James P. COTTON, Jr., Gerald Eickhoff, Plaintiffs-Appellees Cross-Appellants, v. MASSACHUSETTS MUTUAL LIFE INSURANCE COMPANY, a Massachusetts Corporation, Defendant-Appellant Cross-Appellee, 402 F.3d 1267

TJOFLAT, Circuit Judge: Defendant Massachusetts Mutual Life Insurance Co. appeals the judgment of the district court in favor of the plaintiffs, James Cotton and Gerald Eickhoff, on their claim for breach of fiduciary duty under the Employee Retirement Income Security Act (ERISA), 29 U.S.C. §§ 1001 et seq. The district court entered judgment for the plaintiffs after striking Mass Mutual’s answer and entering a default as a sanction for discovery violations.

On appeal, Mass Mutual makes three arguments: first, that the entry of a default was an abuse of discretion; second, that the well-pleaded factual allegations in the amended complaint fail to establish liability under ERISA; and, third, that the district court impermissibly awarded individualized, extra-contractual damages that are not available under ERISA. We agree with Mass Mutual that the amended complaint fails to establish liability under ERISA. Accordingly, we reverse and remand with instructions that the plaintiffs’ ERISA claims be dismissed with prejudice. As we explain in

Part II, infra, the plaintiffs are unable to establish liability under ERISA because Mass Mutual simply is not a fiduciary for any purpose related to the misconduct they allege. Indeed, as we explain in

Part III, infra, our review leads us to conclude that this was never really an ERISA case at all, that it never should have been litigated in federal court, and that the plaintiffs’ motion to remand should have been granted at the outset. In

Part IV, infra, we briefly address the other discovery-related sanctions that the district court imposed in its order entering a default. I. After entering a default, the district court adopted as true the well-pleaded factual allegations contained in the amended complaint. Our statement of the case is, in turn, drawn primarily from the district court’s

summary of these allegations. Cotton and Eickhoff were executive officers of BEI Holdings, Inc. (now known as AMERSCO, Inc.). In 1982, Cotton and Eickhoff entered into a “Wealth-Op Deferred Compensation Agreement” with BEI. Under the agreement, BEI agreed to pay Cotton and Eickhoff, or their beneficiaries, $250,000 annually for fifteen years beginning at age 65 or upon their death.

In December 1986, Mass Mutual agents Ronald Hilliard and Gary Martin proposed arrangements between Cotton and BEI and between Eickhoff and BEI whereby BEI and the employee would split premiums and death benefit proceeds on a permanent whole life insurance policy issued on the employee. Under the pro-posai, BEI would pay all premiums on each policy. A portion of the premiums would be taxable as compensation to Cotton and Eickhoff, while the remainder would be treated as loans from BEI to Cotton and Eickhoff.

According to the proposal, the cash value of each whole life policy would continue to grow until it would cover the annual premium payments — that is, until the premiums would “vanish.” This was projected to occur after only seven years in Eickhoff s case and only ten years in Cotton’s case. Cotton and Eickhoff are required to repay the portion of the premiums treated as loans at a certain time from the cash value of the policies. This was referred to as the “roll-out.” At the rollout, the cash values of the policies will be reduced by the amount of premiums repaid to the employer.

Based on this proposal, Cotton, Eickhoff, and BEI agreed to establish what the amended complaint and the district court refer to as the “Split-Dollar Employee Welfare Benefit Plan Sponsored by AMRESCO, Inc.” As part of the plan, Mass Mutual issued two whole life insurance policies to Cotton in early 1987. Each had a face amount of $5,715,500 and a stated annual premium of $119,697.

Eickhoff was also issued two policies, one of which was later divided into two, leaving him with three policies, one with a face amount of $5,327,500 and a stated annual premium of $76,426, one with a face amount of $8,350,729 and a stated annual premium of $48,079, and one with a face amount of $1,976,771 and a stated annual premium of $28,376. BEI and its successor, AMRESCO, paid the premiums on the policies.

In 1990, the plan’s primary objective was changed to use the insurance policies as the primary source of retirement income for the plaintiffs and to provide greater cash and payout values by committing BEI to continue to pay premiums for additional periods of time — seventeen years, beginning in 1990, in Cotton’s case, and eighteen years, also beginning in 1990, in Eiekhoffs case.

In return, Cotton and Eickhoff relieved BEI of its obligations under the “Wealth-Op Deferred Compensation Agreement.” The plaintiffs claim that in making these amendments they relied substantially on the policy projections Mass Mutual and its agents provided, as well as the agents’ similar oral representations. These analyses projected significant death benefits and cash surrender values and annual retirement income of $250,000. In 1992, Cotton and Eickhoff took out substantial loans against their policies.

Cotton borrowed $910,000, and Eickhoff borrowed $571,000 — the maximum amounts the two were permitted to borrow against their policies. They allege that they did so in reliance upon representations made by Mass Mutual and its agents that the policies were sufficiently strong and had sufficient value to support the loans while still generating retirement income after the rollout. Cotton and Eick-hoff also changed the policies’ dividend option at this time so that dividends would be used to pay interest and principal on the loans rather than to purchase additional insurance, as had been the case previously.

In February 1996, Alexander & Alexander Benefit Services, another Mass Mutual agent, notified Cotton and Eickhoff that the policy illustrations provided in December 1995 overstated their policies’ cash surrender values and death benefits because they did not take into account the rollout- — i.e., the repayments due AMRES-CO in 2006 in Cotton’s case and in 2007 in Eickhoff s case. The plaintiffs allege that illustrations provided to them in 1986, 1990, 1992, 1993, and 1994 included the same error.

Whereas analyses provided in 1994 and 1995 projected death benefits of several million dollars and annual retirement income of at least $200,000, new anal-yses predict that Cotton would receive only nominal retirement income and death benefits of less than $600,000, and that Eickhoff would receive virtually no retirement income and death benefits of no more than $1,000,000. Moreover, the new analyses project that Cotton and Eickhoff will have to pay post-rollout premiums in order to keep the policies in force. In other words, premiums have not “vanished.” In addition, the plaintiffs allege that Mass Mutual (

a) concealed the existence of “modal penalties” that resulted from their employer paying premiums on a monthly rather than annual basis; (

b) used methods of allocating premium payments between income to the plaintiffs and loans from their employer that were less favorable to the plaintiffs than those used in policy projections; (

c) understated the policies’ sensitivity to changes in interest and dividend rates; and (

d) generally obscured actual policy performance by consistently presenting reports in different and misleading formats. In August 1998, Cotton and Eickhoff filed suit in the Superior Court of Dekalb County, Georgia. The state-court complaint asserted state-law claims of innocent misrepresentation, negligent misrepresentation, fraud, and promissory estoppel and sought equitable relief requiring the defendants to ensure that their policies performed as described in policy analyses.

All defendants — i.e., Mass Mutual, Martin, Hilliard, Alexander & Alexander, and several other alleged Mass Mutual agents— joined in removing the case to federal court on the theory that the plaintiffs’ claims were completely preempted by ERISA. The plaintiffs moved to remand the case to state court, arguing that the claims were not completely preempted because the policies and agreements did not constitute an ERISA plan and because the complaint did not seek relief available under ERISA. On May 1999, the district court denied the plaintiffs’ motion to remand.

It first held that an ERISA plan existed under the test established in Donovan v. Dillingham, 688 F.2d 1367, 1371 (11th Cir.1982) (en banc). Relying primarily on our decision in Engelhardt v. Paul Revere Life Ins. Co., 139 F.3d 1346 (11th Cir.1998), it also held that the complaint sought relief available under ERISA. As such, it determined that the plaintiffs’ state-law claims were completely preempted so that removal based on federal question jurisdiction was permissible. In a subsequent order, the court granted the plaintiffs’ motion to file an amended complaint asserting claims under ERISA.

The plaintiffs’ amended complaint asserted four ERISA claims: a § 502(a)(1)(

B) claim to enforce the terms of the plan, a § 502(a)(2) claim for breach of fiduciary duty, a § 502(a)(3) claim for appropriate equitable relief, and a claim under the ERISA federal common law doctrine of equitable estoppel. The amended complaint also reasserted the state-law claims that comprised the plaintiffs’ original complaint.

The same day the plaintiffs filed the amended complaint, they also filed a motion to compel Mass Mutual to produce a number of items that they claimed Mass Mutual had improperly withheld in response to legitimate discovery requests, namely, the computer software and data used to produce the various policy analyses presented to them and, in general, all documents issued or in effect during the preceding fourteen years or so that in any way related to the sale or promotion of, or policies or practices regarding, any Mass Mutual split-dollar whole life insurance plan or policy and also, more particularly, the specific type of plan sold to them.

In response, Mass Mutual argued that this information was not discoverable because it was relevant only, if at all, to the plaintiffs’ completely preempted — and, therefore, no longer viable — state-law claims. At a hearing on February 11, 2000, the court granted the motion to compel. This order, however, did not put an end to discovery-related disputes, and in July 2000, the plaintiffs moved for sanctions against Mass Mutual under Fed.R.Civ.P. 37(b), claiming that Mass Mutual had failed to comply with the compel order. That same day, they also filed a brief in response to Mass Mutual’s motion for

summary judgment in which they appear to have abandoned all of their state-law claims. In a December 28, 2000 order, the district court concluded that Mass Mutual had “acted in bad faith and abused the discovery process” by “delay[ing] the production of responsive documents, only to produce them subsequently in a piecemeal fashion; failpng] to produce knowledgeable [Fed R. Civ.

P.] 30(b)(6) representatives; and attempting] to conceal a history of vanishing premium litigation.” The court further determined that Mass Mutual had “engaged in a systematic effort to frustrate Plaintiffs legitimate attempts at discovery and ... failed to obey” the earlier compel order.

The court found that these violations had prejudiced the plaintiffs because — despite bringing the dispute before the court on two separate occasions — the plaintiffs had yet to receive the relevant information to which they were entitled and had incurred the trouble and expense of traveling outside the state to depose unqualified Rule 30(b)(6) witnesses. Finally, the court determined that imposing sanctions less significant than a default would not sufficiently deter future litigants from engaging in similar conduct.

As such, it ordered, among other things, that Mass Mutual’s answer be stricken and a default entered against it, and that Mass Mutual pay all reasonable costs and attorney’s fees incurred by the plaintiffs in connection with litigation against Mass Mutual since the February 11 compel order. In May 2001, the plaintiffs settled with all defendants other than Mass Mutual and claims against them were accordingly dismissed.

On October 3, 2001, following a bench trial held to examine the legal sufficiency of the amended complaint’s well-pleaded factual allegations and to address the issue of damages, the district court issued an opinion detailing its findings of fact and conclusions of law.

First, the court rejected the plaintiffs’ claim under ERISA § 502(a)(1)(B), which permits a beneficiary “to recover benefits due to him under the terms of his plan, to enforce his rights under the terms of the plan, and to clarify his rights to future benefits under the terms of the plan.” The court determined that the various policy projections presented to the plaintiffs were neither “plan documents” nor “summary plan descriptions”; rather, the only such documents at issue were the insurance agreements and insurance policies.

And because these documents were clear and unambiguous, there was no plan term that needed to be “clarified” or “enforced.” Therefore, the plaintiffs’ § 502(a)(1)(

B) claim failed as a matter of law. Second, the court addressed the plaintiffs’ claim under ERISA § 502(a)(2) “for appropriate relief’ based on Mass Mutual’s alleged breach of fiduciary duty under ERISA § 409(a), 29 U.S.C. § 1109(a). Accepting the amended complaint’s well-pleaded factual allegations as true, it con-eluded (

a) that the plaintiffs sought relief on behalf of the plan, (

b) that Mass Mutual was a fiduciary with respect to the plan, (

c) that Mass Mutual had breached a fiduciary duty to the plan, and (

d) that such breach had caused a loss to the plan. As such, it concluded that the amended complaint established liability under § 502(a)(2). On the issue of damages, the court ordered Mass Mutual to restore annually the “missing” death benefits and cash surrender values — ie., the difference between those shown in the 1990 illustrations and the current policy values — less any portion of the missing values attributable to (

a) the 1990 changes in the dividend option, (

b) the 1992 loans, or (

c) changes in dividend scales/interest rates. The court reasoned that the adjustment for dividend scale/interest rate changes was necessary because Cotton and Eickhoff “were knowledgeable businessmen who took an active role in monitoring the practices governing the Plan and understood that the interest rates and scales would fluctuate over time.” The parties were directed to draft and file an injunction requiring Mass Mutual to pay damages in accordance with the order. Third, the court addressed the plaintiffs’ claim for relief under ERISA § 502(a)(3), which permits beneficiaries to bring suit “(

A) to enjoin any act or practice which violates any provision of this subchapter or the terms of the plan, or (

B) to obtain other appropriate equitable relief (

i) to redress such violations or (ii) to enforce any provisions of this subchapter or the terms of the plan.” The Supreme Court has described § 502(a)(3) as a “safety net, offering appropriate equitable relief for injuries caused by violations that § 502 does not elsewhere adequately remedy.” Varity Corp. v. Howe, 516 U.S. 489, 512, 116 S.Ct. 1065, 1078, 134 L.Ed.2d 130 (1996). In Katz v. Comprehensive Plan of Group Insurance, 197 F.3d 1084 (11th Cir.1999), we held that a plaintiff who has an available claim under another subsection of § 502(

a) cannot proceed under § 502(a)(3), and, moreover, that “the availability of an adequate remedy under the law ... does not mean, nor does it guarantee, an adjudication in one’s favor.” Id. at 1089. In other words, a plaintiff who can state a cognizable claim under either (a)(1)(

b) or (a)(2) may not also rely on the (a)(3) “safety net” even if he is ultimately unable to prevail under both (a)(1)(

b) and (a)(2). Relying on Katz, the district court held that the plaintiffs’ (a)(3) claim was barred as a matter of law. Finally, the district court addressed the plaintiffs’ federal common law equitable estoppel claim. “This circuit has created a very narrow common law doctrine under ERISA for equitable estoppel. It is only available when (1) the provisions of the plan at issue are ambiguous, and (2) representations are made which constitute an oral

interpretation of the ambiguity.” Id. at 1090 (citations omitted). As it had already determined that the terms contained in the plan document were not ambiguous, the district court held that the plaintiffs were not entitled to recover under this doctrine. Although the court’s order included what it thought were “straightforward” instructions regarding the calculation of damages, the parties could not agree as to how the order would be implemented.

Thus, after the plaintiffs filed a proposed injunction and Mass Mutual filed extensive objections to it, the parties ended up back before the court for another hearing on the issue of damages. Shortly thereafter, the court issued a short order requiring Mass Mutual to restore to the plan the rollout amounts — ie., the amounts the plaintiffs’ employer is due to be repaid under the insurance agreements. This appeal followed. II.

To establish liability for a breach of fiduciary duty under any of the provisions of ERISA § 502(a), a plaintiff must first show that the defendant is in fact a fiduciary with respect to the plan. Baker v. Big Star Div. of the Grand Union Co., 893 F.2d 288, 289 (11th Cir.1989) (“non-fiduciaries cannot be held liable under ERISA”). ERISA provides that a person is a fiduciary with respect to the plan to the extent (

i) he exercises any discretionary authority or discretionary control respecting management of such plan or exercises any authority or control respecting management or disposition of its assets, (ii) he renders investment advice for a fee ór1 other compensation, direct or indirect, with respect to any moneys or other property of such plan, or has any authority or responsibility to do so, or (iii) he has any discretionary authority or discretionary responsibility in the administration of such plan. ERISA § 3(21)(A), 29 ' U.S.C. § 1002(21)(A).

Significantly, under this definition, a party is a fiduciary only “to the extent” that it performs a fiduciary function. As such, fiduciary status under ERISA is not an “all-or-nothing concept,” and “a court must ask whether a person is a fiduciary with respect to the particular activity at issue.” Coleman v. Nationwide Ins. Co., 969 F.2d 54, 61 (4th Cir.1992).

Here, although “Mass Mutual has conceded that it is a fiduciary for the limited purpose of making death benefit determinations under the policies,” it maintains that the amended complaint fails to establish that it is “a fiduciary for any other purpose” (brief p. 43). The question whether a party is an ERISA fiduciary is a mixed question of law and fact. In this case, because the district court accepted as true all of the amended complaint’s well-pleaded factual allegations, there are no “findings of fact” in the ordinary sense.

Our review of the district court’s legal conclusions is de novo, and we accord no deference to its legal analysis. Reich v. Lancaster, 55 F.3d 1034, 1045 (5th Cir.1995).

The district court concluded that the plaintiffs had sufficiently alleged that Mass Mutual was wearing its fiduciary hat when the alleged misconduct occurred: The factual allegations in the Complaint clearly state that Mass Mutual and its agents helped establish the Plan and “administered, managed and controlled [the insurance policies governing the Plan] since they were first issued.” Mass Mutual retained “discretionary authority to administer[,] manage and control” policy assets. It also exercised discretionary authority with respect to plan management and administration.

Finally, Mass Mutual and its agents provided policy projections upon which Plaintiffs relied. D. Ct. Op. at ¶ 41 (quoting First Am. Compl.) (citations omitted). As a result of the discovery sanctions the district court imposed, it accepted all of the well-pleaded factual allegations in the amended complaint as true. But even assuming for the purpose of this appeal that the district court’s sanction was appropriate, the amended complaint does not state a set of facts under which Mass Mutual is a fiduciary for the purposes of this case. Therefore, we must reverse.

The passage from the district court’s opinion reprinted above summarizes the plaintiffs’ basic allegations regarding Mass Mutual’s fiduciary status. In other words, the amended complaint simply recites portions of the statutory definition. The plaintiffs argue that this is enough to establish Mass Mutual’s fiduciary status because the factual allegations of their admitted complaint are deemed admitted. But while a defaulted defendant is deemed to “admit[] the plaintiffs well-pleaded allegations of fact,” he “is not held to admit facts that are not well-pleaded or to admit conclusions of law.” Nishimatsu Constr.

Co. v. Houston Nat'l Bank, 515 F.2d 1200, 1206 (5th Cir.1975). Therefore — again, assuming for the sake of the argument that a default was an appropriate sanction' — although Mass Mutual “may not challenge the sufficiency of the evidence, [it] is entitled to contest the sufficiency of the complaint and its allegations to support the judgment.” Id. Plaintiffs allege that Mass Mutual (a) “sold Policies and helped establish the Plan,” (b) “administered, managed, and controlled such Policies,” (

c) exercised discretionary authority over plan assets, and (d) “exercised discretionary authority with respect to plan management and administration.” Simple allegations that Mass Mutual falls within the statutory definition of fiduciary, however, are not well-pleaded factual allegations sufficient to establish liability in the wake of an entry of default.

Indeed, because the plaintiffs do not otherwise explain from whence this discretion came or how specifically Mass Mutual “administered, managed, and controlled” the plan, their allegations are really no more than a conelusory assertion that Mass Mutual is an ERISA fiduciary. Thus, the plaintiffs’ “well-pleaded factual allegations” regarding Mass Mutual’s fiduciary status are not well-pleaded factual allegations at all; rather, they are merely the plaintiffs’ own conclusions of law, which Mass Mutual is not deemed to have admitted as a result of default. Cf. Oxford Asset Mgmt. v.

Johans, 297 F.3d 1182, 1188 (11th Cir.2002) (“[On a motion to dismiss,] [t]he plaintiffs factual allegations are accepted as true.... However, ... legal conclusions masquerading as facts will not prevent dismissal.”). “Simply urging the purchase of its products does not make an insurance company an ERISA fiduciary with respect to those products.” Am. Fed’n of Unions Local 102 Health & Welfare Fund v. Equitable Life Assurance Soc’y, 841 F.2d 658, 664 (5th Cir.1988). This is all that the plaintiffs here really allege.

Specifically, they allege that in 1987 they purchased a set of Mass Mutual life insurance policies as a means of funding the split-dollar plan. They claim that the purchase was the result of a presentation by Martin and Hilliard that projected that the policies would yield substantial retirement and death benefits and that premiums would “vanish” within ten years. They allege that in 1990 the plan was amended and that they relieved BEI of its obligations to them under a separate deferred compensation agreement in exchange for BEI’s promise to pay additional premiums under the plan.

They claim that in making this decision they again relied on analyses prepared by Martin and Hilliard. Then, in 1992, the plaintiffs took out substantial loans against the policies — indeed, they borrowed all that the policies permitted.

They allege that they did so “in reliance on the representations of Mass Mutual and its agents that the Policies were sufficiently strong and had sufficient value to support the loans while still generating substantial retirement income after rollout.” Finally, they claim that even after 1992, Mass Mutual continued to provide inaccurate or misleading policy analyses showing that the plan was continuing to perform as projected. Thus, reduced to the size of a pea, this case is really about claims of fraud and misrepresentation in the sale of some life insurance policies.

The plaintiffs attempt to match these allegations with ERISA’s definition of a fiduciary, but they are unable to do so. Mass Mutual has never exercised discretionary authority or control over plan management or the administration of plan assets because the decisions to purchase, amend, and borrow against the policies were made by the plaintiffs themselves. Cases holding that insurers like Mass Mutual are not ERISA fiduciaries are numerous. See, e.g., Flacche v.

Sun Life Assurance Co. of Can., 958 F.2d 730, 733-35 (6th Cir.1992) (insurer did not act as a fiduciary by supplying annuity contracts to fund a plan and performing ministerial services for the plan); Consol. Beef Indus., Inc. v. N.Y. Life Ins. Co., 949 F.2d 960, 965 (8th Cir.1991) (insurer was not acting as a fiduciary when, through its agent, it “simply sold ... annuities” and “did not recommend specific investments beyond its products”); Gallagher Corp. v. Mass. Mutual Life Ins.

Co., 105 F.Supp.2d 889, 893-97 (N.D.Ill.2000) (insurer did not act as a fiduciary by persuading an employer to purchase its whole life insurance policies to fund a prototype plan it designed and providing record-keeping and actuarial services to the plan); Fechter v. Conn. Gen. Life Ins. Co., 800 F.Supp. 182, 196-206 (E.D.Pa.1992) (insurer did not act as a fiduciary by supplying fund assets and actuarial calculations); Assocs. in Adolescent Psychiatry v. Home Life Ins.

Co. of N.Y., 729 F.Supp. 1162, 1191 (N.D.Ill.1989) (insurer was not acting as a fiduciary because “all [it] did was talk the plans into buying its insurance products as a means of funding retirement benefits” and “design ... the prototype plan [the employer] adopted”) aff'd 941 F.2d 561 (7th Cir.1991); Bozeman v. Provident Nat’l Assurance Co., 1992 WL 328804, at *1-4 (W.D.Tenn. 1992) (insurer did not act as a fiduciary by supplying annuity contracts to fund a plan and giving actuarial advice).

The plaintiffs argue that they have alleged that Mass Mutual did more than just sell its insurance products to the plan because it (along with various alleged agents and delegatees) was also responsible for allocating premium payments between the portion counted as loans from their employer and the portion treated as income to them, analyzing policy performance, and communicating with them regarding policy performance. The performance of such ministerial policy-related services, however, does not render an insurer- such as Mass Mutual a fiduciary if it does not otherwise occupy that status.

See, e.g., Gallagher, 105 F.Supp.2d at 895-96; Fechter, 800 F.Supp. at 204-06. Nor does the performance of multiple duties, none of which in and of itself creates fiduciary status, render one a fiduciary. See, e.g., Bozeman, 1992 WL 328804, at *4 (rejecting the suggestion “that a ‘package deal’ of services and products, none of which separately creates fiduciary status, necessarily betokens an unusual degree of influence because of their variety and number”).

Moreover, the plaintiffs’ claims that Mass Mutual acted fraudulently, even if true, do not alter the fact that they have failed to allege facts from which we could infer that Mass Mutual is in fact a fiduciary. “Congress took a functional approach to defining an ERISA fiduciary.” Fechter, 800 F.Supp. at 205. Therefore, while “[wjrongful motivation and conduct” is certainly relevant once a plaintiff demonstrates that a defendant is a fiduciary with respect to the plan, it is “insufficient to establish fiduciary responsibility” to begin with. Id.

And, finally, because the plaintiffs have not shown that Mass Mutual itself ever had any fiduciary duty with respect to the plan, their additional allegation that, “[d]uring relevant times, Mass Mutual delegated certain discretionary authority” is also of no help. In sum, because the plaintiffs cannot establish that Mass Mutual is a fiduciary for any purpose that is relevant to the misconduct they allege, they cannot recover under ERISA § 502(a). III.

Our determination that Mass Mutual is not a fiduciary for any purpose implicated by the amended complaint is reinforced by our additional conclusion that the plaintiffs’ original state-law claims were not completely preempted; thus, removal was improper, as federal question jurisdiction was lacking, and the plaintiffs’ motion to remand should have been granted at the outset. On appeal, neither party has raised this issue. Ordinarily, if removal was improper, the court lacks subject matter jurisdiction and must raise the issue sua sponte and then dismiss on that ground, see Rembert v.

Apfel, 213 F.3d 1331, 1333 (11th Cir.2000) (“As a federal court of limited jurisdiction, we must inquire into our subject matter jurisdiction sua sponte even if the parties have not challenged it.”); however, because the post-removal amended complaint asserted claims under ERISA, we have jurisdiction even if removal was initially improper, see Pegram v.

Herdrich, 530 U.S. 211, 215 n. 2, 120 S.Ct. 2143, 2147 n. 2, 147 L.Ed.2d 164 (2000) (“[The plaintiff] does not contest the propriety of removal before us, and we take no position on whether or not the case was properly removed.... [The] amended complaint alleged ERISA violations, over which the federal courts have jurisdiction, and we therefore have jurisdiction regardless of the correctness of the removal.”); see also Grubbs v. Gen. Elec.

Credit Corp., 405 U.S. 699, 700, 92 S.Ct. 1344, 1346, 31 L.Ed.2d 612 (1972) (holding that when a district court has jurisdiction at the time it enters judgment “the validity of the removal procedure followed may not be raised for the first time on appeal”). We nonetheless address this issue here because it helps to clarify that Mass Mutual was not wearing its fiduciary hat when it allegedly misled the plaintiffs regarding future benefits under them policies.

Under 28 U.S.C. § 1441, a defendant may remove a state-court case to federal court only if the district court would have had jurisdiction over the case had it been brought there originally. As this case was originally filed, diversity jurisdiction was lacking because both plaintiffs and four of the seven defendants were citizens of Georgia. Therefore, if the case was removable it must have been because the plaintiffs’ claims “arise under” federal law for the purposes of federal question jurisdiction. See 28 U.S.C. § 1331.

But because the original complaint asserted only state-law claims, this case also does not arise under federal law under the ordinary operation of the well-pleaded complaint rule. See Kemp v. Int’l Bus. Mach. Corp., 109 F.3d 708, 712 (11th Cir.1997) (“A case does not arise under federal law unless a federal question is presented on the face of the plaintiffs complaint.”).

Therefore, if the case arises under federal law, it must be because it falls within the special category of federal question jurisdiction created by the doctrine of complete preemption. “Under that doctrine, Congress may preempt an area of law so completely that any complaint raising claims in that area is necessarily federal in character and therefore necessarily presents a basis for federal court jurisdiction.” Id. “Congress has accomplished this ‘complete preemption’ in [ERISA § 502(a)], which provides the exclusive cause of action for the recovery of benefits governed by an ERISA plan.

State law claims seeking relief available under [§ 502(a)] are recharacterized as ERISA claims and therefore ‘arise under’ federal law.” Id. (citations omitted). This court’s clearest statement of the ERISA complete preemption rule is found in Butero v. Royal Maccabees Life Insurance Co., 174 F.3d 1207 (11th Cir.1999), where we explained that complete preemption exists only when the plaintiff is seeking relief that is available under [ERISA § 502(a)].

Regardless of the merits of the plaintiffs actual claims (recast as ERISA claims), relief is available, and there is complete preemption, when four elements are satisfied. First, there must be a relevant ERISA plan. Second, the plaintiff must have standing to sue under that plan. Third, the defendant must be an ERISA entity. Finally, the complaint must seek compensatory relief akin to that available under [§ 502(a)]; often this will be a claim for benefits due under a plan. Id. at 1212 (internal citations and quotation marks omitted).

Complete preemption is thus narrower than “defensive” ERISA preemption, which broadly “supersede[s] any and all State laws insofar as they ... relate to any [ERISA] plan.” ERISA § 514(a), 29 U.S.C. § 1144(a) (emphasis added). Therefore, a state-law claim may be defensively preempted under § 514(

a) but not completely preempted under § 502(a). In such a case, the defendant may assert preemption as a defense, but preemption will not provide a basis for removal to federal court. Although we address complete preemption in this Part, we will also discuss several defensive preemption cases. These cases are helpful because claims that are completely preempted are also defensively preempted. Butero, 174 F.3d at 1215 (“If the plaintiffs claims are [completely preempted], then they are also defensively preempted.”).

Thus, if it appears that a claim is not even defensively preempted, then it will not be completely preempted either. As such, defensive preemption cases may inform the complete preemption analysis. “The complete preemption and defensive preemption doctrines are very complicated and the cases are numerous. The facts of the instant case do not fall neatly into any category of case law that allows for an easy or quick answer to be found from Eleventh Circuit case law.” Wilson v. Coman, 284 F.Supp.2d 1319, 1341 (M.D.Ala.2003). Therefore, we will start by briefly reviewing our significant decisions in this area in

section A. In

section B, we will apply those cases to the facts of this case. A. In Butero, the defendant insurer issued a life insurance policy to the husband of the plaintiff (Butero) through his employer; however, sometime after the policy’s putative effective date, the defendant remitted all premiums paid by the employer and informed it that its request for coverage was denied and that no policy existed.

Butero’s husband died that same day, her claim for benefits under the policy was denied on the ground that no policy existed, and she filed suit in state court against the insurer and an insurance agent alleging breach of contract, bad faith refusal to pay, and fraud in the inducement. The defendants then removed the case to federal court, arguing that the state-law claims were completely preempted.

On Butero’s motion to remand, the district court held that the claims against the agent were not preempted, and they were severed and remanded; however, it held that the claims against the insurer were completely and defensively preempted and accordingly dismissed them without prejudice to Butero’s right to file a complaint stating claims under ERISA. See Butero, 174 F.3d at 1210-11. On appeal, Butero challenged the district court’s disposition of her state-law claims against the insurer.

After stating the complete preemption rule set out above, we held that her claims were completely and defensively preempted and accordingly affirmed. As to the third element of complete preemption — whether the defendant is an ERISA entity — we concluded that the insurer was a fiduciary because “[i]t could ‘control ... the payment of benefits’ and the ‘determination of [the plaintiffs] rights’ under [the plan].” Id. at 1213 (quoting Morstein v. Nat’l Ins. Servs., Inc., 93 F.3d 715, 723 (11th Cir.1996) (en banc)).

As to the fourth element — whether the plaintiff seeks relief available under § 502(a) — we first noted that “we have held that claims against an insurer for fraud and fraud in the inducement to purchase a policy are in essence claims ‘to recover benefits due to [the beneficiary] under the terms of the plan,’ ” and we then concluded that the plaintiffs’ additional claims for bad faith refusal to pay and breach of contract essentially sought the same thing — payment of a life insurance benefit. Butero, 174 F.3d at 1213 (quoting ERISA § 502(a)(1)(B)).

Initially, we note two important distinctions between Butero and the instant case. First, the basis of the plaintiffs case in Butero was the insurer’s decision in its capacity as an ERISA entity not to pay a death benefit. Here, in contrast, Mass Mutual emphasizes, and we agree, that it is an ERISA fiduciary for the purpose of making death benefit determinations only, and the plaintiffs do not challenge any decision not to pay benefits under the terms of the plan.

Rather, they allege that Mass Mutual induced them to purchase and maintain vanishing premium life insurance policies as a source of retirement income and death benefits by misrepresenting the level of benefits those policies would provide. In other words, the plaintiffs’ dispute is with Mass Mutual the seller of insurance products, not Mass Mutual the ERISA fiduciary.

The second distinction, which is related to the first, is that while Butero’s claim was one for benefits due under the terms of an ERISA plan, Cotton and Eickhoff claim a loss based on the difference between Mass Mutual’s alleged misrepresentations — perhaps the most important of which predate the formation of the plan itself — -and the terms of the plan.

Thus, while Butero’s statement that “claims against an insurer for fraud or fraud in the inducement to purchase a policy are in essence claims ‘to recover benefits due to [the beneficiary] under the terms of the plan,’ ” id., does, as one district court in this circuit recently put it, “immediately draw[] this court’s attention,” we agree with that court that this generalization should not be automatically extended to cases in which the plaintiffs claims do not actually seek benefits under the terms of the plan. Wilson v. Coman, 284 F.Supp.2d 1319, 1334 (M.D.Ala.2003). In Morstein v.

National Insurance Serv., 93 F.3d 715 (11th Cir.1996) (en banc), the plaintiff (Morstein), the president/sole shareholder of a small business, purchased health insurance for herself and her only employee from an independent insurance agent. According to the complaint, Morstein told the agent that any policy that did not cover any preexisting condition would be unacceptable, and the agent assured her that the policy she purchased would cover all such conditions.

Over a year later, however, the insurer denied a claim for benefits under the policy for the reason that it was based on an undisclosed preexisting condition. Mor-stein then sued the agent and the insurer in state court for negligence, malfeasance, misrepresentations, and breach of contract; in response, the defendants removed the case to federal court under the complete preemption doctrine. The district court then denied Morstein’s motion to remand and granted the defendants’ motions for

summary judgment on the ground that all of Morstein’s state-law claims were both completely and defensively preempted. Morstein thereafter voluntarily dismissed her claims against the insurer and appealed the district court’s grant of

summary judgment in favor of the agent. Id. at 716-17. . . - On appeal, we reversed the district court’s order granting the agent’s motion for

summary judgment and held that Mor-stein’s state-law claims against him were not preempted. Id. at 722-74. In so holding, we relied primarily on the Fifth Circuit’s decision in Perkins v. Time Insurance Co., 898 F.2d 470 (5th Cir.1990), which held that a claim that an insurance agent fraudulently induced an insured to surrender coverage under an existing policy, to participate in an ERISA plan which did not provide the promised coverage, “relates to” that plan only indirectly.

A state law claim of that genre, which does not affect the relations among the principal ERISA entities (the employer, the plan fiduciaries, the plan, and the beneficiaries) as such, is not preempted by ERISA. Id. at 473. The Fifth Circuit in Perkins reasoned — and we agreed — that Congress did not intend to “immunize agents from personal liability for their solicitation of potential participants in an ERISA plan prior to its formation.” Morstein, 93 F.3d at 722 (quoting Perkins, 898 F.2d at 473).

We therefore held that “when a state law claim brought against a non-ERISA entity does not affect relations among principal ERISA entities as such, then it is not preempted by ERISA.” Morstein, 93 F.3d at 722 (emphasis added). This included Morstein’s claims, as the defendant agent plainly was not an ERISA entity. In Morstein, we also explained why we thought it unlikely that Congress intended to preempt the type of claims at issue in that case: [E]conomic impact alone is not necessarily enough to preempt a state law.

Therefore, the possibility that insurance premiums will be higher or that insurance will be more difficult to obtain because independent agents will have less incentive to sell insurance to employers whose employee benefit plans will be governed by ERISA, does not provide a reason to preempt state laws that place liability on agents for fraud. These same agents currently face the threat of state tort claims if they make fraudulent misrepresentations to individuals and entities not governed by ERISA.

To hold these agents accountable in the same way when making representations about an ERISA plan merely levels the playing field. Allowing preemption of a fraud claim against an individual insurance agent will not serve Congress’s purpose for ERISA.... Congress enacted ERISA to protect the interests of employees and other beneficiaries of employee benefit plans. To immunize insurance agents from personal liability for fraudulent misrepresentation regarding ERISA plans would not promote this objective.

If ERISA preempts a beneficiary’s potential cause of action for misrepresentation, employees, beneficiaries, and employers choosing among various plans will no longer be able to rely on the representations of the insurance agent regarding the terms of the plan. These employees, whom Congress sought to protect, will find themselves unable to make informed choices regarding available benefit plans where state law places the duty on agents to deal honestly with applicants. Id. at 723-24 (citations omitted). This reasoning can be applied to the instant case as well.

After all, the plaintiffs allege that Mass Mutual, like the independent agent in Morstein, made misrepresentations in the sale of an insurance policy. And when an insurer is not acting in its capacity as an ERISA entity, we can see no reason that Congress would have sought to immunize it from liability for fraud or similar state-law torts. For conduct such as that alleged by the plaintiffs, insurers currently face the threat of suit under state law by non-ERISA entities.

To hold them accountable in this context, therefore, “merely levels the playing field,” and any indirect economic impact such claims may have on plans governed by ERISA is not by itself sufficient to establish complete preemption. Id. at 723. Indeed, just like the judgment against the independent insurance agent in Morstein, a judgment against Mass Mutual will have no direct economic impact on any ERISA plan. Finally, Engelhardt v. Paul Revere Life Insurance Co., 139 F.3d 1346 (11th Cir.1998), which addressed both defensive and complete preemption, is relevant here.

There, the plaintiff (Engelhardt) disclosed glaucoma as a preexisting condition in his application for disability insurance through his employer. After reviewing Engelhardt’s application, the insurer (Paul Revere) requested that he sign an amendment to the policy excluding disability related to “either or both eyes.” Engelhardt, however, insisted that only glaucoma-related disability should be excluded. The independent agent who sold the policy agreed that the exclusion should be so interpreted, and Paul Revere sent Engelhardt a letter confirming the same.

When Engelhardt suffered a detached retina, however, Paul Revere relied on the exclusion to deny his claim for benefits. Engelhardt then sued Paul Revere in state court for fraudulent inducement, and Paul Revere removed the case to federal court under the complete preemption doctrine. The district court granted Engelhardt’s motion to remand, concluding that complete preemption did not exist. See id. at 1348-50. On appeal, we reversed and held that Engelhardt’s state-law claim was completely and defensively preempted.

On the defensive preemption issue, we distinguished Morstein as follows: “[ujnlike the independent insurance agent and agency in Morstein, Paul Revere is an ERISA fiduciary, and thus an ‘ERISA entity,’ because it had the exclusive authority to determine eligibility for benefits under the plan and to review denied claims.” Id. at 1352. We then elaborated on this distinction: Engelhardt’s suit “affects the relations among principal ERISA entities as such ” and “relates to” an ERISA plan for a number of reasons. First, as noted above, Engelhardt’s suit is against an ERISA entity.

Second, Paul Revere had assumed its role as an ERISA entity and [the] ERISA plan had been established before, albeit only shortly before, Paul Revere made its alleged misrepresentations to an individual regarding the scope of coverage under the ERISA plan. Third, the impetus for Engelhardt’s suit was Paul Revere’s decision in its role as an ERISA entity to deny Engelhardt’s claim for benefits under the ERISA plan. Although Engelhardt initially attempted to challenge the denial through Paul Revere’s appeals process, Engel-hardt ultimately abandoned that approach and filed this lawsuit.

Fourth, as a result of documents produced during the discovery process, Paul Revere determined that its denial of benefits was improper and decided to pay En-gelhardt the ERISA benefits he had been seeking all along. Id. at 1352-53 (emphasis added) (footnote omitted) (quoting Morstein, 93 F.3d at 722). Thus, we emphasized that the suit was between ERISA entities “as such,” that Paul Revere “assumed its role as an ERISA entity” prior to the alleged fraud, and that the suit was really about Paul Revere’s denial of benefits “in its role as an ERISA entity.” Id.

In other words, we did not simply establish that Paul Revere was an ERISA entity at some point or for some purposes; rather, we emphasized that Engelhardt’s complaint alleged misconduct by Paul Revere in that capacity.

Having determined that Engelhardt’s suit related to an ERISA plan and was thus defensively preempted, we also held that it was completely preempted for three reasons: First, the parties were “ERISA entities whose relationship became strained when Paul Revere denied Engel-hardt’s claim for benefits under the ERISA plan” — i.e., “precisely the types of parties who Congress intended to litigate under § 1132(a)” and “precisely the type of dispute that the statute was intended to resolve.” Id. at 1353. Second, Engel-hardt’s claim was “essentially a challenge to Paul Revere’s refusal to pay benefits.” Id. at 1354.

Third, Engelhardt did not seek to rescind his contract but instead sought benefits allegedly owed him under its terms. Id. Thus, the district court had jurisdiction over Engelhardt’s state-law claim and should have dismissed it as preempted under § 514. Thus, again, we did not rely simply on Paul Revere’s general status as an “ERISA entity.” Rather, we thought it necessary to show that Paul Revere was acting in that role when the alleged fraud took place and when it subsequently denied Engelhardt’s claim for benefits, and we noted that the dispute concerned a denial of benefits under an ERISA plan.

In contrast, in the instant case, we agree that Mass Mutual is not a fiduciary for any purpose other than making benefit determinations, a function that is not at issue in this lawsuit. Thus, it was not acting “in its role as an ERISA entity” at the time the plaintiffs allege that it fraudulently induced them to buy the vanishing premium life insurance policies at issue here. As such, this lawsuit does not seem to “affect[ ] ... relations among principal ERISA entities as such ” but instead affects only the relationship between two policyholders and their insurer. Id. at 1352.

Nor is the suit “essentially a challenge to [a] refusal to pay benefits” under an ERISA plan. Id. at 1354. Benefits have not yet become due under the policy, and from the plaintiffs’ perspective the problem is that the terms of the policy themselves do not match Mass Mutual’s earlier representations. Moreover, as noted above, the Engel-hardt court also relied on the fact that “Paul Revere had assumed its role as an ERISA entity and [the] ERISA plan had been established before, albeit only shortly before, Paul Revere made its alleged misrepresentations.” Id. at 1352. Along these lines, in Franklin v.

QHG of Gadsden, Inc., 127 F.3d 1024 (11th Cir.1997), we commented that, “unlike the circumstances in Morstein, [the plaintiff and the defendant] were ERISA entities at the time the alleged fraudulent misrepresentations were made.” Id. at 1029. In Butero, however, we described Franklin’s comment as a mere “suggestion] in dicta ... that an insurance company allegedly obligated to pay benefits under a plan is not considered an ERISA entity if the complaint alleges pre-policy fraud.” Butero, 174 F.3d at 1213 n. 3.

Furthermore, we expressed some “doubt that ERISA status can be so cleanly switched on and off’ when a plaintiff claims to be due benefits under an ERISA plan, and the insurer’s decision denying those benefits was plainly made in its fiduciary capacity. Id.

We agree with Butero’s statement that a plaintiff cannot avoid complete preemption simply by alleging pre-policy fraud only. (The plaintiffs here, of course, allege that misrepresentations were made both before and after the plan was established.) However, we do think that the fact that misconduct is alleged to have begun at a time when the defendant could not possibly have been acting as an ERISA fiduciary at least suggests that the suit may not be against the insurer-qua-fiduciary. Franklin and Hall v.

Blue Cross/Blue Shield of Ala., 134 F.3d 1063 (11th Cir.1998), also require brief mention, but, as a district court in this circuit recently concluded, their analyses are not altogether consistent with Butero, and they are therefore less helpful to us here. See generally Wilson, 284 F.Supp.2d at 1330-42. In Franklin, the plaintiff (Franklin) alleged that she accepted a position with the defendant hospital on the condition that her husband would continue to receive the same home nursing care provided under her former employer’s plan.

Two-plus years later, however, the hospital informed Franklin that it was discontinuing coverage for home nursing care; in doing so, it relied on a clause in the plan that reserved to it the right to reduce or terminate coverage at any time. Franklin then sued in state court, alleging that the hospital fraudulently induced her to leave her former employment by promising to provide the same level of home nursing care provided by her former employer’s plan.

The hospital removed the case to federal court under the complete preemption doctrine, and the district court denied Franklin’s motion to remand and ultimately entered

summary judgment in favor of the hospital on the ground that Franklin’s state-law claims were completely and defensively preempted. Franklin, 127 F.3d at 1026-27. On appeal, we affirmed. This result is consistent with Butero: the hospital, in its role as an ERISA fiduciary, discontinued benefits under an ERISA plan, and Franklin, a beneficiary, responded by filing a lawsuit that essentially sought to recover benefits under the terms of that plan— that is, she sought “compensatory relief akin to that available under § 1132(a).” Butero, 174 F.3d at 1212.

Franklin’s, analysis, however, focused almost entirely on § 514’s defensive preemption test — that is, whether the plaintiffs state-law claims “related to” an ERISA plan. See Franklin, 127 F.3d at 1027-29. The Franklin court did begin its analysis correctly by recognizing that “the jurisdictional issue ... turns on whether the plaintiffs are seeking relief that is available under [ERISA § 502(a)],” id. at 1028 (quoting Kemp, 109 F.3d at 712), but the rest of the opinion addresses only § 514’s “relate to” standard.

And, after noting that a “state law relates to an employee benefit plan ‘if it has a connection with or reference to such a plan,’ ” Franklin, 127 F.3d at 1028 (quoting N.Y. Conference of Blue Cross & Blue Shield Plans v. Travelers Ins. Co., 514 U.S. 645, 656, 115 S.Ct. 1671, 1677, 131 L.Ed.2d 695 (1995)), the court held that the plaintiffs state-law claims were completely preempted because they had such a “direct connection to the administration of medical benefits under an ERISA plan,” Franklin, 127 F.3d at 1029.

This analysis is not consistent with Butero’s complete preemption rule, which holds that a state-law claim is completely preempted only where all four “elements” of complete preemption are present. Butero, 174 F.3d at 1212. Butero, moreover, explicitly recognizes that state-law claims may be defensively preempted under § 514 but still not completely preempted under § 502' — that is, that defensive preemption is broader than complete preemption. Id. Thus, if a court determines that state-law claims are not defensively preempted, it necessarily follows that they are not completely preempted.

But the converse is not true, so a court cannot hold that state-law claims are completely preempted simply because they “relate to” an ERISA plan. In Hall, the plaintiff (Hall) alleged that the defendant insurer (Blue Cross) assured her that although known preexisting conditions would not be covered during the first 270 days of her new medical plan, any other condition that arose during the period would be covered; however, when Hall required surgery to remove an ovarian mass during the 270-day period, Blue Cross denied her claim even though the mass was not a known preexisting condition.

Hall then filed suit in state court, asserting three fraud-related claims, and Blue Cross removed on the basis of complete preemption. The district court denied Hall’s motion to remand, and granted Blue Cross’s motion to dismiss, holding that Hall’s claims were defensively and completely preempted. Hall, 134 F.3d at 1064. On appeal, we affirmed.

This result is also consistent with Butero: Blue Cross, in its role as a fiduciary, denied a claim for benefits under the terms of an ERISA plan, and Hall, a beneficiary, responded by filing a lawsuit that essentially sought benefits under the terms of that plan — that is, she sought “compensatory relief akin to that available under [ERISA § 502(a)].” Butero, 174 F.3d at 1212. The court, however, relied primarily on Franklin, and its analysis focused entirely on § 1144’s defensive preemption standard. See Hall, 134 F.3d at 1065-66.

Thus, Hall and Franklin reach results that are consistent with Butero, but they employ analyses that are inconsistent with the rule it established. In Wilson v. Coman, 284 F.Supp.2d 1319 (M.D.Ala.2003), the court addressed this inconsistency at length and concluded (

a) that while Butero “provides the proper standard for analyzing the question of complete preemption in the Eleventh Circuit,” (

b) that standard is in direct conflict with Franklin, and (c) “since Franklin is an earlier panel decision ..., under the prior panel rule,” he was bound to follow Franklin and not Butero. Id. at 1329. We agree that the analysis of Franklin, the earlier case, conflicts with that of Butero, the later one, but we do not think that the prior panel rule obliges us to follow Franklin. The reason is that our pre-Franklin cases are consistent with Butero. For example, in Kemp v. Int’l Bus. Mach.

Corp., 109 F.3d 708 (11th Cir.1997), we held that the complete preemption question “turns on whether the plaintiffs are seeking relief that is available under [§ 502(a)]” and that “[a]n ordinary ERISA preemption defense, even if valid, is not enough to create federal question jurisdiction,” and we analyzed the complete preemption issue without reference to § 514’s “relate to” standard. Id. at 711-14.

Indeed, in declining the defendant’s invitation to reach the defensive preemption issue, we explained, “We cannot decide whether the plaintiffs’ claims relate to an ERISA plan ... because we have no jurisdiction over this case. The defense of ordinary ERISA preemption, by itself, does not create federal question jurisdiction.” Id. at 714 (emphasis added). Similarly, in Brown v. Conn. Gen. Life Ins. Co., 934 F.2d 1193 (11th Cir.), we distinguished defensive preemption under § 514 and complete preemption under § 502(a). See id. at 1195-96; id. at 1197-99 (Johnson, J., dissenting).

And in our most recent complete preemption case (issued after the district court’s decision in Wilson), we again held that a claim is not completely preempted under § 502(

a) simply because it is defensively preempted under § 514; indeed, we specifically stated that “the district court erred in deciding the [complete] preemption issue by applying the defensive preemption analysis.” Ervast v. Flexible Products Co., 346 F.3d 1007, 1013 (11th Cir.2003). Thus, we are not bound by the analyses of Franklin and Hall; rather, we will apply the complete preemption rule set out in Butero. B.

In this section, we state briefly what is likely clear at this point: The claims of the plaintiffs in this case are unlike those of plaintiffs in past cases in which we have held complete preemption to apply because they “do[ ] not affect ... relations among principal ERISA entities ... as such,” Perkins, 898 F.2d at 473, or concern Mass Mutual’s conduct in its capacity as an ERISA fiduciary.

Nor do the plaintiffs “seek compensatory relief akin to that available under [§ 502(a)],” Butero, 174 F.3d at 1212, because they seek damages based on fraud in the sale of insurance policies, not benefits alleged to be due under the terms of the plan itself.

Because of these differences, this case proves to be an exception to our general rule of thumb that “claims against an insurer for fraud or fraud in the inducement to purchase a policy are in essence claims ‘to recover benefits due to [the beneficiary] under the terms of the plan.’ ” Id. at 1213 (quoting ERISA § 502(a)(1)(B)). “From the Butero opinion, it is apparent that [the plaintiff in that case] sought to recover insurance benefits that she believed she was entitled to as a result of [her husband’s] death.” Wilson, 284 F.Supp.2d at 1334.

In contrast, Cotton and Eickhoff are “very much alive” and are “not seeking actual death benefits” from Mass Mutual. Id. “In essence,” they are suing Mass Mutual for its “alleged fraudulent and negligent conduct, which caused [them] to suffer economic losses.” Id. at 1336. “What [they] are not suing ... for is a refusal to pay benefits under the terms of the ... life insurance policies.... This fact distinguishes the instant case from Butero.” Id. (emphasis added).

It also distinguishes it from cases such as Engelhardt, Franklin, and Hall, because in each of those cases the plaintiff was actually contending to be due a benefit that could be identified within the terms of his or her policy — namely, a disability benefit {Engel-hardt), home nursing care (Franklin), and reimbursement for healthcare expenses (Hall). More important with respect to the ERISA claims at issue here, our case law in this area supports our conclusion that Mass Mutual was not acting in its fiduciary capacity for any purpose relevant to the plaintiffs’ § 502(

a) claims. In Bwtero, Engelhardt, Franklin, and Hall, there were actual decisions denying benefits under the relevant plan. When an insurer makes such a decision, it is plainly wearing its fiduciary hat, and the beneficiary may challenge the correctness of the decision according to the terms of the ERISA plan. Here, in contrast, as the district court correctly concluded, the terms of the plan documents “are clear and unambiguous,” and the plaintiffs are not entitled to relief under them.

Nor do their allegations, if properly characterized and understood, seek relief under the plan or challenge any action by Mass Mutual in its fiduciary capacity. This distinguishes the case from those in which we reasoned that the plaintiff was essentially challenging an insurer’s denial of benefits. The plaintiffs claims are instead more like those asserted against the independent insurance agent in Morstein.

We are, of course, aware that Mass Mutual is in many respects different from such an agent and would be considered a fiduciary for some purposes, but, as we explain above, fiduciary status under ERISA is a functional concept, and in this case Mass Mutual was performing a function similar to that performed by the agent in Mor-stein or professionals who provided ministerial plan-related services in other cases. Therefore, it was not wearing its fiduciary hat in any situation relevant to this lawsuit.

Thus, while the case does in one sense affect relations among parties who are ERISA entities, it “does not affect ... relations among the principal ERISA entities ... as such." Perkins, 898 F.2d at 473. Moreover, just as in Morstein, the plaintiffs’ state-law claims, if proven, should have only had an indirect economic impact on any ERISA plan, and allowing state-law claims against Mass Mutual on the same basis that they are permitted outside of the ERISA context “merely levels the playing field.” Morstein, 93 F.3d at 723. IV.

When it denied the plaintiffs’ motion to remand, the district court held that the plaintiffs’ state-law claims were completely preempted. As a technical matter, “a decision regarding complete preemption does not decide the issue of defensive preemption” because “defensive preemption is a substantive issue that must be decided by a court with competent jurisdiction.” Ervast, 346 F.3d at 1013 n. 7. Thus, if-a district court remands to state court claims that are not completely preempted, the defendant may still attempt to raise ERISA § 514 preemption as a defense in the state court.

As a practical matter, however, because defensive preemption is significantly broader than complete preemption, a disti’ict court’s determination that a claim is completely preempted does resolve the defensive preemption issue. As we held in Butero, “[i]f the plaintiffs claims are [completely preempted], then they are also defensively preempted.” 174 F.3d at 1215. Thus, we agree with Mass Mutual that the district court’s order denying the plaintiffs’ motion to remand is inconsistent with its subsequent decision allowing them to file an amended complaint asserting all of the same state-law claims.

Mass Mutual argues that all of the discovery requests that ultimately led to discovery sanctions related solely to the state-law claims. Therefore, the argument goes, the earlier order to compel was itself an abuse of discretion. When a district court imposes sanctions under Fed.R.Civ.P. 37(b), “the propriety of [the sanctions] depends in large part on the propriety of the earlier compel order.” Chudasama v. Mazda Motor Corp., 123 F.3d 1353, 1366 (11th Cir.1997).

Thus, “[i]n evaluating whether a district court abuses its discretion when it imposes severe sanctions upon a party that violates an order, we believe that an important factor is whether the entry of that order was itself an abuse of discretion.” Id. This is not to say, however, that sanctions based on erroneous discovery orders will never be upheld. “Because we expect litigants to obey all orders, even those they believe were improvidently entered, sanctions will very often be sustained, particularly when the infirmity of the violated order is not clear and the sanctions imposed are moderate.” Id. at 1366 n. 34.

In Chudasama, we noted that a “[f]ailure to consider and rule on significant pretrial motions before issuing dis-positive orders can be an abuse of discretion.” Id. at 1367. We also instructed that “[fjacial challenges to the legal sufficiency of a claim or defense ... should ... be resolved before discovery begins,” id., especially when the challenged claim will significantly expand the scope of allowable discovery, id. at 1368. This is so because every claim has the potential to enlarge the scope and cost of discovery.

Id. “If the district court dismisses a nonmeritorious claim before discovery has begun, unnecessary costs to the litigants and to the court system can be avoided.” Id. Given the relationship between complete and defensive preemption, it was an abuse of discretion for the district court to delay ruling on Mass Mutual’s challenges to the plaintiffs’ state-law claims. Indeed, the court declined to resolve this issue not only when the state-law claims reappeared in the plaintiffs’ amended complaint, but also when Mass Mutual raised it again in the context of the later motion to compel.

That we disagree with the district court’s initial ruling that the state-law claims were completely preempted does not alter our analysis in this Part. We assume the correctness of that ruling for the purpose of addressing the court’s compel order and order imposing sanctions, since those orders were entered on the assumption that the plaintiffs’ state-law claims were, in fact, completely preempted.

The default sanction is, of course, mooted by our determination that the plaintiffs cannot establish liability under ERISA, but the order requiring Mass Mutual to pay a part of the plaintiffs’ reasonable costs and attorneys fees remains and should be revisited on remand. Even if the district court’s failure to rule on Mass Mutual’s preemption defense to the state-law claims was an abuse of discretion, we do not think that necessarily means that the compel order was also an abuse of discretion.

Although Mass Mutual has repeatedly asserted that the discovery requests that led to the compel order related only to the state-law claims, the plaintiffs deny that this was the case, and it does not appear that the district court ever explicitly agreed with Mass Mutual that the ERISA claims did not by themselves justify the compel order. If Mass Mutual had been wearing its fiduciary hat for any purpose relevant to the plaintiffs’ allegations, the material covered by the compel order could have been relevant to the plaintiffs’ breach of fiduciary duty claims.

The central theory of these claims is that “[l]ying is inconsistent with the duty of loyalty owed by all fiduciaries and codified in

section 404(a)(1) of ERISA.” Varity Corp. v. Howe, 516 U.S. 489, 506, 116 S.Ct. 1065, 1075, 134 L.Ed.2d 130 (1996) (quoting Peoria Union Stock Yards Co. v. Penn Mut. Life Ins. Co., 698 F.2d 320, 326 (7th Cir.1983)). In the district court, Mass Mutual argued that the only documents relevant to such a claim are formal plan documents and perhaps any informal documents that the plaintiffs actually reviewed. We think that Mass Mutual’s view of the discovery relevant to such a claim is too narrow. While it is true that the district court could have resolved the plaintiffs’ § 502(a)(1)(

B) claim solely by reference to the plan documents, the plaintiffs’ breach of fiduciary duty theory justified additional discovery into whether Mass Mutual intentionally defrauded or misled them and thus breached the fiduciary “duty of loyalty ... codified in

section 404(a)(1) of ERISA.” In Varity, for example, § 502(a)(3) liability was premised on the fiduciary-employer’s “falsely optimistic forecasts about its new venture’s prospects for success,” which were intended to persuade its employees that they would not undermine the security of their benefits by transferring to the newly formed subsidiary. Varity, 516 U.S. at 537, 116 S.Ct. at 1089 (Thomas, J., dissenting).

To establish that the forecasts were in fact “falsely optimistic,” the Varity plaintiffs apparently introduced evidence that management itself expected the new subsidiary to fail and sought to convince them to switch employment specifically so that the company could rid itself of its ERISA obligations to them. See Howe v. Varity Corp., 36 F.3d 746, 748-50 (8th Cir.1994).

To prevail on a similar theory in this case, the plaintiffs would have needed to show that Mass Mutual, acting in its fiduciary capacity, presented falsely optimistic policy illustrations that it essentially knew overstated the benefits that the policies could be expected to produce.

We express no opinion as to what exactly the plaintiffs would have had to show to prevail on such a claim, but we do not see how they could have even begun to do so without discovery regarding the methodologies, assumptions, and data Mass Mutual used to produce the policy illustrations and Mass Mutual’s internal communications regarding the marketing and economics of vanishing premium policies. From the record and the orders below, it is difficult to discern the precise basis of the compel order.

Because the district court apparently thought that all of the original state-law claims could survive even though completely preempted, it never clarified whether the compel order related to the those claims, the new ERISA claims, or both. On remand, the district court should revisit this issue.

If, as Mass Mutual contends, the plaintiffs had everything they needed to proceed on a Varity/Peoria Union-type breach of fiduciary duty claim, then the compel order must have related solely to the no longer viable (under the district court’s order denying the plaintiffs’ motion to remand) state-law claims and was an abuse of discretion because the failure to dismiss those claims was itself an abuse of discretion.

The question whether a finding of complete preemption compelled an additional finding of defensive preemption was a straightforward question of law, and the district court’s decision not to answer it cannot be justified. Thus, if the court’s decision not to decide significantly increased the burden of discovery, then any significant sanctions it imposed were an abuse of discretion because “the infirmity of the violated order” in that case would be “clear.” Chudasama, 123 F.3d at 1366 n. 34.

If, however, as the plaintiffs’ claim, the compel order related not only to the state-law claims but also to the breach of fiduciary duty claims under ERISA, then it was not an abuse of discretion, at least to the extent that was justified by the ERISA claims. And if the compel order was not itself an abuse of discretion, then sanctions for a willful, “bad faith” violation of it would be well within the district court’s “broad discretion to fashion appropriate sanctions for violation of discovery orders.” Malautea v. Suzuki Motor Co., 987 F.2d 1536, 1542 (11th Cir.1993). V.

For the foregoing reasons, we REVERSE the judgment of the district court in favor of the plaintiffs under ERISA § 502(a)(2). Because the plaintiffs cannot establish that Mass Mutual is an ERISA fiduciary for any purpose relevant to their amended complaint, all ERISA claims should be dismissed with prejudice. We VACATE the district court’s order imposing sanctions under Fed.R.Civ.P. 37(

b) and REMAND with the instruction that the district court should consider what, if any, sanctions are appropriate in light of this opinion. . Because we reverse on this ground, we do not address Mass Mutual's additional arguments. . Mass Mutual maintains that Hilliard, Martin, and the other former co-defendants in this case were not its legal agents, but rather were simply independent agents authorized to sell Mass Mutual products. We assume that they were in fact Mass Mutual agents. . Mass Mutual also denies that Alexander & Alexander was its agent; again, though, we assume that it was. .

This scenario, unfortunately, is not uncommon. Vanishing premium plans are a product of the soaring interest rates of the late 1970s and early 1980s. During this time, "the economics of traditional whole life insurance policies turned unattractive” because the policies generally "earned a rate of return based on the average interest rate of the predominately fixed-rate securities in the company’s investment portfolio, which generally had interest rates that were much lower than the rates then available to consumers.” Daniel R. Fischel & Robert S.

Stillman, The Law and Economics of Vanishing Premium Life Insurance, 22 Del. J. Corp. L. 1, 5 (1997). The industry responded to this phenomenon by offering consumers several new types of interest-sensitive policies. These new policies differed from traditional whole life insurance in that their returns were based on current interest rates rather than the interest and dividend income produced by the insurance company's historical investments. Id. at 5-6. The vanishing premium plan was one common payment option for these new types of policies; Professors Daniel R. Fischel and Robert S.

Stillman describe the economics of its rise and fall as follows: In a vanishing premium plan, the policyholder pays higher-than-normal premiums in the early years of the policy. By making higher payments in early years, a higher fraction of premium dollars is distributed into the policy's savings account (i.e., accumulation fund), allowing the cash value of the policy to accumulate faster.

The goal of a vanishing premium plan is to set premiums at a level where, after a certain number of years, enough cash value has accumulated within the policy so that future administrative and insurance costs can be paid out of the accumulation fund, with no further out-of-pocket payments by the policyholder. In the mid-1980s, when the new policies were marketed most aggressively, the assumption of most vanishing premium "illustrations” was that no further out-of-pocket premiums would be required after five or ten years. Vanishing premium plans have not worked out as initially contemplated.

For many, vanishing premiums have not vanished. The primary cause of this problem has been low interest rates.... Although rates rose to as high as twelve percent in the mid-1980s, the early 1990s reflected a low of three percent. These declining rates upset the economics of vanishing premium plans. With the economy-wide decline in interest rates, interest-credit rates have also dropped and cash value has not grown as assumed in the initial illustrations. In many cases, cash value in vanishing premium plans has become insufficient to pay expected future insurance and administrative costs.

As a result, many consumers who bought insurance on a vanishing-premium basis will soon be forced to make additional out-of-pocket premium payments or else have their insurance terminated or death benefits reduced. Id. at 7-8 (footnotes omitted). . These ERISA sections are at 29 U.S.C. §§ 1132(a)(1)(b), (a)(2), and (a)(3). . Fed.R.Civ.P. 37(b)(2) provides in relevant part: If a party ... fails to obey an order to provide or permit discovery, including an order made under subdivision (

a) of this rule ..., the court in which the action is pending may make such orders in regard to the failure as are just, and among others the following: (

C) An order striking out pleadings or parts thereof, ... or dismissing the action or proceeding or any part thereof, or rendering a judgment by default against the disobedient party; In lieu of any of the foregoing orders or in addition thereto, the court shall require the party failing to obey the order or the attorney advising that party or both to pay the reasonable expenses, including attorney’s fees, caused by the failure, unless the court finds that the failure was substantially justified or that other circumstances make an award of expenses unjust. .

In June of 2001, the court denied Mass Mutual's motion to set aside the default under Fed.R.Civ.P. 55(c). . On cross-appeal, the plaintiffs contest the district court’s determination that the policy projections were neither plan documents nor

summary plan descriptions (SPDs). However, the projections certainly are not plan documents of the type described in ERISA § 402(b), 29 U.S.C. § 1102(b), as they do not (1) “provide a procedure for establishing and carrying out a funding policy,” (2) "describe ... the operation and administration of the plan,” (3) "provide a procedure for amending [the] plan” or "for identifying the persons who have authority to amend the plan,” or (4) "specify the basis on which payments are made to and from the plan.” See, e.g., Cinelli v. Sec. Pac. Corp., 61 F.3d 1437, 1442 (9th Cir.1995).

Nor are they "other instrument[s] under which the plan was established or is operated.” ERISA § 104(b)(2), 29 U.S.C. § 1024(b)(2); see, e.g., Faircloth v. Lundy Packing Co., 91 F.3d 648, 653 (4th Cir.1996) ("the language ‘other instruments under which the plan is established or operated' encompasses formal or legal documents under which a plan is set up or managed”). The policy projections do not govern plan management, amendment, or administration; rather, they simply project future benefit levels based on certain economic assumptions.

Therefore, they are not plan documents and provide no basis for a § 502(a)(1)(

B) claim unless they can qualify as

summary plan descriptions. In Hicks v. Fleming Cos., 961 F.2d 537 (5th Cir.1992), the Fifth Circuit held that to qualify as a

summary plan description a document must "contain[] all or substantially all categories of information required under 29 U.S.C. § 1022(

b) and the DOL’s regulations at 29 C.F.R. § 2520.102-3.” Hicks, 961 F.2d at 542; accord Palmisano v. Allina Health Sys., Inc., 190 F.3d 881, 887-88 (8th Cir.1999); Pisciotta v. Teledyne Indus., Inc., 91 F.3d 1326, 1329-30 (9th Cir.1996) ("[I]t is undisputed ... that neither ... booklet contained all twelve of the required statutory elements. Accordingly, the district court correctly concluded ... that, as a matter of law, the booklets did not constitute SPDs.”); see also Coleman v. Nationwide Life Ins. Co., 969 F.2d 54, 62 n. 3 (4th Cir.1992) ("ERISA provides a lengthy, detailed list of what must be included in a

summary plan description. 29 U.S.C. § 1022(b). An agreement to provide a description of ‘the main benefits and requirements’ cannot be reasonably regarded as the equivalent of the

summary plan description, which the statute requires to be far broader in scope and detail.”). The Third Circuit appears to hold that additional factors may be relevant to this issue, but it does agree that a document that is missing significant statutorily required information is unlikely to qualify as an SPD. See Gridley v. Cleveland Pneumatic Co., 924 F.2d 1310, 1316-17 (3d Cir.1991). One district court, in contrast, has stated, We do not believe that Congress intended to exclude from the term "summary plan description” every document which lacks some of the information required in

section 1022(b). Rather, any document a plan distributes to its participants which contains all or substantially all of the information the average participant would deem crucial to a knowledgeable understanding of his benefits under the plan shall be deemed a

summary plan description. Kochendorfer v. Rockdale Sash and Trim Co., 653 F.Supp. 612, 615 (N.D.Ill.1987). We have perhaps suggested that we would adopt the majority view stated in Hides, but we have never unequivocally said so. In Alday v. Container Corp. of Am., 906 F.2d 660 (11th Cir.1990), we concluded that a benefit

summary booklet did "not fulfill the requirements ... set out at 29 U.S.C. § 1022” was not an SPD. Alday, 906 F.2d at 665. The plaintiff there

[…]