Becker, C.J.
Five women professors, who are or were employed by the School of Dentistry, brought this gender discrimination and equal pay lawsuit against the University of Washington. At issue is whether the trial court properly granted their motion to certify it as a class action. The plaintiffs contend that intentional discrimination against the female faculty is a common course of conduct at the University, as evidenced by the administration’s failure to close a statistically-modeled gender gap in compensation. The certification order is before us on discretionary review.
We first hold that the Legislature’s waiver of the State’s sovereign immunity to tort actions encompasses class actions in tort. The trial court correctly ruled that Lacey Nursing Center, Inc. v. Department of Revenue, 128 Wn.2d 40, 905 P.2d 338 (1995), presents no obstacle to certification. We also hold that when a tort action against the State is properly initiated by a plaintiff who has timely filed a notice of claim as required by RCW 4.92.100-110, additional plaintiffs later added to the action when it is certified for class treatment need not separately fulfill the claim filing requirement.
Nevertheless, we ultimately reverse the class certification order. We conclude that the plaintiffs, whose allegations of discriminatory treatment by the School of Dentistry clearly may proceed as individual cases, have not presented common issues that would entitle them to represent all the rest of the women faculty in a class action. A statistical model showing gender disparities in faculty pay, without more, will not prove that the University has a discriminatory motive. Evidence of intentional discrimination in a single department will not prove a common course of intentionally discriminatory conduct that is fairly attributable to decision making by central administration.
SOVEREIGN IMMUNITY
The Washington Constitution authorizes the Legislature to “direct by law, in what manner, and in what [84]*84courts, suits may be brought against the state.” Const, art. II, § 26. In 1961, the Legislature waived the state’s sovereign immunity with respect to tort actions:
The state of Washington, whether acting in its governmental or proprietary capacity, shall be liable for damages arising out of its tortious conduct to the same extent as if it were a private person or corporation.
RCW 4.92.090. This statute is “one of the broadest waivers of sovereign immunity in the country.” Savage v. State, 127 Wn.2d 434, 444, 899 P.2d 1270 (1995). It makes the State presumptively liable for its alleged tortious conduct “in all instances in which the Legislature has not indicated otherwise.” Savage, 127 Wn.2d at 445. Discrimination is a tort. Blair v. Wash. State Univ., 108 Wn.2d 558, 576, 740 P.2d 1379 (1987). Therefore, the waiver of sovereign immunity in RCW 4.92.090 applies to this gender discrimination action brought under the Washington Law Against Discrimination, chapter 49.60 RCW
The University contends that the above statute, in the absence of additional language so stating, cannot be deemed to have waived the State’s immunity with respect to class actions in tort. For this argument, the University relies on Lacey Nursing.
In Lacey Nursing, nursing homes attempted to maintain a class action suit asserting an exemption from the business and occupation tax for the sale or rental of real estate. By way of relief, they sought refunds for a four-year period. They proceeded under RCW 82.32.180. That statute imposes specific conditions upon taxpayers seeking excise tax refunds. It requires that “[a]ny person” who is aggrieved must “keep and preserve books, records, and invoices” as a prerequisite to filing an application for refund in the superior court of Thurston County. RCW 82.32.180; Lacey Nursing, 128 Wn.2d at 49-50. Taxpayers must identify themselves, state the correct amount of tax each concedes to be the true amount, state the reasons why the tax should be reduced or abated, and prove that the tax paid is incorrect. RCW 82.32.180; Lacey Nursing, 128 Wn.2d at 50.
[85]*85None of the Lacey Nursing claimants had stated facts meeting all of these specific statutory conditions. Obviously, unnamed and unidentified class plaintiffs had not done so either. The trial court reasoned that the action did not need to be treated as an individual taxpayer protest when the key issue was the legal basis for application of the tax rather than the computation of individual refunds. Lacey Nursing, 128 Wn.2d at 53.
The Supreme Court agreed that the order certifying the case for class treatment satisfied the requirements of Civil Rule 23, including commonality and numerosity. The court nevertheless held that the trial court did not have a tenable basis for allowing an excise tax refund lawsuit to proceed as a class action. Lacey Nursing, 128 Wn.2d at 51-52. Tax statutes allowing refunds are narrowly construed. The court concluded that the Legislature intended excise tax refunds to be made only as prescribed by the statute, and the statute requires each individual taxpayer to satisfy the conditions specified before maintaining an appeal. “RCW 82.32.180 contains no express language authorizing class actions in suits for tax refunds. Since the state waives sovereign immunity only to the extent provided in the statute, the statute must expressly authorize class actions.” Lacey Nursing, 128 Wn.2d at 53-54.
The statute waiving sovereign immunity for damages arising out of the State’s tortious conduct — RCW 4.92.090— does not expressly authorize class actions. On that basis, the University contends that state agencies remain immune from tort cases brought as class actions.
We read Lacey Nursing as requiring express authority for class actions only with respect to excise tax refund suits, not with respect to all types of lawsuits against the State. In Lacey Nursing the court was concerned with the State’s narrow waiver of sovereign immunity to one specific type of tax appeal authorized by RCW 82.32.180. By contrast, the waiver of sovereign immunity in tort in RCW 4.92.090 is broad. Neither RCW 4.92.090 nor any other statute dictates a specific format for a tort action [86]*86against the State comparable to the limitations with which the Legislature has circumscribed the initiation of an excise tax refund appeal.
The University argues that the Legislature’s waiver of sovereign immunity to tort actions is found not in RCW 4.92.090, but rather in RCW 4.92.010. RCW 4.92.010 provides that “[a]ny person or corporation having any claim against the state of Washington shall have a right of action against the state in the superior court.” The statute then sets forth the venue for such actions. The University argues that this statute excludes class action suits because it uses the singular “any person,” thus implying only an individual suit. The University claims that because both the tax statute at issue in Lacey Nursing and RCW 4.92.010 use the same “any person” language to describe who can sue the State, there is no basis for distinguishing tort actions and tax suits. We are unpersuaded. The waiver of sovereign immunity is found in RCW 4.92.090, and it is broad. The tax statute has unique aspects not shared by tort actions.
The Legislature has made the State liable for tortious conduct “to the same extent as if it were a private person or corporation.” RCW 4.92.090. This language satisfies the Legislature’s constitutional obligation to direct the manner in which suits may be brought. The Constitution does not require the Legislature to go further and specifically waive immunity with respect to class actions.
A class suit is a valuable procedure “for vindicating claims which, taken individually, are too small to justify individual legal action but which are of significant size and importance if taken as a group.” Brown v. Brown, 6 Wn. App. 249, 253, 492 P.2d 581 (1971). This is no less true for claims against the State than for claims against other defendants. Because private persons and corporations are subject to class actions for damages arising out of their discriminatory conduct, the State is also.
[87]*87TORT CLAIMS
The University contends that a class action for tort damages may be maintained only if all members of the class have complied individually with the statutory requirements for filing tort claims with the office of risk management or, at minimum, if at least one properly filed tort claim gives notice that the claimant intends to seek class certification. The original plaintiff in this case, Dolphine Oda, timely presented a properly verified tort claim to the office of risk management, but her notice did not indicate an intent to maintain a class action. Of course, none of the potential members of the class have presented notices of claim.1
The Tort Claims Act requires that a tort claimant, before commencing an action against the State, must present a verified claim to the risk management office at least 60 days before filing suit. RCW 4.92.100-110. Among other things, the claim “shall accurately describe the conduct and circumstances which brought about the injury or damage, describe the injury or damage, state the time and place the injury or damage occurred, state the names of all persons involved, if known, and shall contain the amount of damages claimed.” RCW 4.92.100. Strict compliance with the requirements of notice of claim statutes is generally a condition precedent to recovery. Hardesty v. Stenchever, 82 Wn. App. 253, 259, 917 P.2d 577 (1996). Dismissal of the action is the proper remedy for a plaintiff’s failure to comply. Hardesty, 82 Wn. App. at 259.
The statute requires a claimant to provide specifics about the claim before filing suit. It does not require a claimant to anticipate and describe future procedural developments that may occur in the lawsuit once it is filed. The names of other persons involved are to be provided, but only “if known.” Dr. Oda’s tort claim, filed in 1998 at least [88]*8860 days before she commenced the present action, did not announce a class action and did not name any other potential plaintiffs. But it fully complied with all the statutory notice requirements that constitute a condition precedent to commencing a lawsuit.
It is the very nature of a class action to gather into a single lawsuit a large number of individuals whose names may be unknown to the original parties. The mechanisms that fully identify and notify the members of the class are not available until someone commences the action and then obtains permission to proceed under Civil Rule 23. To require dismissal of all class plaintiffs who had not filed a verified tort claim at least 60 days before commencement of the suit would make it virtually impossible to proceed with a class tort action against the State. That result would be at odds with our earlier conclusion that the broad waiver of sovereign immunity in RCW 4.92.090 subjects the State to class actions in tort cases.
We conclude that the claim filing statute does not compel reversal of the class certification order. “No action shall be commenced against the state for damages arising out of tortious conduct until sixty days have elapsed after the claim is presented to and filed with the risk management office.” RCW 4.92.110. In this case, the action was commenced by Dr. Oda. Her individual lawsuit is now past the commencement stage and has moved on procedurally to a later stage, a motion for certification as a class action. The statute does not require the court to revert to the prefiling stage and apply the prefiling requirements to claimants whose presence in the action is the result of a postfiling procedural development.
We hold that once a tort action is properly commenced against the State, a CR 23 certification motion may not be defeated by the fact that the claimants to be added as plaintiffs have not previously filed a tort claim.
[89]*89COMMONALITY
The question to which we now turn is whether the class of all the women faculty shares a common issue with the original claimants so as to justify allowing the original claimants to proceed as representative parties.
Commonality is one of the four prerequisites for a class action listed in CR 23(a):
One or more members of a class may sue or be sued as representative parties on behalf of all only if (1) the class is so numerous that joinder of all members is impracticable, (2) there are questions of law or fact common to the class, (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class, and (4) the representative parties will fairly and adequately protect the interests of the class.
The requirements of commonality and typicality tend to merge, and are often addressed as a single issue. Gen. Tel. Co. of the S.W. v. Falcon, 457 U.S. 147, 157, 102 S. Ct. 2364, 72 L. Ed. 2d. 740 (1982). Additionally, a class action must also satisfy one of the three requirements in CR 23(b).2
Commonality supporting class certification exists where facts indicate that the defendant was engaged in a “ ‘common course of conduct’ in relation to all potential class members.” King v. Riveland, 125 Wn.2d 500, 519, 886 P.2d 160 (1994) (quoting Brown, 6 Wn. App. at 255).
The trial court certified a class consisting of all women faculty employed after 1994 (excluding staff, clinical, affiliate, visiting, adjunct, temporary or extension faculty, medical residents, teaching assistants, research assistants or librarians). The court found that the claims of the plaintiffs [90]*90involved questions of law and fact common to the class, including but not limited to: (1) whether there is a pattern or practice of intentional discrimination against women faculty at the University of Washington in compensation, promotion and other kinds of employment; (2) whether female faculty are promoted less quickly than their male counterparts; (3) if so, is the perceived disparity in promotion rates due to intentional gender discrimination; (4) whether the University’s central administration knows of the disparities; (5) whether it has a duty and authority to remedy those disparities; (6) whether the University has intentionally failed to remedy gender discrimination; and (7) whether members of the class have sustained injuries and damages and, if so, what is the most appropriate mechanism for assessing and awarding damages and administering injunctive relief.
Evidence reviewed by the trial court shows that decisions about pay and promotion of faculty are made at the departmental level on criteria specific to that department’s academic discipline. The University argues that due to decentralization, there is no common course of intentional discriminatory conduct to which disparities in compensation can fairly be attributed. The University maintains that the claims of the plaintiffs must be handled as individual gender discrimination actions.3
A trial court’s decision to certify a class is reviewed for abuse of discretion. Eriks v. Denver, 118 Wn.2d 451, 466, 824 P.2d 1207 (1992). The plaintiffs first contend that the trial court’s decision in this case is insulated from further review once it is clear that the court considered all [91]*91the factors set forth in CR 23. For this proposition, Dr. Oda cites a statement in Eriks that “a trial court’s class certification will be upheld if it appears from the record that the court considered all of the criteria of CR 23.” Eriks, 118 Wn.2d at 467. However, the authority cited in Eriks is Washington Education Ass’n v. Shelton School District No. 309, 93 Wn.2d 783, 793, 613 P.2d 769 (1980) (WEA). The Supreme Court in WEA held that the trial court abused its discretion in refusing to certify a class “without appropriate consideration and articulate reference to the criteria of CR 23.” WEA, 93 Wn.2d at 793. WEA does not hold that a certification decision must be upheld if the trial court explicitly considers the CR 23 factors. As is true in all types of cases, a court abuses its discretion when its decision is based on untenable grounds or is manifestly unreasonable or arbitrary. Weyerhaeuser Co. v. Commercial Union Ins. Co., 142 Wn.2d 654, 683, 15 P.3d 115 (2000). The record does indeed show that the trial court here expressly considered the factors set forth in CR 23 before deciding to certify the class. Whether that decision rests on tenable grounds remains a question to be decided by this court.
The plaintiffs also argue that a decision certifying a class should be reviewed liberally, citing Brown, 6 Wn. App. 249. Plaintiffs in Brown alleged that a public utility company was arbitrarily and discriminatorily cutting off services to customers in certain situations. The trial court denied a motion to certify a class action. Reversing, this court stated that it favored “a liberal interpretation of CR 23, rather than a restrictive one,” and quoted from a federal case:
“It cannot be denied that the resolution of the class action issue in suits of this type places an onerous burden on the trial court. But if there is to be an error made, let it be in favor and not against the maintenance of the class action, for it is always subject to modification should later developments during the course of the trial so require.”
Brown, 6 Wn. App. at 256 (quoting Esplin v. Hirschi, 402 F.2d 94, 99 (10th Cir. 1968)).
[92]*92Notwithstanding the rule of liberal interpretation, it is clear that the class action rule does not contemplate automatic affirmance whenever a trial court certifies a class. “Class actions are specialized types of suits, and as a general rule must be brought and maintained in strict conformity with the requirements of CR 23.” DeFunis v. Odegaard, 84 Wn.2d 617, 622, 529 P.2d 438 (1974). The United States Supreme Court, in its most recent decision addressing the analysis to be given a class certification order, acknowledged that class action treatment can save the resources of the courts and the parties by permitting an issue potentially affecting every class member to be litigated in an economical fashion. Falcon, 457 U.S. at 155. The Supreme Court, like the Brown court, recognized that the tentative, modifiable nature of a certification order adds useful flexibility to the class action device. Falcon, 457 U.S. at 160. Nevertheless, the Supreme Court reversed the class certification order in Falcon, stating that “actual, not presumed, conformance with Rule 23(a)” remains indispensable, and that a “rigorous analysis” is called for. Falcon, 457 U.S. at 160-61; E. Tex. Motor Freight Sys., Inc. v. Rodriguez, 431 U.S. 395, 405-06, 97 S. Ct. 1891, 52 L. Ed. 2d. 453 (1977).
In Falcon, a Mexican-American employed as a lineman was refused a promotion. His suit alleged that the company, in keeping with a policy of discrimination, had passed him over because of his national origin. The trial court, without conducting an evidentiary hearing on the allegations, certified an “across the board” class including not only the company’s current Mexican-American employees, but also Mexican-American applicants for employment who had not been hired. The Supreme Court acknowledged its awareness that suits alleging discrimination are often by their very nature class suits. Falcon, 451 U.S. at 157. The court held, however, that Falcon’s complaint “provided an insufficient basis for concluding that the adjudication of his claim of discrimination in promotion would require the decision of any common question concerning the failure of [93]*93petitioner to hire more Mexican-Americans.” Falcon, 457 U.S. at 158.
Falcon is instructive in its emphasis that a class certification in a class suit maintained by private parties should be ordered only after a “rigorous analysis” to ensure that the prerequisites of Rule 23 have been satisfied. Falcon, 457 U.S. at 161. In this case the court’s findings as to prerequisites of CR 23(a) do not reflect a rigorous analysis, in that they do little more than recite the language of the rule. The court rejected the University’s request for an evidentiary hearing on the certification issues, and commented that it “would be difficult, if not impossible, to separate the merits from certification.”4 The court reasoned, “I am not to decide the merits. I am to, in effect, take the substantive evidence as it’s pleaded, unless it is so unreasonable that it can’t be true, or unless there is something directly refuting it.”5
In support of the trial court’s reasoning, the plaintiffs cite the United States Supreme Court’s statement that a trial court should not “conduct a preliminary inquiry into the merits of a suit in order to determine whether it may be maintained as a class action.” Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 177, 94 S. Ct. 2140, 40 L. Ed. 2d 732 (1974). The plaintiffs fail to acknowledge the limited context of this statement. The order on review in Eisen imposed upon a class defendant the cost of notifying members of the plaintiff class, based on the district court’s finding that the plaintiff was “ ‘more than likely5 to prevail.” Eisen, 417 U.S. at 177. It was the district court’s assessment as to which [94]*94side would ultimately prevail that the Supreme Court objected to. Here, the question is whether the requirements of CR 23 have been met. To answer that question, a trial court may certainly look past the pleadings. “Going beyond the pleadings is necessary, as a court must understand the claims, defenses, relevant facts, and applicable substantive law in order to make a meaningful determination of the certification issues.” Castano v. Am. Tobacco Co., 84 F.3d 734, 744 (5th Cir. 1996) (citing Fed. Judicial Ctr., Manual For Complex Litigation § 30.11 (3d ed. 1995).
Rigorous analysis of the commonality issue in the present case requires discussion of the theory of the plaintiffs’ case as well as consideration of the statistical model with which they intend to prove it. The plaintiffs allege discrimination by disparate treatment. Disparate treatment “ ‘is the most easily understood type of discrimination. The employer simply treats some people less favorably than others because of their race, color, religion, sex, or national origin.’ ” Shannon v. Pay ‘N Save Corp., 104 Wn.2d 722, 726, 709 P.2d 799 (1985) (quoting Int’l Bhd. of Teamsters v. United States, 431 U.S. 324, 335 n.15, 97 S. Ct. 1843, 52 L. Ed. 2d 396 (1977)). In a disparate treatment case, as contrasted to a case of disparate impact, the plaintiff has the burden of proving that a discriminatory motive more likely than not motivated the employer’s practices. Shannon, 104 Wn.2d at 727.
The plaintiffs propose to follow the approach to proof in “pattern or practice”6 suits as discussed in Teamsters. In that case, the decision on review found the defendants, a transportation company and a union, had engaged in a “pattern or practice” of discriminating against minorities in hiring so-called line drivers. Statistical data showed that out of 1828 positions, only five had been held by minorities (and none by African-Americans) until after the litigation began. Most minorities who worked for the company held [95]*95lower paying, less desirable jobs. Teamsters, 431 U.S. at 337-38. In addition, there was testimony from individuals who recounted over 40 specific instances of discrimination. Numerous qualified Black and Spanish-surnamed applicants who sought line driving jobs at the company over the years “ ‘either had their requests ignored, were given false or misleading information about requirements, opportunities, and application procedures, or were not considered and hired on the same basis that whites were considered and hired.’ ” Teamsters, 431 U.S. at 338 (quoting District Court opinion).
Once the plaintiffs established this discriminatory pattern or practice in the first or liability stage of the case, the employer became a “proved wrongdoer,” and the litigation went into the second or individual relief stage with the burden upon the employer to provide, in each individual case, “a nondiscriminatory explanation for the apparently discriminatory result.” Teamsters, 431 U.S. at 360 n.46. The Supreme Court approved of this approach to proof in a disparate treatment case because, even if the first stage of the case has not brought forth specific evidence of discrimination against each individual, the proved existence of an intentionally discriminatory pattern or practice creates a greater likelihood that any single employment decision was a component of the overall pattern. Teamsters, 431 U.S. at 359-61.
The plaintiffs in this case envision a first stage in which they would prove, using a statistical regression model, that compensation for male members of the tenured faculty tended on average to be higher than for their female counterparts.7 The University’s awareness of the statistically-modeled disparity would prove the existence of a policy of intentional discrimination against women. Having proved the University to be a wrongdoer in stage one, the [96]*96class would become entitled to class-wide prospective and injunctive relief and salary adjustments. In stage two, the University would bear the burden in proceedings as yet undefined to show that each past individual pay decision affecting a female faculty member was not the product of the discriminatory policy.8
Teamsters was not a case based on statistics alone. There was also testimony from 40 individuals whose detailed accounts of their personal encounters with discrimination in the company “brought the cold numbers convincingly to life.” Teamsters, 431 U.S. at 339. Moreover, the statistical proof was not particularly complex. The “fine tuning of the statistics could not have obscured the glaring absence of minority line drivers.” Teamsters, 431 U.S. at 342 n.23.
The “ ‘inherently slippery nature’ ”9 of statistical evidence has caused many courts to be cautious about allowing the necessary inference of a discriminatory motive to arise solely from disparities shown by a statistical model. Thus, our Supreme Court has said that an individual plaintiff can support a disparate treatment case with statistics, but cannot establish a prima facie case of disparate treatment with statistics. Shannon, 104 Wn.2d at 735. If a plaintiff presents only the circumstantial evidence of statistics in lieu of all other proof of discriminatory intent, the case ordinarily becomes a case of disparate impact, not disparate treatment. See Spaulding v. Univ. of Wash., 740 F.2d 686, 703 (9th Cir. 1984), overruled on other grounds by Atonio v. [97]*97Wards Cove Packing Co., 810 F.2d 1477 (9th Cir. 1987); Am. Fed’n of State, County & Mun. Employees (AFSCME) v. State, 770 F.2d 1401, 1407 (9th Cir. 1985).
The statistical regression model used by both sides in this case was created in the course of a salary study performed by a faculty advisory committee in response to a request by the University president in 1996. The purpose of the study was to determine whether individual faculty members were being inappropriately paid based on race or sex rather than factors related to performance.10 The model examined the salaries of 1,820 full-time faculty, of whom about 25 percent were female. The model predicted what each faculty member should be paid, after using regression analysis to adjust for differences produced by certain factors thought to be nondiscriminatory. The model then compared predicted salaries to actual salaries. The study found what it described as small but statistically significant average salary differentials — for the sample of Caucasian females, $271 less per month than for their male counterparts, and a greater disparity for Asian females. An earlier draft of the study characterized the disparity as female faculty earning, on average, 96.5 cents for each dollar paid to the male faculty. 11
An update of the study in 1998 analyzed the effects of salary increases implemented in July of 1997. Instructions for the salary increase had stressed that salary inequities should receive a high priority.12 The updated study found that a statistically significant disparity still existed for Caucasian female faculty, but not for Asian female faculty.13
The primary problem with attempting to draw from this data an inference that the University intentionally promotes or condones a policy of discrimination against women [98]*98is that the very factors which all sides agree should be weighed most heavily in determining salary — performance and productivity — are also the factors least easily incorporated into a statistical regression model. The study found that up to three-quarters of the original large variance between male and female faculty compensation could be eliminated by adjusting for objective variables such as highest degree earned, years of experience, and market factors that make salaries in some fields higher than others. The analysis then attempted to incorporate into the model such indicators of performance as teaching awards, salary adjustments in response to competitive offers, and research money. The study committee, however, recognized serious limitations with this approach. For example, there was no data on the quality and quantity of publications by individual faculty members, and no way to define and compare the different forms that productivity takes across disciplines.14 Even objectively measurable factors such as a particular degree, years of experience, research grants, or rank, are not given the same weight in each academic discipline. According to an expert witness for the University, there was less than a one-percent probability that all schools in the University valued these variables in a sufficiently similar way so as to justify combining them in a University-wide regression model.15
Pointing out that the study “describes overall trends and not individual salaries,” the study stated, “Review of individual records is necessary to draw conclusions about individual salaries.”16 Thus, the study itself does not furnish glaring evidence of system-wide intentional discrimination in pay practices.17
[99]*99In connection with the present litigation, an expert witness for the plaintiffs, Dr. Mary Gray, used the University’s model to conduct her own regression analysis. She concluded that the actual average disparity between male and female faculty, unaccounted for by the objective factors in the University’s database, was as great as 19 percent— clearly indicating, in her opinion, systematic and institutionalized discrimination. In her declarations, Dr. Gray stated that the University’s study incorporates variables that tend to mask gender bias and minimize the unexplained gap in average salaries.18 Like the University’s study but to a greater degree, Dr. Gray’s analysis shows that for reasons unexplained by the model after it has used regression analysis to screen out quantifiable factors that legitimately affect salary, women are more likely than men to be paid less than the predicted salary for a given position.
In general, the courts will not allow the inference of intentional gender discrimination to be drawn solely from the fact that such statistical disparities are otherwise unexplained. The weight to be accorded to such proof is determined by the existence of independent corroborative evidence of discrimination, such as testimony of specific incidents of discrimination. AFSCME, 770 F.2d at 1407 (citing Teamsters, 431 U.S. at 339-40).19
The plaintiffs offer no testimony of specific incidents of discrimination except for their own experiences of allegedly unfair treatment in the School of Dentistry. Their [100]*100accounts cannot be assumed to represent a common and typical course of conduct across the entire University because evaluations of merit are localized at the departmental level. For example, the School of Dentistry uses a compensation system unique to that school.
The weight of authority indicates that in a large, decentralized university, where departments have great autonomy in personnel decisions, class action treatment for a disparate treatment case is inappropriate. See, e.g., Rosenberg v. Univ. of Cincinnati, 654 F. Supp. 774, 780 (S.D. Ohio 1986); Mich. State Univ. Faculty Ass’n v. Mich. State Univ., 93 F.R.D. 54 (W.D. Mich. 1981). Where personnel decisions are decentralized, plaintiffs who may be able to prove in an individual lawsuit that they have encountered intentional discrimination in their own departments are frequently unable to show that the same discriminatory motive is afoot in the institution as a whole. The fact that numerous individual decisions are made by a large number of department heads and deans means that there are “individually tailored justifications” for the alleged discrimination in the case of each faculty member. Merrill v. S. Methodist Univ., 806 F.2d 600, 608 (5th Cir. 1986).
The University of Washington fits into the model of the decentralization cases. It is one of the nation’s largest research universities. It has 18 separate schools and colleges, including 120 departments or divisions, with campuses in three cities and two teaching hospitals. Each school or college has a dean, who reports to the provost, who in turn reports to the president and the Board of Regents. By law, governance is shared between the Board of Regents and the president and other faculty. RCW 28B.20.200.
The University employs approximately 6,200 faculty members. Each member of the faculty has an individual appointment agreement and an individually negotiated salary. Compensation and promotion decisions are determined under the Faculty Code. The Code defines excellence in terms of teaching, research and service. The Faculty Code recognizes the broad diversity of disciplines to which [101]*101it applies, and provides for a variety of ways to measure faculty scholarship — including:
the quality of their published or creative work; the range and variety of their intellectual interests; the receipt of grants, awards, and fellowships; their success in directing productive work by advanced students and in training graduate and professional students in scholarly methods; their participation and leadership in professional associations and in the editing of professional journals.[20]
The faculty of each department has primary responsibility for deciding the pay and promotion of its personnel in accordance with the Faculty Code, including the allocation of funds for merit increases. Departmental recommendations regarding compensation and promotions are transmitted to the provost. The provost reviews each college’s compensation decisions for compliance with budgetary constraint, and not individual merit. Annually, the provost reviews over 200 promotion and tenure files, and accepts the recommendation from the local faculty level in more than 95 percent of cases. No evidence was presented to show a pattern of adverse treatment of women in the review of departmental decisions by the central administration.
The plaintiffs attempt to avoid the decentralization argument by means of various reports tending to show that the University’s central administration has long recognized unequal treatment of the female faculty to be a concern worthy of study. According to a 1997 report by the President’s Advisory Committee on Women, faculty members often responded to surveys by identifying salary inequity as a crucial issue affecting women faculty.21 Meeting minutes and agendas from the University’s Special Committee on Faculty Women showed that the committee made the issues of promotion and salary disparity the focus of its agenda from 1991 to 1996.22 A draft of the faculty salary study [102]*102recommended that each school review salaries with an awareness of a school-wide problem facing female faculty of all races.23
The plaintiffs argue that it is legitimate to infer from this evidence that decision making about faculty employment at the University, far from being decentralized, is actually controlled from the top down. The trial court accepted their theory of “a top-down decision-making policy at the University” as the basis of the certification order. The court stated, “whether or not the central administration really does make definitive policy regarding pay and promotion for faculty at the University of Washington, there was some belief in the central administration that they could affect policy in those instances.”24
There is an element of bootstrappery in this reasoning. The disparities shown by the statistical model are the core of the plaintiffs’ case. If the statistical model is insufficient by itself to prove a discriminatory motive or policy, then it is also insufficient to impose upon the University a duty to eliminate the disparities shown by the statistical model. Under those circumstances, evidence that top-level decision makers at the University were aware of the statistically-modeled disparities is not proof of a top-down policy of intentional discrimination.
As the Supreme Court observed in regard to the Title VII lawsuit in Falcon, the law “prohibits discriminatory employment practices, not an abstract policy of discrimination.” Falcon, 457 U.S. at 159 n.15. The use of an entirely subjective process in personnel decisions is a practice from which a discriminatory motive can be inferred when it results in the exclusion of a protected class. Falcon, 457 U.S. at 159 n.15; Shannon, 104 Wn.2d at 733 (“discriminatory motivation can be readily inferred when solely subjective employment evaluations result in the exclusion of a protected class”). But here, although they describe the [103]*103University’s central administration as having “orchestrated” a “subjective scheme,” the plaintiffs have not alleged that the University intended to use delegated decision making as a means of discrimination. See Reap v. Cont’l Cas. Co., 199 F.R.D. 536, 544 (D.N.J. 2001). Whether subjective hiring decisions at the departmental level is a discriminatory practice is not one of the common issues identified in the certification order.
The only allegedly discriminatory conduct that we can identify in the trial court’s findings of common issues is the University’s failure to eliminate the statistical disparities. Plaintiffs argue that once the administration became aware of the disparities shown by its own studies, the University had an obligation to raise pay immediately for all women faculty, and its failure to do so is competent proof that the University intentionally discriminates against its women faculty.
In fact, the evidence shows that the University actually did respond to the salary studies by instructing departments to use the 1997 pay raise to address inequities, and by a directive from the President to all deans and department heads to individually review salaries with special attention to the possibility of gender bias.25 Implementing an immediate pay raise for all women faculty is a step that would have brought its own set of intractable problems. As Dr. Gray has recognized, “Many men have salaries below those predicted by the model and might challenge the results of a process that remedies the salaries of their similarly situated female colleagues but leaves their own salaries untouched.”26 See Maitland v. Univ. of Minn., 155 F.3d 1013 (8th Cir. 1998).
[104]*104As Dr. Gray acknowledges, the use of a statistical model to determine individual salaries is “problematical.”27 Because the statistical model does not explain what to do in individual cases, the University’s alleged failure to respond to the salary study by raising the pay of every female professor cannot fairly be categorized as an “employment practice.”
In short, once the plaintiffs’ case moves outside the School of Dentistry, its focus is the abstract and theoretical world of statistics rather than employment practices— decisions made by and about individual human beings. The direction in which the class action is headed is to substitute an official statistical model for the Faculty Code as a guide to the University’s personnel decisions. As Dr. Gray has acknowledged, there is no guarantee that such a remedy would end all unfairness; “all that the procedures described can guarantee is that gender is not the basis of the inequities.”28 But in order to protect the University from further gender discrimination lawsuits, the administration would have to make salary decisions centrally and by formula so as to achieve a salary structure with no gaps unaccounted for in the regression model. Statistical gender parity in compensation would become the primary concern; assessments of each professor’s contributions to teaching, research and service would become secondary. Because performance and productivity are difficult to measure with numbers, they would become less important values.29 The plaintiffs’ allegations do not warrant sending the University down that uninspiring path, given the absence of tangible proof that the University intentionally uses its current decentralized faculty compensation system to discriminate against women.
[105]*105To summarize, the plaintiffs have not shown that a discriminatory motive can be inferred from the University’s disinclination to use the salary study as the determinant for individual decisions about compensation. There is not a common course of conduct that will support certification of a class action. The plaintiffs may proceed with their lawsuits as individual cases. The order certifying the class is reversed.
Coleman, J. and Webster, J. Pro Tern., concur.
Review denied at 147 Wn.2d 1018 (2002).