightened scrutiny, and no sense that the specified classifications are suspect as “socially unacceptable.” Majority opinion, supra p. 910, n. 2. The hearing officer, incidentally, may have preferred yet another interpretation of subsection (h), that stated Eighth.
The court employs several external aids to interpretation — legislative history, parallel statutory passages, past administrative interpretation, and so on — in arguing that the court’s interpretation of subsection (h) is exclusively correct and, in conclusion, that “these statutes do not authorize the prohibition of the use of sex, marital status and scholastic achievement as rating factors.” Supra p. 910, text at n. 4. I doubt that external interpretative devices can ever properly be employed to override the responsible agency’s legitimate interpretation of the statutory words themselves. But here, at any rate, the interpretative devices produce no evidence that the Department’s conception of subsection (h) is impermissible, nor that the court’s is exclusively correct.
The legislative history of subsection (h) “irrefutably shows,” it is said, “that the legislature expressly considered, but rejected, provisions that would prohibit the use of these factors as unfairly discriminatory.” Supra, p. 911, n. 5. On the contrary, the Senate committee referred to simply rejected an amendment that would have enacted the interpretation of subsection (h) stated Fourth above, characterizing as unfair per se all insurance rates based solely on “age, sex, school grades, education or lack of driver education.” That was in no sense an advance rejection of Rule 4-43.-03, which does not proscribe classifications based on age, education or lack of driver education. The senators who killed that amendment may well have done so because it included “age” as a per se unfair classification. More than likely, those senators and all the legislators referred to in the court’s discussion were content to write the statute in general terms, allowing the Department to implement those terms within the “fairness” degree enacted.
The court also finds it “highly significant,” to the prejudice of Rule 4-43.03, that the rule represents a change in the Department’s “own interpretation” of another statute, section 627.0651(6). That statute generally regards as “unfairly discriminatory” a rate that, compared to another, “clearly fails to reflect equitably the difference in expected losses and expenses” underlying the two rates. Supra, p. 913. The court also attaches “great significance” to the hearing officer’s observation that “the Department did not offer evidence or testimony sufficient to establish that factual changes of any nature have occurred, or that the Department has become aware of new factual information, which would sup*926port a deviation from their historic interpretation of the Florida Insurance Code.”
These statements are troubling on two scores. First, they seem to disapprove agencies changing their minds on questions of policy, as though the law preferred that “foolish consistency” of which Emerson wrote, and which today we call bureaucratic drift. Further, this calling for evidence verifies my earlier observation that the policy of Rule 4-43.03 has been judged, both by the hearing officer and by the court, as though the rule does not exist; as though these were adjudicatory proceedings for a final order determining the substantial interests of a party, in which the agency’s emerging or changing policy must be supported by a record foundation placed before the hearing officer. E.g., Florida Cities Water Co. v. Florida Public Service Comm’n, 384 So.2d 1280, 1281 (Fla.1980). No decision I am aware of requires that kind of showing when an agency’s policy is changed by rulemaking. The rule itself stands in the place of that showing and is the preferred vehicle for changing “established industry-wide policy.” Ibid.; see also Anheuser-Busch, supra, 393 So.2d at 1181.
Furthermore, if this new rulemaking required record justification in the form of new information or insights sufficient to “change the agency’s mind,” that is fully provided by the promulgating document itself. App., infra.
But all this debate overlooks the critical fact that subsection (h) of section 626.-9541(15) was newly enacted by chapter 77-468, section 19, effective Sept. 1,1977. The “historic” policy to which the court would bind the Department was developed under a considerably earlier statute, which since 1977 is numbered section 627.0651(6) but was previously, since 1971, numbered section 627.082(l)(e)3. I cannot imagine a more compelling reason to re-examine closely the “historic” policy than that the 1977 Legislature enacted subsection (h), calling, as in Tenth above, for closer scrutiny of the claimed actuarial accuracy of classifications based on “age, sex, marital status and academic achievement.” The promulgating document recites that the rulemaking study began in February 1978 (App. 932), within seven months after subsection (h) became effective. The promulgating order plainly emphasizes the 1977 law.
The court’s invalidating decision, reinstating the Department’s “historic” policy, not only judges the policy of Rule 4-43.03 as though the rule were never promulgated; it also judges the fairness of discriminations based on gender, marital status and scholarship as though section 626.9541(15)(h), Florida Statutes (1977), were never enacted.
While Rule 4-43.03 should be permitted to earn its vindication simply by rebutting the several arguments advanced unsuccessfully against it, I sense that several of those arguments — what the hearing officer thought after a seven-day trial, what the senate committee did to a substitute amendment, how the Department construed section 627.0651(6), nee 627.082(l)(e)3, before subsection (h) of 626.9541(15) became law — are but servants to an underlying objection that would yet disapprove the rule *927after all the lesser objections have been met and scattered.
Fundamentally, I take it, the court is unwilling to concede the broad range of permissible conceptual interpretations that the words of subsection (h) permit. This is no objection to any particular version, Tenth or any other, but is rather an assertion that the judiciary exclusively is to declare the one right answer to this question of how subsection (h) is to be conceived.
Though the dominant theme of our APA decisions from McDonald forward has been that this interpretative power resides principally in the executive branch, which is disciplined and spurred to responsibility through APA processes including those of rulemaking; and though our decisions claim judicial superintendence only over “flagrant” abuses of rulemaking, and they endorse agency interpretations “within the range of permissible interpretations of the statute” notwithstanding that “other interpretations are possible and may even seem preferable,” yet no precedent can compel the mind to conceive of wider possibilities than it is willing to conceive. The judicial will must be convinced, else its reason is not free to honor even our own precedents.
At this stage in the maturity of our judicial system, a modest disclaimer of judicial hegemony in matters of statutory interpretation would seem to be required by a decent respect for the executive as a coordinate branch of government, made more responsive to the public, and more disciplined, by APA processes. But if neither those considerations nor our decisions endorsing them convince us of that, we should be persuaded to that end by the needs of our own institution, this court, for adjudicative processes that operate with fair uniformity. For as we judges grow more numerous, forty-six and counting in the appellate courts of Florida, and twelve on this court alone, the folly of three judges or a majority of them declaring the “one right answer” to a question of statutory interpretation, when the executive branch has spoken another permissible answer through rulemaking, becomes more evident and more dangerous.
If I should insist that Tenth is the only correct conceptual interpretation of subsection (h), and my colleagues say instead that Sixth is exclusively correct, then Sixth it is. But I readily concede that Sixth is a permissible choice, and my colleagues in charity and intellectual modesty will say that Tenth is permissible too, but not as good as Sixth. From this it is apparent that the shades of meaning and emphasis differentiating our choices of Sixth and Tenth are quite small, ultimately a judgment call.
Ordinarily, judgment calls are no cause for embarrassment to sound adjudication; adjudication carries its questions into articulated resolution as far as possible, then trusts the judges. But when we insist upon selecting from several permissible interpretations of a generally-worded statute the single “correct” one, and the validity of a solemn act of a coordinate branch hangs in the balance, then we test the legitimacy of our inarticulable judgment calls to the very utmost. In such a judicial posture we are at risk of deciding how governmental policy and private fortune shall fare, in areas of considerable public importance, by random vote of preference.
*928Then, to a great but incalculable degree, the numerousness of judges from whom the decisive judgment must come adds to the problem of legitimacy.
Consider first the array of twelve permissible interpretations previously suggested for subsection (h):
Consider next the array of twelve judges, each conscientious and fully informed, but each with different perceptions m matters of choice:
If my colleagues are judges 2 and 5, say, and I am judge 10, then their judgment calls for Sixth override mine for Tenth. But if the panel were instead judges 2, 10 and 11, would the same “one right answer” prevail?
As the number of judges increases on a court that decides in shifting panels of three, certain destructive tendencies, among them the fragmentation of principle and the ascendancy of preference, are set loose. Remembering Edinburgh’s court of session, which he saw abolished in the year 1808 after three centuries as a Scottish institution, Henry Cockburn wrote:
[I]t was extinguished at last by defects which no modern tribunal could survive. Its radical defect was its numerousness. A bench of fifteen judges can only be “a learned crowd.”
Because “nobody could ever predict of whom the Court would be composed at any one moment” and “they were too many to confer harmoniously together,” Cockburn wrote, “[e]ach acting independently was tempted to stand up for every particle of his own notion.” The resulting “hubbub, jockeyship, and uncertainty had always to be settled by a vote; the putting and settling the result of which gave rise every hour to as keen conflicts as the original matter of the dispute. I cannot conceive that this system ever did well ....”
These perils of judicial numerousness are by no means peculiar to Scotland, or to courts of the early 1800’s. A judicial veteran of the multifarious and recently bifurcated United States court of appeals for the fifth circuit recently commented upon the modern perils of numerousness of appellate judges:
This tremendous potential for instability in the rule of law creates a great deal of litigation. So you have a situation where you add judges to dispose of more cases, and at the court of appeals level, at least, the new judges may well cause more litigation than they can terminate.
For this court of twelve, the proper response to the modern perils of numerousness is neither bifurcation nor handwring-ing; rather, the proper response is coherence, which is to be achieved only by sublimating preference to principle and, in cases of this sort, by acknowledging the rightful power of the executive branch to select the most propitious of several permissible statu*929tory interpretations, broadly conceived. In this we not only acknowledge the executive’s rightful place in government, depending on the legislature and the electorate to apply corrective forces where necessary. By our deference we also resist the seductive notion, fatal to a court of many judges, that our individual judgment calls, the small preferences of each of us, are infallible.
The economic impact statement
Judge Joanos and I concur that the economic impact statement is quite sufficient. Any inadequacies do not impair “the fairness of the proceedings or the correctness of the action.” Secs. 120.54(2)(a), 120.-68(8), Fla.Stat. (1979). The objections to it “command adherence to form over substance” and impose “a transparent technicality” upon the rulemaking process. See FIorida-Texas Freight, Inc. v. Hawkins, 379 So.2d 944, 946 (Fla.1980); Plantation Residents’ Association, Inc. v. School Board of Broward County, 424 So.2d 879, 881 (Fla. 1st DCA 1982); Cortese v. School Board of Palm Beach County, supra, 425 So.2d at 558, fn. 12.
The Department’s economic impact statement is a seven-page document that systematically complies, paragraph by paragraph, with the four substantive requirements of section 120.54(2)(a), Florida Statutes (1979). First the Department estimated at $6,000 “the cost to the agency of the implementation of the proposed action.” Next, the Department generally predicted decreased insurance premiums directly benefiting single male operators under age 25, and increased premiums directly costing single female operators and married male operators under 25, though “it is not intended that there will be an overall statewide increase or decrease in premium volume.” Thirdly, for the required “estimate of the impact of the proposed action on competition,” the Department opined that those getting smaller insurance premiums “may have difficulty in securing coverage,” while those paying larger premiums would benefit by “increased competition for their coverage.” Fourth and finally, the Department set out at length the “data and method used” in the prior estimates, even exhibiting “an overall distribution of the various affected classes as well as the estimated premium effect by class.” The Department noted, however, that the premium estimates were as presented by one company only, Allstate Insurance Company, in the public hearings upon the rule.
The Department’s economic impact statement cannot properly be faulted for its failure to estimate the “expenses, possibly substantial in nature,” that the industry and individual companies may incur in devising rate classification plans complying with Rule 4r-43.03. In order to make such an estimate, either for the industry or for individual insurers, one would have to know both their costs in administering the present classifications, varying from classification to classification and from company to company, and their costs in administering the classifications which each company, variously, will now put into effect. One would have to estimate how efficiently each company will proceed. If as the Department predicts many companies will begin with simple rough averaging and leveling, collecting the same gross income from essentially the same insureds, the added costs will be inconsequential. See Division of Workers’ Compensation, etc. v. McKee, 413 So.2d 805, 806 (Fla. 1st DCA 1982).
*930But more fundamentally, section 120.54(2)(a) cannot reasonably be read as requiring the Department to speculate on the amount of decreased or increased cost each company will incur, from this day forward, as a result of revising its classifications in a way the Department cannot predict and does not control. It is the industry, not the Department, that sets rates; and their operating costs in setting rates and collecting premiums are nothing more or less than their cost of doing business. The statute requires an estimate of costs and benefits to persons “directly affected by the proposed action.” Those, the Department most reasonably determined, áre the insureds who pay the premiums, which premiums reflect not only the carriers’ insured losses but also their operating costs.
The Department’s impact statement is faulted also for not estimating the assumed “potential cost” to the Department, over some unstated period, of reviewing rate filings affected by Rule 4-43.03. This too would be a meaningless speculation, and the Department is to be credited for not engaging in such a charade. The Department rightly construed the statutory language, “costs ... of the implementation of the proposed action,” as meaning the cost of putting the rule into effect, not the cost of performing all regulatory functions affected by the rule in the decades ahead. Moreover, the Department’s chief of insurance rating testified that review of immediate rate filings could be staggered to avoid unusual additional personnel costs.
Finally, the hearing officer criticizes the impact statement because it does not estimate the cost impact on students now receiving discounts for scholarship. Again, no meaningful estimate could be given without knowing both the extent to which those individual insureds, by the thousands, benefited from such discriminatory premium rates, and the extent to which their individual insurers will now raise or lower their rates based on factors other than good scholarship. The insurer whose comparative premium studies were exhibited in the impact statement, Allstate, did not give a “good student” discount because it was considered indefensible actuarially. Obviously, then, Allstate’s insureds will not suffer from its deletion, though perhaps they will if they were previously insured by another insurer; though perhaps not if the other insurer’s new classifications offset the loss of the “good student” advantage with other new advantages for some insureds; and on and on the speculation continues, gratifying those who oppose the rule or admire pointlessly complex hypothetical impact statements, but surely disappointing those who strive for intelligible APA processes. Significantly, “no insurance company submitted any actuarial data linking scholastic achievement to accident avoidance.” App., 938. It would be especially ironic, then, if the Department’s failure to make wild guesses, with a straight face, on the economic impact of prohibiting this palpably unfair and undefended discrimination, should now become a ground for perpetuating that discrimination.
A majority of the panel has thus determined that the economic impact statement is not inadequate. For the reasons previously stated, I would sustain Rule 4-43.03 in its entirety and so would reverse the order of the hearing officer.
Appendix to dissenting opinion
FLORIDA’S AUTOMOBILE INSURANCE RATE CLASSIFICATION SYSTEM:
Report to the Insurance Commissioner and The Commissioner’s Orders and Findings
October 1, 1979
Commissioner’s Preface
The Florida Department of Insurance with this Report has concluded one of the most exhaustive studies of the automobile insurance rate classification system undertaken in recent years, certainly in this state.
Contained within these pages is the distillation of more than thirty hours of testimony from witnesses representing consumer groups, university research institutes, public *931officials, the insurance industry and Florida citizens from all walks of life.
My powers as Insurance Commissioner are expressed clearly by statute. I do not, for example, have the power to set rates charged by insurance companies doing business in Florida. Nor do I seek such authority for reasons which are discussed in appropriate sections of this Report.
However, I have express authority under Florida Statutes to reject any existing rate classifications that are unfairly discriminatory. This authority I am exercising today.
Bill Gunter State Treasurer and Insurance Commissioner
I. Introduction
Automobile insurance is unlike most products. The cost is predicated to a large degree on the characteristics of the person who is making the purchase.
Insurers use classifications, such as a person’s age, sex, marital status, and place of residence, to estimate how much expected losses may vary from one driver to another. Ideally, the price of insurance should reflect the risk involved in insuring the driver. When it fails to meet this goal, the question becomes whether existing classifications can be improved further or eliminated in favor of a more equitable and acceptable measure of risk.
This inspection was not limited solely to the testimony given by witnesses at hearings held in Tampa, Miami, and Orlando. We also examined inquiries made in other states and by a special committee of the National Association of Insurance Commissioners. We reviewed the Risk Classification Report prepared expressly for the Florida Insurance Department by the University of Florida Insurance Research Center, as well as other relevant publications.
At the outset, it is important to understand that changes in the rate classification system will not reduce the aggregate cost of automobile insurance to Florida’s seven million licensed motorists.
The purpose of this study is to determine whether a more equitable redistribution of premium charges can be made by eliminating or tempering existing rate classifications.
Differences in premiums for the vast majority of companies depend on the primary rating factors of age, sex, marital status, use of the automobile, and place of residence. Secondary factors utilized in rating include the application of discounts for insuring more than one car, driver training, driver record, and type of automobile. Surcharges may be applied for certain traffic violations and at-fault accidents.
The Dade County Petitions
The high cost and shrinking availability of automobile insurance in Dade County led to the extraordinary impaneling of a Grand Jury in 1977 to study the problem....
The Grand Jury was charged to determine whether rates charged to Florida motorists for private passenger automobile liability insurance was reasonable and justified. The Grand Jury heard extensive testimony from numerous witnesses throughout the year, including the Insurance Commissioner and members of his staff. On January 5, 1978, the Grand Jury filed its report characterizing auto insurance rates in Florida as “discriminatory, biased and patently unfair.”
Rates based on age, sex, and marital status, the Grand Jury concluded, should probably be prohibited. A study of territorial based rating should be undertaken to determine if it, too, should be abandoned. The Grand Jury suggested rates based on driving experience and miles driven as acceptable substitutes.
On January 10, 1978, Dade County attorneys formally petitioned the Florida Department of Insurance to change the automobile insurance classification system. *932Dade County, in two petitions, proposed four rules and requested that the Insurance Commissioner set administrative hearings to review them.
The first rule would prohibit companies from establishing premium rates for automobile insurance based on age, sex, marital status, and scholastic achievement. The second proposal would prohibit insurers from establishing premium rates on the basis of arbitrary territorial boundaries. The third rule would prohibit insurers from calculating company administrative costs as a percentage of the premium paid by individual policyholders. Finally, the fourth proposed rule would prohibit insurers from calculating agents’ commissions as a percentage of premiums.
On February 10,1978, the Department of Insurance announced its intention to hold public hearings with respect to each of the rules. At the same time, it contracted with the Florida Insurance Research Center at the University of Florida to prepare a report evaluating Florida’s risk classification system. The purpose of the report was to review the existing state of the art in order to provide a framework for the hearings. Specifically, the researchers were asked to evaluate the predictive accuracy of the existing class system when compared with alternatives such as driver experience and miles driven, suggested by the Dade County Grand Jury and others.
The Florida Insurance Research Center Report was completed and submitted to the Department in December, 1978. Generally speaking, it concluded that the existing system was superior in predictive accuracy to the suggested alternatives. However, it suggested that further refinement would be necessary to eliminate inequities within the risk classification system.
The public hearings on the proposed rule to prohibit the use of age, sex, marital status, and scholastic achievement in rate classifications were held in Tampa, January 31, 1979, and in Miami on February 14, 1979. These hearings attracted witnesses from throughout the country who represented insurers, consumer organizations, the academic community, and interested citizens.
II. The Risk Classification System
Modern risk classification is based on the primary factors of age, sex, marital status, and principal place of garaging, better known as territory. The use of the automobile — for pleasure, commuting, business, and farm as well as the type of operator— principal, occasional, resident student — are secondary considerations along with annual mileage. Other factors include the type of vehicle — standard, intermediate, high performance, sport — its make and model, its age, and its value. There are discounts for scholastic achievement, driver training, carrying more than one car on a policy, and for compact cars. There are surcharges for inexperience, and for chargeable accidents and violations as permitted by law. These factors are used in varying degrees and combinations in an effort to predict expected losses.
Although not all companies use all the rating factors, age, sex, marital status, and territory — the factors under review — are fairly common. Scholastic achievement, also considered in this Report, is used by a minority of companies.
Age is the common thread that is woven through all class plans used in Florida. It is the primary determinant used to classify Florida drivers, 16 and older, for private passenger automobile insurance. Sex and marital status are used extensively from age 16 to 24 for females and age 16 through 29 for males. However, by the time females reach 25 and males reach 30, sex and marital status are used infrequently, if at all. There are, however, two notable exceptions.
A few insurers continue a sex distinct class for females age 30 through 64 if they are the only operator in the household. And, at least one company continues a combined sex and marital status class for single males 30 through 49 if the vehicle is found *933to be driven in excess of 50% of the estimated annual mileage.
III. The issues
The evaluation of the effectiveness and utility of the rating factors of age, sex, marital status, scholastic achievement, and territory, must consider two basic questions. First, are the factors accurate as predictors of loss propensity? And, second, are the factors equitable, in that they do not unfairly discriminate against individuals or groups of policyholders?
The Accuracy of the System
The complex classification plans now used by automobile insurers were developed from an initially simple system. As data collection and analytical methods became more sophisticated with the development of high speed computers and the availability of a more extensive data base,, class plans became more refined. But even as the plans reach greater levels of refinement, the foundation on which they are built has come under increasing challenge.
Predictive Accuracy
In the ratemaking process, the level of losses for a future period of time must be forecast. After the loss estimate is prepared, the statewide average premium level must be established that also takes into account administrative expenses and a reasonable profit margin. When the overall premium level has been set, it is distributed to various groups and subgroups of insureds through the risk assessment process. Risk assessment involves risk selection which is subjective acceptance or rejection, and rating, which is based on classification systems, to arrive at the final rate or premium for the risk.
The accuracy of the risk assessment, and, more particularly, of rating are difficult to measure quantitatively. The complexity of testing for accuracy increases significantly when the accuracy of one individual’s rate, as opposed to broad groupings or class accuracy, is examined.
Dr. Joseph Ferreira, a principal witness for the petitioners, stated that you achieve such accuracy when “the class average cost for a group is close to the expected cost for each and every member of that group.” This problem has arisen because for rating purposes insurers are not really concerned about specific individual losses. Rather, their interest lies in determining whether aggregate losses for a class of individuals approach the expected aggregate losses for the class as a whole. The practical effect is that all individual members of a class are charged the class average cost. This approach is known as class average pricing.
Those who challenge the accuracy of risk classification most often direct their criticisms at the class average pricing method. Certainly, within any class there will be individuals who are charged more than their individual expected loss, just as there will be individuals who are charged less than their individual expected loss. What complicates the pricing mechanism is that individual expected loss must be projected from past actual losses. Those actual losses reflect two distinct elements: the random or fortuitous nature of most accidents in combination with a statistical likelihood that one group of drivers with similar characteristics is more likely to have more accidents than another.
The predictive accuracy — the ability of class plans and ratemaking processes to identify the degree of inherent risk in each individual motorist — within the current risk assessment process is difficult to achieve because it explains only a fraction of the variation among inherent risks. The elements of fortuity and inherency cause expectant costs for many members of a class to vary from the class average cost and thus their rate, the class average rate, fails as an accurate predictor of individual expected loss. The result is that under the current system, many individuals are either overcharged or undercharged.
Our hearings and the testimony when objectively appraised, clearly indicate that *934further adjustment in the risk assessment process is desirable.
The Equity of the System
The question of equity in risk classification also does not lend itself to facile quantitative evaluation. Analysis of equity turns on the concept of unfair discrimination. A classification system, by its nature, must discriminate in order to arrive at the appropriate cost for the product. It is clear that discrimination between policyholders is fundamental to the insurance mechanism. What is less clear, and indeed exceedingly elusive, is where lawful discrimination ends and where unfair discrimination begins. It is only upon a showing of unfair discrimination that the Florida Insurance Commissioner has the statutory authority and obligation to make a finding.
Where to draw the fine line between necessary discrimination and unfair discrimination is at the core of the controversy over the use of age, sex, marital status, territory, and scholastic achievement as rating variables. Few issues have prompted greater debate among insurance companies, regulators, the news media, and individual policyholders. Few issues have produced as much misinformation.
There probably is not a driver in Florida who believes that the raising of his or her rate will contribute to an equitable solution. Just as surely, no motorist would turn down a rate reduction or hesitate to switch companies even if their action made pricing more equitable.
For their part, industry witnesses were almost unanimous in defending the use of current rating variables. Their confidence of the current system was closely accompanied by statistical scenarios of massive market dislocation should the current system be altered.
When is Discrimination Unfair?
The apportionment of automobile insurance cost among consumers requires strict attention to the equity of cost distribution. Although there is a presumption of equity in the risk classification system, if the system lacks predictive accuracy, it is by definition inequitable. The risk classification system is inherently discriminatory, because it depends upon the adequate identification of different risk characteristics among the population. Discrimination is reflected by different charges to individuals that creates a spread of total cost that is intended to be equitable. But, when it appears that this system of cost apportionment creates gross inequities, the essential underpinnings of the risk classification system are dangerously weakened.
There are two equity questions to address. First: What is the legal basis for discrimination? Second: What is the social acceptability of existing potential risk classifications, and the accompanying public policy considerations?
Discrimination and The Law
Another important legal aspect of the use of discrimination in insurance rates is the recent decision in City of Los Angeles, Department of Water and Power v. Manhart, 435 U.S. 702 (1978). The Manhart Decision essentially held that to make benefits for families different than for males, even though as a class females have different life expectancies, was unfair discrimination. This decision is notable because it raises important legal counter-arguments to the presentations by the insurance industry which defend the equity of the current classification system. The Manhart Decision raises questions about the continued use of sex distinct rates.
These are the underlying issues that must be examined in a discussion of the concept of equity in automobile classification. It stretches credulity to believe there is equity in charging a youthful male in Miami, Florida, ten times more for full coverage than a 26-year-old woman in Orlando or Tallahassee. In times of rapid social change, the burden is on the insurance industry to clearly demonstrate that sex or marriage has any appreciable impact on the propensity toward accidents or inherent risk characteristics on the part of an individual. It is far more likely, as Dr. Ferreira suggests, that such distortions result from using adminis*935tratively and historically convenient rating measures which in fact are outmoded or in need of significant adjustment....
Social Acceptability and the Role of the Regulator
Charging more or less to policyholders on the basis of race, religion, or economic status would be socially unacceptable regardless of any statistical justification a company might offer. Several witnesses at the hearings argued that it is socially unacceptable to continue the use of sex and marital status to justify lower premiums to women and married couples. Even, the witnesses who spoke most earnestly to abolish two rating categories, represented themselves and others who would be economically disadvantaged by such a change.
It was apparent from the hearings that the insurance industry does not have such reservations with respect to any existing rate classifications. Any determination of social acceptability must come, if appropriate, from the Insurance Commissioner.
However, there were indications from the testimony that fears of being placed at a competitive disadvantage might in part be responsible for a lack of industry introspection as to the social acceptability of current rating factors.
Therefore, it is clear that the power and responsibility to make public policy in this area must rest with the regulator. . The private sector has neither the incentive nor the will to make such judgments. To leave the determination of social acceptability to the private sector would serve only to further retard the already slow movement in the direction of ending discrimination based on certain classifications which the Supreme Court has found to be invidious.
Rate classifications based on sex and marital status have not historically shared the same onerous connotations of discrimination by race or religion. However, legitimate questions now have been raised both in this regard and as to whether or not these two classes pass the test of predictive accuracy.
Equitable Allocation of Costs
Three primary variables used by automobile insurers for rating purposes — age, sex, and marital status — are essentially a maturity index. Insurers contend that the age of an individual has a significant relationship to his or her maturity. They argue that women mature earlier than men, and that marriage forces people to “settle down.” Finally, it is said, students with good grades are more mature. For the most part, these presumptions are highly subjective and inaccurate measures of maturity. They create innumerable inequities.
Insurance industry witnesses warned that disallowing continued use of these often questionable rating variables would create severe and unexpected market-place dislocation and suppress rather than enhance competition.
Yet it is lack of competition and innovative inertia that led to recent availability and affordability problems in Florida.
That competition is stagnant in parts of the state is evidenced by the all too rapid growth of the Florida Joint Underwriting Association. What was intended as a mechanism to provide last resort coverage to motorists has now become the state’s second largest insurer. '
A shake-up in the rate classification system with imprecise rating factors being phased out in favor of enhanced and more socially acceptable predictors may be just the right prescription to encourage a new burst of market competition.
IV. Addressing the Dade County Petitions
Is the industry’s current use of age, sex, and marital status as primary classifications designed with equity in mind? Or are they simply the most easily verifiable and administratively convenient factors?
*936In making these assessments, one point must be kept firmly in mind. Under Florida law, the Insurance Commissioner has neither the power to set rates nor to dictate what classification plans should be employed.
What the Commissioner can do, is review rates and rate classification systems in Florida and reject those he finds to be unfairly discriminatory....
Age
Generally speaking, classification by age charges young and less experienced drivers considerably more than adults for automobile insurance. It is justified on the grounds that young drivers, especially young male drivers, have a much higher accident rate and thereby draw a far greater proportionate share of the total premium dollar to cover claims.
While virtually all witnesses agreed that the use of age as a rating variable has a high correlation to loss expectancy, many agreed that age is actually an administratively convenient conglomerate variable measuring factors such as maturity, experience, and amount and type of driving. Age also suffers the defect of providing no incentive value since it is beyond the policyholder’s control.
On the other hand, if age is eliminated without substituting more accurate and equitable predictors of loss, the result will be a system with even greater inequities.
Industry spokesmen readily advanced doomsday theories as to what would happen to rates charged senior citizens, young married couples, and women if age, sex, and marital status were eliminated as rating factors. However, that is only half of the equation. The industry should not be permitted to box motorists into either/or situations seeking to keep the vast majority of drivers satisfied at the expense of good drivers whose only sin is their youth.
Though there are potential alternatives, the dilemma is that the industry simply has not attempted to develop a replacement vehicle for age more accurate and equitable than the current system.
The inescapable conclusion is that no workable alternative to age as a primary risk assessment factor was presented by either the industry or the petitioners at our hearings. Absent such alternatives, the elimination of age as a rating variable would lead only to greater inequities.
Commissioner’s Decision:
The record of our hearings clearly establishes that there is a strong correlation between age classifications and loss experience resulting from vehicle accidents. It is also true that a classification such as age is an actuarial strait-jacket providing no differentiation among risks within a high rated age bracket.
A major impediment discovered through the hearing process is that the insurance industry simply lacks statistical correlation between accident data and alternatives suggested for age classification. Absent such correlation, it is improbable that a more equitable replacement classification variable will be discovered.
Therefore, I will direct companies writing automobile insurance in Florida to collect additional data to identify similarities among drivers involved in accidents. This data is to be reported to the Department in a manner to be prescribed. The purpose of maintaining and examining such data is to improve or find a substitute for age classifications.
Admittedly, this is a long term project. However, the fact is that the industry simply does not now keep loss records in a manner conducive to comparing alternatives to age classification. This is not just a Florida problem; it is a national problem.
The Florida Department of Insurance will be plowing new ground in the search to find an equitable alternative to age
*937
classification or, at least, improvements to it.
Sex
It is obvious from testimony that automobile insurers in Florida use gender as a proxy variable to measure other factors. Unlike age, however, gender rating attempts to go beyond measuring maturity and driving experience that separates youthful from adult classifications. It is one thing to argue that maturity and experience have predictive value in risk assessment. It is quite another thing to argue that there is a similar distinction based on gender. A careful examination of the testimony offered by industry spokesmen clearly reveals that sex discrimination among youthful operators is in reality a proxy vehicle for other measurements.
There can therefore be no justification for the continued use of sex discrimination for automobile insurance rating purposes. It is unfair discrimination in that it continues to prolong a stereotype and results in many youthful males being overcharged when compared with youthful females.
Industry representatives forecast significant increases in youthful female driver premium rates should gender be eliminated as a rating factor. Indeed, that would be the case if a company chose to do nothing more but transfer premium charges from male to female drivers within youthful driver categories.
However, such redistribution would raise new questions concerning unfair discrimination since it is recognized that mileage, vehicle ownership, and use, rather than gender, accounts for the difference in accident statistics between male and female drivers.
The innovative company will use the opportunity to develop new youthful driver categories that seek to make these distinctions. The Insurance Commissioner has the authority and the inclination to disapprove rate filings from companies which choose not to make such distinctions.
Commissioner’s Decision:
There is ample and convincing evidence on the record to conclude that rates based on sex distinctions are unfairly discriminatory. Therefore, the declaration sought by the petitioners is granted.
It is apparent that classification by gender, a factor obviously beyond an individual’s control, does not provide useful incentives for loss prevention.
Therefore, all automobile insurers doing business in Florida are required by March 1, 1980, to submit for Departmental approval a rate plan that eliminates price distinctions based solely on sex.
Marital Status
Marital status, like gender, is a proxy variable for other measurements. Again, like sex, it is an imprecise and inadequate measure of maturity, miles driven, and vehicle use.
In Florida, marital status is used most frequently in rating young male and female motorists under the age of 25.
Again, like gender, it generates misclassi-fication since there is no direct causal relationship between marriage and accident frequency. We are dealing once again with proxy variables that insurers, given sufficient inclination and commitment, can improve upon.
Marital status is a poor proxy for maturity in driving experience. It offers no incentive for loss prevention. It is no measure of driving ability. It lacks social acceptability.
Once again, company officials forecast extreme adverse price dislocations if marital status was found to be unfairly discriminatory. Once again, however, price adjustments can and should be tempered by devising a system based on measurement for which marital status is a poor surrogate. Commissioner’s Decision:
There is ample and convincing evidence on the record to conclude that rates based on marital status distinctions are unfairly discriminatory. Therefore, the declaration sought by the petitioners is granted.
*938
It is apparent that classification by marital status does not provide meaningful incentives for loss prevention.
Therefore, all automobile insurers doing business in Florida are required by March 1, 1980, to submit for Departmental approval a rate plan that eliminates price distinctions based solely on marital status.
Scholastic Achievement
Some companies provide scholastic achievement discounts on the premise that a good student is a careful driver who spends more time behind books and less time behind a wheel of a car than others in his peer group.
There was absolutely no evidence presented at the hearings to support such an assumption. In fact, there was much evidence to recommend against its continued use as a rating variable.
A scholastic achievement discount discriminates against part-time students and against those who must work to pay for their studies. The assumptions that under-, lie the scholastic achievement discount are easily rebuttable. Good students do not necessarily avoid excessive driving, just as poor students do not necessarily make more frequent use of automobiles. In fact, a counter-argument can be made that good grades actually result in greater access to the family automobile.
In thirty hours of testimony, it is impossible to find any support for the discount. Most convincingly, no insurance company submitted any actuarial data linking scholastic achievement to accident avoidance. As one industry spokesmen said:
“Allstate does not use scholastic achievement as a rating variable. We don’t think it has any merit.”
Neither do we.
Commissioner’s Decision:
The request by the petitioners to find rates or discounts based on scholastic achievement unfairly discriminatory in establishing private passenger motor vehicle insurance rates is hereby granted effective March 1, 1980. Any company offering such discount or reduced rate is encouraged to discontinue the practice prior to the effective date of this Rule. All automobile insurance companies doing business in Florida which offer discounts for scholastic achievement are required by March 1, 1980, to submit for Departmental approval a rate plan that eliminates such price distinctions.
V. Improving the System
Driving Experience
Driving experience used in combination with one or more other factors ultimately may prove to have more predictive accuracy than age as a rating variable. Companies should be required to gather accident loss statistics by driving experience and other variables which is unrecorded and unavailable at the present time.
Miles Driven
The use of the number of miles driven by an individual during a designated time period was presented by several witnesses as another attractive alternative classification variable.
Miles driven scores well on causality, homogeneity, and social acceptability. There is little doubt that the number of miles a person travels increases exposure to loss.
Merit Rating
A pure merit rating plan would use a driver’s record of chargeable accidents and traffic violations to determine an appropriate insurance rate. Realistically, such a system would be impractical.
