Opinion

The Secretary of the Health and Welfare Agency and the Director of the Department of Benefit Payments 1 (hereinafter defendants) appeal from a judgment granting writs of mandate and declaratory relief invalidating a portion of a welfare regulation because of conflict with federal and state law. Plaintiffs cross-appeal from certain restrictions placed on the scope of the relief.

Plaintiffs are recipients of benefits under the program for aid to families with dependent children (AFDC). Established by the federal Social Security Act ( 42 U.S.C. § 601 et seq.), the AFDC program is intended to provide financial assistance to needy dependent children and the parents or other relatives with whom they reside. The program is financed in substantial part by the federal government and is administered by the states, under a scheme of “cooperative federalism.” (King v. Smith (1967) 392 U.S. 309, 316 [ 20 L.Ed.2d 1118, 1125 , 88 S.Ct. 2128 ].) State participation is elective, but the receipt of federal funds is conditioned on compliance with the Social Security Act and applicable federal regulations. (45 C.F.R. pt. 233 (1979).) California participates in the AFDC program under the provisions of the Burton-Miller Act. (Welf. & Inst. Code, § 11200 et seq.) The State Department of Social Services (see fn. 1, ante) is charged with administering the program to “secure full compliance with the applicable provisions of state and federal laws” (Welf. & Inst. Code, § 10600), and is required to promulgate regulations “not in conflict with the law” (id., § 10604).

The operation of the program is well known. As we explained in Conover v. Hall (1974) 11 Cal.3d 842, 847 [ 114 Cal.Rptr. 642 , 523 P.2d 682 ], “California has adopted a ‘flat grant’ system for determining the actual payments to be made to AFDC recipients. Section 11450 of the Welfare and Institutions Code establishes a schedule of ‘maximum aid’ payments, with maximum payments varying according to the *132 number of eligible needy persons in the same home. The amount of a recipient’s actual aid payment is computed by subtracting from the appropriate maximum aid figure all nonexempt income of the aid recipient. [Citation.] [1Í] In determining the amount of nonexempt income of a recipient to be subtracted from the maximum figure, federal law requires that the state consider not only the recipient’s gross income but also the off-setting expenses which a recipient incurs in earning that income. In particular, the specific federal requirement at issue in this case provides that ‘the State agency shall, in determining need, take into consideration any other income and resources of any child or relative claiming aid ... as well as any expenses reasonably attributable to the earning of any such income.’ (Italics added.) (Social Security Act, § 402(a)(7), 42 U.S.C. § 602 (a)(7).)” (Fn. omitted.)

Plaintiffs filed this action primarily to challenge the compliance of a state welfare regulation with the foregoing emphasized language of the Social Security Act. The regulation in issue is EAS 44-113.24. 2 It enumerates certain work-related expenses allowable as deductions from earned income to determine eligibility for and amount of AFDC assistance. The listed expenses are: (a) mandatory withholding deductions for income taxes, social security, retirement and similar contributions; (b) child care costs; (c) costs of food and clothing required for employment; (d) work-related transportation costs; (e) nonpersonal expenses such as tools, licenses, and union dues; and (f) business expenses incurred by self-employed persons.

In particular, the portion of the regulation prescribing travel expenses (EAS 44-113.241(d)) declares that the recipient’s “necessary costs of transportation to and from work” shall be allowed; but in the case of a recipient who is compelled to use his or her own automobile for that purpose because public transportation is either unavailable or inappropriate, the regulation limits those costs to a flat rate of 15 cents per mile. 3

*133 Plaintiff Green alleged that she owned an automobile required for transportation to and from work, and that defendants should have taken into consideration the actual costs of operating that vehicle in determining her AFDC grant. She was permitted only a mileage-based deduction of $10.08 for the month of October 1974 when, in fact, her actual costs included $105 for transmission repair. Defendants refused to allow the latter deduction, and her monthly grant was reduced accordingly.

Plaintiff Fingers similarly alleged that she required the use of the family automobile in her employment as a visiting nurse. In July and August 1975 her actual cost of owning and operating that vehicle far exceeded her mileage-based deduction, and as a result her monthly grant was reduced below the amount to which she was otherwise entitled.

The gravamen of the complaint, comprising three causes of action, was that EAS 44-113.24 imposes an impermissible limitation on the recognition of expenses reasonably attributable to the earning of income. In the first cause of action, plaintiff Green sought individual relief by writ of administrative mandate (Code Civ. Proc., § 1094.5; Welf. & Inst. Code, § 10962) to review defendants’ decision to deny her transportation expenses under this regulation. In the second cause of action, plaintiffs Green and Fingers applied for an ordinary writ of mandate (Code Civ. Proc., § 1085) on behalf of themselves and all persons similarly situated to compel defendants to allow all work-related expenses in determining eligibility and benefit levels, and to obtain retroactive relief for benefits wrongfully withheld. In the third cause of action, also framed as a class action, plaintiffs prayed for a declaration that the regulation is invalid on the foregoing ground, i.e., because it does not allow deduction of all expenses reasonably attributable to the earning of income.

Plaintiffs moved for class certification and for summary judgment. The court granted the motion and made an order certifying the action as a class action; declaring EAS 44-113.24 invalid in its entirety; granting a writ of administrative mandate to Green; and issuing a peremptory writ of mandate, the terms of which were to be determined after an evidentiary hearing, against enforcement of the regulation. On defendants’ subsequent motion for decertification, however, the court limited the class to present and future AFDC recipients, thereby denying retroactive relief. Additionally, after hearing argument as to the *134 proper scope of its decision, the court ruled that only the portion of the challenged regulation relating to transportation expenses was in issue; it limited the evidentiary hearing accordingly, and in the final judgment awarded plaintiffs relief only as to EAS 44-113.241(d), the transportation-expense limitation.

Defendants’ Appeal

I.

In its declaratory judgment the trial court ruled that EAS 44-113.241(d) is invalid because “by establishing a per mile standard allowance for expenses of driving a motor vehicle to and from work and on the job, [the regulation] imposes a maximum limit on the recognition of these expenses in violation of section 602(a)(7) of Title 42 of the United States Code.”

Defendants contend, as the “major issue” in the case, that the federal statute permits a state to impose a maximum limit of “reasonableness” on the amount of work-related transportation expense it must recognize in calculating AFDC grants. They then urge that the mileage allowance provided in EAS 44-113.241(d) is such a limit: it was assertedly fixed after statistical studies demonstrated that the average cost of owning and operating a private automobile for work-related purposes was 15 cents per mile. Defendants offered to prove the reasonableness of that figure, and now claim the court erred in ruling such evidence irrelevant to the legal question presented.

The contention is refuted by our unanimous decision in County of Alameda v. Carleson (1971) 5 Cal.3d 730 [ 97 Cal.Rptr. 385 , 488 P.2d 953 ]. There, certain county welfare authorities complained that state welfare regulations “improperly permitted AFDC applicants to deduct from their earned income all work-related expenses actually incurred, without regard to the reasonableness of the amount expended.” (Id. at p. 747.) The trial court agreed, and ordered that the regulations be amended “to provide for the deduction of only a reasonable amount of work-related expenses actually incurred in producing income, up to certain maximum levels” (ibid.). We reversed the judgment, squarely holding such a limitation to be a violation of section 602(a)(7). We reasoned that “The legislative history underlying the work-related expenses deduction provision indicates that it was the intent of Congress to permit the exclusion of all such expenses, without regard to the *135 amount expended, provided that such expenses were reasonably related to employment. ‘The committee [Senate Committee on Finance] believes that it is only reasonable for the States to take these expenses [of earning income] fully into account. Under existing law if these work expenses are not considered in determining need, they have the effect of providing a disincentive to working since that portion of the family budget spent for work expenses has the effect of reducing the amount available for food, clothing and shelter. The bill has, therefore, added a provision in all assistance titles requiring the States to give consideration to any expenses reasonably attributable to the earning of income.’ (Italics added; 1962 U.S. Code Cong. & Admin. News, pp. 1959-1960.)

“In addition to the legislative history, HEW regulations provide that in determining eligibility for aid, ‘only such net income as is actually available for current use on a regular basis will be considered . . . . ’ (Italics added; 45 C.F.R., § 233.20 , subd. (a)(3)(ii)(c).) The legislative history, together with these regulations (which have been adjudged to ‘clearly comport’ with the Act [citations]), compel the conclusion that all work-related expenses must be considered in determining eligibility, for those expenses necessarily reduce the net income ‘actually available’ for current support needs. To disallow a portion of those expenses as ‘unreasonable’ would undermine the primary purpose of the deduction to provide further incentives toward employment.” (Italics in original; id. at pp. 747-748 .)

Defendants urge that a different result is compelled by Shea v. Vialpando (1974) 416 U.S. 251 [ 40 L.Ed.2d 120 , 94 S.Ct. 1746 ], but we do not so read that opinion. Shea dealt with AFDC regulations promulgated by the State of Colorado. Prior to 1970 those regulations permitted the deduction from income of all expenses reasonably attributable to employment, including the actual cost of work-related transportation. In 1970, however, Colorado substituted for such regulations a flat allowance of $30 per month for all work-related expenses, including transportation, except mandatory payroll deductions and child care costs. The figure was based on a statistical survey showing it to be the statewide average of such expenses.

Under the pre-1970 regulations the plaintiff in Shea had been allowed to deduct actual work-related automobile expenses totalling $110 per month. When the new regulation took effect that deduction was reduced to the flat $30 figure, and the corresponding increase in net income terminated her eligibility for the AFDC program. She filed *136 a class action in federal district court, contending that Colorado’s standardized work-expense allowance violated section 402(a)(7) of the z Social Security Act. ( 42 U.S.C. § 602 (a)(7).) The court agreed, granting a motion for summary judgment and enjoining enforcement of the regulation.

The United States Supreme Court affirmed in a unanimous opinion. The court began by observing ( 416 U.S. at p. 253 [ 40 L.Ed.2d at p. 125 ]) that a principal purpose of the AFDC program, as declared by Congress, is to help the parents or other relatives who care for needy dependent children “to attain or retain capability for the maximum self-support and personal independence consistent with the maintenance of continuing parental care and protection” ( 42 U.S.C. § 601 ). 4 The court then focused on the terms of section 402(a)(7) of the Social Security Act, stressing that it explicitly requires the recognition of “any” work-related expenses in determining net income. ( 416 U.S. at p. 260 [ 40 L.Ed.2d at p. 129 ].) Reviewing the legislative history of that section, the court then relied on the same Senate report that we found persuasive in County of Alameda v. Carleson (cf. 416 U.S. at p. 263 [ 40 L.Ed.2d at p. 131 ] with 5 Cal.3d at p. 748 ). From this history, moreover, the court drew the same conclusion—that “Congress thus sought to encourage AFDC recipients to secure and retain employment by requiring the States to take into account fully any expenses attributable to the earning of income in determining eligibility for assistance,” and failure to allow that deduction “might well discourage the applicant from seeking or retaining employment whereby such expenses are incurred. Section 402(a)(7) was aimed at removing this disincentive.” ( 416 U.S. at p. 264 [ 40 L.Ed.2d at p. 131 ]; cf. 5 Cal.3d at p. 748 .) The court held that the flat $30 limit on work-related expenses imposed by the Colorado regulation resulted in just such a disincentive, and hence was invalid because in conflict with federal law. ( 416 U.S. at p. 265 [ 40 L.Ed.2d at p. 132 ]; accord, Conover v. Hall (1974) supra, 11 Cal.3d 842, 847-849 .)

Defendants nevertheless seize on isolated language in the Shea opinion to support their claim that County of Alameda v. Carleson is no longer the law. Thus at several points the Shea opinion speaks of *137 “reasonable expenses.” (E.g., 416 U.S. at pp. 260, 264-265 [40 L.Ed.2d at pp. 129, 131-132].) In our view, however, this usage is simply a shorthand reference to the entire clause of section 402(a)(7) that requires the states to recognize “any expenses reasonably attributable to the earning of any such income” by an AFDC recipient. In that clause, of course, “reasonably” modifies “attributable” rather than “expenses”; to read it as defendants would have us do—i.e., that the states need recognize only “reasonable expenses attributable to the earning” of such income—would be to rewrite the plain wording of the statute in a manner contrary to the entire thrust of Shea .

Defendants also quote the high court’s summary of its holding, to wit, that “no limitation, apart from that of reasonableness, may be placed upon the recognition of expenses attributable to the earning of income.” (Italics added; id. at p. 260 [ 40 L.Ed.2d at p. 129 ].) On the ground just stated, however, the emphasized language must be taken to refer not to the amount of such expenses but to whether they are properly “attributable to the earning of income”: the states need not recognize any expenses, regardless of their amount, that would be unreasonable to view as work-related. Indeed, we made precisely the same point in County of Alameda v. Carleson ( 5 Cal.3d at p. 749 ): “Of course, our interpretation of the Act and regulations thereunder does not foreclose the state from disallowing in whole or in part those expenses which are not reasonably attributable to the production of income. For example, if an automobile were not required by the nature of one’s employment, or if other feasible means of transportation were available, the state could properly disallow at least a portion of the expense of acquiring and maintaining an automobile, not because the amount expended was unreasonable, but because the type of expenditure was not a bona fide work-related expense subject to the statutory deduction.” (Italics in original.) 5

Our reading of the emphasized language of Shea is confirmed by the very next sentence of the high court opinion, in which the court concludes: “Accordingly, a fixed work-expense allowance which does not permit deductions for expenses in excess of that standard is directly contrary to the language of the statute.” (Italics added; 416 U.S. at *138 p. 260 [ 40 L.Ed.2d at p. 129 ].) The “standard,” in Shea as here, is the statistical average expense figure that defendants now claim is both “reasonable” and serves the state’s interest in administrative efficiency.

The court returned to this point several times in Shea . Thus it carefully distinguished its earlier rule (Rosado v. Wyman (1970) 397 U.S. 397, 419 [ 25 L.Ed.2d 442, 459 , 90 S.Ct. 1207 ]) that it is permissible to use statistical averages in fixing the statewide standard of need: “the discretion granted the States by Congress in determining need [citation] contrasts sharply with the statutory requirement of § 402(a)(7) that any expenses reasonably attributable to the earnings of income be considered. In the face of that statutory command and the clear statement of congressional purpose, we must also reject petitioners’ claims of administrative efficiency or convenience.” ( 416 U.S. at p. 266, fn. 13 [ 40 L.Ed.2d at p. 132 ].) Finally the court explained, in words we find dispositive of defendants’ contention in the case at bar: “It is, of course, not the adoption of a standardized work-expense allowance per se which we hold to be violative of § 402(a)(7) of the Act, but the fact that the standard used by Colorado is in effect a maximum or absolute limitation upon the recognition of such expenses. As the Court of Appeals correctly observed, a standard allowance would be permissible, and would substantially serve petitioners’ interests in administrative efficiency, if it provided for individualized consideration of expenses in excess of the standard amount. See 475 F.2d, at 735.” (2d italics added; id. at p. 265 [ 40 L.Ed.2d at p. 132 ].) 6 In short, under federal law a standardized allowance can serve as a base of calculation, but it cannot be used as a ceiling on the amount ultimately deductible.

Defendants argue that the state welfare regulation here challenged does not impose a “standard allowance” within the meaning of Shea because the amount it permits each AFDC recipient using a private automobile to claim as transportation expense varies according to the number of miles the car is driven each month. It is true that such a method of calculation produces a rough approximation of at least some *139 of the expenses that the recipient incurs: the more the car is driven, obviously, the higher are the costs of gasoline, oil and lubrication, the greater the wear on the tires and other parts of the vehicle, and the more frequent the need for repair or replacement. Nevertheless, in two significant respects the method fails to take into consideration the actual expenses of the recipient.

First, because the per-mile figure is admittedly an average it cannot, by definition, include every justifiable expense of each individual car used for work-related travel. The expense of owning and operating a hypothetical “average” car is never the same as the actual cost for any given vehicle: some will cost less to run, others will unavoidably cost more. In particular, the cars used by AFDC recipients to get to work are typically older second-hand vehicles; 7 in comparison to new cars, such vehicles tend to be less fuel-efficient, burn more oil, require more frequent lubrication, and break down more often. An average mileage rate ignores these realities. It thus results in out-of-pocket losses to many AFDC recipients in the operation of their cars for travel to and from their jobs, and for the same reason it constitutes a disincentive to seek or maintain employment.

Secondly, certain major costs of owning an automobile are wholly unrelated to the number of miles it is driven: e.g., monthly payments on the car loan, depreciation, insurance premiums, and licensing and registration fees. The owner must bear these costs whether he drives one mile or a hundred miles. Previously, a number of such expenses were specifically deductible against income (former EAS 44-114.223 (eff. Aug. 1, 1969)), but the regulation now in issue provides for none of them. It is apparent that many AFDC recipients who must use their cars to reach their place of employment but need drive only a relatively short distance to do so—as is often true of inner-city residents—do not recoup their actual work-related travel expenses when their allowance is calculated solely by the number of miles covered: a round trip of six miles to and from work, for example, produces a total allowance of merely 90 cents a day. Again the limitation results in a disincentive to hold a job. 8

*140 It follows that the regulation under attack must be viewed as a standardized allowance governed by the express language of Shea . It was therefore correctly declared invalid by the trial court, because it fails to provide for “individualized consideration of expenses in excess of the standard amount.” ( 416 U.S. at p. 265 [ 40 L.Ed.2d at p. 132 ].) The large majority of federal courts that have ruled on the issue are in accord with this reading of Shea and its statutory foundation. (See, e.g., Chambly v. Freeman (W.D.Mo. 1979) 478 F.Supp. 1221 , affd. without opn. (8th Cir. 1980) 624 F.2d 1108 [17 cents per mile]; Fones v. Percy (W.D.Wis. 1978) No. 78-C-191, affd. without opn. (7th Cir. 1979) 601 F.2d 600 [same]; Perez v. Chang (D.Hawaii 1977) 438 F.Supp. 238 [15 cents per mile]; Byrnes v. Beal (W.D.Pa. 1976) No. 75-1336 [Pov.L.Rep. (CCH) ¶ 22,488] [7 cents per mile]; McDougal v. Davis (D.Ore. 1978) No. 77-862 [ 12 Clearinghouse Rev. 378 ] [consent decree; 8 cents per mile]; contra, Weaver v. Page (M.D.Fla. 1979) No. 78-458-Orl-Civ-R [14 cents per mile].) 9

II.

Defendants next contend the trial court erred in ordering them to restore to the two named plaintiffs all AFDC benefits unlawfully withheld from them because of the transportation-expense limitation since October 1, 1971, the date on which the predecessor to the present regulation took effect. It is urged that plaintiffs’ retroactive relief *141 should have been restricted to the period of the statute of limitations preceding the filing of their complaint on October 23, 1975. With respect to their cause of action for mandate, that period was three years. 10

The contention is meritorious. Our courts have viewed the obligation of a governmental entity to pay welfare benefits pursuant to law as a debt due to the recipient as of the date he first became entitled to them. (Bd. of Soc. Welfare v. County of L.A. (1945) 27 Cal.2d 81, 86 [ 162 P.2d 630 ]; Lowry v. Obledo (1980) 111 Cal.App.3d 14, 24 [ 169 Cal.Rptr. 732 ]; Hypolite v. Carleson (1975) 52 Cal.App.3d 566, 583 [ 125 Cal.Rptr. 221 ]; Leach v. Swoap (1973) 35 Cal.App.3d 685, 689-690 [ 110 Cal.Rptr. 62 ]; cf. Tripp v. Swoap (1976) 17 Cal.3d 671, 682-683 [ 131 Cal.Rptr. 789 , 552 P.2d 749 ].) Such a debt is analogous to the obligation of a governmental entity to make payments of back wages unlawfully withheld from its workers or of pension benefits unlawfully withheld from its retirees: each such payment becomes a debt due to the employee as of the date he was entitled to receive it. It is settled that in such cases each deficient payment constitutes a separate violation triggering the running of a new period of limitations, and hence that the employee can recover only those payments which accrued within the period of the applicable statute of limitations preceding the filing of his complaint. (Jones v. Tracy School Dist. (1980) 27 Cal.3d 99, 103-107 [ 165 Cal.Rptr. 100 , 611 P.2d 441 ] [salary]; Dryden v. Board of Pension Commrs. (1936) 6 Cal.2d 575, 580-582 [ 59 P.2d 104 ] [pension].) The same rule governs here, and the court should therefore have ordered defendants to restore to the two named plaintiffs only those AFDC benefits unlawfully withheld from them by reason of the transportation-expense limitation during the three years preceding October 23, 1975. 11

*142 Even that period, however, may be excessive for the purpose of calculating retroactive AFDC payments to the class that plaintiffs represent. In distinction to a suit brought by an individual claimant for his own account, an action by an entire class to recover past welfare benefits withheld pursuant to an invalid regulation or statute might impose, in some circumstances, a disproportionate clerical and financial burden on the governmental entity if such benefits were ordered paid for the entire period of limitations. This could occur, for example, if the regulation or statute had been in force for a number of years at the time of the judgment, if the class were particularly large, or if the potential individual recovery of each class member were small. Accordingly, on such a showing the trial court, acting as a court of equity, has discretion to fix a more realistic starting date for the payment of retroactive benefits to class members.

Precisely this distinction was drawn in Hypolite v. Carleson, supra, 52 Cal.App.3d 566 , when the trial court declined to order full retroactive payment of benefits because an invalid welfare regulation had been in effect for many years; instead, the court limited recovery to the period beginning with the filing of the complaint. The Court of Appeal held that the record supported the court’s exercise of its discretion to fix a proper starting date, explaining that “According to the ‘debt’ theory which the trial court correctly applied in principle, any individual claimant’s eligibility for retroactive relief could be held to have commenced on the actual date when he was ‘first entitled to receive the aid.’ (Bd. of Soc. Welfare v. County of L.A., supra, 27 Cal.2d 81 at p. 86 .) On the other hand, a practical application of the theory to a class of claimants requires that retroactive relief be granted back to a single date which has some relevance and which is feasible, in practical fact, when applied to the class under the realities of the situation.” ( 52 Cal.App.3d at p. 585 .)

*143 In the case at bar, as will appear (Part IV, post), plaintiffs are entitled on remand to have the class they represent reconstituted to include past recipients of and applicants for AFDC benefits, as originally certified. At that time, however, the trial court will have discretion, upon motion and showing, to reconsider pursuant to the foregoing principles its determination of the appropriate starting date for payment of retroactive benefits to the class plaintiffs. 12

Plaintiffs’ Cross-appeal

III.

Plaintiffs first contend the trial court erred in limiting the scope of its judgment to the portion of EAS 44-113.24 that provides an allowance for transportation expenses alone (i.e., EAS 44-113.241(d)), rather than ruling on the validity of the regulation in its entirety. The point is well taken.

Despite defendants’ claim to the contrary, a fair reading of the complaint reveals that plaintiffs challenged the validity of EAS 44-113.24 as a whole. They alleged, inter alia, that their class of AFDC recipients and applicants is entitled to a redetermination of benefits for each month in which defendants refused to consider all work-related expense exemptions because of EAS 44-113.24 or its predecessor regulations. In turn, the prayer explicitly asked for a declaration that the regulation is invalid because it imposes “fixed, maximum limits on the recognition of the actual expenses reasonably attributable to the earning of income .. . . ” It is thus apparent that the complaint raised the issue of the entire regulation’s validity. 13 Any other reading would be unjustifiably *144 restrictive and contrary to the well-established rule of liberal construction in favor of the pleader. (Code Civ. Proc., § 452; Skopp v. Weaver (1976) 16 Cal.3d 432, 438 [ 128 Cal.Rptr. 19 , 546 P.2d 307 ]; Harman v. City and County of San Francisco (1972) 7 Cal.3d 150, 157 [ 101 Cal.Rptr. 880 , 496 P.2d 1248 ], and cases cited.) As this court stated in Terry Trading Corp. v. Barsky (1930) 210 Cal. 428, 438 [ 292 P. 474 ], “to deny the party his right to a trial, there must be an obvious failure of the pleadings to state a cause of action or defense.” No such failure is shown here.

Defendants next argue that plaintiffs’ evidence showed a dispute only with regard to the transportation-expense portion of the regulation, and hence that plaintiffs had no standing to seek mandatory or declaratory relief as to the regulation as a whole.

It is true that ordinarily the writ of mandate will be issued only to persons who are “beneficially interested.” (Code Civ. Proc., § 1086.) Yet in Bd. of Soc. Welfare v. County of L.A. (1945) 27 Cal.2d 98 [ 162 P.2d 627 ], this court recognized an exception to the general rule ‘“where the question is one of public right and the object of the mandamus is to procure the enforcement of a public duty, the relator need not show that he has any legal or special interest in the result, since it is sufficient that he is interested as a citizen in having the laws executed and the duty in question enforced’” (id. at pp. 100-101). The exception promotes the policy of guaranteeing citizens the opportunity to ensure that no governmental body impairs or defeats the purpose of legislation establishing a public right. (Id. at p. 100.) It has often been invoked by California courts. (See, e.g., Hollman v. Warren (1948) 32 Cal.2d 351, 357 [ 196 P.2d 562 ]; American Friends Service Committee v. Procunier (1973) 33 Cal.App.3d 252, 256 [ 109 Cal.Rptr. 22 ]; Fuller v. San Bernardino Valley Mun. Wat. Dist. (1969) 242 Cal.App.2d 52, 57 [ 51 Cal.Rptr. 120 ], and cases cited.)

Properly read, our recent decision in Carsten v. Psychology Examining Com. (1980) 27 Cal.3d 793 [ 166 Cal.Rptr. 844 , 614 P.2d 276 ], does not hold to the contrary. There, a member of an administrative board filed an action for writ of administrative mandate to compel her own board to comply with a certain statute. We affirmed a judgment of dismissal, holding that the plaintiff lacked standing because she was not a party “beneficially interested.” We recognized (at p. 797) that “there are circumstances under which a citizen-taxpayer may compel a governmental instrumentality to comply with its constitutional or statutory *145 duty.” But we found “persuasive legal and policy reasons why an exception to the statutory requirement of beneficial interest should not be carved out for a person in this posture.” (Italics added; id. at p. 798.) Among those reasons, we emphasized (at p. 799) a number of “policy issues which militate against permitting disgruntled governmental agency members to seek extraordinary writs from the courts.” And we concluded (at p. 801) that “a board member is not a citizen-taxpayer for the purpose of having standing to sue the very board on which she sits.”

As these excerpts make clear, Carsten was not intended to sound the death knell of the “public right/public duty” exception to the requirement of beneficial interest for a writ of mandate. Rather, its ratio decidendi is simply that the policy underlying the exception may be outweighed in a proper case by competing considerations of a more urgent nature—there, the dangers consequent upon allowing an administrative board member to sue her own agency.

No such considerations are present in the case at bar, and plaintiffs were therefore entitled to claim the exception. There can be no question that the proper calculation of AFDC benefits is a matter of public right (Diaz v. Quitoriano (1969) 268 Cal.App.2d 807, 811 [ 74 Cal.Rptr. 358 ]), and plaintiffs herein are certainly citizens seeking to procure the enforcement of a public duty. (American Friends Service Committee v. Procunier (1973) supra, 33 Cal.App.3d 252, 256 .) It follows that plaintiffs have standing to seek a writ of mandate commanding defendants to cease enforcing EAS 44-113.24 in its entirety. The trial court erred in ruling otherwise, and in limiting the scope of the evidentiary hearing accordingly. Plaintiffs are therefore entitled to a new hearing on their cause of action for writ of mandate, and to a determination of the validity of the remainder of the regulation. 14

IV.

Plaintiffs next contend the trial court erred in partially decertifying the class action—to exclude past recipients of and applicants for AFDC benefits—after it had decided the merits by its order granting plaintiffs’ motion for summary judgment. The contention is meritorious. It is well established that in the absence of relevant state precedents our *146 trial courts are urged to follow the procedures prescribed in rule 23 of the Federal Rules of Civil Procedure for conducting class actions. (Vasquez v. Superior Court (1971) 4 Cal.3d 800, 821 [ 94 Cal.Rptr. 796 , 484 P.2d 964 , 53 A.L.R.3d 513 ]; accord, Occidental Land, Inc. v. Superior Court (1976) 18 Cal.3d 355, 360 [ 134 Cal.Rptr. 388 , 556 P.2d 750 ]; City of San Jose v. Superior Court (1974) 12 Cal.3d 447, 453 [ 115 Cal.Rptr. 797 , 525 P.2d 701 , 76 A.L.R.3d 1223 ]; La Sala v. American Sav. & Loan Assn. (1971) 5 Cal.3d 864, 872 [ 97 Cal.Rptr. 849 , 489 P.2d 1113 ]; Daar v. Yellow Cab Co. (1967) 67 Cal.2d 695, 708-709 [ 63 Cal.Rptr. 724 , 433 P.2d 732 ]; Travelers Ins. Co. v. Superior Court (1977) 65 Cal.App.3d 751, 759 [ 135 Cal.Rptr. 579 ]; Home Sav. & Loan Assn. v. Superior Court (1974) 42 Cal.App.3d 1006, 1012 [ 117 Cal.Rptr. 485 ].) The California class action statute (Code Civ. Proc., § 382) is silent on the question of when a court may order a class to be certified or decertified. Federal rule 23, however, explicitly provides in subdivision (c)(1) that “As soon as practicable after the commencement of an action brought as a class action, the court shall determine by order whether it is to be so maintained. An order under this subdivision may be conditional, and may be altered or amended before the decision on the merits.” (Italics added.) Giving effect to the emphasized qualification, our courts have held that procedural class-action issues—including the composition of the class—must ordinarily be resolved before a decision on the merits is reached. (Home Sav. & Loan Assn., supra, 42 Cal.App.3d at pp. 1012-1014; Home Sav. & Loan Assn. v. Superior Court (1976) 54 Cal.App.3d 208, 211-214 [ 126 Cal.Rptr. 511 ]; cf. Hypolite v. Carleson (1975) supra, 52 Cal.App.3d 566, 581-583 .)

Although this rule has thus far been applied only for the benefit of defendants, no reason appears why plaintiffs should not also enjoy its benefits within the limitations discussed below. We recently reviewed the rationale for invoking the rule to protect defendants: “unless a decision on the merits is postponed until after the class issues are decided, a defendant is subject to ‘one-way intervention,’ which would allow potential class members to elect whether to join in the action depending upon the outcome of the decision on the merits. Thus, if the merits were decided favorably to the class, and notice followed such determination, most class members would join in the action, whereas they would decline if the determination was against the class.

“From a defendant’s viewpoint, this is said to result in ‘an open-ended lawsuit that cannot be defeated, cannot be settled, and cannot be adju *147 cheated.’ [Citation.] Failure to require notification of the class before a decision on the merits prevents a binding adjudication against the class because members of the class who were not notified are not barred by the determination in the defendant’s favor since they were not parties. On the other hand, a defendant who loses an action brought by individual class members may be estopped under the doctrine of collateral estoppel to deny the binding effect of the judgment against him in a subsequent action brought by other class members. Moreover, a defendant is entitled to know, before final determination of the substantive issues in a class action, the full potential consequences and liability that may attach to a judgment against him.” (People v. Pacific Land Research Co. (1977) 20 Cal.3d 10, 16-17 [ 141 Cal.Rptr. 20, 569 , P.2d 125].)

A similar rationale warrants the protection of plaintiffs against the abuse of a delayed motion to decertify by defendants. Without the requirement that class issues be resolved prior to a decision on the merits, a defendant could take advantage of decertification by a strategy similar to that of “one-way intervention.” Thus he could appear to acquiesce in the plaintiff’s motion to certify the class, holding back his evidence and arguments on the issue. If the judgment on the merits then goes in his favor, it will bind all members of the class who were notified and bar further lawsuits against him on the same cause of action by all such unnamed class members; 15 indeed, the larger the class, the more he will be insulated from such litigation. Yet if instead he loses on the merits, he can undo most of the damage by bringing out his evidence and arguments and mounting a belated attack on the certification order. To be sure, if he fails in his effort to decertify the class he will have given away the chance to enter into a settlement for potentially less than the liability flowing from a judgment on the merits. But that is all he loses. The strategy of acquiescence thus affords the defendant the inexpensive option of seeking to bind as large a plaintiff class as possible while taking no commensurate risk. 16

Rule 23(c)(1) was designed to prevent a party from obtaining the advantage of such a no-lose litigation strategy. While defendants have an additional reason for the rule’s protection—i.e., their interest in *148 knowing the consequences of a judgment against them—there is no reason to interpret the plain language of the rule so that it will not apply equally to plaintiffs and defendants. Although a defendant must take some risk in order to implement the strategy of acquiescence, we will not give rule 23(c)(1) a strained reading to permit the possibility of such abuse: whether the motion to certify or decertify be made by the plaintiff or the defendant, the rule directs that it be determined “before the decision on the merits.”

Nonetheless, we decline to fashion an iron-clad standard removing all jurisdiction from a trial court to decertify a class or part thereof after such a decision. We have always recognized that it is desirable for the trial court to retain some measure of flexibility in handling a class action. (Vasquez v. Superior Court, supra, 4 Cal.3d at p. 821 .) It would be both unduly rigid and unjust to force the maintenance of such an action even when there is a proper reason for decertification after judgment. (See Link v. Mercedes-Benz of N. Am., Inc. (3d Cir. 1977) 550 F.2d 860 , 864.) Before judgment, a class should be decertified “only where it is clear there exist changed circumstances making continued class action treatment improper.” (Sley v. Jamaica Water & Util., Inc. (E.D.Pa. 1977) 77 F.R.D. 391, 394 .) A fortiori, a similar showing must be made to warrant decertification after a decision on the merits. This standard will prevent abuse on the part of the defendant while providing the trial court with enough flexibility to justly manage the class action.

Applying this standard to the case at bar, we conclude that defendants cannot escape application of the general rule that decertification must occur, if at all, prior to a decision on the merits. By failing to contest plaintiffs’ motion for class certification when it was filed, and by waiting until after a determination on the merits before acting, defendants in effect waived any right to move for decertification. (Cf. Civil Service Employees Ins. Co. v. Superior Court (1978) 22 Cal.3d 362, 371-374 [ 149 Cal.Rptr. 360 , 584 P.2d 497 ].) Defendants’ arguments in support of their belated motion were not based on changed circumstances, nor did they adduce new evidence that became available to them only after the merits were decided. 17 It follows that the trial court *149 erred in partially decertifying the class after it had granted plaintiffs’ motion for summary judgment. 18

Conclusion

For the foregoing reasons, we conclude that (1) EAS 44-113.241(d) is invalid; (2) plaintiffs’ retroactive relief should be restricted to the period of limitations preceding the filing of the complaint; (3) plaintiffs are entitled to challenge the remainder of the regulation by writ of mandate; and (4) plaintiffs are entitled to represent the class first certified by the trial court. The judgment is reversed for further proceedings consistent herewith. Plaintiffs shall recover their costs on appeal.

Bird, C. J., Tobriner, J., Richardson, J., and Newman, J., concurred.

The Department of Benefit Payments is now the State Department of Social Services. (Stats. 1977, ch. 1252, § 725, p. 4635.)

Whcn the action was filed, the regulation was numbered EAS 44-113.23. It was later renumbered and amended in minor respects. The judgment originally referred to the superseded regulation; it was subsequently corrected to show the current numbering, and we shall so refer to it herein.

When the action was filed the figure was 12 cents per mile; it was subsequently fixed at 15 cents per mile by an amendment to the regulation. The regulation also provides that the resulting total is to be adjusted by (1) adding the actual cost of tolls and parking fees and (2) subtracting all amounts contributed by persons riding with the recipient.

Our Legislature declared an identical purpose in the Burton-Miller Act: “The employment and self-maintenance of parents of needy children shall be encouraged to the maximum extent and this chapter shall be administered in such a way that needy children and their parents will be encouraged and inspired to assist in their own maintenance.” (Welf. & Inst. Code, § 11205; see also id., § 11008.)

This statutory limitation is the obvious answer to defendants’ alleged fear that under the trial court’s ruling section 402(a)(7) will require them to recognize as a transportation cost a hypothetical AFDC recipient’s expense of “hot-rodding” the car that he or she uses to get to work every day. They may fairly conclude, rather, that such an expense is not “reasonably attributable to the earning” of the recipient’s income.

At the cited page the court of appeals stated: “a reasonable interpretation of the statute permits 'a state to use a standard deduction for [employment] expenses in the interest of efficient administration of the [AFDC] program, provided the allowance is adequate to cover all actual expenses'” (Italics in original.)

Federal administrative authorities likewise construe section 402(a)(7) and Shea to require that in calculating AFDC eligibility or benefits “The State may designate a flat amount [allowable as a work-related expense] but any expenses exceeding that amount will also be deducted from the earned income.” (Italics added.) (Dept. of Health, Ed. & Welf., Soc. Sec. Admin., Action Transmittal SSA-AT-78-27 (OFA) (June 20, 1978) p. 2.)

For example, a previous version of the transportation-expense regulation did permit a recipient to deduct the actual cost of buying a car for work-related travel, but placed a ceiling of $770 on the purchase price of the vehicle. (Former EAS 44-114.231 (eff. Aug. 1, 1969).) Even in 1969, $770 would buy no more than an older second-hand car.

Defendants' claim that their current rate of 15 cents per mile is adequate to reimburse AFDC recipients for all work-related car expenses may also be viewed in the context of the real world of the automobile rental business. The Hertz Corporation reg-

*140 ularly calculates the actual cost of owning and operating various-sized cars driven for 10,000 miles a year for 3 years, not including tolls and parking fees. For 1979, the most recent year available, the resulting figures were as follows: full-size cars, 43 cents per mile; mid-size cars, 38 cents per mile; and compact cars, 31.9 cents per mile. Of course we cite these figures as illustrative only, not as defining standards the state must meet in operating the AFDC program.

Finally, defendants point to the fact that the State of California uses a per-milc method of reimbursing those of its employees who find it necessary to drive their personal cars on state business when state vehicles arc unavailable. The analogy is unpersuasive. First, such reimbursement is more flexible and somewhat more generous than the AFDC regulation in issue, as the rate varies from 18.5 to 21 cents per mile according to the circumstances. (State Admin. Manual, § 0755 (as revised Apr. 1980).) More importantly, reimbursement of certain slate employees for this limited purpose does not implicate the important social policy, declared by statute and emphasized in Shea , of encouraging welfare recipients to hold jobs in order to become as self-supporting as possible.

The trial court declared that the challenged regulation is also invalid on a second ground—to wit, that it violates Flealth and Safety Code section 11008—and defendants complain of the ruling. The point raises complicated questions of statutory construction, not all of which are addressed by the parties. Because we hold that the trial court correctly ruled the regulation invalid under controlling federal law, we need not and do hot reach this alternate ground of decision.

lt is often difficult to decide which statute of limitations governs an action for writ of mandate. The code provisions authorizing this action are silent as to the time within which it must be filed. (See Code Civ. Proc., § 1085 et seq.) Accordingly, the courts have developed the rule that the question is to be resolved not by the remedy prayed for but by the nature of the underlying right or obligation that the action seeks to enforce. (Allen v. Humboldt County Board of Supervisors (1963) 220 Cal.App.2d 877, 884 [ 34 Cal.Rptr. 232 ].) Because plaintiffs’ cause of action for mandate in the case at bar seeks to enforce “a liability created by statute,” it is subject to a three-year limitations period. (Code Civ. Proc., § 338, subd. 1; Dillon v. Board of Pension Commrs. (1941) 18 Cal.2d 427, 429-430 [ 116 P.2d 37 , 136 A.L.R. 800 ].)

Defendants also argue that plaintiffs did not introduce any evidence of the benefits they were denied prior to the particular dates mentioned in the complaint (i.e., Oct. 1974; July and Aug. 1975). No such evidence, however, was necessary or appropriate *142 in the proceeding below. Plaintiffs showed that the regulation had used an invalid standard for calculating allowable work-related transportation expenses, and that they had been harmed thereby. In the circumstances, the trial court correctly ordered defendants to cease enforcing that standard against plaintiffs, and to determine the amounts due to them by reason of its past operation. The court had no duty to fashion the proper standard in its place—that remains the prerogative of the Legislature and, by delegation, the administrative authorities (cf. Pitts v. Perluss (1962) 58 Cal.2d 824, 832-833 [ 27 Cal.Rptr. 19 , 377 P.2d 83 ])—or to calculate the amounts payable thereunder (Bila v. Young (1942) 20 Cal.2d 865, 869-870 [ 129 P.2d 364 ]). On this question, accordingly, the effect of the judgment is simply to give plaintiffs the opportunity to show, in an appropriate forum and under a standard yet to be devised, the amount of benefits they were denied by reason of the invalid regulation and its predecessor.

Lastly, defendants contend that insofar as the judgment grants plaintiff Green a writ of administrative mandate because of the invalidity of the regulation here in issue it violates the rule against using that procedure to review legislative acts of an administrative agency. (See, e.g., Strumsky v. San Diego County Employees Retirement Assn. (1974) 11 Cal.3d 28, 34-35, fn. 2 [ 112 Cal.Rptr. 805 , 520 P.2d 29 ].) The act here reviewed, however, was adjudicatory: Green exercised her statutory right to demand a “fair hearing” of the decision to reduce her AFDC grant (Welf. & Inst. Code, § 10950); and when that hearing resulted in an adverse decision, her “exclusive remedy” (id., § 10962) was to file an action for a writ of administrative mandate in superior court pursuant to Code of Civil Procedure section 1094.5. In both proceedings she was entitled to challenge the validity of the underlying regulation. (Woods v. Superior Court (1981) 28 Cal.3d 668 [ 170 Cal.Rptr. 484 , 620 P.2d 1032 ].)

PIaintiffs’ dissatisfaction with- the entire regulation is also shown by the fact that almost six months before instituting this action they filed a formal petition (former Gov. Code, § 11426) asking the state welfare director to amend the regulation to conform to section 602(a)(7) of title 42 of the United States Code. The petition was denied but plaintiffs asked that the ruling be reconsidered. It was only when that request was also denied that plaintiffs filed the present complaint.

In these circumstances plaintiffs’ cause of action for declaratory relief duplicates their cause of action for writ of mandate. We therefore need not and do not reach the question of plaintiffs' standing to seek declaratory relief as to the entire regulation, and on remand the trial court may treat the question as moot.

See, e.g., Daar v. Yellow Cab Co. (1967) supra, 67 Cal.2d 695, 704 , and cases cited.

When the defendant is a governmental body, as here, the risk involved in acquiescence may be further reduced because the fear of defending numerous individual suits challenging the same regulation may outweigh the fear of liability on a class scale.

Two of their claims—i.e., that the class as certified was unascertainable and unmanageable—clearly were not premised on changed circumstances or evidence available only after the merits were decided. Defendants also alleged that subsequent to the original certification plaintiffs attempted to “redefine” the class by proposing to the court alternate ways of describing its membership. However, that proposal cannot *149 amount to a changed circumstance when, as here, plaintiffs never formally moved to expand the original class and the alternate verbal formulations they suggested to the court described essentially the same class as originally certified. Furthermore, the court apparently never considered recertifying an expanded class of any kind.

Defendants have filed a motion to dismiss for untimeliness plaintiffs’ cross-appeal on the class action issue, arguing that plaintiffs should have appealed directly from the order of partial decertification. We have recognized an exception to the “one final judgment rule’’ so as to allow a direct appeal from an order denying certification to an entire class, on the theory that “the order is tantamount to a dismissal of the action as to all members of the class other than [the individual] plaintiff.” (Daar v. Yellow Cab Co. (1967) supra, 67 Cal.2d 695, 699 .) Here, by contrast, the portion of the class that was decertified (past recipients of and applicants for AFDC benefits) is not separate and distinct from the portion that remained in the case (present and future recipients of and applicants for the same benefits). Rather, because the cut-off date used by the court—the date of entry of judgment in this particular lawsuit—was unrelated to any event in the ongoing administration of the regulation in issue, there is doubtless a significant overlap between the two classes: many persons who received benefits before that date continued to receive them afterwards. Thus the order of partial decertification was not “tantamount to a dismissal of the action as to all members of the class," and was not directly appealable. (Cf. Vasquez v. Superior Court (1971) supra, 4 Cal.3d 800, 806-807 [order dismissing one of two causes of action in a class action, held not appealable].) It follows that plaintiffs’ cross-appeal from the final judgment on this issue was both proper and timely. The motion is denied.