Spanish Speaking Citizens' Foundation, Inc. v. Low

Opinion HANLON, P. J. The ballot pamphlet arguments for Proposition 103 passed by the voters in 1988 promised among other things that the initiative would “force[] insurance companies to base your [automobile insurance] rates on your driving record first, rather than on where you live. This means good drivers throughout the state will pay less than they do now, while bad drivers will pay more.” (Ballot Pamp., Gen. Elec. (Nov. 8, 1988) rebuttal to argument against Prop. 103, p. 101.) To those ends the following provisions were added to the Insurance Code: “Rates and premiums for an automobile insurance policy . . . shall be determined by application of the following factors in decreasing order of importance: “(1) The insured’s driving safety record. “(2) The number of miles he or she drives annually. “(3) The number of years of driving experience the insured has had. “(4) Such other factors as the commissioner may adopt by regulation that have a substantial relationship to the risk of loss. The regulations shall set *1185 forth the respective weight to be given each factor in determining automobile rates and premiums. Notwithstanding any other provision of law, the use of any criterion without such approval shall constitute unfair discrimination.” (Ins. Code, § 1861.02, subd. (a).) While these provisions may appear straightforward enough, devising regulations to implement them has been very difficult. Three Insurance Commissioners struggled with the problem before the present regulations were promulgated by former Commissioner Quackenbush in 1996. Interpretation of these regulations was contested in an administrative proceeding that concluded in 1998. Then this case was filed, alleging that the regulations as so interpreted contravene the statute. The trial court set aside one of the regulations, and interpreted the statute in a manner that may require yet another round of rulemaking. “ ‘Proposition 103 [has] proved to be a problem child from its inception’ ” (20th Century Ins. Co. v. Garamendi (1994) 8 Cal.4th 216, 247 [32 Cal.Rptr.2d 807, 878 P.2d 566]), and that is perhaps nowhere more apparent than in this area of automobile insurance rating factors. This case potentially may affect everyone’s auto insurance premiums but it will not affect the overall cost of auto insurance in the state; what is at stake is only the distribution of premium payments among policyholders. It is anticipated that drivers in urban areas will see their auto insurance premiums decrease and that drivers in nonurban areas will see their premiums increase if the judgment below is affirmed. Thus, the cities of Los Angeles, San Francisco, and Oakland are among the parties on one side of the dispute, and the counties of Fresno, Madera and Mariposa are among the amicus curiae on the other. However, it does not appear that any interpretation of the statute will uniformly benefit “urban” drivers at the expense of “rural” drivers or vice versa. For example, evidence has been presented that under the trial court’s ruling premiums paid by drivers in seven counties (Los Angeles, San Francisco, Sacramento, Orange, San Bernardino, Riverside and Ventura) will decrease, while premiums in the other 51 counties will increase. That evidence is challenged, but even if it correctly predicts the result of an affirmance the benefit or detriment would not be parceled out along strictly urban/nonurban lines. The 51 “rural” counties where premiums would increase include four of the state’s 10 largest urban centers, and at least two of the seven “urban” counties where premiums would decrease have considerable rural areas. Thus, while there is talk in the record and briefs about drivers in certain areas being made to unfairly “ ‘subsidize’ ” the premiums of those in others under different constructions of the statute, it is not always *1186 clear who would be subsidizing whom. It is also impossible to predict how different methods of calculating rating factor weights will affect the premiums of individual policyholders. The ambiguities in this case begin with the language of Insurance Code section 1861.02, subdivision (a). Part of the problem is that this statute has been and may remain the only one in the country that refers to the “importance” and “weight” of auto insurance rating factors. Further difficulty is created by the need to harmonize the statute’s weight ordering mandate with its requirement that rating factors be substantially related to the risk of loss, and with Proposition 103’s overall aim of eliminating arbitrary insurance rates. The ballot pamphlet promises that rates would be based on driving record “first” rather than a driver’s residence, and that factor weight ordering would reduce good drivers’ premiums, must also be taken into account. (Ballot Pamp., Gen. Elec., supra, rebuttal to argument against Prop. 103, p. 101.) In view of Proposition 103’s unprecedented and potentially conflicting demands with respect to auto insurance rating factors, it is perhaps not surprising that different Insurance Commissioners have advanced different interpretations of the factor weight ordering mandate, or that the parties challenging the current regulations have themselves taken inconsistent positions on what the law entails. There may be no one single correct interpretation in this instance. We nevertheless conclude that the current regulations, which were adopted after years of study and debate, lawfully implement Insurance Code section 1861.02, subdivision (a). We therefore reverse the judgment below. I. Background A. Calculation of Auto Insurance Premiums (1) Base Rate Auto insurance premiums are determined in two stages. The company first calculates a base rate for a particular type of coverage which is the same for each policyholder and reflects the total annual premium the company must charge all policyholders to cover its projected losses and expenses and obtain a reasonable rate of return. This first stage in the process is not at issue in the case. The base rate is then modified by applying a series of “rating factors” for each policyholder which determine how much the policyholder is charged and how the total premium the company receives is divided among the policyholders. This case concerns this second stage in the process. *1187 (2) Rating Factors Proposition 103 requires that driving safety record, annual miles driven, and years of driving experience be applied as rating factors (mandatory factors), and permits the application of other factors adopted by regulation, which have a substantial relationship to the risk of loss (optional factors). (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, § 3, p. 99; Ins. Code, § 1861.02, subd. (a); Cal. Code Regs., tit. 10, § 2632.5.) 1 The regulations permit the use of the following optional factors: (1) type of vehicle; (2) vehicle performance capabilities; (3) type of use of vehicle (pleasure only, commute, etc.); (4) percentage use of the vehicle by the rated driver; (5) multivehicle households; (6) academic standing of the rated driver; (7) completion of driver training or defensive driving course by the rated driver; (8) vehicle characteristics (engine size, repairability, etc.); (9) gender of the rated driver; (10) marital status of the rated driver; (11) persistency (years insured by the company); (12) nonsmoker; (13) secondary driver characteristics; (14) multipolicies with the same or an affiliated company; (15) relative claims frequency; and (16) relative claims severity. (Regs., § 2632.5, subd. (d).) Relative claims frequency and severity are “territorial” factors based on where the vehicle is garaged and the average number and cost of claims in that zip code or census tract. (Id., § 2632.5, subd. (d)(15), (16) .) (a) Categories Each rating factor is divided into two or more categories which determine whether the policyholder receives a discount or a surcharge. For example, the mileage rating factor could be divided into categories for high, average, and low. Those in the high category would be surcharged, those in the low category would receive discounts, those in the average category would see no change in their base premium. (b) Relativities To accomplish these adjustments, each category is given a number known as a “relativity.” Continuing the foregoing example and assuming that the rating factors are applied through a series of multiplications, the high mileage category could be assigned a relativity of 1.5, the average mileage category could be assigned a 1.0, and the low category a 0.5. If the base premium were $800, the premium of those in high mileage category would be increased to $1,200 ($800 x 1.5), the premium of those in the low *1188 category would be decreased to $400 ($800 x 0.5), and the premium of those in the average category would remain unchanged at $800 ($800 x 1.0). This process is repeated for all of the rating factors to arrive at the final premium. The process can be described as the multiplicative algorithm: premium = base rate x factor 1 x factor 2 x factor 3, etc., with high risk factors having a value greater than 1, average risk factors equal to 1, and low risk factors below 1. (c) Sequential Analysis The regulations require that the company’s relativities be initially determined through a “sequential analysis” of the rating factors. (Regs., § 2632.7.) Sequential analysis is a complex process which accounts for the fact that different rating factors (e.g., driving safety record and miles driven) may bear on risk in overlapping ways (e.g., the more one drives, the greater the likelihood of an accident). State Farm Mutual Automobile Insurance Company actuary Ina Becraft has summarized the process as follows: “[F]or each rating factor the loss and expense costs are determined for every category within that factor, and a relativity calculated based upon the relative cost to insure insureds within those categories. The analysis is performed in a prescribed sequence of factors, and the cost explained by prior factors is removed in the analysis of each subsequent factor. In that way, if there is a correlation of the risk of loss between two or more factors, the first factor analyzed will explain all of the risk of loss that it can and the second factor is confined to explaining the residual risk, to the extent the second factor explains that residual risk.” The regulations establish the order in which the rating factors are to be analyzed (Regs., § 2632.7, subd. (b)), and direct that the sequential analysis “remove the variation in loss costs already explained by prior factors” (Id., § 2632.7, subd. (a)). The prescribed sequence tracks the order of mandatory and optional factors set forth in Insurance Code section 1861.02. The first mandatory factor, driving safety record, is to be analyzed first, followed next by miles driven, then years of driving experience, and then the optional factors the company uses. (Regs., §§ 2632.7, subd. (b), 2632.5, subd. (c).) The territorial factors of claims frequency and severity must be analyzed last, but the optional factors may otherwise be analyzed in any order the company chooses. (Id., § 2632.7, subd. (b)4.) Farmers Insurance Exchange actuary Jonathan Adkisson has testified that “the order in which the factors are analyzed is very important. If there is [a] correlation of the risk of loss for two factors, the factor analyzed first will be assigned the majority of the variation in costs. The second factor will receive only the residual (or uncorrelated) variation, after the data are adjusted.” *1189 A sample adjustment had been prepared by consulting actuary Michael Miller. In Miller’s example, insureds in the category of 0 to 4 years of driving experience have a relativity of 1.95 for that rating factor before the application of sequential analysis. The relativity is greater than 1, which means that these drivers are considered a relatively high risk. “[T]he next step is to back-out the effects of the rating factors previously analyzed.” Under the regulations, safety record and miles driven are analyzed before driving experience. (Regs., §§ 2632.7, subd. (b), 2632.5, subd. (c).) In the example, drivers with 0 to 4 years experience have an average relativity of 1.1 after analysis of these 2 higher order factors. Since drivers in this category would be considered a high risk (relativity greater than 1) even before their driving experience is considered, a “downward adjustment” is made to the driving experience factor to “avoid[] double-counting the risk.” The unadjusted relativity of 1.95 for the driving experience factor is divided by the 1.1 average relativity from the previous factors to yield a lower relativity of 1.77 for driving experience. (3) Weight After the initial relativities are determined according to this process, the rating factors are given a “weight.” The weight of an individual rating factor is calculated with what is called a “single omit” formula, which measures the factor’s average influence on the premiums of the company’s policyholders. The difference between the total premium for each car and that premium without the single factor being weighed is calculated as an absolute number (i.e., without regard to whether it is positive or negative), the results of those calculations are added together, and the total is divided by the number of calculations to yield the factor’s “weight” or average effect on the premiums of all policyholders. (Regs., § 2632.8, subd. (c).) For example, if the company has three policyholders, the first has his or her premium increased by $75 by virtue of the driving safety factor, and the other two have their respective premiums decreased by $50 and $25 on account of that factor, the weight of the factor would be 50 ([$75 + $50 + $25] -s- 3). (a) Alignment of Weights The next step is to align weights in accordance with Proposition 103. Section 2632.8, subdivision (d) of the regulations provides that the factor weights “must align in decreasing order of importance as follows: driving safety record must have the most weight followed by annual miles driven followed by years of driving experience followed by the weight for the optional factor.” Under regulations section 2632.8, subdivision (a), one weight is to be calculated “for all the optional factors . . . taken together as *1190 a single factor weight.” Each factor weight is the factor’s average effect on the company’s premiums, and the single factor weight of the optional factors is the average of those averages for the optional factors. If the company uses 10 optional factors, two with weights of 6 and eight with weights of 1, the single factor weight of the optional factors would be 2 (the combined factor weight or premium effect of 20 [(2 x 6) + (8 x 1)] divided by the number of factors [10]). 2 The regulations thus permit the use of optional factors which, both individually and collectively, may have greater weight than any of the mandatory factors. A company, for example, could combine the optional factors posited in the preceding paragraph, with mandatory factors of driving record, miles driven, and driving experience having weights of 5, 4, and 3, respectively. Even though two of the optional factors would have weights of 6 (more than the weight of any mandatory factor), and. the optional factors would have an even greater combined weight or premium effect of 20, the weights would be considered in proper alignment because the single average weight of the optional factors would be 2 (less than the weight of each mandatory factor). This result is exemplified by the weights of the rating factors used by State Farm, the state’s largest automobile insurer, for its bodily injury and property damage liability (BIPD) coverage. 3 Insurers submitted for approval by the Insurance Commissioner 4 “class plans” for their private passenger auto rates and premiums which identified the rating factors they used and the weights of those factors. (Regs., §§ 2632.3, 2632.8, subd. (a), 2632.10, 2632.11.) The weights of the three mandatory factors for BIPD coverage in State Farm’s approved class plan are 20.65 (safety record), 13.64 (mileage), and 10.51 (driving experience). The single factor weight for all of the optional factors is 9.82. Since the weights of the mandatory factors are aligned in the requisite order of importance, and the single optional factor weight (9.82) is less than the weight of the least important mandatory factor (driving experience—10.51), the factors are considered to be properly aligned under the regulations. {Id., § 2632.8, subd. (d).) State Farm uses the following 10 optional factors: (1) the territorial factors of relative claims severity and frequency combined as one factor with a weight of 34.6; (2) gender and marital status combined as one factor with a *1191 weight of 25.10; 5 (3) persistency (years insured) with a weight of 15.51; (4) multi/single car 14.52; (5) academic standing 2.70; (6) percentage use by driver 2.22; (7) driving safety education 1.51; (8) usage of vehicle 1.46; (9) mature driver improvement ,31; 6 and (10) second driver characteristics .24. These individual optional factor weights add up to 98.2, far more than the weight of any of the mandatory factors. Some of the individual optional factor weights also exceed the individual weights of mandatory factors (the territorial factor and gender/marital status factor each exceed every mandatory factor, persistency exceeds mileage and driving experience, etc.). Optional factors both individually and collectively are allowed to have a greater weight than any of the mandatory factors because the average weight of the 10 optional factors is only 9.82 (98.2 4- 10), less than that of any mandatory factor. (b) Pumping and Tempering An additional step must be taken if the three mandatory factor weights and the single optional factor weight calculated both with sequential analysis and the single omit formula do not align in the requisite order (safety record-mileage-driving experience-optional factor). In that event, it is necessary to adjust the weight of factors by “ ‘pumping’ ” or “ ‘tempering’ ” their relativities. “ ‘Pumping’ ” increases the factor’s weight by moving the relativities further from 1; “ ‘tempering’ ” reduces the factor’s weight by moving the relativities closer to 1. A previous example posited a three-category mileage factor, with a high mileage relativity of 1.5, an average mileage relativity of 1.0, and a low mileage relativity of .5. In that example, where a base rate of $800 was *1192 assumed, these relativities produce premiums for high, average, and low mileage drivers of $1,200 ($800 x 1.5), $800 ($800 x 1.0) and $400 ($800 x .5), respectively. If there are four drivers, one in the high mileage category, two in the average category, and one in the low category, the mileage factor would have a weight of 200. Two drivers would have their premiums increased or decreased by $400, two drivers would see no change in their premium, and thus the factor, on average, would change the premium by $200 ([400 + 0 + 0 + 400] + 4). If the relativities for the high and low mileage categories were “ ‘tempered’ ” (moved closer to 1) from 1.5 and .5 to 1.2 and .8, respectively, then taking the same four drivers the weight of the factor would decrease from 200 to 80. The premium for the high mileage driver would be $960 ($800 x 1.2, an increase of $160), the premiums for the average drivers would be unchanged, and the premium for the low mileage driver would be $640 ($800 x .8, a decrease of $160). The four drivers collectively would see their premiums increased or decreased by $320 (160 + 0 + 0+ 160), and thus the average premium effect or weight of the factor would be 80 ($320 + 4). The opposite result would occur if the relativities were “ ‘pumped.’ ” B. King v. Meese In 1987, the year before Proposition 103 was enacted, King v. Meese (1987) 43 Cal.3d 1217 [240 Cal.Rptr. 829, 743 P.2d 889] rejected a challenge to enforcement of financial responsibility laws requiring drivers to have insurance. Portions of the majority and concurring opinions in that case help to set the context for the provisions of the Proposition at issue here. The majority wrote: “California is a so-called ‘open rate’ state, that is, rates are set by insurers without prior or subsequent approval by the Insurance Commissioner (Commissioner). (Ins. Code, § 1850.) This is not to say, however, that there is absolutely no regulation of the rates. California law does require that rates not be ‘excessive, inadequate or unfairly discriminatory.’ (Ins. Code, § 1852.) No rate is excessive unless: ‘(1) such rate is unreasonably high for the insurance provided and (2) a reasonable degree of competition does not exist in the area with respect to the classification to which such rate is applicable.’ (Ibid.) Risk classifications are permissible if based on any reasonable (i.e., actuarially sound), and not prohibited, ground. (Ins. Code, § 1852, subd. (d).) Although the term ‘unfairly discriminatory’ is not defined in Insurance Code section 1852, section 11628 of that code prohibits discrimination by an insurer with regard to issuance of policies, or the terms of such policies, on the basis of ‘race, language, color, religion, national origin, ancestry, or location within the same geographic area.’ ‘Geographic area’ is defined as an area ‘not less than 20 square miles,’ and *1193 is made up by combining a series of contiguous zip code zones. Under the statutory scheme, different geographic areas may be treated differently.” {King v. Meese, supra, 43 Cal.3d at pp. 1221-1222.) In a concurring opinion, Justice Broussard, joined by Justice Mosk, addressed the “serious problem [of] insurance pricing practices which make automobile liability insurance prohibitively expensive for many of the urban poor.” {King v. Meese, supra, 43 Cal.3d at p. 1237.) The concurrence cited “exhibits verifying] that private insurance rates in South Central Los Angeles are two to three times as high as rates in other areas of the state, with the result that good driver rates in Los Angeles often exceed rates charged drivers with bad records in other areas.” {Id. at p. 1240.) The opinion noted that the antidiscrimination provisions of the auto insurance law (§ 11628 et seq.) had “been interpreted to authorize territorial rate differentials, so long as rates are uniform within 20-square-mile blocks,” with the result that “insurers can draw lines which have the practical effect of discriminating between applicants on the basis of race.” {King, at p. 1243.) The concurrence also observed that there was no statutory definition of “ ‘unfairly discriminatory’ ” rates, and that the Commissioner had issued no regulations or decisions explaining when a rate would be considered unreasonable or discriminatory. {King v. Meese, supra, 43 Cal.3d at pp. 1240, 1242.) However, the Commissioner appeared to “assume that so long as a rate is actuarially sound it cannot be unfairly discriminatory or unreasonably high” {id. at p. 1241), and the concurring opinion thought that this assumption was “open to challenge. One can argue that it is unfairly discriminatory to use classifications which result in charging good drivers in some areas much more than bad drivers in other[] parts of the state; it could be considered unreasonable to price liability insurance at levels many cannot afford. Rates which took affordability into account, and weighted driving record more than residence, would go far to alleviate the problem caused by the financial responsibility laws” {id. at pp. 1241-1242). C. Proposition 103 Section 1 of Proposition 103, under the heading “Findings and Declaration,” states that: “Enormous increases in the cost of insurance have made it both unaffordable and unavailable to millions of Californians. [H] The existing laws inadequately protect consumers and allow insurance companies to charge excessive, unjustified and arbitrary rates. fl[] Therefore, the People of California declare that insurance reform is necessary.” (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, p. 99.) Among the reforms then listed is that “automobile insurance rates shall be determined primarily by a *1194 driver’s safety record and mileage driven.” {Ibid.) Section 2, under the heading “Purpose,” indicates that the Proposition was intended “to protect consumers from arbitrary insurance rates and practices, to encourage a competitive insurance marketplace, to provide for an accountable Insurance Commissioner, and to ensure that insurance is fair, available, and affordable for all Californians.” {Ibid.) In addition to the rating factor provisions of section 1861.02, subdivision (a), the Proposition added section 1861.05, subdivision (a), which provides that: “No rate shall be approved or remain in effect which is excessive, inadequate, unfairly discriminatory or otherwise in violation of this chapter. In considering whether a rate is excessive, inadequate or unfairly discriminatory, no consideration shall be given to the degree of competition and the commissioner shall consider whether the rate mathematically reflects the insurance company’s investment income.” In their rebuttal argument in the ballot pamphlet for the November 1988 election, the Proposition’s opponents maintained that: “RATES WILL INCREASE by an average 22% for two-thirds of the state’s drivers, according to the State Department of Insurance, because PROP 103 eliminates rating based on the driving safety record of your neighborhood and forces suburban and rural drivers to subsidize motorists in high-risk areas.” (Ballot Pamp., Gen. Elec., supra, rebuttal to argument in favor of Prop. 103, p. 100.) These claims were repeated in the ballot pamphlet argument against the Proposition, which alleged that the Proposition would: “Raise insurance premiums, in the long term, for the majority of California drivers. PROP 103 forces insurers to ignore the driving safety record of where you live and, instead, forces you to subsidize drivers in areas that have the highest insurance losses. For example, a 55-year-old suburban driver will end up paying more for insurance so that a young urban driver can pay less. A State Department of Insurance study recently predicted that this aspect of PROP 103 will raise rates for two-thirds of the state’s drivers—by an average 22%!” {Id., argument against Prop. 103, p. 101.) We quoted the proponents’ rebuttal argument at the outset of the opinion. Harvey Rosenfield, the author of Proposition 103, wrote as chair of the group Voter Revolt to Cut Insurance Rates that: “103 forces insurance companies to base your rates on your driving record first, rather than on where you live. That means good drivers throughout the state will pay less than they do now, while bad drivers will pay more.” (Ballot Pamp., Gen. Elec., supra, rebuttal to argument against Prop. 103, p. 101.) D. The Tempered Regulations The rating factor provisions of section 1861.02, subdivision (a) became operative in November 1989; the Commissioner was directed by the statute *1195 to adopt regulations which implemented them (§ 1861.02, subd. (e)). The history of those regulations was recounted by the Department of Insurance (hereafter Department) in its May 22, 1996, Final Statement of Reasons, No. RH-338 (Statement) for adoption of the last of the regulations in their current form. In 1989 and 1990, former Commissioner Gillespie adopted what came to be known as the “Tempered Regulations,” which “required that the weights assigned to various rating factors be ‘tempered’ so that—notwithstanding the weight that would be assigned to a rating factor if calculated purely on the basis of a sequential analysis of the mandatory and optional factors: the second mandatory factor in Proposition 103 (number of miles driven annually) would account for less of the premium than is accounted for by the first mandatory factor (driving safety record); the third mandatory factor (number of years of driving experience) would, in turn, account for less than the second mandatory factor; and, the weight of any and all optional factors used by an insurer would in turn be less than that accounted for by the third mandatory factor.” (Statement, supra, at pp. 2-3.) The Tempered Regulations further provided that premiums calculated with the tempered approach could not exceed either “the premium that would have been charged if calculated by a sequential analysis of all mandatory and optional factors,” or “the premium that was charged or would have been charged in the immediately preceding calendar year, as adjusted to reflect any increase in the consumer price index.” (Statement, supra, at p. 3.) In an August 13, 1990, note appearing in the annotated Barclay’s California Code of Regulations (Register 91, No. 14 (Apr. 5, 1991) pp. 728.10 to 728.12) following title 10, section 2632.1 (Note), the Department explained the considerations that led to adoption of the Tempered Regulations. The Note indicated that the Commissioner had appointed an actuarial advisory committee with three members representing insurers, consumer groups, and the Department. The committee members agreed that the three mandatory rating factors should be applied in the statutory order, followed by optional factors designated by regulation, using a sequential analysis. (Note, supra, at p. 728.11.) The committee disagreed as to whether premiums should ultimately be set through what was called a “cost-based approach,” or rather through a “ ‘tempered’ ” approach which “would artificially strengthen the first three factors and give them more weight than they would otherwise be given in a cost-based system.” (Id. at p. 728.12.) “The Committee members agreed that if the cost-based approach is utilized, even with the sequential analysis methodology, then legitimate Optional Factors will likely call for premium rates in certain urban areas, such *1196 as the Los Angeles inner city area, to be higher than those in most non-urban areas. The Committee members also agreed that if the cost-based approach is ‘tempered’ by overriding the ‘weight’ for the Optional Factors, then premium rates in non-urban areas will likely increase while rates in certain inner city areas will likely decrease.” (Note, supra, at p. 728.12.) Commissioner Gillespie found it “very likely, probably to the point of certainty, that these Optional Factors would, on a pure cost-based system which bases premium on the risk of loss, account for a greater proportion of the risk of loss than would any one of or possibly all three Mandated Factors.” (Note, supra, at pp. 728.11 to 728.12.) Nevertheless, the Commissioner found “that the three Mandated Factors must be given the most weight, regardless of whether the result would otherwise be viewed as discriminatory, arbitrary or inadequate based upon actuarial science and a cost-based approach because the statutory language states that this must be done.” {Id. at p. 728.12.) “As a result,” the Note continues, “the Commissioner has determined that she cannot implement Proposition 103 according to its terms unless she ‘tempers’ the rating factors to provide that no Optional Factor may be given more weight than any of the Mandatory Factors. Further, the combined weight of the optional factors may not have more weight than any Mandatory Factor.” {Ibid.) The Commissioner thought that the applicable provisions of Proposition 103 were in conflict, and that the significantly higher premiums in nonurban areas likely to result from the required tempering were a “serious problem.” (Note, supra, at p. 728.12.) Since “rates for many Californians will very likely actually increase and this result would be precisely the opposite of what the promoters and Proposition 103 promised it would accomplish,” the Commissioner found that she could “only harmonize the provisions of Proposition 103 which call for lower, not higher rates” by “forbid[ding] substantial rate increases as a result of such ‘tempering.’ ” {Ibid.) In so doing, she observed that: “Proposition 103 was voted on as a measure to lower insurance rates, not to raise them. Proposition 103 was intended to avoid arbitrary rates, not to impose them. Proposition 103 was intended to encourage competition, not discourage it. Proposition 103 was supposed to avoid excessive rates, not impose them. Proposition 103 was intended to avoid unfair discrimination, not to encourage it. Proposition 103 was intended to make insurance more available, not less available.” {Ibid.) The Tempered Regulations were challenged in an action in Los Angeles Superior Court, in which Judge Miriam Vogel issued a preliminary injunction against their enforcement in May of 1990 shortly before her elevation to the Court of Appeal. (Judicial Council of Cal., Coordination Proceeding No. *1197 JCC 002419 (1990) Proposition 103 Implementation Cases (Proposition 103 Cases).) In her statement of decision, Judge Vogel reasoned that “the overriding concern in adopting Proposition 103 was to ensure that rates are neither excessive nor unfairly discriminatory,” and thus that “[a]ny rating factor adopted by the Commissioner must be consistent with the mandate of Section 1861.05(a).” She noted the Commissioner’s concession that the tempered regulations would raise rates in nonurban areas, and the Commissioner’s decision “to rectify this problem by imposition of a totally artificial cap on all insurance rates.” She also noted the Commissioner’s finding that “under a tempered approach some drivers would be required arbitrarily to subsidize other drivers’ rates in a manner that unfairly discriminates against them.” (Id., Statement of Decision.) She concluded: “There is nothing in Proposition 103 that requires abandonment of a cost-based approach; to the contrary, the requirement that no rate be inadequate, excessive or unfairly discriminatory, considered in light of the evidence before this Court, compels the conclusion that only a cost-based approach will afford consumers the relief they sought when they voted for Proposition 103. As with other provisions of the [Tempered Regulations], it appears that the Commissioner’s reason for rejecting a cost-based approach is her effort to solve social problems which simply are not subject to cure by her regulations or orders of this Court. Artificial weighting of rating factors is invalid; it is not actuarially sound and it results in unfairly discriminatory, inadequate and excessive rates.” (Proposition 103 Cases, supra, Statement of Decision.) E. Interim Regulations and the Impact Analysis Commissioner Gillespie appealed from the order granting the preliminary injunction, and to comply with the injunction adopted interim regulations in August of 1990 which provided for application of the rating factors through a sequential analysis. Commissioner Garamendi took office in 1991 while the appeal from the injunction was pending, and indicated that he did not intend to reinstate the Tempered Regulations even if the appeal succeeded. In view of that development, the appeal was dismissed by Division Seven of the Second Appellate District in 1992, and the case was remanded to the trial court where it was eventually dismissed. (Allstate Ins. Co. v. Gillespie (Jan. 22, 1992, B050439) mod. Feb. 20, 1992 [nonpub. opn.].) 7 The Court of Appeal declined to reach “abstract questions” of rating methodology before Commissioner Garamendi had developed additional *1198 actuarial data and new regulations. (Allstate Ins. Co. v. Gillespie, supra, No. B050439.) The court thought that “[t]he pertinent portions of Proposition 103 are ambiguous to an extent which demands reasoned construction by the Commissioner beyond that contained in the record compiled by former Commissioner Gillespie as a prelude to intelligent judicial action.” One issue presented was; “Should [section 1861.02, subdivision (a)] be interpreted to maximize the [rate] reduction to a class consisting only of urban drivers or to maximize the number of ‘good drivers’ who benefit from a lesser reduction per driver?” In the court’s view, this was “a question best left to determination after Commissioner Garamendi has developed the predicate for his new regulations and promulgated them.” The opinion added that “[t]he present Commissioner’s eventual interpretation of the relevant statutory provisions will be entitled to great deference by this court if it is supported by the record and is compatible with the wording and purpose of Proposition 103.” (Ibid.) While the appellate court thus declined to resolve the substantive issues, it ventured various observations about Proposition 103 and the parties’ positions to support its suggestion that the issues were appropriate for mediation. The court characterized the Tempered Regulations adopted by Commissioner Gillespie, as well as the sequential, or “ ‘cost-based,’ ” analysis advocated by the insurers as “extreme positions.” (Allstate Ins. Co. v. Gillespie, supra, No. B050439.) On the one hand, the opinion noted Commissioner Garamendi’s argument that “the very concept of ‘cost-based pricing’ as the term has been employed in this case ‘is at best only a rough approximation to neutral, scientific process, if not a complete fiction.’ ” On the other hand, the court found that arguments for the Tempered Regulations based on statements in the ballot pamphlet were not “nearly as clear cut” as they purported to be. Although the proponents of Proposition 103 had indicated that rates would be determined by driving record first, rather than residence, and opponents had warned that rates for nonurban drivers would increase, the court thought it a “fair inference” that “the voters in favor of the initiative rejected the arguments in opposition and accepted the rebuttal to these arguments to the effect that automobile insurance rates would be reduced for all good drivers.” (Ibid.) Interim regulations providing for sequential analysis of the rating factors were readopted from 1990 to 1994. At the end of Commissioner Garamendi’s tenure in December 1994, the Department released an “Impact Analysis of Weighting Auto Rating Factors to Comply with Proposition 103” (Impact Analysis) based on the collection and examination of over 10 million records from the state’s 11 largest insurers over an 18-month period. (Id. at pp. v & vi.) The Impact Analysis covered approximately 80 percent of the state’s private passenger auto insurance market, and its analytical scope was “revolutionary in auto rate setting.” (Id. at pp. 29 & vi.) Among other things, *1199 “[e]very category used by every rating factor for every insurer and the rate associated with it was identified and programmed into a computer.” (Id. at p. 2.) The results disclosed “wide variations ... in how different insurers created and used the same rating factors,” and “[t]hese substantial variations from insurer to insurer appealed] to be arbitrary.” (Ibid.) It also appeared that “most companies were unable to link their loss data with their classification data, except in the most limited way.” (Id. at p. 31.) The Impact Analysis found that “[n]one of the insurers analyzed [were] currently complying with the requirements that auto premiums be determined by the safety record, mileage, and driving experience rating factors in that rank order.” (Id. at p. 3.) The Impact Analysis compared the premiums currently charged with those produced by different models using rating factors with weights which were presumed to comply with Proposition 103. In order to derive current premiums which could be compared with those produced by different weighting methods, “it was necessary to transform each company’s current rating practices into a rating plan utilizing standardized factors.” (Impact Analysis, supra, at p. 31.) The difference or “dislocation” between the existing premiums and those produced by the models was considered to be “positive” (intended) to the extent that “[g]ood,” low mileage and experienced drivers paid less for their insurance, or “bad,” high mileage and inexperienced drivers paid more. The opposite results were viewed as “negative” (unintended) dislocation. (Id. at p. 39.) The Impact Analysis rejected sequential analysis as a method of weighing factors because it “is not a method for measuring how much the rating factor influences the premium charged to consumers .... A factor could be the second factor developed and still have a very minimal influence on premium. ... [H] ... If a company had not previously performed a proper sequential analysis, just performing the proper sequential analysis could slightly increase the influence of the three required factors on premium. However, there is no guarantee that they will have more influence on premium than any other rating factor.” (Impact Analysis, supra, at p. 21.) Two of the alternative models analyzed used weights determined by the “single omit” method. Ninety percent or more of the dislocation produced by these models was either positive, or considered “nil” (+/- 10 percent of current premium); only 5 percent or less was negative. (Impact Analysis, supra, at pp. 43-46.) Thus, the “major approaches examined . . . resulted] in primarily positive or nil dislocation.” (Id. at p. 55.) The Impact Analysis also found that “for most consumer groups there is not that much change in average premium [under the alternative models] from what is currently being *1200 charged.” (Id. at p. 42.) In Commissioner Garamendi’s view, the Impact Analysis thereby “demonstrate[d] that it is possible to meet the requirements of Proposition 103 without massive variations in premiums.” (Id. at p. vi, emphasis omitted.) Of particular interest is what was meant by “meeting] the requirements of Proposition 103.” The Impact Analysis assumed that Proposition 103 “intended that the pricing of auto insurance be restructured so that the safety record, mileage rating, and driving experience factors have the greatest impact on the amount that consumers are charged for auto insurance.” (Impact Analysis, supra, at p. 7.) Tables on pages 110 and 111 of the Impact Analysis showed the optional factors used for the “single omit” models and listed different weights for those factors under various company plans which were deemed to be “in [c]ompliance” with Proposition 103. 8 The weights of the three mandatory factors were aligned in the requisite descending order, and the weight of each individual optional factor was less than that of any mandatory factor. 9 This alignment was evidently regarded as sufficient under Proposition 103. Although the Impact Analysis stated that “Proposition 103 requires territory to have less influence than safety record, mileage, and years licensed,” the “single omit” models used two territorial factors, “frequency” and “severity,” and the combined weight of those factors was allowed to exceed the individual weights of mandatory factors. (Id. at pp. 9, 110-111.) The Impact Analysis noted that the use of two territorial factors rather than one reduced the dislocation the models produced. (Id. at pp. 35, 74.) The process of formulating regulations for the calculation of factor weights continued after Commissioner Quackenbush took office in January of 1995. Public investigatory hearings on the Impact Analysis were held in San Diego, Los Angeles, and Sacramento in April 1995. Regulations providing for sequential analysis of rating factors were approved and readopted as emergency regulations in February and June of 1995. In December 1995, the Commissioner held a public hearing and accepted written comments on proposed permanent regulations. (Statement, supra, at p. 5.) F. Current Regulations The relevant current regulations became operative in 1996. They include: Regulations section 2632.4, which provides in subdivision (a) that “[n]o insurer shall use a rating factor which is not set forth in these regulations,” *1201 and states in subdivision (b) that “[n]o insurer shall use a rating factor . . . in a manner that does not bear a substantial relationship to loss.” Regulations section 2632.5, which further defines the mandatory factors of safety record, mileage and driving experience (subd. (c)), and approves the use of the 16 optional rating factors we have previously listed (subd. (d)). The two territorial factors of relative claims frequency and severity in subdivision (d)(15), (16) consolidated more numerous territorial factors which had been used in the previous version of this section. 10 Regulations section 2632.7, which directs the manner in which sequential analysis is to be performed to “ensure[] the influence of a rating factor is properly counted.” The previous version of this section merely specified the order in which the factors were to be analyzed and did not otherwise “describe or define ‘sequential analysis[,’] thereby allowing each insurer to implement different interpretations and use different procedures.” (Statement, supra, at p. 8.) Regulations section 2632.8, which provides for the factor weight determination challenged herein. Subdivision (a) states in full: “For each type of coverage, four factor weights shall be calculated, one weight for each of the three mandatory factors listed in Section 2632.5(c)(1) through (3) and one for all the optional factors (from Section 2632.5(d)) taken together as a single factor weight.” Regulations section 2632.8, subdivision (c) directs how these calculations are to be made. This subdivision reads in full: “For every insured vehicle in the data set and each rating factor utilized in the class plan: ft¡] 1. First, calculate the premium using the initial relativities from Section 2632.7(c); ft[] 2. Second, calculate the premium excluding the rating factor being analyzed; ftQ 3. Third, calculate the absolute value of the difference between subdivision (c)(1) and subdivision (c)(2); [f] 4. The weight for the rating factor being analyzed is the summation of the amounts in subdivision (3) divided by the number of calculations.” Regulations section 2632.8, subdivision (d) stipulates that “[t]he weights of the factors, as calculated in subdivision (c), must align in decreasing order of importance as follows: driving safety record must have the most weight followed by annual miles driven followed by years of driving experience followed by the weight for the optional factor.” The balance of this subdivision covers the process of pumping or tempering to adjust the factor weights if they do not initially align in this order. *1202 G. The Administrative Proceeding Fanners Insurance Exchange, Mid-Century Insurance Company, and Truck Insurance Exchange (collectively Farmers) and State Farm filed applications in February of 1997 for the Commissioner’s approval of their class plans under the current regulations. The Proposition 103 Enforcement Project (Project), a nonprofit organization founded and led by Harvey Rosenfield, filed petitions for a hearing regarding these class plans and that of the Allstate companies. The petitions were granted and the hearings were consolidated into one proceeding on the issue of the methodology for calculating the one combined weight of the optional factors. Consumers Union of U.S., Inc., and the Southern Christian Leadership Conference of Greater Los Angeles, Inc., (collectively Consumers Union) intervened on the side of the Project. The dispute boiled down to two different ways of calculating the single weight for the optional factors under regulations section 2632.8. Under the “single omit” formula described above and in section 2632.8, subdivision (c), factor weights are determined by calculating the absolute value of the difference between the total premium for each car and the premium without the factor being weighed, adding those amounts together and dividing that total by the number of “calculations.” There was no dispute about the method of calculating the individual weights of the three mandatory factors under this formula. The absolute difference for each factor was determined for each car, those differences were added together, and this “sum of the absolute differences” was divided by the number of cars to yield the weight of the factor. To return again to the State Farm example set out in part I.A.(3), (ante, pp. 1189-1192), the sum of the absolute differences for the safety record factor for the company’s 2,751,975 insured cars was $56,833,135 (policyholders in the aggregate had their premiums increased or decreased by that total amount by virtue of their safety records), and the weight of the factor was thus 20.65 ($56,833,135 h- 2,751,975). 11 The average policyholder, in other words, had his or her premium go up or down $20.65 on account of his or her safety record. The sums of the absolute differences for mileage and driving experience were $37,534,388 and $28,916,715, respectively, and thus the weights of those factors were 13.64 ($37,534,388 -r 2,751,975), and 10.51 ($28,916,715 -r 2,751,975), respectively. This approach, as previously explained, measures the average premium effect of the rating factor on individual policyholders. As for the fourth factor weight for all of the optional factors combined, the insurers followed the foregoing procedure for each of the individual optional *1203 factors for each car, then totaled the sums of the absolute differences for all of those factors, and then divided that total by the number of vehicles multiplied by the number of factors. In the State Farm BIPD example, the sum of the absolute differences for the “Usage of Vehicle” optional factor was $4,013,849, the sum of the absolute differences for the “Academic Standing” factor was $7,424,429, and so on, for each of the 10 optional factors State Farm used. The total of those 10 sums ($270,127,062) was divided by the number of cars (2,751,975) multiplied by the number of factors (10)—a denominator of 27,519,750—to yield an optional factor weight of 9.82 ($270,127,062 + 27,519,750). Another way of reaching this figure would be to total the individual weights of the optional factors, and divide that total by the number of factors. In the State Farm case, the weight of the usage of vehicle factor would have been 1.46 ($4,013,849 -f 2,751,975), the weight of the academic standing factor would have been 2.70 ($7,424,429 -f 2,751,975), and the total of those weights, plus the individual weights of the other optional factors would have been 98.17. That total divided by the number of factors (10) would have been 9.82, the same number achieved in the preceding paragraph. The optional factor weight under the insurers’ method was thus the average weight of the individual optional factors. As previously noted, this method allows the weights of individual optional factors to exceed those of individual mandatory factors. Consumers Union argued that if the insurers’ method of calculating the optional factor weight complied with regulations section 2632.8, then that regulation as so interpreted violated the factor ordering scheme of Insurance Code section 1861.02, subdivision (a). Consumers Union took the position that the statute required the weights of individual optional factors to be less than those of individual mandatory factors. The Project went a step further and submitted that the aggregate weight of the optional factors could not exceed the weight of any mandatory factor. The Project’s posthearing brief argued that the insurers’ method “subverts the legislative intent of [section 1861].02 by allowing optional rating factors, either individually or combined, to have greater importance in terms of dollar impact on premium than one or more of the mandatory factors.” (Italics added.) The Project’s expert witness, Bimy Bimbaum, likewise testified that the insurers’ method improperly “allow[ed] rating factors and relativities for optional factors, which in some cases individually but in all three cases taken as together have far greater impact, far greater importance in the determination of premium than one or more of the mandatory factors.” Consumers Union argued that the insurers’ calculation was improper because it yielded only an average weight for all of the optional factors. *1204 Consumers Union pointed out that whereas individual factor weights measured the average effect of the factor on the premium, the single optional factor weight calculated by the insurers was an average of those averages. Consumers Union observed that “[t]he total premium effect of the ten optional factors [used by State Farm] taken together adds up to $270,127,062—the sum of the total premium effects of the ten optional factors.” Consumers Union noted that in the State Farm example “the total premium effect of all optional factors [$270,127,062] is 4.75 times greater than the total premium effect of the first mandatory factor, driving safety record [$56,833,135].” Another way of stating this would have been to say that the combined weight of all of the optional factors (98.2) was 4.75 times greater than the weight of the safety record factor (20.65). Thus, according to Consumers Union, the problem was that “the insurers are not calculating the average premium effect of the optional factors, and instead are substituting the average of the average premium effects.” To ensure that weights were consistently measured in accordance with this argument and that the weights of optional factors individually and as a whole were less than that of any mandatory factor, Consumers Union could have endorsed a system where the single optional factor weight was simply the sum of the individual optional factor weights. This number in the State Farm case would have been 98.2, what Consumers Union described as the “total premium effect” of the optional factors ($270,127,062) divided by the number of vehicles (2,751,975). In that event, State Farm’s mandatory factor weights of 20.65, 13.64 and 10.51 would have had to be “‘pumped’” to more than 98.2, or the 98.2 optional factor weight would have had to be tempered to less than 10.51, to achieve the requisite alignment of factor weights. Instead, Consumers Union endorsed a different method proposed by Project expert Bimbaum. Rather than calculating the premium effect of each optional factor separately for each car, Bimbaum proposed only one optional factor calculation per car: the absolute difference between the total premium and the premium without all of the optional factors. Bimbaum took the sum of those absolute differences for all cars and divided that figure by the number of cars to yield the single optional factor weight called for in the regulation. The resulting optional factor weight was higher than the one produced by the insurers’ method, but lower than the figure produced by simply adding up individual factor weights. In the State Farm example, as we have noted, the individual optional factor weights added up to 98.2, and the optional factor weight produced by the insurers’ method was 9.82 (98.2 divided by *1205 the number of optional factors [10]). In her testimony at the administrative hearing, State Farm actuary Becraft estimated that under the Bimbaum method the optional factor weight for State Farm’s class plan would have been approximately 55.8. The difference between the 98.2 sum of the individual factor weights and the 55.8 weight derived from the Birnbaum method resulted apparently because the Bimbaum method measured the net effect of all the optional factors combined on the premiums for particular cars. Since some optional factors may increase the premium while others decrease it, the optional factors may cancel each other out to some extent when their effects are combined for each car. For example, if there were two optional factors, one that increased the premium by $10 and the other that decreased the premium by $10, those optional factors would be deemed to have a single factor weight of zero under the Bimbaum method. If there were two cars that fit this same profile, the optional factor weight would still be 0 ([0 + 0] * 2). The two factors would not “count” because they completely cancelled each other out for each car in the sample. However, these same factors would have weight in a system that determined individual weights for optional factors in the same manner as those weights are calculated for the mandatory factors. Again, that method is to determine the premium effect of the factor for each car, add those numbers up, and divide by the number of cars to measure the average premium effect of the factor per car. In the foregoing two-car example, each of the optional factors would have individual weights of 10 ([10 + 10] ~ 2), and the sum of those individual weights would be 20 (10 + 10), as opposed to the zero weight derived from the Bimbaum method. 12 In any event, what is of most interest is how the results produced by the Birnbaum method differed from those required under different interpretations of section 1861.02, subdivision (a). As previously noted, Consumers Union was of the view that the statute required individual optional factors to weigh less than any mandatory factor. Birnbaum acknowledged in his *1206 testimony that the optional factor weight calculated under his method could exceed the weight of a mandatory factor even if the individual weights of every optional factor were less than those of any mandatory factor. These calculations, he said, were “not comparable.” It thus appears that Birnbaum’s method went further in suppressing optional factor weights than a system which merely required that the weight of each individual optional factor be less than that of any mandatory factor. On the other hand, Birnbaum’s method evidently did not go so far as to ensure that the sum of the individual optional factor weights did not exceed that of any mandatory factor. Birnbaum conceded that he had not studied the effects of his method on actual premiums, and that he had never gone through the process of pumping or tempering factors in a class plan to ensure that the optional factor weight as he calculated it was less than the weight of any mandatory factor. The insurers prepared charts for a sample of 10 cars which showed the results of the pumping or tempering required when the optional factor weight was calculated according to Birnbaum’s method. These examples showed that the sum of the individual optional factor weights could continue to exceed the weights of individual mandatory factors even after the weights of the optional factors were lowered relative to those of the mandatory factors by the pumping and tempering. 13 The insurers argued that the pumping or tempering of factors required under Bimbaum’s method would be so extreme that the resulting premiums would no longer “have a substantial relationship to the risk of loss” within the meaning of section 1861.02, subdivision (a)(4), and would be “unfairly discriminatory” within the meaning of that statute and section 1861.05, subdivision (a). The insurers further argued that the Project’s constmction would produce rates that were “arbitrary” and “[unjfair” within the meaning of sections 1 and 2 of Proposition 103, and would be more likely than the insurers’ interpretation to raise the premiums paid by good drivers. (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, p. 99.) The testimony of State Farm actuary Becraft was representative of the evidence presented in support of the insurers’ arguments. Becraft testified *1207 that, because of the pumping or tempering of factors required under Birnbaum’s method, good drivers in most of the state would pay higher BIPD premiums under that method than they would under State Farm’s interpretation. Conversely, under Bimbaum’s method more drivers who did not meet good driver criteria would have premium decreases, and fewer of those drivers would have premium increases, than under State Farm’s interpretation. Becraft also testified that there would be substantial differences in the effects of Bimbaum’s interpretation on the premiums in different counties. She said that “[u]nder the Project’s proposal, if the relativities for the optional factors were compressed to achieve the required order of weights, our policyholders in most California counties would pay at least ten percent higher premiums than under our approved Class Plan filings [, w]hile no more than three counties, including Los Angeles and San Francisco, would pay at least ten percent less. The cost of insuring the policyholders in those counties paying less remains the same, but is shifted to the policyholders in other counties. That is, policyholders in most of California have to pay for the cost of insuring policyholders in a few urban centers. This effect is not dependent upon a policyholder’s ability to pay—wealthy policyholders in Beverly Hills benefit just as much as the Los Angeles urban poor, and the rural poor in counties like Humboldt County pay significantly higher rates to subsidize policyholders no worse off than they are.” Birnbaum admitted that pumping and tempering of relativities would change premiums without changing risks of loss. However, he thought “what you see here [with Proposition 103] is [sic] different ideas about what rating factors actually reflect the risk of loss, and you also see public policy decisions about which rating factors should be given the most weight in the determination of premium.” Whereas Becraft opined that pumping and tempering of relativities resulted in premiums unrelated to the risk of loss, Birnbaum thought “the fact that you are modifying the relativities through the pumping and tempering . . . does not alter the fact that [a] rating factor has a substantial relationship with the risk of loss. That’s immutable.” Birnbaum also disputed Becraft’s assumption that relativities would be “cost based” apart from any pumping and tempering the statutes or regulations required. He had seen many instances in which insurers had departed from “cost-based pricing” for “other than statutory or administrative purposes.” Becraft conceded that State Farm would sometimes deviate from cost-based pricing to avoid “sharp [rate] increases for a large number of policyholders.” In December 1997 the administrative law judge (ALJ) rendered a detailed decision in favor of the insurers’ method of calculating the optional factor *1208 weight, and the decision was adopted without change by Commissioner Quackenbush in February 1998. The ALJ interpreted the “clear and unambiguous language” of regulations section 2632.8, subdivision (c), to preclude the Project’s alternative method. The decision at the outset defined the “sole issue” to be whether the insurers’ method for “calculating] the weight of the optional factors [was] in compliance with . . . Section 2632.8,” and later reiterated that the “sole issue” was “whether the methodology which the [Department] instructed the Insurers to use complies with . . . Section 2632.8.” The decision was thus confined to interpretation of the regulation. While the decision also described the “sole issue” at one point as whether the insurers’ method “complied with the insurance code and the regulations” (italics added), and the conclusion stated that the insurers’ method “complied with the statutory and regulatory requirements of Proposition 103” (italics added), it does not appear that the ALJ ever reached the issue of whether regulations section 2632.8, as she interpreted it, was consistent with the Proposition. We note in this regard that during the hearings the ALJ found a question to a Department witness “asking what the statute requires” to be. “irrelevant.” The ALJ explained that “this proceeding, I think, is about whether or not the regulation was followed as it is written,” not whether the “regulation is, in fact, . . . proper.” (Punctuation from the transcript adjusted to reflect the apparent sense of the statement.) H. This Lawsuit On March 26, 1998, Consumers Union of U.S., Inc., and the Southern Christian Leadership Conference of Greater Los Angeles, Inc., together with the Spanish Speaking Citizens’ Foundation, Inc., the City of Los Angeles, the City of Oakland, and the City and County of San Francisco (collectively Consumers) filed a verified petition for writ of mandate and complaint for declaratory relief against Commissioner Quackenbush. The Project filed its verified petition for writ of mandate and complaint for declaratory relief against the Commissioner on the same date. Consumers and the Project (hereafter collectively Petitioners) both asserted causes of action for. traditional mandamus (Code Civ. Proc., § 1085), administrative mandamus (Code Civ. Proc., § 1094.5) and declaratory relief (Gov. Code, § 11350), alleging that regulations section 2632.8 violated Insurance Code section 1861.02, subdivision (a). State Farm and Farmers (hereafter collectively Insurers) were granted leave to intervene in the actions. The causes of action for administrative mandamus and declaratory relief were eventually dismissed without prejudice by stipulation, and the two cases were eventually consolidated for decision. Petitioners’ original prayers for relief asked that the Commissioner be compelled to perform alleged duties under section 1861.02, subdivision (a), *1209 to “set forth the respective weight of each optional factor, rather than one weight for all optional factors combined”; and to “ensure that no optional factor has greater importance in the determination of rates and premiums than any of the three mandatory factors.” Petitioners argued in support of their motions for peremptory writ that these duties were plainly spelled out in the statute, and cited the rating factors in the State Farm class plan described above as examples of optional factors which impermissibly outweighed mandatory factors. The Project emphasized that, contrary to the asserted purpose of Proposition 103, “ZIP Code [the territorial factors of claims frequency and severity] [r]emain[ed] the [pjrimary [f]actor in [d]etermining [State Farm’s] [insurance [r]ates.” Consistent with the prayer for relief in their petition, Consumers’ initial points and authorities filed below stated that, under the terms of the statute, each optional factor could have no greater weight than any of the mandatory factors. However, Consumers concluded their memorandum with the claim that the Commissioner could not lawfully “approve any automobile class plans, premiums or rates with the weight of any one or more individual optional rating factors being greater than the weight of any one or more of the three mandatory factors.” (Italics added.) The Project’s amended notice of motion for peremptory writ likewise asked that the Commissioner be directed “not to approve in the future any automobile insurance class plans, premiums or rates if the weight of any one or more individual optional rating factor used pursuant to section 1861.02(a)(4) is greater than the weight of any one or more of the three mandatory factors specified in section 1861.02(a)(l)-(3).” (Italics added.) With the words we have placed in italics, Petitioners expanded their initial claims for relief by asking for a writ precluding the use of optional factors that in the aggregate, and not just individually, outweighed any mandatory factor. 14 Insurers argued, as they had in the administrative proceeding, that to accept Petitioners’ position would create arbitrary and unfair insurance rates in violation of Proposition 103. Insurers also argued, among other things, that the process of sequential analysis required under regulations section 2632.7 by itself ensured that mandatory and optional factors were applied in the “order of importance” dictated in Insurance Code section 1861.02, *1210 subdivision (a). Commissioner Quackenbush argued among other things that his construction of a statute he was charged to administer was entitled to great weight. The Insurers supported their arguments about arbitrary and unfair rates with evidence presented in the administrative proceeding on the effects of “the method promoted by the Project and endorsed by Consumers” in that proceeding. State Farm also filed a new declaration from actuary Becraft on the effects of “Petitioners’ proposal in these lawsuits.” Although, as we have observed, Petitioners’ position in the lawsuit evolved into the contention that the total weight of all optional factors could not exceed the weight of any mandatory factor, Becraft understood Petitioners to suggest only that the weight of each optional factor could not exceed that of any mandatory factor. Becraft applied that assumption to a random sample of 200,000 State Farm policies and compared the premiums for BIPD coverage under the company’s approved class plan with those produced by Petitioners’ proposal. The results were similar to those Becraft reported in the administrative proceeding. She found: that if Petitioners’ proposal were implemented, substantial pumping and tempering would be required to make the individual factor weights align in the requisite order; that the premiums of 65 percent of good drivers would increase, and that 45 percent of good drivers would have increases of 10 percent or more; that the premiums of 63 percent of all drivers would increase, and that 43 percent of all drivers would have increases of 10 percent or more; that the premium changes would range from an average decrease of 18 percent in Los Angeles County to an average increase of 85 percent in Inyo County; that most good drivers in 42 counties would pay at least 10 percent higher premiums, while most drivers in Los Angeles and San Francisco would pay at least 10 percent less; that average premiums would increase in 51 of the 58 counties in the state, and thus that “the policyholders in these 51 counties would be subsidizing the policyholders in the remaining 7.” The court’s order granting the motions for peremptory writ did not refer to the evidence but included the following findings: “1) Contrary to the requirement of . . . Section 1861.02(a)(4), [the Commissioner’s] regulations (. . . Section 2632.1 et seq.) do not set forth the respective weight to be given each optional rating factor in determining automobile rates and premiums. Instead, . . . Section 2632.8 requires the averaging of all optional rating factors to arrive at a single weight for the optional factors (see . . . Section 2632.8(a) and (c)(4)), and the task of assigning ‘weight’ is delegated to insurers ([Ibid.]). *1211 “2) . . . Section 1861.02(a) requires that each optional factor have a lesser effect on premium than any of the mandatory factors. Contrary to the requirements of. . . Section 1861.02(a), . . . Section 2632.8 permits insurers to use individual optional factors that have a greater impact in the determination of rates and premiums than one or more of the three mandatory factors (i.e. the insured’s driving safety record, the number of miles he or she drives annually, and the number of years of driving experience the insured has had). “3) Since . . . Section 2632.8 of [the Commissioner’s] regulations is inconsistent with and violates the requirements of . . . Section 1861.02(a), Section 2632.8 is void to the extent that it requires and permits insurers to calculate only one average weight for all of the optional rating factors combined. Government Code Section 11342.2.” The court indicated that the writ would command the Commissioner “[n]ot to utilize, enforce, or attempt to enforce that portion of . . . Section 2632.8(a) providing that the weight for the optional factors shall be ‘taken together as a single factor weight.’ ” However, the regulation would “remain operative pending further proceedings to be taken promptly by [the Commissioner] in accordance with this ruling.” The remaining relief requested by Petitioners was denied. Petitioners were directed to prepare a proposed statement of decision. In it they provided that after the “further proceedings” contemplated in the court’s order the Commissioner “(a) [shall] not approve any automobile class plans, premiums or rates with the weight of any one or more individual optional rating factors being greater than the weight of any one or more of the three mandatory factors; and (b) shall not allow any current automobile class plans, premiums or rates to remain in effect with the weight of any one or more individual optional rating factors being greater than the weight of any one or more of the three mandatory factors.” (Italics in text added.) Petitioners acknowledged that this quoted language, which also appeared in their proposed preemptory writ of mandate, added to the terms of the court’s order, but submitted that the additional language was “entirely consistent with and effectuate[d]” the order. The Commissioner objected to the italicized words “or more” because they “could be read to mean that not only must each of the individual optional rating factors be less than any of the mandatory rating factors, but that the weight of ‘more’ (i.e. the sum of all) of the optional factors in an *1212 insurer’s class plan cannot exceed any of the three mandatory rating factors.” The Commissioner argued that this result was contrary to the court’s findings that the “statute requires that ‘each optional factor have a lesser effect on premium than any of the mandatory factors’ and that the regulation ‘permits insurers to use individual optional factors that have a greater impact in the determination of rates and premiums than one or more of the three mandatory factors.’ ” In the Commissioner’s view, this language in the order “made it clear that while an individual optional rating factor could not exceed any of the mandatory factors, the sum of the optional rating factors could exceed any of the mandatory factors.” The court declined to add the “or more” words to the statement of decision, judgment or writ. The statement of decision retained the findings in the original order that “permitting] insurers to use individual optional factors that have a greater impact in the determination of rates and premiums than one or more of the three mandatory factors” violated the statutory “require[ment] that each optional factor have a lesser effect on premium than any of the mandatory factors.” The statement of decision added the observation, not included in the original order, that, under the regulation, “[w]hile the average [of the optional factors] may fall below the weight of the third mandatory factor, individual optional factors can have a weight greater than one or more of the mandatory factors.” The statement of decision also retained the original order’s findings that the regulations improperly failed to set forth individual optional factor weights, and improperly delegated the assignment of weights to Insurers. The judgment and writ filed in October 1998 directed the Commissioner not to utilize or attempt to enforce regulations section 2632.8, subdivision (a)’s provision that the weight of the optional factors be “taken together as a single factor weight.” This provision was to remain operative, however, “pending further proceedings to be-taken promptly by [the Commissioner].” Based on the Commissioner’s representation that he would appeal, he was directed to file a return to the writ within 30 days of service of notice of issuance of the remittitur or a court order effectively removing any stay of enforcement. In November 1998, the court ordered that Consumers and the Project recover attorney fees of $152,744.85 and $144,950.71, respectively, from the Commissioner. These attorney fees were awarded under section 1861.10, subdivision (b), which provides for recovery of “reasonable advocacy and witness fees and expenses” by persons representing the interest of consumers who have “made a substantial contribution to the adoption of any order, regulation or decision by the commissioner or a court” in an action to enforce Proposition 103. *1213 Commissioner Quackenbush and Insurers each appealed from the judgment (A084024, A085376). The Counties of Fresno, Madera and Mariposa, along with the City of Hughson, have filed an amicus curiae brief in support of the Commissioner and Insurers, as has the American Insurance Association, a national trade organization. The Los Angeles County Board of Supervisors has appeared as an amicus curiae in support of Consumers and the Project. Commissioner Quackenbush also appealed from the attorney fee order (A085713). Commissioner Quackenbush resigned while this appeal was pending, and after a period in which J. Clark Kelso served as interim Commissioner, Harry W. Low was sworn in as Commissioner within a week after oral argument herein. Through counsel at the oral argument we invited Commissioner Low to advise us if he had any comments on issues in the case before we rendered a decision. Having heard nothing further, we will assume for purposes of this opinion that the arguments that have been advanced in this appeal continue to reflect the views of the Department. 15 II. Discussion A. Statutory Construction and Scope of Review The issue is whether the factor weight regulation, regulations section 2632.8, as interpreted in the administrative proceeding, is consistent with the factor ordering provisions of Insurance Code section 1861.02, subdivision (a). To resolve that issue, we are called upon to construe the statute and the regulation, and to review the constructions made by the Department and the trial court. We therefore begin with the pertinent principles of statutory construction and judicial review. “ ‘The fundamental purpose of statutory construction is to ascertain the intent of the lawmakers so as to effectuate the purpose of the law. [Citations.] In order to determine this intent, we begin by examining the language of the statute. [Citations.]’ ” (Calatayud v. State of California (1998) 18 Cal.4th 1057, 1064 [77 Cal.Rptr.2d 202, 959 P.2d 360].) Although the intent ultimately prevails over the letter of the law (id. at p. 1065), the statutory language may be sufficiently clear and unambiguous to obviate the need for further inquiry (Board of Supervisors v. Lonergan (1980) 27 Cal.3d 855, 866 [167 Cal.Rptr. 820, 616 P.2d 802]). *1214 The rules governing interpretation of statutes generally apply also to initiatives and regulations. (Whitman v. Superior Court (1991) 54 Cal.3d 1063, 1072 [2 Cal.Rptr.2d 160, 820 P.2d 262] [initiatives]; Industrial Indemnity Co. v. City and County of San Francisco (1990) 218 Cal.App.3d 999, 1008 [267 Cal.Rptr. 445] [regulations].) “The legislative history of a statute, as well as the wider historical circumstances of its enactment, may be considered in ascertaining legislative intent.” (Watts v. Crawford (1995) 10 Cal.4th 743, 753 [42 Cal.Rptr.2d 81, 896 P.2d 807].) In the case of initiatives, it may be appropriate to consult statements to the voters (People v. Knowles (1950) 35 Cal.2d 175, 182 [217 P.2d 1]), including arguments in the ballot pamphlet (People v. Anderson (1987) 43 Cal.3d 1104, 1143 [240 Cal.Rptr. 585, 742 P.2d 1306]; People ex rel. Lungren v. Peron (1997) 59 Cal.App.4th 1383, 1393 [70 Cal.Rptr.2d 20]). “ ‘[W]e do not construe statutes in isolation, but rather read every statute “with reference to the entire scheme of law of which it is part so that the whole may be harmonized and retain effectiveness.” ’ ” (Calatayud v. State of California, supra, 18 Cal.4th at p. 1065.) We “ ‘must consider the consequences that might flow from a particular construction and should construe the statute so as to promote rather than defeat the statute’s purpose and policy.’ ” (Escobedo v. Estate of Snider (1997) 14 Cal.4th 1214, 1223 [60 Cal.Rptr.2d 722, 930 P.2d 979].) Interpretation of a statute or regulation is, of course, an issue of law for the court (Evans v. Unemployment Ins. Appeals Bd. (1985) 39 Cal.3d 398, 407 [216 Cal.Rptr. 782, 703 P.2d 122]; Home Depot, U.S.A., Inc. v. Contractors’ State License Bd. (1996) 41 Cal.App.4th 1592, 1599 [49 Cal.Rptr.2d 302]), as is the question whether a regulation is consistent with the authorizing statute (California Assn, of Psychology Providers v. Rank (1990) 51 Cal.3d 1, 11 [270 Cal.Rptr. 796, 793 P.2d 2]; see Gov. Code, § 11342.2). Thus, we must review the interpretations of the Department and the trial court de novo, and come to our own independent conclusions on these issues. (20th Century Ins. Co. v. Garamendi, supra, 8 Cal.4th at pp. 271-272; Morris v. Williams (1967) 67 Cal.2d 733, 748 [63 Cal.Rptr. 689, 433 P.2d 697].) While the “ ‘final responsibility for the interpretation of the law rests with the courts’ ” (Morris v. Williams, supra, 67 Cal.2d at p. 748), “the construction of a statute by officials charged with its administration ... is entitled to great weight” (ibid.). A helpful discussion of the weight accorded agency interpretations of law appears in Asimow, The Scope of Judicial Review of Decisions of California Administrative Agencies (1995) 42 UCLA *1215 L.Rev. 1157, 1192-1209 (hereafter Asimow). (See Yamaha Corp. of America v. State Bd. of Equalization (1998) 19 Cal.4th 1, 12 [78 Cal.Rptr.2d 1, 960 P.2d 1031] [citing article]; id. at pp. 17-18 (conc. opn. of Mosk, J.) [same].) According to the article’s analysis of California cases on the subject, unless “a statute demonstrably delegates to the agency the power to interpret particular statutory language” (Asimow, at p. 1198), the deference given to the agency’s interpretation depends on the extent to which “the agency has a comparative interpretive advantage over courts” with respect to the subject matter, and the extent to which it appears that the agency’s interpretation is “probably correct” (id. at p. 1195). If deference is appropriate in light of these considerations “an agency’s view is given more weight by the court than the interpretation suggested by other litigants,” and if “the interpretive question is close, the scales are likely to tip in the agency’s direction.” (Id. at pp. 1194-1195.) With regard to the agency’s potential interpretive advantage, the article notes that: “In many situations, the legal text to be interpreted is technical, obscure, complex, or open-ended, or it may be entwined with issues of fact, policy and discretion. Frequently, agencies have developed the sort of expertise and technical knowledge that gives them a comparative advantage in interpreting such texts over a generalist court that lacks such qualifications. fl[] Similarly, agencies are often immersed in administering a particular statute. Such specialization gives those agencies an intimate knowledge of the problems dealt with in the statute and the various administrative consequences arising from particular interpretations. In contrast, a generalist court that visits a particular regulatory statute only infrequently lacks the advantage arising out of specialization. In short, if by reason of expertise, specialization or both, an agency demonstrably has qualifications to interpret a particular text that are superior to the court’s, deference is appropriate.” (Asimow, supra, 42 UCLA L.Rev. at pp. 1195-1196, fn. omitted.) As for whether the agency’s interpretation is “probably correct,” the article opines that: “The degree to which the agency’s interpretation appears to have been carefully considered by responsible agency officials is an important factor. . . . [IQ Deference is called for if the agency has consistently maintained the interpretation in question, especially if the interpretation is long-standing. . . . [ft] ... Of course, an agency can abandon an old interpretation and substitute a new one, provided that the agency explains and rationally justifies the change. However, the new interpretation would lack the deference called for by a consistently maintained interpretation.” (Asimow, supra, 42 UCLA L.Rev. at pp. 1196-1198.) The article contrasts what it describes as the rule of “weak deference” applied in this state with a “strong deference” rule, which would require that *1216 any reasonable agency interpretation of an ambiguous statute be upheld. (Asimow, supra, 42 UCLA L.Rev. at pp. 1194-1195 [citing Chevron U. S. A. v. Natural Res. Def. Council (1984) 467 U.S. 837 [104 S.Ct. 2778, 81 L.Ed.2d 694] for the “strong deference” principle].) The article acknowledges that there are California cases with language suggesting a “strong deference” approach, but persuasively contends that the rule is otherwise. (Asimow, at pp. 1199-1201.) The opinions in Yamaha Corp. of America v. State Bd. of Equalization, supra, 19 Cal.4th 1, confirm that it would not be enough for us simply to determine that the statute is ambiguous and that the Department’s interpretation is reasonable; we would still be required to independently assess whether the Department’s construction is lawful and whether it is deserving of deference under the circumstances. (See id. at p. 11, fn. 4 [“even quasi-legislative rules are reviewed independently for consistency with controlling law”]; id. at pp. 16, 17 (conc. opn. of Mosk, J.), quoting Morris v. Williams, supra, 61 Cal.2d at p. 748 and California Assn, of Psychology Providers v. Rank, supra, 51 Cal.3d at p. 11 [same].) B. Arguments on Appeal In this appeal, Petitioners, Insurers, and the Commissioner essentially reiterate and amplify their arguments to the trial court. Petitioners’ main argument is that section 1861.02, subdivision (a) is clear that optional factors, and in particular territory, cannot have greater weight in determining premiums than any of the mandatory factors the statute identifies, and that the regulation improperly allows results that violate this clear mandate. The Insurers’ arguments are more complex. They are based primarily on: (1) the asserted need to harmonize the factor ordering provisions of section 1861.02, subdivision (a) with that statute’s requirement that optional factors “have a substantial relationship to the risk of loss,” and with the requirements of other provisions of Proposition 103; and (2) evidence that interpretations of the statute that differ from the regulation produce effects which conflict with those other requirements. The Commissioner shares the Insurers’ concerns about the competing requirements of the Proposition, and characterizes the regulation as an “admittedly result-oriented” attempt to balance them. The American Insurance Association shares the Insurers’ concerns generally, and emphasizes in particular the evidence assertedly showing that territory is more closely related to the risk of loss than any of the statute’s mandatory factors. The Los Angeles County Board of Supervisors contends among other things that “zip code rating” constitutes unlawful discrimination prohibited by the Unruh Civil Rights Act (Civ. Code, § 51). The counties of Fresno, Madera et al., contend among other things that the trial court’s decision violates the equal protection rights of nonurban drivers. *1217 The foregoing outline greatly oversimplifies the contentions set forth in the some 500 pages of briefing in this case. Points raised in the briefs and the various requests for judicial notice include: citations to numerous authorities on statutory construction and scope of review to opposite effects; minute parsing of the statutory language from every conceivable angle; examples of anomalous effects of different factor weighting schemes on hypothetical individual drivers; discussion of foreign precedents on the meaning of “unfair discrimination”; and statistics on population and income. 16 There is even some debate in the briefs about the nature of “insurance” in the abstract. One point does stand out among the rest: the Project’s submission that if “one focuses simply and properly upon the plain language of the single, voter-enacted statute and single regulation here at-issue, this case becomes — respectfully — an easy one.” Unfortunately this suggestion is untenable. As explained below, we find no “plain meaning” in the statute that resolves any significant issue. We cannot ignore claims that particular interpretations of the statute and regulation in question would be inconsistent with other provisions of Proposition 103 and other rating factor regulations, which bear on the same subject matter. (Calatayud v. State of California, supra, 18 Cal.4th at p. 1065 [statutes are to be construed in light of the overall scheme they comprise; conflicts must be harmonized if possible].) We cannot overlook the consequences of different interpretations and must therefore grapple with the evidence presented on their effects. (Escobedo v. Estate of Snider, supra, 14 Cal.4th at p. 1223.) We cannot disregard the legislative history of Proposition 103, or the long and tortured history of the rating factor regulations. {Watts v. Crawford, supra, 10 Cal.4th at p. 753; Industrial Indemnity Co. v. City and County of San Francisco, supra, 218 Cal.App.3d at p. 1008.) Lost to some extent among the myriad points we are invited to consider is what the trial court actually decided. Part of the difficulty in this case is that a statement like the following appears midway through the Project’s brief: “the optional factors together, and each, cannot have a greater impact on premium than the mandatory factors, or any of them.” As we attempted to highlight in describing the trial court proceedings, and as we will reiterate below, this is not what the trial court decided. If that is indeed what the Project believes the statute requires, then we do not understand why it did *1218 not appeal from the judgment. In any event, we will begin our analysis by focusing on the trial court’s decision. The judgment and the writ of mandate leave the regulation nearly intact, striking down only the part of regulations section 2632.8, subdivision (a) which directs that the weights of the optional factors be “taken together as a single factor weight.” However, both sides evidently anticipate that another round of rulemaking will follow if the court’s decision is upheld. The Project contends that the Insurers’ forecasts of premium increases are premature because there will be a new “as yet unknown and unpromulgated regulatory plan.” State Farm interprets the court’s decision “to permit the Commissioner to attempt to devise an alternative to utilizing the set of regulations created by excising the identified language,” while conceding that “[a]ny alternative . . . would be restricted by the Superior Court’s interpretation of [Insurance Code section 1861.02, subdivision (a)].” We must thus address all of the court’s conclusions as to the meaning of the statute. As we read the statement of decision, the court found that the regulation conflicts with the statute in four respects: (1) the regulation improperly delegates the task of assigning weight to insurers; (2) the regulation erroneously fails to set forth the weight of each optional factor; (3) the regulation improperly allows insurers to calculate only one average weight for all of the optional rating factors combined; and (4) the regulation improperly permits insurers to use individual optional factors which have a greater effect on premiums than one or more of the mandatory factors. The issue of improper delegation merits separate, but only brief, discussion before the other conclusions are addressed. C. Unlawful Delegation of Responsibility The first two problems the court identified—failure to specify individual optional factor weights, and unlawful delegation of factor weight assignment to insurers—imply that the Department must promulgate numerical weights for the optional factors, and that there has been an abdication of regulatory responsibility on this score. Neither of these implications is correct. Petitioners do not go so far as to advocate the use of standardized factors, much less standardized weights. Even if companies were required to use the same rating factors with the same categories and relativities, the weights of the factors would vary according to the distribution of drivers among the categories. As the Project’s expert acknowledged in the administrative *1219 proceeding, the weight of a rating factor depends to some extent on the distribution of insureds among the relativities. 17 As Farmers points out in its briefs, “[t]o have the Commissioner designate one weight for all insurers is nonsensical and is akin to setting one price for all insurers.” Petitioners respond that this observation misrepresents the trial court’s decision. Consumers note that “Petitioners did not challenge Section 2632.8(c)’s use of a method or formula, instead of a fixed number, to determine a weight, and the regulation’s formula for determining weights, Regulation § 2632.8(c), was not an issue.” In the same vein, the Project states that “[a]ll [it] argued, and all the Superior Court held, was that the weight had to be determined by regulation, not left to insurers. Currently, weights for the mandatory factors are determined by, for example, § 2832.8(c) (the ‘Single Omit’ method), which establishes a regulatory formula for determining weights for optional factors, rather than a regulation for each ‘exact’ factor. No Petitioner has ever challenged the use of this formula and the narrow Superior Court’s ruling does not disturb the use of such a formula.” The Project then points out that use of regulatory formulas was approved in the context of Proposition 103’s rate rollback provisions in 20th Century Ins. Co. v. Garamendi, supra, 8 Cal.4th at page 287. Thus, it is conceded that weights may properly be established by a formula, which is just what the regulation provides for the single optional factor as well as the mandatory factors. If, as is also conceded, there is no problem of improper “delegation] to insurers” with the use of a formula for the calculation of mandatory factor weights, then there is no such problem with respect to the optional factor weight either. In both cases, the regulations tell insurers how the weights are to be calculated, and the insurers do the calculating. Therefore, the problem, if any, is with the formula, and not with the process of fixing weights. The regulation does not unlawfully delegate responsibility to insurers to determine factor weights. *1220 D. Ambiguities and Potential Conflicts in the Proposition The trial court’s other conclusions encompass two kinds of determinations. One determination involves the manner in which the factor weights are calculated. On this subject, the court concluded that individual weights must be calculated for each of the optional factors, as opposed to one weight for all of them. The other determinations involve how much weight the optional factors may have relative to the mandatory factors. On that subject, the court concluded that individual optional factors cannot have greater weight than any of the mandatory factors, and that the regulation violates the statute by requiring that only the average weight of the optional factors need be less than that of any mandatory factor. Before examining how the words of section 1861.02, subdivision (a) and other provisions of Proposition 103 bear on these issues, we note that the court’s ruling on the permissible weight of the optional factors is more important than its ruling on the manner in which optional factor weights are calculated. The Petitioners who appeared in the administrative proceeding would have been satisfied with only one weight for all of the optional factors, provided that single weight had been calculated in the manner they advocated. Thus, the crux of the case is not whether there is one optional factor weight or several, it is the relative weight of the optional factors versus the mandatory factors. (1) The Factor Weight Statute Section 1861.02, subdivision (a), once again, reads as follows: “Rates and premiums for an automobile insurance policy, as described in subdivision (a) of Section 660, shall be determined by application of the following factors in decreasing order of importance: “(1) The insured’s driving safety record. “(2) The number of miles he or she drives annually. “(3) The number of years of driving experience the insured has had. “(4) Such other factors as the commissioner may adopt by regulation that have a substantial relationship to the risk of loss. The regulations shall set forth the respective weight to be given each factor in determining automobile rates and premiums. Notwithstanding any other provision of law, the use of any criterion without such approval shall constitute unfair discrimination.” *1221 We agree with Petitioners that the words “importance” and “weight” as used in the statute are synonymous, and thus that the language “decreasing order of importance” means that optional factors are to have less weight than any mandatory factor. State Farm notes that when different terms are used in a statute they are generally presumed to have different meanings {American Airlines, Inc. v. County of San Mateo (1996) 12 Cal.4th 1110, 1137-1138 [51 Cal.Rptr.2d 251, 912 P.2d 1198]), but Insurers identify no persuasive distinction between “importance” and “weight” in this instance. In our view, “respective weight” simply connotes a quantitative measurement of the “importance” of factors, and the regulation is right to treat the terms as referring to the same concept. (Regs., § 2632.8, subd. (d) [“weights of the factors . . . must align in decreasing order of importance as follows: driving safety record must have the most weight followed by annual miles driven”].) However, while section 1861.02’s basic mandate, that optional factors are to weigh less than any mandatory factor, is reasonably clear from the statutory language, precisely what that means is not. The length of time it took to promulgate a permanent factor weighting regulation is strong circumstantial evidence of problems that cannot be resolved by any statutory “plain meaning.” There are indeed a number of difficulties which the statutory provisions do not solve. One basic problem is that the statute specifies no means of measuring “weight.” It is one thing to say that “weight” in a general sense means the “importance” of a factor, and quite another to derive an acceptable method for quantifying that importance. As has been noted, the “single omit” formula, which all sides now appear to accept as a suitable means of calculating individual factor weights, 18 only emerged after years in which other measures yielding different numbers were studied and debated. The record indicates that at least seven other alternatives were considered before *1222 the single omit formula was adopted. The Project observes that Justice Broussard broached the concept of “weight” in his concurrence in King v. Meese, supra, 43 Cal.3d at page 1242, when he ventured that problems caused by the financial responsibility laws could be ameliorated by rates which “weighted driving record more than residence.” However, the record shows that the term “weight” as applied to insurance rating factors had no established meaning, actuarial or otherwise, before the passage of Proposition 103. Project expert Bimbaum acknowledged in the administrative proceeding that no other state requires a weighting calculation. The voters plainly could not be said to have endorsed a “sum of the absolute differences divided by the number of calculations” system, as opposed to any other plausible mathematical system, of measuring a rating factor’s “weight.” The precise meaning of “weight” is one of the ambiguities in section 1861.02, subdivision (a). The statute is also ambiguous on the specific points the trial court decided. As for whether one combined weight, or separate individual weights, must be determined for the optional factors, the difficulty stems from the location of the factor weight sentence in the statute. The requirement that a “respective weight ... be given each factor” means that multiple factor weights must be determined. The requirement is set forth in section 1861.02, subdivision (a)(4). Since this subdivision appears to refer solely to optional factors, the placement suggests that weights need only be determined for those factors. However, Petitioners do not dispute that mandatory factor weights are also required. If the words “each factor” extend beyond subdivision (a)(4) to cover the mandatory factors in subdivision (a)(l)-(3), then the multiple factor weights contemplated by the statute could just as well be all of the mandatory factor weights, plus a single optional factor weight, as the regulations provide, or weights for each and every mandatory and optional factor, as the trial court decided. The words of the statute, in their acknowledged operation, do not clearly distinguish between these two approaches. Petitioners cannot in any event plausibly maintain that the statute plainly precludes a single optional factor weight calculation because some of them advocated such a calculation in the administrative proceeding. The statute is also ambiguous on the crucial question of the weights the optional factors can have relative to the mandatory factors. Various answers have been given. Under the Tempered Regulations, optional factors, in the aggregate, could have no more weight than any mandatory factor. Under the trial court’s decision, optional factors, individually, can have no more weight than any mandatory factor. Under the current regulations, optional factors, on average, can have no more weight than any mandatory factor. We interpret the statute to provide that optional factors must weigh less than *1223 mandatory factors, but nothing in the text dictates whether the weights of optional factors collectively, individually, or on average, must be less than that of any mandatory factor. If there were any doubt that section 1861.02, subdivision (a) is ambiguous on the permissible weight of the optional factors relative to the mandatory factors it would be dispelled by the conflicting positions Petitioners themselves have taken on this point. As we have observed, Petitioners’ initial position in the trial court was that the weight of individual optional factors could not exceed that of any mandatory factor. They eventually maintained below that the aggregate weight of the optional factors could not exceed any mandatory factor. The Petitioners who appeared in the administrative proceeding endorsed what we have referred to as the Bimbaum method of calculating a single weight for the optional factors which appeared to produce results somewhere in between the other two interpretations Petitioners advocated in the trial court. If the statute were crystal clear on the subject, we doubt that Petitioners would have vacillated like this on its meaning. (2) Other Provisions of Proposition 103 Interpretation of section 1861.02, subdivision (a)’s mandate that optional factors are to have less weight than any mandatory factor must take into account other relevant provisions of Proposition 103, as well as the legislative history of the initiative, which may shed light on what was intended. ('Calatayud v. State of California, supra, 18 Cal.4th at p. 1065; Watts v. Crawford, supra, 10 Cal.4th at p. 753.) Most of the debate on this facet of the case is a dispute about the extent to which auto insurance rates and premiums must be cost-based, and the principal focus is on the section 1861.05, subdivision (a) standard that insurance rates must not be “excessive, inadequate, unfairly discriminatory or otherwise in violation of this chapter.” (See Calfarm Ins. Co. v. Deukmejian (1989) 48 Cal.3d 805, 822 [258 Cal.Rptr. 161, 771 P.2d 1247] [this provision is the “general rate-setting standard” of Proposition 103].) The “excessive, inadequate, unfairly discriminatory” language previously appeared in a section of the McBride-Grunsky Insurance Regulatory Act of 1947 (former § 1850 et seq. added by Stats. 1947, ch. 805, § 1, p. 1896; id., § 7, p. 1908) which Proposition 103 replaced (Calfarm Ins. Co., at p. 822; see also Karlin v. Zalta (1984) 154 Cal.App.3d 953, 964, 968 [201 Cal.Rptr. 379]), and the parties’ positions echo observations on the former law in the pre-Proposition 103 case, King v. Meese, supra, 43 Cal.3d 1217, described above. *1224 The majority in King v. Meese noted that the words “ ‘unfairly discriminatory’ ” were not defined in the prior law, but that risk classifications were considered reasonable if they were “actuarially sound.” (King v. Meese, supra, 43 Cal.3d at pp. 1221-1222.) In a similar vein, Insurers contend that rates which are not cost-based from an actuarial standpoint are “unfairly discriminatory” within the meaning of Proposition 103. The concurrence in King v. Meese questioned whether actuarially sound rates were necessarily reasonable and nondiscriminatory, suggesting that affordability could be taken into account, and that the use of classifications which “charg[ed] good drivers in some areas much more than bad drivers in other[] parts of the state” could be unfairly discriminatory. (Id. at pp. 1241-1242 (cone. opn. of Broussard, J.).) In the opinion of the two concurring justices, “[t]he fact that current territorial rates may be actuarially justified does not mean that a rate which placed less weight on residence would be unsound, or would be unfair to residents of low-risk territories.” (Id. at p. 1242, fn. 14.) In a similar vein, Petitioners submit that section 1861.02, subdivision (a) was intended to subordinate optional factors to mandatory factors without regard to actuarial considerations like the cost of providing coverage. In support of this view, Petitioners also cite 20th Century Ins. Co. v. Garamendi, supra, 8 Cal.4th at page 289, which held in the context of a challenge to rate rollback regulations under Proposition 103 that the “ ‘excessive’ ” and “ ‘inadequate’ ” standards in section 1861.05, subdivision (a), are “ ‘unique’ and without ‘precedent’ among ‘similar statutes,’ ” despite the appearance of those same words in the former law. Petitioners note that 20th Century rejected an argument that the “ ‘excessive’/‘inadequate’ standard as defined in the initiative should be interpreted in accordance with the insurance industry’s or the actuarial profession’s understanding of its operative terms.” (20th Century Ins. Co., at p. 289.) Based on King v. Meese and 20th Century and the fact that Proposition 103 was enacted by a “lay electorate,” Petitioners reason that the concept of “ ‘fairness’ ” in the “unfairly discriminatory” standard of section 1861.05, subdivision (a), should be read to “invoke notions of equity, not actuarial ‘science.’ ” Petitioners also make the unchallenged observation that there is legislative power to entirely prohibit the use of particular rating factors, even though Proposition 103 does not go that far. Part of the problem here is that certain provisions of Proposition 103 which appear potentially applicable to the issues in this case do not militate in any clear way in favor of the parties’ opposing positions. Both sides invoke the “ ‘excessive’/‘inadequate’ ” standards of section 1865.01, subdivision (a) to some extent, but those standards appear to be aimed more at the base rates insurers can charge than at the distribution of premiums among *1225 policyholders. (See 20th Century Ins. Co. v. Garamendi, supra, 8 Cal.4th at p. 289 [referring to those standards in connection with claims about reasonable rates of return]; Calfarm Ins. Co. v. Deukmejian, supra, 48 Cal.3d at pp. 822-823 [same].) As we stated at the outset, base rates for auto insurance are not at issue in this case. Accordingly, we do not find Proposition 103’s “ ‘excessive’/'inadequate’ ” standards or the discussion of those standards in the 20th Century case to be helpful here. The same is true of the Proposition’s stated goal of making insurance “affordable for all Californians.” (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, § 2, p. 99.) To the extent that our ruling raises anyone’s premiums it will lower those of others. The overall cost, as we have stated, will remain the same. There is no decision we can render which will make auto insurance in “all” of California any more or less “affordable.” Likewise unhelpful are the Proposition’s references to “fair[ness],” and Petitioners’ appeal to “equity” in the abstract. Although another of the stated purposes of Proposition 103 is to “ensure that insurance is fair,” whatever benefit is afforded by any ordering of rating factor weights will be offset by a corresponding detriment. (Ibid.) “[E]quity” in this context depends on your point of view, or more precisely, your own pocketbook. Petitioners evidently take “ ‘fairness’ ” and “equity” to mean something other than actuarial standards, but they do not otherwise attempt to define those terms apart from citations to the concurring opinion in King v. Meese. The King v. Meese concurrence invited factor weight legislation which would suppress the influence of territory so as to make auto insurance more available and affordable for the urban poor. (King v. Meese, supra, 43 Cal.3d at. p. 1242.) There is indeed evidence in this case, as there was in King v. Meese, that this group may have special problems (e.g., a vicious circle where uninsured motorist coverage is more expensive in urban areas because there are more uninsured drivers there; there are more uninsured drivers in urban areas because the insurance there is more expensive). But while there may be policy arguments for such legislation, they were not made to the voters when they enacted Proposition 103. 19 The ballot pamphlet argument did not state that factor weight ordering was intended to assist the urban *1226 poor, but rather that it would “mean[] good drivers throughout the state will pay less than they do now, while bad drivers will pay more.” (Ballot Pamp., Gen. Elec., supra, rebuttal to argument against Prop. 103, p. 101.) This specific promise about the effect of factor weight ordering is pertinent to construction of section 1861.02, subdivision (a). (People v. Anderson, supra, 43 Cal.3d at p. 1143; People ex rel. Lungren v. Peron, supra, 59 Cal.App.4th at p. 1393.) The same cannot be said for the Proposition’s standards of “excessive[ness],” “inadequacy,” “affordab[ility],” and “fair[ness],” which do not suggest any particular outcome. Another general standard we decline to apply is that of “availability].” Although Proposition 103 was meant expressly to ensure the availability of insurance to all Californians, we find no sufficient basis in the record to assess Insurers’ isolated warnings that auto insurance may be unavailable if the current factor weighting regulation is not upheld. The only general standard which bears clearly on the issue of factor weight ordering is the Proposition’s stated aim of “protecting] consumers from arbitrary insurance rates and practices.” (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, §§ 1, 2, p. 99.) We can conceive of no interpretation of “arbitrary insurance rates,” and Petitioners offer none, as anything other than rates which do not reflect the cost of providing insurance. The concept of relation to cost is also embodied in section 1861.05, subdivision (a)’s prohibition against “unfairly discriminatory” rates, at least as that prohibition applies to the ordering of auto insurance rating factor weights. The parties have searched far and wide for authority on the proper interpretation of “unfair discrimination” in the context of this case, including published actuarial standards, the legislative history of the McBride-Grunsky Insurance Regulatory Act of 1947, and cases from other jurisdictions. However, we agree with State Farm’s argument that the answer lies within *1227 section 1861.02 itself, specifically the provisions in subdivision (a)(4), that optional factors cannot be approved unless they “have a substantial relationship to the risk of loss,” and that “the use of any criterion without such approval shall constitute unfair discrimination.” These provisions link the concept of “unfair discrimination” with that of “substantial relationship to the risk of loss.” Petitioners contend that the substantial relationship to risk of loss requirement attaches only to the Commissioner’s approval of optional factors for use, and does not extend to the operation of those factors. However, we agree with Insurers that it would make little sense to direct that optional factors may be used only if they have a substantial relationship to the risk of loss, and then ignore whether that relationship continues to exist when the factors are actually applied. Petitioners contend that to the extent section 1861.05, subdivision (a)’s prohibition against “unfairly discriminatory” rates conflicts with section 1861.02, subdivision (a)’s factor weight ordering mandate, the former must yield under the maxim that “ ‘[a] specific provision relating to a particular subject will govern in respect to that subject, as against a general provision . . . .’” (San Francisco Taxpayers Assn. v. Board of Supervisors (1992) 2 Cal.4th 571, 577 [7 Cal.Rptr.2d 245, 828 P.2d 147].) Petitioners point out that section 1861.05, subdivision (a) prohibits rates “in violation of this chapter,” as well as those that are unfairly discriminatory. However, a prohibition against unfair discrimination is rooted right in section 1861.02, subdivision (a). Moreover, in choosing among the possible interpretations of the factor weight ordering requirement of this statute we must try to harmonize all of the relevant provisions of Proposition 103. {Calatayud v. State of California, supra, 18 Cal.4th at p. 1065.) Interpretations that preserve a substantial relationship between premiums and the risk of loss (§§ 1861.02, subd. (a)(4), 1861.05, subd. (a)) are therefore to be favored over those that would produce arbitrary rates (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, § 2, p. 99). In view of the proponents’ ballot pamphlet representations about how the weight ordering mandate would operate, interpretations which reduce the premiums of good drivers, and raise those of bad drivers, throughout the state are to be favored over those that do not. {People v. Anderson, supra, 43 Cal.3d at p. 1143; People ex rel. Lungren v. Peron, supra, 59 Cal.App.4th at p. 1393.) However, we must not lose sight of the proponents’ related statement in the ballot pamphlet that the Proposition would “force[] insurance companies to base your rates on your driving record first, rather than on where you live,” and be mindful of the Proposition’s command that “automobile insurance rates shall be determined primarily by a driver’s safety record and mileage driven.” (Ballot Pamp., Gen. Elec., supra, rebuttal to argument against Prop. 103, p. 101; id., text of Prop. 103, § 1, p. 99.) *1228 E. Permissible Weight of the Optional Factors We turn finally to the central issue presented: the allowable weight of the optional factors relative to the mandatory factors. We will focus on the territorial factors, because those optional factors are Petitioners’ main concern. The centerpiece of Petitioners’ case is that the current regulations unlawfully permit territory to have a greater influence on premiums than driving record and the other mandatory factors. It must be pointed out, however, that the trial court decision Petitioners would have us affirm also permits that result. Because there are two territorial factors (claims frequency and claims severity) rather than one, the combined effect of those factors must be considered. Under the trial court’s ruling, the only limit on optional factor weights is that they must individually be less than any mandatory factor. As we have emphasized, the court declined to rule that the combined weights of optional factors could not exceed that of any mandatory factor. The combined weights of the claims frequency and severity factors can thus exceed that of any mandatory factor. For example, the weights of the mandatory factors could be 10, 9, and 8, and the weights of the two territorial factors could each be 7. Although the combined weight of the territorial factors would be 14, more than that of any mandatory factor, this result would be permissible because the individual weights of the territorial factors are each less than any mandatory factor. In a system like the one envisioned by the trial court where the only limitation is on individual optional factor weights, the number of optional factors, and the extent to which they may or must be combined—subjects nowhere addressed in the voluminous briefing—become important. Under this scheme, the total allowable weight of territory will vary according to the number of territorial factors. If there were 10 territorial factors, as there used to be in the prior regulations (see fn. 5, ante), instead of two, territory could have five times greater weight. To continue the foregoing example, there could be 10 individual territorial factor weights of 7, adding up to 70, as opposed to two adding up to 14. And yet we hear from the Los Angeles Board of Supervisors that the evil of “zip code rating” is its “composite or compound nature,” and that “each of the myriad of individual factors that are represented by so-called zip code rating should be individually and independently rated and weighed.” Under the trial court’s system, this argument is directly contrary to Los Angeles’s perceived interests. With respect to optional factor combinations, we have previously noted that State Farm’s class plan combines the claims frequency and severity factors into one single factor, and that this combination is not required by the regulations. (See fn. 3, ante.) In fact, Consumers Union objected to this *1229 combination in their briefing in the administrative proceeding. Under the trial court’s system, these positions will presumably now be reversed. State Farm will want to use two separate territorial factors rather than one to maximize the premium effect of territorial considerations. Petitioners, presumably, would now prefer that the two factors be combined to produce the opposite result. With these preliminary thoughts in mind, we will address three interpretations of the factor weight ordering requirement: that the (1) aggregate, (2) individual, or (3) average weight of the optional factors cannot exceed that of any mandatory factor. (1) Aggregate Weight The interpretation that would most suppress the weight of the optional factors is the one where their combined weight could not exceed that of any mandatory factor. The Petitioners who appeared in the administrative proceeding advocated one way of ensuring this result where only one weight is calculated for all of the optional factors: what we have described as the “Birnbaum method,” which determined the total premium effect of the optional factors car by car, added up the results, and divided by the number of cars. In court, Petitioners have taken the position that individual weights must be determined for each of the optional factors. Thus, when Petitioners have from time to time argued that the weights of “one or more” of the optional factors cannot exceed that of any “one” mandatory factor, we take them to mean that the sum of the individual optional factor weights must be less than that of any mandatory factor. We further assume that in such an arrangement the individual optional factor weights would be calculated under the regulation’s “single omit” formula. (See fn. 18, ante.) Therefore, to return to the example at the outset of this part of the opinion, where the mandatory factor weights were 10, 9, and 8, and the weights of the two territorial factors were each 7, the aggregate weight of those optional factors would be 14. Because this aggregate weight exceeds those of individual mandatory factors, pumping or tempering would be required to decrease the optional factor weights relative to the mandatory factor weights to achieve the requisite alignment. Perhaps the strongest argument against this approach is that Petitioners fail to advocate it on appeal. Although the trial court refused to rule as Petitioners urged below that no one “or more” of the optional factor weights could exceed that of any mandatory factor, Petitioners are evidently content to accept the trial court’s decision. They did not appeal from the judgment, and apart from a single passing line in the Project’s brief, there is no trace on appeal of Petitioners’ no one “or more” stance in the trial court. *1230 No litigant, of course, can effectively “abandon” the correct reading of a statute; we must independently determine the best interpretation of the factor weight ordering mandate regardless of what the parties argue. The “in the aggregate” alternative merits consideration because it is the only one that fulfills the Proposition’s command that rates “be determined primarily by a driver’s safety record and mileage driven,” and the ballot pamphlet’s apparent promise that auto insurance rates would be based more on “your driving record . . . than on where you live.” (Ballot Pamp., Gen. Elec., supra, text of Prop. 103, § 1, p. 99; id., rebuttal to argument against Prop. 103, p. 101.) As we have explained, if optional factor weights in the aggregate can outweigh any mandatory factor, the two territorial factor weights added together could exceed the weight of each mandatory factor, including driving record. Where you live, in other words, could have more influence on your premiums than how well you drive. Apart from any doubts cast on an “in the aggregate” approach by Petitioners’ failure to insist that it be followed, the most serious problems with that approach are created by unrefuted evidence that the approach would raise the premiums of most good drivers, undermine the relationship between rating factors and risks of loss, and produce premiums which are arbitrary. That evidence was presented in the administrative proceeding on the results of the Bimbaum method. As we have explained, the Bimbaum method of calculating the aggregate weight of optional factors did not go as far in suppressing the weight of those factors as a system like the one we are considering where the aggregate weight of the optional factors is the sum of their individual weights. The “in the aggregate” alternative we are examining, therefore, would be likely to produce even more extreme dislocations than the Bimbaum method. It follows that all of the evidence on the results of the Bimbaum method is relevant in assessing the “in the aggregate” alternative. That evidence would if anything understate the effects of that alternative. As previously noted, State Farm actuary Becraft testified in the administrative proceeding that more good drivers, and fewer “poor” drivers, would pay more for BIPD coverage if the Birnbaum method were followed instead of the current regulations. These results are plainly contrary to the intent of Proposition 103. Becraft also predicted that the premiums of State Farm policyholders in most counties would increase by at least 10 percent if Birnbaum’s method were followed. Although we have observed that Proposition 103’s “affordab[ility]” standard is not a clear criterion for evaluating alternative constructions of the factor weight ordering statute, Becraft’s prediction of large premium increases for some insureds was consistent with *1231 the findings years earlier that led to the imposition of limits on premium increases when an aggregate cap on optional factor weights was first tried in the Tempered Regulations. (Note, supra, at p. 728.12.) Those earlier findings at the least give added credence to Becraft’s conclusions. Premium changes resulting from the Birnbaum method would be caused by the need to pump or temper factor relativities to bring the factor weights into alignment if that method were followed. While it may be debatable whether the original numbers generated by sequential analysis are truly “cost based,” it was conceded in the administrative proceeding that pumping and tempering produces premium changes for individual cars without any changes in the underlying risk of loss. Pumping and tempering therefore mn counter to section 1861.02, subdivision (a)’s requirement that rating factors be substantially related to the risk of loss. Becraft testified without contradiction in the administrative proceeding that the Birnbaum method created a need for more “extreme” pumping or tempering than the method provided under the current regulations. It thus appears that the current regulations are preferable to an “in the aggregate” system insofar as they maintain a closer relationship between the operation of the rating factors and the risk of loss. (§ 1861.02, subd. (a)(4).) The Insurers also presented uncontradicted calculations in the administrative proceeding showing that when the initial weights produced by Birnbaum’s method were adjusted into alignment, the premiums for individual insureds would rise or fall in different ways depending solely on whether the adjustments were made through pumping or tempering. In the 10-car example we have previously cited, premiums for individual cars were compared before and after pumping and tempering. The results showed premiums for individual cars which rose when optional factors were tempered but fell when mandatory factors were pumped, and vice versa. 20 Neither pumping nor tempering is the preferred method of adjusting factor weights; they are just different means of accomplishing the same thing. The different results flowing solely from which method is used appear to be entirely arbitrary. *1232 (2) Individual Weight The system envisioned by the trial court, where the individual weights of optional factors must be less than that of any mandatory factor would suppress individual optional factor weights less than the “in the aggregate” approach just discussed, but more than the approach to be discussed below where individual optional factor weights need only be less on average than those of any mandatory factor. In this respect the trial court’s interpretation represents a middle ground between other plausible constructions of the statute. However, the trial court’s ruling does not ensure that territory has less effect on premiums than any of the mandatory factors, and it places no effective limit on the permissible weight of the optional versus the mandatory factors. As has been pointed out, under the trial court’s system the combined weight of the two territorial factors can exceed the weight of any mandatory factor. Moreover, nothing in the trial court’s decision would prevent regulations which divided the two territorial factors currently used into a greater number of territorial factors with a potentially greater premium effect. As has been noted, if 10 territorial factors were once again used instead of the current two, territory’s influence on premium would further dwarf that of any mandatory factor. Petitioners contend that the 1994 Impact Analysis outlined above shows that the trial court’s approach would yield results consistent with the aims of Proposition 103, including the ballot pamphlet’s stated goal of reducing good driver premiums through the factor ordering mandate. As we have indicated, the Impact Analysis attempted to measure the difference between the premiums companies were charging with those that would be produced under different models. One of the models studied was a “[sjingle [o]mit method without standardized factors” model which apparently assumed in line with the trial court’s system that individual factor weights would be calculated with the single omit method, that there would be two territorial rating factors for claims frequency and claims severity, and that the individual, but not the aggregate, weight of each of those territorial factors would be less than that of any mandatory factor. Ninety-seven percent of the change or “dislocation” under this model was “positive” (i.e., good, low mileage, and experienced drivers paid less, while “bad,” high mileage and inexperienced drivers paid more) or “nil” (+/- 10 percent of current premium). The Impact Analysis found no “massive variations in premiums” under the models studied; instead, there was “not that much change” in the average premiums “[f|or most consumer groups.” These findings contrast sharply with those reported to the trial court by State Farm actuary Becraft, who studied a sampling of 200,000 State Farm *1233 policies and found that a system where individual optional factor weights had to be less than those of mandatory factors would “dramatically change premiums.” Becraft’s study showed BIPD premium increases for 65 percent of good drivers, and large premium increases in certain areas, ranging up to an average increase of 85 percent in Inyo County. Petitioners characterize the Becraft study as speculative in a number of respects, and submit that it is “squarely rebutted” by the Impact Analysis. Petitioners contend that the Impact Analysis should be given more credence than the Becraft study because it was far larger in scope, and was not prepared to support any particular litigation position. State Farm and the Commissioner argue on various grounds that the Impact Analysis is of limited relevance in assessing the trial court’s decision. The Commissioner points out among other things that the current regulations provide for the use of different rating factors than those analyzed in the Impact Analysis. State Farm notes among other things that the current premiums the Impact Analysis used to measure dislocation were based on standard factor modeling rather than the actual premiums insurers charged. It is difficult to draw any firm conclusions from this conflicting evidence. On the one hand, it appears to us that Becraft’s study may overstate the effect of the trial court’s decision. Becraft determined that pumping or tempering would be required because the individual weights of four of State Farm’s optional factors were greater than those of one or more of the mandatory factors. The optional factors of (1) cost/frequency bands, (2) gender/marital status, (3) persistency, and (4) multi/single car had weights of 34.33, 25.08, 15.46, and 14.46, respectively, while the weights of the mandatory factors of safety record, mileage, and driving experience were 20.55, 13.60, and 10.47, respectively. 21 The most extreme pumping or tempering would be required to account for the optional factors of cost/ frequency bands and gender/marital status, because those two optional factors outweighed every mandatory factor. However, those two optional factors in State Farm’s class plan each combined optional factors which are separate under the regulations. (Regs., § 2632.5, subd. (d)(9) [gender], (10) [marital status], (15) [claims frequency], & (16) [claims severity].) Nothing compels the use of a single territorial factor instead of two. (Fn. 5, ante.) Presumably, the individual weights of the two territorial factors would have been less than their combined weight. Thus, less pumping or tempering would have been required if State Farm had used two territorial factors instead of one. The same reasoning would apply to the gender and marital status factors. On the other hand, even if State Farm’s apparently arbitrary decisions to combine separate optional factors under its class plan exaggerated the *1234 dislocations which would result from the trial court’s decision, Becraft’s study showed that pumping or tempering would be required under the trial court’s system. For example, even if the 34.33 weight of State Farm’s single cost/frequency bands factor were divided evenly in two, the resulting weights (17.2) would still exceed those of the mandatory factors of annual mileage (13.60) and driving experience (10.47). Moreover, it is questionable whether the Impact Analysis provided any better prediction of the likely effects of the trial court’s decision than Becraft’s study. The Commissioner is in the best position to evaluate the Department’s own study, and he contends that it is of little relevance to the trial court’s ruling. The Impact Analysis conceded that the low incidence of negative dislocation under the “[s]ingle [o]mit method without standardized factors” model, which has been cited as a good proxy for the trial court’s system, resulted in part because “in some instances the companies’ current practices were modified to make compliance with Prop. 103 easier.” This admission supports State Farm’s argument that the Impact Analysis underestimated the difference between the premiums payable under alternative models and the “premiums actually paid by policyholders” at the time. (Emphasis omitted.) (3) Average Weight Under regulations section 2632.8 as interpreted. in the administrative proceeding, only the average weight of the individual optional factors must be less than that of any mandatory factor. As explained above, this interpretation allows individual optional factors to outweigh mandatory factors and thus, like the trial court’s decision, permits territory to have a greater influence on premiums than any mandatory factor. The Commissioner argues that his regulations strike the best overall balance between Proposition 103’s weight ordering mandate and its other requirements that rates and premiums be substantially related to cost and not arbitrary. The Commissioner’s view is entitled to great weight. (Morris v. Williams, supra, 67 Cal.2d at p. 748.) While we do not believe that Proposition 103 demonstrably delegates to the Commissioner the power to interpret the weight ordering statute (Asimow, supra, 42 UCLA L.Rev. at p. 1198; § 1861.02, subd. (e) [the Commissioner is merely empowered to “implement[]” that statute]), it may nonetheless be appropriate to defer to the Commissioner’s judgment on the point if it appears that he has a “comparative interpretative advantage” with regard to the subject matter, and that his interpretation is “probably correct.” (Asimow, at p. 1195.) There can be no question that the Commissioner is in a better position than this court to evaluate the technical points in this case. This is perhaps *1235 most apparent on the subject of sequential analysis. Not being actuaries it is very difficult for us to evaluate the competing claims about this regression equation used for determining the initial relativities of rating factor categories. As will be recalled, the process requires that factors be analyzed in a particular sequence beginning with the mandatory factors in their statutory order and ending with the territorial factors, and that the results be adjusted at each stage to account for factors’ overlapping effects. Insurers submit that sequential analysis alone fulfills the weight ordering mandate by ensuring that each of the mandatory factors will influence premiums to the fullest extent of their statistical significance. Petitioners respond that sequential analysis has nothing to do with factor weight. In support of that argument, Petitioners quote the statement in the Impact Analysis that sequential analysis “ ‘is not a method for measuring how much the rating factor influences the premium charged to consumers (i.e., determining the rating factor’s weight).’’ ” The Commissioner currently takes the intermediate position that sequential analysis, “at least in part,” accomplishes the requisite factor weighting. While this statement apparently contradicts the earlier one in the Impact Analysis, it may be that the process of sequential analysis has become more sophisticated in the interim. We note that it was not until the current regulations were adopted that the process was defined beyond a specification of the order in which factors were to be analyzed. The Commissioner also persuasively maintains that sequential analysis “alone eliminate[s] the problem of arbitrary rates and good drivers subsidizing bad drivers in the same zip code that Proposition 103 was designed to address.” Apart from whether sequential analysis accomplishes factor weighting, a broader and more fundamental question associated with sequential analysis is whether it produces premiums which are cost based. The Commissioner sides with Insurers on this question, describing sequential analysis as “an objective, scientific process” which “ensures that the rating factors are ‘substantially related to the risk of loss.’ ” Petitioners, on the other hand, scoff at any attempt to “paint a landscape of calm, mathematical ratemaking, where insurers apply immutable scientific principles, and premiums and rates are never disassociated from the underlying risk of loss and cost of insurance.” Petitioners warn that “insurers massage numbers in rate making hearings with numerous hidden assumptions and tactical mathematical choices.” As Farmers only slightly exaggerates, Petitioners argue “as if actuarial studies were somehow to be avoided like the plague.” Actuary Becraft acknowledged in the administrative proceeding that State Farm occasionally departs from cost-based pricing, and the Department has *1236 made statements which could be interpreted to validate Petitioners’ concerns to some extent. For example, the Impact Analysis noted that doubts had been expressed about the relationship between risks of loss and some rating factors such as persistence, smoking, and academic standing. In the prior court action, Commissioner Garamendi characterized the “ ‘cost-based pricing’ ” identified as an alternative to the Tempered Regulations as “ ‘at best only a rough approximation to neutral, scientific process, if not a complete fiction.’ ” {Allstate Ins. Co. v. Gillespie, supra, No. B050439.) However, these statements from the Department did not refer to sequential analysis as practiced under the current regulations, and the Commissioner is in any event more qualified than this court to assess whether that analysis yields premiums which are cost based. If premiums produced by sequential analysis are in fact cost based, then there is another reason to reject alternative constructions of the statute which require pumping or tempering of factor relativities. As has been shown, pumping and tempering make individual premium changes which are both arbitrary and unrelated to the risk of loss. If the initial relativities established by sequential analysis are cost based, then the overall premium changes caused by pumping and tempering will also weaken the required relationship between rating factors and the risk of loss. For example, sequential analysis indicates that territory is the single most important determinant of the risk of loss. If that is true, then to the extent that pumping or tempering is required to suppress territory’s effect on premiums, premiums as a whole will not reflect the cost of providing insurance. As to whether the Commissioner’s interpretation of the weight ordering mandate is “probably correct,” one relevant consideration is whether the interpretation has been “consistently maintained.” (Asimow, supra, 42 UCLA L.Rev. at pp. 1196-1197.) There has been no single consistent construction in this instance. The Department originally adopted the “in the aggregate” alternative in its Tempered Regulations, but eventually abandoned that approach. The change of heart was understandable in light of the reception the Tempered Regulations initially received in court. Another equally relevant criterion is whether “the agency’s interpretation appears to have been carefully considered” (Asimow, at p. 1196), and this standard certainly militates in the agency’s favor here. The current construction emerged only after litigation and extensive study. The process has included adoption of the Tempered Regulations, the court challenge thereto, the Impact Analysis, public hearings on that study and the current regulations, and the administrative action which preceded this case. It appears that all competing views have been thoroughly aired along the way. The results of this process cannot lightly be set aside. *1237 Given the Department’s technical expertise and the time and effort it has devoted to the problem, its interpretation of the weight ordering mandate has a strong claim to deference. (4) Conclusions A line can be drawn through Proposition 103’s declarations of purpose, the provisions of the factor weight ordering statute, and the ballot pamphlet representations about how that statute would operate. On one side: (1) Proposition 103 declares that auto insurance rates are to be determined primarily by a driver’s safety record and mileage driven; (2) section 1861.02, subdivision (a) provides that these factors and the other mandatory factor of driving experience are to be more important in determining auto insurance rates than optional factors the Commissioner may adopt; and (3) the ballot pamphlet represents that the law would force rates to be determined by driving safety record “first,” rather than the driver’s residence. On the other side: (1) Proposition 103 declares that it will protect consumers from arbitrary insurance rates; (2) section 1861.02, subdivision (a) provides that factors used to determine auto insurance rates are to be substantially related to the risk of loss; and (3) the ballot pamphlet represents that factor weighting would decrease the auto insurance premiums of good drivers. The shared assumption underlying all of these declarations, provisions and representations is that safety record and other mandatory factors are more indicative of the insurance risk drivers pose than where they live. The line between these declarations, provisions and representations marks a conflict because that assumption is false. Unrefuted evidence establishes that territory is a more important determinant of the risk of loss than any other single factor. Petitioners cite no evidence to support their position other than the Impact Analysis which, as we have shown, divided territory into two optional factors and thereby allowed territory to have a greater effect on premiums than any mandatory factor. Thus, there is a conflict in the law, which any set of implementing regulations must attempt to reconcile. The current regulations manage to implement most of the law’s conflicting demands. The regulations create no necessity for pumping or tempering of relativities developed by sequential analysis, thereby avoiding a process which produces premiums that are individually and collectively arbitrary, and thereby preserving a substantial relationship between rating factors and risks of loss. The regulations produce lower premiums for more good drivers than the- other alternatives under consideration. The regulations ensure that optional factors will weigh less, at least on average, than any of the *1238 mandatory factors. The regulations’ provisions for sequential analysis ensure that mandatory factors will be considered “first” in determining premiums, and that those factors will account for all of the risks of loss they represent. What the regulations do not do is ensure that rates will be determined primarily by driving safety record and mileage driven. But as Judge Vogel put it when she looked at this problem over a decade ago, because “[t]his is not a perfect world and Proposition 103 is not a perfect system,” the Commissioner is left to “choose [among] imperfect results.” (Proposition 103 Cases, supra, Statement of Decision.) The current regulations constitute a lawful choice among imperfect options. In reaching this conclusion we are relying to some extent on the Commissioner’s assessment of the evidence, in particular the evidence that sequential analysis produces premiums which are cost based—a consideration at the very heart of the case. On technical points like the value of sequential analysis, deference to administrative expertise is warranted. To borrow a phrase from separation of powers cases, “ ‘confidence [must be] reposed in some one’ ” (Board of Supervisors v. Superior Court (1995) 33 Cal.App.4th 1724, 1741 [39 Cal.Rptr.2d 906]) with respect to such matters. We have nonetheless reviewed the entire record, including the technical aspects, and drawn our own conclusions as to what the law requires. In our independent judgment the regulations strike a proper balance between the Proposition’s competing demands. To observe that this conclusion is consistent with the evidence does not respond to Petitioners’ argument, because it is Petitioners’ position that the evidence in this case is irrelevant. Their argument in this appeal is that the Proposition is clear, that it dictates only one result, and that it can only be interpreted to require that individual optional factors not outweigh individual mandatory factors. Although Petitioners question the evidence that such a system would produce arbitrary premiums and raise the premiums of most good drivers, that evidence ultimately does not matter to Petitioners because they discern a clear statutory directive which must be applied regardless of the consequences. That directive stated in its simplest form is that territory cannot have a greater effect on premiums than an insured’s driving record. However, there is no single clear directive in the factor weight statute, much less Proposition 103 as a whole or its legislative history. There are instead competing requirements, and the consequences of interpretations must be evaluated to determine the extent to which they fulfill those requirements. The interpretation Petitioners advocate does not implement the directive they identify in any event. As we have repeatedly observed, when there are multiple territorial factors, a limitation on individual optional factor weights *1239 does not guarantee that territory will have less effect on premiums than any mandatory factor. Since section 1861.02, subdivision (a)(4) does not limit the number of optional factors the Commissioner may adopt, a limitation on the individual weights of territorial factors does not limit territory’s total effect on premiums. Since there are currently two territorial factors, the trial court’s interpretation does not ensure that territory will have less effect on premiums than the mandatory factors. The only rationale offered for the trial court’s decision is therefore untenable. The most that could be said for a rule that the weight of each optional factor must be less than that of each mandatory factor is that this may be a permissible interpretation of the statute. Evidence on the consequences of that interpretation could be seen as somewhat equivocal. However, we find no reason to prefer that interpretation to the one reflected in the current regulations. We therefore hold that the trial court erred in concluding that the weights'of individual optional factors must be less than that of any mandatory factor. We further hold that the trial court erred in drawing the related conclusion that individual optional factor weights must be calculated in lieu of a single optional factor weight which reflects the average weight of all of the optional factors. 22 Petitioners attempt to stand on principle but fail to advocate the only rule that would produce the result they claim the statute requires: a rale that the aggregate weight of all of the optional factors must be less than that of any mandatory factor. We have considered this rale sua sponte because it alone would ensure that rates are determined primarily by a driver’s safety record and mileage driven. However, uncontradicted evidence establishes that this rule would produce rates which are individually and collectively arbitrary, undermine the relationship between rating factors and risks of loss, and result in higher premiums for most good drivers. Thus, while this rule alone would implement one of Proposition 103’s aims, it would contravene several others. Therefore, we hold that the aggregate weight of the optional factors need not be less than that of any mandatory factor. F. Other Issues (1) Unruh Civil Rights Act The Los Angeles Board of Supervisors (the Board) contends in its amicus curiae brief that “the retention of zip code rating under [regulations *1240 section] 2632.8” violates the Unruh Civil Rights Act (Civ. Code, § 51) because it “creates a subtle, but pernicious form of unlawful discrimination against less affluent minority communities in urban areas.” Presumably, “zip code rating” means the two territorial factors described in regulations section 2632.5, subdivision (d)(15), (16), which provide for categories where insureds are grouped by zip codes with similar average claims frequency and claims severity. However, it is unclear what the Board means by “retention” of zip code rating under regulations section 2632.8. Regulations section 2632.8 addresses how factor weights are calculated, not which optional factors may be used. Ambiguity is not the least of the problems with the Board’s argument. This is a case about regulations, not business practices. Since the Department is not a “business establishment” within the meaning of the Unruh Civil Rights Act (see Taormina v. California Dept, of Corrections (S.D.Cal 1996) 946 F.Supp. 829, 833), that law has no bearing on any regulation the Commissioner may adopt. If it is the Board’s position that the Commissioner has unlawfully approved the use of claims frequency and severity as optional factors in auto insurance rate setting, that claim is outside the scope of the issues raised in the case. (See Costa v. Workers’ Comp. Appeals Bd. (1998) 65 Cal.App.4th 1177, 1187-1188 [77 Cal.Rptr.2d 289] [amici curiae generally cannot raise new issues].) Petitioners do not contend that the frequency and severity factors cannot be used; their concern is with the allowable weight of those territorial factors. There is no dispute that those factors are substantially related to the risk of loss. (See § 1861.02, subd. (a)(4) [criterion for approval of optional factors].) Thus, even if the policies reflected in the Unruh Civil Rights Act were relevant here, there is a legitimate business reason for the use of those factors which would preclude any finding of unlawful discrimination. (Harris v. Capital Growth Investors XTV (1991) 52 Cal.3d 1142, 1162-1169 [278 Cal.Rptr. 614, 805 P.2d 873] [risk-oriented distinctions based on economic as opposed to personal characteristics do not violate the Unruh Civil Rights Act]; see also § 11628, subd. (a) [“[differentiation in rates between geographical areas shall not constitute unfair discrimination”].) (2) Attorney Fees Reversal of the judgment dictates that the attorney fee awards to Petitioners also be reversed. Whether and to what extent Petitioner’s are entitled to attorney fees for making a “substantial contribution” within the meaning of section 1861.10, subdivision (b), to the decision in this case must be reassessed in light of the result of this appeal. The issues of entitlement to *1241 fees and the amount of any award are properly left to the determination of the trial court on Petitioners’ motions after the remittitur is filed. (See generally Laurel Heights Improvement Assn. v. Regents of University of California (1988) 47 Cal.3d 376, 426 [253 Cal.Rptr. 426, 764 P.2d 278]; Milman v. Shukhat (1994) 22 Cal.App.4th 538, 546 [27 Cal.Rptr.2d 526].) III. Disposition The judgment and the postjudgment orders for attorney fees and costs are reversed, with costs on appeal to the Commissioner and Insurers. Reardon, J., and Poché, J., * concurred. A petition for a rehearing was denied January 25, 2001, and the opinion was modified to read as printed above. The petitions of all respondents for review by the Supreme Court were denied March 28, 2001. George, C. J., Mosk, J., and Kennard, J., were of the opinion that the petitions should be granted. *1242 Appendix A [[Image here]] *1243 Appendix B [[Image here]] *1244 [[Image here]] *1245 Appendix C [[Image here]] *1246 [[Image here]] *1247 [[Image here]] *1248 Appendix D [[Image here]] Footnotes [1] Unless otherwise indicated, all future statutory references are to the Insurance Code, and all future references to regulations are to title 10 of the California Code of Regulations. [2] The regulations prescribe a different means of calculating the single optional factor weight which yields the same result as the calculation we have shown adding up the individual factor weights and dividing by the number of factors. (Regs., § 2632.8, subd. (c).) [3] These weights are shown in appendix A attached to this opinion. [4] References to “the Commissioner” in this part of the opinion are to the Office of Insurance Commissioner and/or the Insurance Commissioner then in office, as the context requires. [5] The factor weight calculation in State Farm’s plan thus combines the two claims frequency and claims severity factors into one optional factor, and the two gender and marital status factors into one optional factor. The regulations do not address factor combinations except to permit the optional factor of second driver characteristics to be composed of a combination of other factors (Regs., § 2632.5, subd. (d)(13)), and to direct that the “three mandatory factors may not be combined with any other factor, except Percent Use, Academic Standing, Gender, Marital Status, and Driver Training may be combined with number of years of driving experience” (id., § 2632.5, subd. (e)). Combinations of optional factors with other optional factors do not appear to be restricted or required. The record includes testimony of a department official confirming that State Farm’s optional factor combinations were acceptable to the department. We note also that “Frequency Bands” and “Severity Bands” are listed as separate factors on the sample “Factor Weights Summary Table” included with the department’s class plan filing instructions, and have been treated as separate factors by other insurers. Thus, it does not appear that State Farm was required to treat the two territorial factors as one optional factor for purposes of the weight calculation. [6] Although “mature driver improvement” is not included in the regulations’ list of optional factors, the department’s class plan filing instructions state that section 11628.3 requires “mature driver discounts where actuarially justified.” [7] We are not relying on the unpublished appellate opinion in the case, which is nonbinding in view of the dismissal of the appeal, or on the trial court’s decision as precedents; they are discussed to put the challenged regulation in historical perspective. [8] Copies are attached as appendix B. We are looking at table (2) on each of the pages, under the heading “Factor Weights in Compliance.” [9] The sole exception for “Company C” on the second page of appendix B (frequency) appears to have been an aberration. [10] The previous territorial factors were: average repair garage labor rates; average medical and hospital costs; average wage and income levels; litigation rates; population density; vehicle density; accident/claims frequency; theft rates; number of uninsured vehicles; and average claims cost. (Statement, supra, at p. 7.) [11] We are again referring to appendix A. [12] This paragraph attempts to account for Becraft’s testimony on the results of the Bimbaum method under State Farm’s plan, and for the difference between two numbers that appear on page 1 of appendix C hereto. This chart compares the results produced by the Bimbaum method with those yielded by the insurers’ calculations for a universe of 10 cars. In the middle of the last line there is a figure of 1881.97 for the “Sum of Abs. Diff. for Optional Factors.” This is the total premium effect of the optional factors when their individual premium effects are added up (the individual premium effects appear on the line above: 202 for factor 1, 220.47 for factor 2, etc.). This total premium effect of 1881.97, divided by the number of cars (10), yields a combined weight of 188.20 when the individual weights are simply added up. In the far right-hand column, the number at the bottom is the combined weight of the optional factors under the Bimbaum method. This number, 69.11, is less than 188.20. The difference appears to result from the “canceling out” process just described. [13] We are looking here at pages 2 and 3 of appendix C. Page 2 shows the results of tempering, page 3 shows the results of pumping. After those adjustments are made, the number at the bottom right-hand comer of the page (the single combined weight of the optional factors under the Bimbaum method), is less than each of the three individual mandatory factor weights shown at the bottom left-hand corner of the page. The Bimbaum number on page 2 is 27.29; the mandatory factor weights are 36.62, 28.65 and 27.47. These numbers are in the same descending order on page 3. (Compare these numbers on pp. 2 and 3 with those on p. 1 of appen. C, before the pumping and tempering.) However, even after the pumping and tempering, the combined weight of the individual optional factors shown on pages 2 and 3 (the “Sum of Abs. Diff. for Optional Factors” divided by 10, as explained in the prior footnote) remains greater than any individual mandatory factor. Those numbers would be 72.82 on page 2 (728.21 * 10) and 191.10 on page 3 (1911.03 10). [14] At the oral argument in this appeal Petitioners denied advocating an aggregate limit on optional factor weights in their briefing to the superior court, but whatever Petitioners’ intention may have been, we can only go by the meaning of the words they used in their papers. We note further that, at this point in the superior court proceedings, Petitioners also expanded their initial claims for relief by asking the court to order refunds of “excess premiums collected in violation of [section] 1861.02(a).” The court declined to order any refunds and they are not an issue in the appeal. [15] Further references to “the Commissioner” are to the Office of Insurance Commissioner and/or Commissioner Low as the context requires. [16] State Farm’s request for judicial notice of Proposition 103 campaign literature disseminated by Voter Revolt is denied. (See 20th Century Ins. Co. v. Garamendi, supra, 8 Cal.4th at p. 269, fn. 8.) The other requests for judicial notice are granted. [17] For example, if companies A and B used the same three categories—high, average, and low—with the same relativities for the mileage factor, and company A happened to have a higher percentage of drivers who fell into the “average” category than company B, the weight of the mileage factor in A’s plan would be lower than B’s simply by virtue of that fact. This can be illustrated by assuming that each company insures two drivers, has a base rate of $1,000, and uses relativities of 1.2, 1.0, and 0.8 for the high, average and low categories. A’s drivers are both in the average category, while one of B’s drivers is in the high category and the other is in the low category. The premiums of A’s drivers are both unchanged by the mileage factor and remain at $1,000 (1000 x 1.0). The weight of the mileage factor in A’s plan would thus be 0 ([0 + 0] * 2). The premiums of B’s drivers would each change by $200: the premium of the high mileage driver would increase by that amount ($1,000 x 1.2 = $1,200); the premium of the low mileage driver would decrease by that amount ($1,000 x 0.8 = $800). The weight of the mileage factor in B’s plan would thus be 200 ([$200 + $200] -h 2). [18] We have previously quoted statements in Petitioners’ briefs where they appear to acknowledge that the “single omit” formula is an appropriate way to determine the weights of individual optional as well as mandatory factors. In its briefs, State Farm likewise assumes that the single omit methodology would be used to calculate individual optional factor weights if the trial court’s decision is affirmed. We note that at one point in its briefs, the Project characterizes the single omit method as only a “ ‘possible’ ” approach to determining individual optional factor weights, and states that it is not “endorsing any one approach” to that determination. However, the Project has clearly endorsed the single omit formula for setting individual mandatory factor weights, and it is not apparent why a different formula would be necessary or proper for the individual optional factors. Use of a different formula for the optional factor weights would be subject to the same criticism Petitioners level against the current regulations: it would be an inconsistent way of comparing the relative weights of the optional and mandatory factors. In any event, in discussing allowable weights of optional factors relative to mandatory factors, both individually and in the aggregate, this opinion will assume that individual optional factor weights, like individual mandatory factors weights, are determined by the single omit method under regulations section 2632.8, subdivision (c). [19] The concurrence in King v. Meese quoted the following passage from Los Angeles Dept, of Water & Power v. Manhart (1978) 435 U.S. 702, 710 [98 S.Ct. 1370, 1376, 55 L.Ed.2d 657], which concerned an agency requirement that female employees make larger pension fund contributions than male employees because women generally live longer than men: “ ‘[W]hen insurance risks are grouped, the better risks always subsidize the poorer risks. Healthy persons subsidize medical benefits for the less healthy; unmarried workers subsidize the pensions of married workers; persons who eat, drink, or smoke to excess may subsidize pension [benefits] for persons whose habits are more temperate. Treating different classes of risks as though they were the same for purposes of group insurance is a common practice that *1226 has never been considered inherently unfair. To insure the flabby and the fit as though they were equivalent risks may be more common than treating men and women alike; but nothing more than habit makes one “subsidy" seem less fair than the other.’ ” (King v. Meese, supra, 43 Cal.3d at p. 1242, fn. 14 (cone. opn. of Broussard, J.).) Petitioners refer to this passage to thwart any criticism that the factor weight ordering they advocate would require some drivers to unfairly “ ‘subsidiz[e]’ ” others. Insurers persuasively distinguish Manhart on the ground that this case was talking about group insurance where premiums are based on a group’s average risk, whereas auto insurance premiums are based on different risk classifications for individuals. With respect to group insurance, Proposition 103 provides that “rates shall not be considered to be unfairly discriminatory, if they are averaged broadly among persons insured under the group plan.” (§ 1861.12.) As State Farm points out, this provision implies that the cross-subsidies inherent in group insurance would otherwise be subject to disapproval under section 1861.05, subdivision (a), as unfairly discriminatory. In any event, we find no indication that Proposition 103 was intended to subsidize any particular class of drivers. [20] We are back now to pages 2 and 3 of appendix C, looking at the numbers in the far right-hand columns in the upper right-hand comers under the heading “% Change.” The change shown on page 2 in car l’s premium after optional factor tempering was plus 84.8 percent. The change shown on page 3 in car l’s premium after mandatory factor pumping was minus 35.4 percent. The opposite results happened with car 8. Tempering reduced that car’s premium 2.6 percent, while pumping increased its premium by 59.8 percent. [21] These numbers are shown on appendix D hereto. [22] The County of Fresno et al. contend that the trial court’s interpretation should be rejected because it raises equal protection concerns. Because we conclude that the decision below must be reversed for other reasons, we do not reach this argument. [*] Retired Associate Justice of the Court of Appeal, First Appellate District, assigned by the Chief Justice pursuant to article VI, section 6 of the California Constitution.