Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
Concurring opinion filed by Circuit Judge WALD.
STEPHEN F. WILLIAMS, Circuit Judge:
This case deals with a so-called “retrospective rating” insurance policy, evidently a type common for workers’ compensation. The insured employer pays an estimated premium, which is typically based on data about the insured’s payroll and the classifications of its employees for risk purposes, and which is subject to later correction. The classifications of course vary radically by activity; here, for example, the initial rate per $100 of payroll for “athletic team or park — contact sports” — the policy was issued to the owner of the Washington Redskins — was nearly 40 times the premium for “clerical office employees.” The rates, fixed by law, also vary markedly according to the jurisdiction where employees may routinely seek compensation, depending on benefit levels and likelihood of recovery in the jurisdiction. They are, for example, far higher in the District of Columbia than in Virginia — more than double in this case. Both jurisdictions allow recovery by an employee who is injured within their respective borders or whose employment is principally located there.
The policy calls for initial payment of an estimated premium, to be followed by adjustments to reflect actual circumstances. Here, during the third of three successive one-year policies, the District of Columbia Court of Appeals affirmed the decision of the D.C. Department of Employment Services that the players performed “the principal services for which they were hired” in the District, where they played their home games (R.F.K. Stadium), rather than in Redskins Park in Virginia, where they spend a majority of their time at practice. Pro-Football, Inc. v. D.C. Department of Employment Services, 588 A.2d 275 (D.C.1991) (“Anderson” — so-called after one of the player claimants). The players were thus entitled to invoke D.C. law as a basis for recovery for injuries occurring anywhere (as the players had sought throughout the period of the three policies).
[1113]*1113The parties agree that the provisions on premium adjustment allow the insurer to make a retrospective premium change to reflect changes in the employer’s payroll or in the job classifications of particular employees. The question is whether the terms of the policy also permit a premium adjustment for a change in the jurisdiction whose law is available to employees, such as resulted from the Anderson decision. The district court read the policy as denying the insurer such a power. We reverse.
Hartford Accident and Indemnity Company provided the Washington Redskins’ owner/operator, Pro-Football, Inc. (“PFI”), with a workers’ compensation insurance policy for three successive annual policy periods, from July 14, 1988 to July 14, 1991. The policy was a standard form, assigned risk policy administered by the National Council on Compensation Insurance (“NCCI”), designed for employers like PFI who cannot purchase coverage on the voluntary workers’ compensation market and who cannot or are not willing to self-insure. See generally 9 Arthur Larson & Lex K. Larson, Larson’s Workers’ Compensation Law §§ 92.53-92.65 (1997) (describing assigned risk, retrospective rating policies). Under the policy, PFI’s premiums were initially based on the parties’ use of Virginia as the expressly assumed principal location of the players’ services.
Under both the District of Columbia and the Virginia workers’ compensation insurance plans (“WCIPs”), the NCCI directs insurers in the state pool to issue coverage to employers eligible for assigned risk insurance. When PFI submitted an application for Virginia assigned risk coverage, NCCI assigned the application to Hartford, which was obligated to issue a policy. The policy issued by Hartford consisted of manuals (by reference), standard forms, an Information Page (actually several pages) of figures specific to PFI, and rates; the forms and manuals were approved, and the rates set, by NCCI. (The policy forms for the District and Virginia WCIPs are identical in all material respects.) The terms and rates for the policy were not negotiated; neither Hartford nor PFI could legally have altered them.
Retrospective rating plans of the sort embodied in this policy are used when the size of the insured’s risks is difficult to measure at the beginning of the policy period, see Lee R. Russ & Thomas F. Segalla, 5 Couch on Insurance § 69.15 (3d ed.1996) (“Couch”). Courts routinely enforce the retrospective provisions in such plans. See, e.g., L.C. Worley v. Travelers Indemnity Co., 124 Ga.App. 64, 183 S.E.2d 91 (1971); Great American Ins. Co. v. Nova-Frost, Inc., 362 N.W.2d 358 (Minn.Ct.App.1985); Texas Soap Mfg. Co. v. American Auto. Ins. Co., 227 S.W.2d 376 (Tex.Civ.App.1950). Workers’ compensation in general, and professional football in particular, present the kind of uncertainty that makes retrospective rating appropriate, because the insured’s activities and the size of its payroll are likely to vary considerably over the course of the policy term.
Premiums under the policy are calculated as the product of the work classification rate for a specific jurisdiction and the amount of payroll allocated to employees in that classification and jurisdiction (the “premium basis”) up to a regulated maximum amount. Initial premiums (at least for years other than the first one) also incorporate an “experience modification factor” (or “mod”), a prospective adjustment to take account of prior years’ claims experience for the particular employer. All the factors other than actual payroll — the rates, classifications, premium basis maxima, and mod' — are set by NCCI.
Following attempts by several injured Redskin players to collect the higher District benefits for injuries received outside the District, the Director of the District of Columbia Department of Employment Services ruled on or about July 10, 1989, just before the end of the first policy year, that the players’ place of principal employment was the District rather than Virginia. The D.C. Court of Appeals issued its decision in Anderson, affirming the Director, in March 1991.
Hartford, relying upon the policy provision that allowed calculation of the “final premium” after the policy’s expiration, then wrote PFI that it had reclassified Redskin players and coaches as District of Columbia employees for all three policy years. For the then-[1114]*1114current policy year, 1990-91, Hartford issued an endorsement, or formal amendment, implementing the reclassification. Hartford then billed PFI for the difference in premium levels — a difference, after adjustments, of $5,350,762. PFI refused to pay, and Hartford filed suit under the federal courts’ diversity jurisdiction. PFI counterclaimed for breach of contract and fiduciary duty, fraud, bad faith, and negligent misrepresentation.
On cross-motions for summary judgment the district court granted summary judgment for PFI, ruling that the policy did not permit Hartford to change the jurisdictional basis of the premium calculation retroactively. Hartford Accident & Indemnity Co. v. Pro-Football, Inc., d/b/a The Washington Redskins, No. 94-2266 (D.D.C. Aug. 21, 1996) (“Mem. Op.”). The district court also suggested that any reclassification might have to be by formal endorsement, which Hartford had failed to issue except for the final policy year at issue. Finally, the district court struck the affidavit of a Hartford expert because it was submitted after the close of discovery, and rejected PFI’s counterclaims as time-barred under the District’s statute of limitations. All of these decisions are on appeal.
I.
The district court noted, and the parties agree, that the task in contract interpretation is to decide “what a reasonable person in the position of the parties would believe the language meant.” Mem. Op. at 8 (citing 1010 Potomac Assoc, v. Grocery Mfrs. of America, Inc., 485 A.2d 199, 205 (D.C.1984)). In the particular context of insurance, under District of Columbia law,
[sjince insurance contracts are written exclusively by insurers, courts generally interpret any ambiguous provisions in a manner consistent with the reasonable expectations of the purchaser of the policy. However, when such contracts aré clear and unambiguous, they will be enforced by the courts as written, so long as they do not ‘violate a statute or public policy.’
Smalls v. State Farm Mut. Auto. Ins. Co., 678 A.2d 32, 35 (D.C.1996) (citations omitted). See also GEICO v. Fetisoff, 958 F.2d 1137, 1141 (D.C.Cir.1992) (“Under District of Columbia law, ‘[cjlear and unambiguous language [in an insurance policy] should be construed according to its everyday meaning.’ ”) (quoting Continental Casualty Co. v. Cole, 809 F.2d 891, 896 (D.C.Cir.1987)). Thus, as we understand District law, no preference for the insured’s reading arises unless the contract is ambiguous, and even then the preference involves no more than accepting what the insured might reasonably believe over an alternative reasonable interpretation offered by the insurer.
The parties also wrestle with the issue of whether the special status of this insurance contract — imposed on the parties as NCCI’s standard form' — calls for any modification of the standard rule for interpreting insurance contracts. Commentators have urged, and many (perhaps most) jurisdictions have agreed, that since the basis for the standard rule — the insurer drafted the contract without negotiating it with the insured — is absent, the presumption in favor of the insured’s reasonable interpretation should be relaxed: “The rule of construction against the insurer does not apply where ... the language was prescribed by statute and controlled by Division of Insurance rather than the individual insurer.” 2 Couch § 22:15. See also John A. Appleman & Jean Appleman, 13 Appleman, Ins. L. & P. § 7407 (1981) (“While North Carolina and a few other states still apply the doctrine of construing policies against the insurer to instances of standard policies, the better authority is to the effect that the doctrine of liberal construction has no place under those circumstances.”).1 There appears to be no District of Columbia precedent on the construction of state-mandated policy language or forms. In any event, because we do not find the language ambiguous, or as reasonably permitting PFI’s proposed interpretation, we need not resolve that contest.
[1115]*1115We now turn to interpreting the policy itself. (The text of the key printed provisions of the policy — the “boilerplate” — is set forth in the Appendix.)
Did the Final Premium Clause entitle Hartford to adjust the premium for jurisdictional change?
The Final Premium Clause explicitly describes the premium shown on the Information Page as “an estimate” and states that the “final premium will be determined after this policy ends by using the actual, not the estimated, premium basis and the proper classification and rates that lawfully apply to the business and work covered by this policy.” Hartford’s decision to charge District rates for the players and coaches in 1991 in the wake of Anderson tracks these terms. Thereafter, only the D.C. classifications and rates would qualify as the “proper classification and rates that lawfully apply” to the affected employees. The logic of any insurance policy, prospective or retrospective, requires premium levels to track expected benefits.2 A retrospective-rating policy allows post-signature events to affect both sides of the balance. Here Hartford accounted for a large, legally mandated, rise in benefit levels through an appropriate adjustment to its premium calculation, as the policy contemplated.
The district court came to a contrary conclusion — here urged on us by PFI — by treating the Final Premium Clause as limited by the Classifications Clause. It read the latter as contemplating only changes to business and work classifications (and the rate and premium basis changes that follow a classification change). It then read the Final Premium Clause as permitting Hartford to make only two types of independent changes to the policy factors: changes to premium. basis (payroll) determined under the Audit Clause, and changes to work classifications determined under the Classifications Clause; rate adjustments would be restricted to the consequences of changes in these two types of factors. Mem. Op. at 9-11. This interpretation left no room in the Final Premium Clause for a change in rates corresponding to a change in the workers’ jurisdictional basis, whether the change arose from an exogenous legal event such as the Anderson decision, or from changes in actual work location (or new information about work location) such as might be revealed during the audit process.
This reading sacrifices the fundamental principles of the policy to symmetry — and a false symmetry at that. The Final Premium Clause calls comprehensively for retrospective adjustment of the premium to conform to “the actual ... premium basis and the proper classification and rates that lawfully apply.” Here the D.C. Court of Appeals reached a conclusive (and retroactive) determination that the jurisdictional predicate of the players’ benefit eligibility was not, in fact, Virginia, but rather the District of Columbia. Hartford correctly concluded that the change in the jurisdictional predicate for benefits entailed a matching change in the “actual” rates. PFI’s reading, by contrast, rewrites the phrase “proper classification and rates” to eliminate any independent meaning for “rates.”
We see no basis for the assumption that the Audit Clause — -authorizing inspection of the insured’s books for up to three years after the policy period for information to “be used to determine final premium” — must refer solely to changes in premium basis (i.e., payroll limited by the NCCI-imposed maxi-ma). Mem. Op. at 10. An audit might reveal, as Hartford’s apparently did, see Appendix for PFI on Cross-Appeal at 106, that all claims by players are being filed in another jurisdiction. Since it is the function of the Audit Clause to gather data for the correct calculation of the premium, it would be arbitrary to perform the final premium calculation in disregard of a highly relevant fact so gathered, simply because the scope of the Classifications Clause is limited to changes in classifications.
In fact, the Information Page states, “All information required below is subject to verification and change by audit.” The information set forth “below” included rates calculated for various job classifications on a [1116]*1116jurisdiction-specific basis, so that the natural, indeed inescapable, reading is that the provision for change applies to the jurisdictional as well as the other aspects of the premium calculation. By contrast, on a narrow reading of the Audit Clause’s scope, it remains mysterious how information relevant to the Classifications Clause will be gathered except through the insured’s good-faith disclosure. But under a harmonious reading of the three clauses, the Audit Clause has potential to generate data for all aspects of the Final Premium calculation, as well as for reclassifications under the Classifications Clause.
Granted, our construction of these clauses makes partially redundant the separate clause governing ordinary, nonjurisdictional, changes to work classifications. Why should such changes be addressed not only in the Audit and Final Premium Clauses, but also in a special clause of their own, the Classifications Clause? One plausible answer is that changes sought by the insurer to work classifications may be likely to engender the greatest resistance by the insured, and may be the most disputable, so the function of the special Classifications Clause is just to underscore the insurer’s power to make those changes. In any event, PFI’s reading of the clause to suggest that it puts jurisdictional changes entirely off the map presents a far greater anomaly, for under PFI’s reading the insurer would be unable to adjust for a change in one of the most significant elements in its exposure.
PFI appears to believe that somewhere in the policy Hartford promised that the rates for players would be calculated on the assumption that Virginia was their principal place of employment, regardless of whatever might prove the case. It is not clear just where PFI finds any such promise. One possibility is the page of the Information Page that covers the players (as well as the clerical staff working at Herndon and the athletic team operational staff), which says, as a heading: “Named Insured and Location Address of Operations Covered by this Schedule,” together with the address of the Herndon facility. (A similar page covers those understood to be working in the District.)
Far from representing any promise, however, the heading appears to be no more than a device for organizing information, framing the then-estimated classifications, rates and payroll for the workers thought to be located at the places shown on the particular pages. Furthermore, as Hartford rightly points out, PFI’s restrictive reading would illogically bind the insurer to the insured’s declaration of address despite misrepresentation, a simple out-of-state move during the policy year, or an honest error. PFI offers no construction of the policy that applies the retroactive adjustment clause to any of those scenarios. Nor does it explain why the parties would limit the insurer to such alternative remedies as a fraud claim (which would encompass only some misrepresentation eases) or a rise in the mod (which would be prospective only).
Courts in other jurisdictions, faced with similar or identical policy language, have held that the retrospective rate-changing power encompassed erroneous jurisdictional assignments. In D.A.X., Inc. v. Employers Ins. of Wausau, 659 N.E.2d 1150 (Ind.Ct.App.1996), the insurer discovered that the policy’s jurisdictional assumption was wrong. The insured, an employee-leasing company, had declared that its operations were entirely in Indiana, because of its office location. Because the insurer found that claims were being filed in Illinois, it launched an audit under the standard Audit Clause, concluded that under rules promulgated by NCCI the correct site was Illinois, and accordingly adjusted the premium. The trial court enforced the insurer’s claim to the higher premium. On appeal the insured claimed that the trial court judgment had wrought an unconstitutional “impairment” of contract and an unjustified contract reformation. The appellate court rejected both theories, finding that the trial court “simply enforced the contract the way it was written.” Id. at 1156. The court read the Final Premium Clause (identical to the clause at issue here) as permitting the insurer to use the “rates and classifications from such other states as may lawfully apply,” which, given the NCCI rating rule, meant the Illinois rates. Id. at 1157.
[1117]*1117And in a New Hampshire case also involving a retrospective rating policy, with language similar though not identical to ours, the insurer discovered during its post-policy audit that many of the insured’s employees, classified as working in New Hampshire, had in fact been working in Massachusetts, and accordingly recalculated the premium. Continental Ins. Co. v. Seppala & Aho Constr. Co., Inc., 121 N.H. 374, 430 A.2d 157 (1981). A clause — the counterpart of our Final Premium Clause — -stated that “the earned premium,” as opposed to the initial “estimated premium,” “shall be computed in accordance with the rule, rates, rating plans, premiums and minimum premiums applicable to this insurance.” Id. 430 A.2d at 158. The court held that this clause clearly and unambiguously gave the insurer the power to re-rate the policy with a corrected jurisdictional classification — indeed, to the point of making clear that the insurer’s agent had no authority to represent that only New Hampshire rates would be charged. Id. at 159. We agree with these decisions.
Must Hartford’s power to correct the premium be exercised through an endorsement?
There are two provisions that might give rise to an obligation by Hartford to exercise its premium-revising power by means of an endorsement. First, PFI relies on Subsection A of the “General Section,” which provides that the “terms of this policy may not be changed or waived except by endorsement issued by us to be part of this policy.”
But we do not see the insurer’s exercise of the power expressly granted by the Final Premium Clause as the “change” of a term of the Policy: it is an exercise of a power that a specific policy term — the Final Premium Clause — expressly grants the insurer to use “the actual, not the estimated, premium basis and the proper classification and rates that lawfully apply to the business and work covered by this policy” (emphasis added). That, of course, was the view of the courts in D.A.X. and Continental Insurance. Moreover, the Final Premium Clause expressly says that the final premium “will be determined after this policy ends.” It seems odd to say that an activity explicitly slated to occur after the policy “ends” constitutes a “change” of the then-expired policy’s terms. (PFI’s theory, incidentally, further entails— implausibly, given the Audit Clause — that application of the Final Premium Clause to correct premium basis would also require an endorsement.) Because PFI’s reading of Subsection A fails to accord independent force to the Final Premium Clause, it fails to give “a reasonable, lawful, and effective meaning to all [the policy’s] terms.” 1010 Potomac Assoc., 485 A.2d at 205.
Moreover, PFI offers no rationale whatsoever for applying the General Section endorsement requirement to adjustments under the Final Premium Clause. Hartford, by contrast, suggests that endorsements are issued only during a policy’s term, and that the purpose of the requirement is to provide notice to the insured to enable it to decide either to cancel or at least not to renew the policy. We do not feel qualified to embrace that view, but it is, at least, a plausible one.
We can imagine two ways that a post-expiration endorsement requirement might possibly have some purpose beyond that contributed by the final billing itself.3 First, an endorsement requirement might make it easier for insurance regulators to monitor assigned risk policies in their jurisdictions, as is apparently the regulators’ charge under the D.C. WCIP. See D.C. WCIP at 9 (“The Plan Administrator shall monitor and review servicing carrier performance by ... (3) conducting on-site audits; and (4) reviewing any other information that relates to the servicing carrier.”). Both D.C. and Virginia require that copies of the policy declarations “and all endorsements” must be filed with the appropriate administrator, see D.C. WCIP at 7, Virginia WCIP at 2.
But if regulators want information on ultimate premiums paid, a requirement of endorsements for jurisdictional changes under the Final Premium Clause is a strangely incomplete solution. We know that the final bill will reflect changes to estimated payroll, but even PFI does not contend that a post-[1118]*1118expiration, ordinary correction of the payroll constitutes a “change” to the policy, such as to warrant formal endorsement. Either the regulators do not depend on a formal endorsement process for their information-gathering needs, or they don’t monitor the policies as closely as this suggestion assumes.
A second possible justification for requiring post-policy jurisdictional adjustments to take the form of endorsements might look to the interests of employee-claimants. If the jurisdictional possibilities open to claimants were governed by the insurance contract, specification of jurisdictional change in an endorsement might make it easier for claimants to ascertain their appropriate jurisdictional bases. Of course, the premise is faulty. Claimants’ jurisdictional options depend on state law, as Anderson shows, not on arrangements between the employer and its insurer.
Thus an endorsement attached to a final premium billing, issued after the expiration of a policy, would perform no notice function beyond that of the billing itself. It would be a purely formal gesture.
Unsurprisingly, PFI has shown no reason to think that it was harmed by Hartford’s non-issuance of an endorsement when it made its final billing for the expired terms.4 By contrast, Hartford’s behavior in issuing an endorsement only to the current 1990-91 policy year is consistent with the rationale for the requirement: it provided PFI with notice in ease it should wish to find other insurance or self-insure in the future (as apparently it chose to do).
An alternative source of an endorsement requirement would be the Classifications Clause’s statement that if exposures are different from those described by the estimated classifications, “we will assign proper classifications, rates and premium basis by endorsement to this policy” (emphasis added). Since PFI’s position rests heavily on a notion that the Classifications Clause does not cover jurisdictional change, and Hartford agrees on that point, it seems plain that it cannot be a source of any endorsement obligation that would be applicable here.
Accordingly, we reverse the grant of summary judgment in favor of PFI and remand for the district court to enter summary judgment for Hartford.
II.
The district court rejected PFI’s counterclaims as time-barred by the District of Columbia statute of limitations. Since PFI’s claims arise under District law, the applicable statute of limitations is also that of the District. See Guaranty Trust Co. v. York, 326 U.S. 99, 65 S.Ct. 1464, 89 L.Ed. 2079 (1945); Kuwait Airways Corp. v. American Security Bank, N.A., 890 F.2d 456, 460 (D.C.Cir.1989). This is true even though Rule 12(a)(4) of the Federal Rules of Civil Procedure extends the time for a responsive pleading to ten days after the denial of a motion under Rule 12, and PFI advanced its counterclaim in an answer timely filed under that rule. See 6 Charles Alan Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice & Procedure 1419 (2d ed.1990, Supp.1997).
An agreement between the parties tolled .the statute until October 22, 1994; PFI’s answer and counterclaim were filed a year later. But Hartford filed its complaint on October 20, 1994, and defendant argues that this further tolled the statute. While there is considerable law (and sensible policy) on PFI’s side, the controlling law — that of the District — is against it.
In Sears, Roebuck and Co. v. Goudie, 290 A.2d 826 (D.C.1972), the court considered whether a counterclaim filed too late (assuming no tolling) could be saved by the doctrine of relation back to a prior, timely counterclaim that had been dismissed on substantive grounds. The court found no relation back because the new counterclaim was insufficiently related to the dismissed one. Id. at 830. A necessary premise of the entire dis[1119]*1119cussion was that the filing of the complaint did not toll the statute on the counterclaim. And the court made this premise specific, saying that a genuine counterclaim, i.e., one going “beyond matters of defense” such as recoupment, “must be viewed as an affirmative cause of action and should therefore be tested apart from the primary claim in determining whether the statute of limitations would bar the counterclaim.” Id. (emphasis added). This superseded the contrary, more lenient, rule of De Vito v. Hoffman, 199 F.2d 468 (D.C.Cir.1952). Because PFI’s counterclaims concededly go “beyond matters of defense,” they must be assessed separately, and thus fail.
III.
In the course of its ruling on the cross motions for summary judgment, the court held that the parties’ failure to observe discovery rules — specifically failure to request extension of a discovery deadline — barred use of experts’ affidavits submitted to bear upon the construction of the policy. Mem. Op. at 12 n.10. Hartford contests the exclusion of its expert’s affidavit.
Hartford’s contention appears moot — at least in the sense that, as we have construed the portions of the policy relevant to this appeal without consideration of any extrinsic evidence, such evidence can make no contribution to resolution of the case. Hartford suggests, however, that we should address the court’s discovery ruling because, even if it wins on the basic issue of its power to adjust the premium by reference to jurisdictional mistake, as it has, “[djamages would still need to be addressed.”
At the time the district court exercised its discretion in ruling on this discovery issue it was simultaneously making summary judgment rulings that brought the case to a complete end (subject, of course, to appellate review). We do not know if that circumstance colored the court’s ruling on the issue. Rather than entangle ourselves in what is surely a messy issue, and perhaps a completely unnecessary one, we simply note that if indeed there are remaining factual issues to be resolved on remand, the court may wish to re-examine its discovery ruling.
Accordingly, the judgment of the district court is reversed in part and affirmed in part, as stated above.
So ordered.
APPENDIX
General Section
A.The Policy
This policy includes at its effective date the Information Page and all endorsements and schedules listed there____ The only agreements relating to this insurance are stated in' this policy. The terms of this policy may not be changed or waived except by endorsement issued by us to be part of this policy.
Part Five — Premium
A. Our Manuals
All premium for this policy will be determined by our manual of rules, rates, rating plans and classification. We may change our manuals and apply the changes to this policy if authorized by law or a governmental agency regulating this insurance.
B. Classifications
Item 4 of the Information Page shows the rate and premium basis for certain business or work classifications. These classifications were assigned based on an estimate of the exposures you would have during the policy period. If your actual exposures are not properly described by those classifications, we will assign proper classifications, rates and premium basis by endorsement to this policy.
C. Remuneration
Premium for each work classification is determined by multiplying a rate times a premium basis....
E. Final Premium
The Premium shown on the Information Page, schedules, and endorsements is an estimate. The final premium will be determined after this policy ends by using the actual, not the estimated, premium basis and the proper classification and rates that lawfully apply to the business and work covered by this policy. If the final premium is more than the premi[1120]*1120um you paid to us, you must pay us the balance....
G. Audit
You will let us examine and audit all your records that relate to this policy. These records include ledgers, journals, registers, vouchers, contracts, tax reports, payroll and disbursement records, and programs for storing and retrieving data. We may conduct the audits during regular business hours during the policy period and within three .years after the policy period ends. Information developed by audit will be used to determine final premium....