Peter Collins sued his former employer, Golden Cat Corp., in state court; he claimed the company--which was sold to Ralston Purina in April 1995--reneged on its promise in a retention agreement to pay Collins money and benefits if he was fired by the new owner before December 31, 1995. Golden Cat removed the case to federal court, claiming that ERISA preempted his claims. Collins moved to remand the case under the theory that no federal question jurisdiction existed, but the district court accepted jurisdiction of the case. After some discovery, Golden Cat moved for partial summary judgment relating to Collins' claim that the change in ownership itself triggered a payout under the retention agreement. The district court granted that motion. Collins also claimed that Ralston Purina substantially reduced his job responsibilities, which again would have triggered a payout. In a bench trial, at the conclusion of Collins' case-in-chief, the district court granted judgment on partial findings in favor of Golden Cat. Collins appeals the district court's denial of his motion to remand, the court's grant of partial summary judgment to Golden Cat, and the court's grant of judgment on partial findings. In all respects, we affirm.

Golden Cat made kitty litter, and Peter Collins served as its director of customer marketing in the grocery division. Collins was in charge of marketing the litter throughout the eastern United States; by all accounts he had a good job with a lot of responsibility. But by late 1994 the buzz at Golden Cat was that the company was looking for a buyer; inevitable rumors of job insecurity arose.

To discourage the key employees from seeking job opportunities elsewhere--a prospect that could diminish the company's market value--Golden Cat executed several "retention agreements" with its upper-level managers. (In the district court, Golden Cat claimed there were at least 60 agreements; for his part, Collins conceded that multiple agreements existed.) Under the terms of those agreements, Golden Cat agreed to pay a manager like Collins six months salary (and a year of COBRA1 benefits) if he was terminated by Golden Cat's prospective acquirer, Ralston Purina. The agreement stated that termination of employment included "any substantial reduction of duties or responsibilities," or any failure on the acquirer's part to "continue to offer existing or substantially similar employee benefits," or a "transfer outside the metropolitan South Bend, IN area." The agreement was to expire on December 31, 1995.

Collins received his copy of the proposed retention agreement in October 1994 and signed it almost immediately. Ralston bought Golden Cat in April 1995, and it wasn't long before the new owner wanted to make some changes. The changes are not important here, except to say that they reflected Ralston's marketing style--unlike Golden Cat, which spent a good deal of money on marketing a high-quality (and more expensive) product, Ralston spent less, preferring instead to cut the prices of their products.

In July 1995, just a few months after the takeover, Collins claims that he began to see sure signs that his job responsibilities were going to be altered significantly. Bob Watt, a Ralston vice president, met with Collins to discuss his future; Watt followed up the conversation with an e-mail in which he offered Collins a position as Ralston's regional sales manager in charge of the company's Dunkirk region. The Dunkirk region covered much of the east coast and upstate New York, but that included considerably less territory than the entire eastern United States, which is what Collins was used to overseeing. The memo did not say that Collins had to relocate in order to take the position, only that Collins would be expected to help Golden Cat "transition" into a Ralston company, which may have included sales for pet food as well as litter.

Collins was not pleased with the offer, and he was even less pleased when he saw a copy of a memo in which his name already was "penciled in" as Ralston's new Dunkirk regional sales manager. On August 7, Collins told Watt that he believed the new position would constitute a substantial reduction in his responsibilities; Watt disagreed. Shortly thereafter, Collins approached Golden Cat CEO Franklin Krum about the new position; Collins tried to persuade Krum that the "substantial reduction of ... duties or responsibilities" clause in his retention agreement had been satisfied. Krum responded by telling Collins that he had less discretion in these matters than Collins believed. On August 22, Collins handed in his letter of resignation, effective September 1. The letter was short--three sentences, the first of which tells us why he quit: "my job responsibilities with the company have been substantially reduced."

Though the parties do not say so, we assume that Collins demanded his payout under the retention agreement (six months salary and benefits) and that Golden Cat refused, because Collins sued Golden Cat (now a division of Ralston) in Indiana State court. Golden Cat, claiming that ERISA preempted Collins' claims because the retention agreement constituted an employee benefit plan, removed the case to federal court. Collins contested the removal and moved to remand, but the district court accepted jurisdiction. After landing the case in federal court, Golden Cat conducted some discovery (it took Collins' deposition), and then moved for partial summary judgment on Collins' claim that the acquisition by Ralston itself constituted a "termination" under the retention agreement. The court granted partial summary judgment to Golden Cat on this claim. The only remaining claim--that Golden Cat was liable under the agreement because Ralston substantially reduced Collins' job responsibilities--went to trial, but it didn't get very far. Golden Cat moved for judgment after the plaintiff's case, and the court (though it found Collins to be a very credible witness) granted the motion, dismissing the case and entering judgment for Golden Cat.

There are three issues on appeal: (1) whether the district court had federal question jurisdiction over Collins' claim under ERISA; (2) whether Golden Cat was liable under the agreement by virtue of Ralston's acquisition of the company; and (3) whether the district court appropriately decided against Collins' claim that Golden Cat had to perform under the agreement because Ralston had substantially reduced his duties and/or moved him outside of South Bend.

Because ERISA preemption is the sole basis of federal jurisdiction here, we must first determine whether the retention agreement is an employee benefit plan under ERISA. Congress enabled ERISA to preempt state law claims affecting benefit plans so that employers would have a uniform set of procedures and regulations when establishing and maintaining such plans. Fort Halifax Packing Co. v. Coyne, 482 U.S. 1, 11, 107 S. Ct. 2211, 96 L. Ed. 2d 1 (1987). ERISA thus preempts a state law claim if the claim requires the court to interpret or apply the terms of an employee benefit plan, which the Supreme Court has defined as "benefits whose provision by nature requires an ongoing administrative program to meet the employer's obligation." Id. In Fort HalifaxI, the Court was faced with a claim brought under a Maine statute that required employers who shut down operations to make one-time severance payments to their employees. The Court held that while the onetime payment system constituted a benefit, it did not constitute a benefit plan. "The requirement of a one-time, lump-sum payment triggered by a single event requires no administrative scheme whatsoever to meet the employer's obligation. The employer assumes no responsibility to pay benefits on a regular basis, and thus faces no periodic demands on its assets that create a need for financial coordination and control." 482 U.S. at 12, 107 S. Ct. 2211 (emphasis added). In short, the payments in Fort Halifax did not require the company to monitor its financial health and budget for future disbursements; the employer could satisfy its duty under the statute all at once by "making a single set of payments to employees at the time the plant closes." Id. The Court concluded that ERISA was not implicated because the "theoretical possibility of a one-time obligation in the future simply creates no need for an ongoing administrative program for processing claims and paying benefits." Id. (emphases added).

There are similarities between Collins' retention agreement and the statute facing the Court in Fort HalifaxI. Both required the employer to issue lumpsum disbursements upon the occurrence of an event. But the differences are clear--at most Fort Halifax had to disburse its lump sum payments once, to all of its employees at the same time. As the Court noted, "[t]o do little more than write a check hardly constitutes the operation of a benefit plan." Id. By contrast, Golden Cat faced the prospect of multiple payments to various managers, at different times and under different circumstances. Golden Cat could not satisfy its obligation by cutting a single check and making a "single set of payments" to all of its managers at once. The individual retention agreements required the company to budget for the prospect of paying out disbursements of varying amounts to its managers and at varying times. Accordingly, what Fort Halifax was spared--the potential for "periodic demands on its assets that create a need for financial coordination and control," 482 U.S. at 12, 107 S. Ct. 2211--Golden Cat was not. Unlike Fort Halifax's potential one-time liability, Golden Cat faced a year in which various managers might demand payouts under the individual agreements, and perhaps more than one time each if not successful on their first attempt to show that their job responsibilities had been "substantially reduced."

Even the triggering event prompting a payout in this case presupposes careful claims processing, in other words, an ongoing administrative scheme. The agreement required Golden Cat to pay only if a manager's job responsibilities were "substantially reduced," which sets a standard, but hardly an easily discernible one. The result is that Golden Cat had to assess the new job duties of Collins and compare them to his previous responsibilities; even then the company somehow had to determine if the difference between the two was "substantial." For example, Collins may have ended up in charge of less territory but with added products, such as pet food. The level of responsibility could be similar or even increased. Records would be important-indeed, vital--because Collins' agreement was not the only one the company executed with its employees. Presumably the company would be expected to process claims like Collins' in a consistent manner, but that could only be achieved by maintaining records as to who received benefits and why. And, as noted above, even if only one manager demanded his payout under the retention agreement, he could do so on several different occasions until successful. Only an ongoing administrative scheme would allow the company to develop a working definition of "substantial reduction of duties or responsibilities," such that it could be consistently applied either to a single employee on multiple occasions or multiple employees on multiple occasions. It is exactly this prospect of multiplicity and record-keeping that distinguishes Collins' case from the one-time, routine disbursement facing the Court in Fort Halifax.

Finding Golden Cat's retention agreements to implicate an ongoing administrative scheme (and thus ERISA) is consistent not only with Fort Halifax, but the decisions of other circuit courts of appeals facing similar benefit agreements. For example, in Bogue v. Ampex Corp., 976 F.2d 1319, 1322 (9th Cir.1992), ten executives of defendant Ampex were informed that the company would be sold. In order to retain those executives (at least temporarily), Ampex's parent company, Allied-Signal, established a program whereby it would pay the managers severance benefits if the purchaser did not offer them "substantially equivalent employment." While the plaintiff argued that ERISA did not preempt his claim under the severance program, the Ninth Circuit disagreed, relying in large part on Fort Halifax. "In this case, Allied-Signal, the program's administrator, remained obligated to decide whether a complaining employee's job was 'substantially equivalent' to his pre-acquisition job. Although the program, like the plan[ ] in Fort Halifax ... was triggered by a single event, that event would occur more than once, at a different time for each employee." 976 F.2d at 1323. In contrast to the one-time, routine payout facing the employer in Fort Halifax, the disbursements in Bogue would be discretionary and recurring. The court concluded:

There was no way to carry out that obligation with the unthinking, one-time nondiscretionary application of the plan administrators in Fort Halifax. Although its application was uncertain, its term was short, and the number of its participants was small, the program's administration required a case-by-case, discretionary application of its terms.... We hold that Allied-Signal was obligated to apply enough ongoing, particularized, administrative, discretionary analysis to make the program in this case a "plan."

Id.

Similarly, in Simas v. Quaker Fabric Corp., 6 F.3d 849 (1st Cir.1993), the First Circuit faced a Massachusetts "tin parachute" statute that required employers to pay severance benefits to employees who lost their jobs within 24 months of a corporate takeover. The payments were conditioned on the employee's eligibility for unemployment benefits; in other words, if an employee had been terminated for cause, he would not be entitled to severance benefits under the statute. The First Circuit held that ERISA preempted the "tin parachute" statute. "It may be that in some instances, a determination of eligibility would be straightforward and, in others, the employer would have to make its own judgment.... But in all events for at least two years after the takeover, ... the employer would have to maintain records, apply the 'for cause' criteria, and make payments or dispute the obligation." 6 F.3d at 853. While recognizing that Fort Halifax established no bright lines to determine whether ERISA preempted a state statute or employer promise, the court ultimately landed on the side of preemption because, "as in Bogue, the time period [24 months] is prolonged, individualized decisions are required, and at least one of the criteria [the 'for cause' standard] is far from mechanical." Id. at 854. See also Fontenot v. NL Industries, Inc., 953 F.2d 960, 963 (5th Cir.1992) (distinguishing single, non-ERISA-preempted "golden parachute" payment from ERISA-preempted plan requiring employer to analyze "the circumstances of each employee's termination ... in light of certain criteria"); James v. Fleet/Norstar Fin'l Group, Inc., 992 F.2d 463, 468 (2d Cir.1993) (employer's promise to give 60 days of additional pay did not implicate ERISA in part "because the simple arithmetical calculations and clerical determination that Fleet was required to make to ascertain each appellee's severance pay was a far cry from the 'ongoing, particularized, administrative, discretionary analysis' required ... in Bogue").

There is no middle ground in these cases; either Golden Cat's plan is preempted by ERISA or it is not. While it is not easy to draw a line in a case such as this one, "line drawing ... is necessary and close cases will approach the line from both sides." Simas, 6 F.3d at 854. Ultimately, what makes Golden Cat's plan subject to ERISA in this case is no different from those factors that convinced the courts in Simas and Bogue to find in favor of preemption, and, for that matter, that convinced the Supreme Court in Fort Halifax to find against it. Golden Cat's plan required it to exercise discretion on an ongoing basis; using certain criteria, the company could determine that some managers experienced substantial reductions in their job duties, while others did not. The plan required it to make that decision as often as once per manager, and potentially many additional times because each manager conceivably could make a claim more than once. As we noted above, we need not determine exactly how many agreements (similar to Collins') Golden Cat executed with its managers. Golden Cat tells us the number is at least 60; Collins conceded in the district court that there were others in addition to his own. The important point is not the exact number, but the fact that Golden Cat executed agreements (of whatever number) requiring it to make nonclerical "judgment calls." See Delaye v. Agripac, Inc., 39 F.3d 235, 238 (9th Cir.1994) (stating that a welfare plan subject to subjective criteria, rather than arithmetical calculations or clerical determinations, implicates ERISA).

One additional point convinces us that the district court properly exercised jurisdiction over Collins' claim. According to Collins, the company was obligated to pay a manager like him money under the agreement simply by virtue of the sale to Ralston, but only when and if the manager left the company. That might not occur for years, or as Collins' attorney characterized it at oral argument, sometime "down the road." It is unreasonable to conclude that Golden Cat could face such potential protracted liability without the advantage of an ongoing administrative scheme to satisfy its obligations. Prolonged individualized decision-making concerning benefits describes a plan subject to ERISA, and preempted by it.

B. Whether the Acquisition Itself Triggered a Payout Under the Agreement

Prior to trial, the district court granted Golden Cat's motion for partial summary judgment on Collins' claim that nothing more than the consummation of the sale of Golden Cat constituted his "termination" under the retention agreement (which in turn would have made him eligible to collect his lump sum benefits under the agreement). Collins' claim under the retention agreement is a matter of contract interpretation, Hickey v. A.E. Staley Mfg., 995 F.2d 1385, 1389 (7th Cir.1993), making it "particularly suited to disposition by summary judgment." Metalex Corp. v. Uniden Corp. of Am., 863 F.2d 1331, 1333 (7th Cir.1988). The district court determined that the language in the retention agreement was ambiguous; parts of it supported Collins' view that Ralston's acquisition itself triggered Golden Cat's duty to pay, while other parts supported Golden Cat's argument that Collins was entitled to the monies only if Ralston terminated him after the acquisition. The court resolved the ambiguity in favor of Golden Cat by relying on extrinsic evidence--mainly an affidavit from Golden Cat's president Frank Krum concerning his intent in creating the retention program and Collins' own concession in his deposition that he did not consider himself terminated at the time of the acquisition.

Our de novo review of the issue convinces us that the retention agreement is indeed ambiguous and unable to resolve Collins' claim by its own terms. Favoring Collins' interpretation is this language from the agreement: "the Company wishes to provide you certain benefits in case there is a Change In Control of the Company." The language implies that Collins would receive his benefits as soon as Golden Cat was sold (in this case to Ralston). But favoring Golden Cat is this language: "In the event that your employment with the Company terminates following a Change in Control ..." Collins wants to read "following" to mean "on account of" or "by reason of," while Golden Cat insists it is a timing mechanism, meaning the termination would occur after (not because of) the company's sale.

Golden Cat's interpretation avoids the prospect of paying its managers lump sum benefits simply by reason of the company's sale, a financial burden it insists it never intended to assume. Under Collins' theory, on the other hand (expressed at oral argument), each manager who signed a retention agreement would have to wait until he left the new company (Ralston) before collecting his payout, but he would nevertheless be entitled to the full amount at that time. That might not be for years. This means that Golden Cat would have to plan ahead for the prospect that it would owe a manager like Collins six months' salary (and other benefits), but it would not know when or even if that would happen. At the outset, we have our doubts that the parties intended Golden Cat to be subject to such protracted and uncertain liability. It would be administratively impractical and expensive. See Kennedy v. Electricians Pension Plan, IBEW No. 995, 954 F.2d 1116, 1123 (5th Cir.1992) ("An interpretation that would result in substantial unanticipated costs may be less likely to be legally correct.") (quoted in Krawczyk v. Harnischfeger Corp., 41 F.3d 276, 279 (7th Cir.1994)). It would also render the short-term job insecurity theory meaningless if an employee is retained for years at full pay before being terminated by the purchasing company.

But while practicality seems to rule out Collins' theory, the terms of the retention agreement are less decisive. Like the district court, we are unable to say that the retention agreement necessarily rules out either party's interpretation, making it unenforceable unless extrinsic evidence resolves the uncertainty. See Hickey, 995 F.2d at 1390 (facing ambiguity in ERISA plan and stating "[b]ecause the Plan language is ambiguous, we must look to the extrinsic evidence to determine whether the meaning of the term[s] ... remains ambiguous").

Extrinsic evidence convinces us that neither Golden Cat nor Collins intended that the company's sale itself would trigger Golden Cat's liabilities under the retention agreement. First, we note that Golden Cat's president, Frank Krum, who oversaw the drafting of the agreements, testified (by affidavit) that the agreements were intended to induce certain managers to remain with the company through the contemplated acquisition and afterwards. Golden Cat surely would be a less attractive purchase if its key managers (who could provide a period of transition for Ralston) quit on the day of the sale. On the other hand, managers might be inclined to take the initiative and seek other (perhaps more secure) jobs if they feared the new company would need fewer of them. According to Krum, the agreements perfected a tradeoff--the managers would accept the uncertainty involved in Ralston's acquisition in exchange for Golden Cat's commitment to compensate them in the event they were let go before a certain time. As Krum put it: "the retention agreements guaranteed those employees benefits to buffer the financial consequences of unemployment caused by any discharge on or before December 31, 1995."

Collins does not contest Krum's affidavit. Instead, his principal argument on appeal is that the company's sale legally severed his employment relationship with Golden Cat, which in turn triggered the company's duty to pay under the retention agreement. He offers no legal or documentary authority supporting his theory that Golden Cat's sale constituted a legal "termination" of his employment, but even assuming it did, he does not explain why this entitles him to a payout under the terms of the agreement. Curiously, Collins' theory demonstrates that he could be terminated in one of two distinct ways--via either Golden Cat's sale or the more obvious method, i.e., the buyer's decision to fire him at some time after the sale. This split contention only confuses the issue. Collins' argument does nothing to help us determine which interpretation the parties intended at the time they contracted. Does he claim to have been terminated twice? If so, which termination triggered a payout by Golden Cat?

Fortunately an additional piece of extrinsic evidence persuades us (as it did the district court) that the parties did not intend Golden Cat's sale to implicate the retention agreements. In his deposition, Collins testified that before mid-July 1995 (months after Golden Cat's sale), he did not consider himself "terminated" from either Golden Cat or Ralston. In other words, not even Collins considered the sale to constitute a "termination," as that term is used in the agreement. See National Diamond Syndicate, Inc. v. UPS, 897 F.2d 253, 261 (7th Cir.1990) ("A court may look to the contracting parties' own conduct in order to determine their understanding of ambiguities in the agreement."). In short, we conclude that extrinsic evidence (in particular, Krum's affidavit coupled with Collins' deposition testimony) resolves the ambiguity contained in the retention agreement. The parties did not intend that Golden Cat would be liable under the agreement simply by virtue of the sale to Ralston. Collins' argument that a sale constitutes a legal severance of the employment relationship, even if accurate, does not rebut this extrinsic evidence. In particular, it does not answer the question as to whether the parties intended to compensate Collins at the point of the sale, or at a point in the future if he was terminated by the buyer. See Hickey, 995 F.2d at 1391 (stating that an ambiguity is not "sufficient to establish a genuine issue of triable fact"; "the plaintiffs must refute the defendant's affidavits" rather than resting on their pleadings). The unrebutted external evidence discussed above does answer this question, which is why we reject Collins' argument that he was entitled to a payout under the retention agreement simply by virtue of Golden Cat's sale. If Golden Cat owed Collins benefits under the retention agreement, it could only be because Ralston terminated his employment before December 31, 1995 by substantially reducing his job responsibilities.

C. Whether Ralston Substantially Reduced Collins' Duties

The final issue on appeal concerns the district court's grant of judgment on partial findings pursuant to Rule 52(c). See Fed.R.Civ.P. 52(c). The district court granted judgment in favor of Golden Cat after Collins completed his case-in-chief. We review the court's factual determinations under the clearly erroneous standard, giving "due regard to the opportunity of the trial court to judge the credibility of the witnesses." Zeige Distributing Co. v. All Kitchens, Inc., 63 F.3d 609, 612 (7th Cir.1995). Under these circumstances, we do not lightly set aside the district court's judgment; rather, we reverse "only if, after reviewing the record, we are left with the firm belief that the district court made a mistake." Id.

Collins had two theories at trial as to why he was entitled to a payout under his retention agreement. First, Collins maintained that Ralston substantially reduced his job responsibilities (meaning he was "terminated" as that term is defined in the agreement). Without more we will assume the Dunkirk position offered to Collins would have meant a lesser job and fewer responsibilities. The Dunkirk region covered much of the east coast and upstate New York, but considerably less than the entire eastern United States, which is what Collins was used to overseeing. There is a possibility that although he would be responsible for less territory he might have taken on new products (pet food), thus substantially increasing sales volume. But that is not clear. Still, the problem for Collins is that he never accepted the Dunkirk offer in the first place (instead, he resigned), and there is no evidence in the record that the company offered him the position on a "take it or leave it" basis. On appeal, Collins claims that the transfer to Dunkirk was a "fait accompli," but nothing in the record supports this position. While he does refer us to a July 1995 memo linking his name and phone number to the company's Dunkirk region (the same memo that refers to pet food as an added sales product), it would be a leap to say that the memo proves the transfer was completed before he resigned. We might as easily interpret the memo to suggest that Collins would be the Dunkirk region's contact person (according to the memo, he could be reached at a South Bend, Indiana phone number), at least until he responded to the company's formal Dunkirk offer.

At bottom, while it looks likely that Ralston intended to make Collins its Dunkirk regional sales manager, the fact remains that he did not stay around long enough to find out. He anticipated his reassignment, but he avoided the possible step backwards in his career by resigning before the company's offer could become an ultimatum. Leaving the company at that point may have made perfect sense, but the agreement was conditioned on an actual--not anticipated--reduction in his job responsibilities. Like the district court, we might agree that Ralston's intimations suggest it was headed that way, but it would be just a hunch, and we cannot rewrite the parties' agreement to eliminate an important condition precedent. See Edwards v. Great-West Life Assurance Co., 20 F.3d 748, 749 (7th Cir.1994) (finding surviving spouse not entitled to death benefit under ERISA plan because deceased had not satisfied condition precedent under plan; namely, deceased was not "actively at work" on the date his insurance was to take effect).

Finally, Collins claims that his transfer to the Dunkirk region constituted a termination under the retention agreement. Once again, the theory fails because Collins cannot (and did not) prove Ralston transferred him. Indeed, there is almost no evidence that the company actually decided Collins would be transferred to the east coast, which may be why Collins himself concedes the relocation had only occurred in "practical effect." The difference between being actually transferred and only "practically" transferred is the difference between a viable claim and one subject to Rule 52(c). That was the conclusion of the district court, too, which is why we affirm its decision in this respect and all others.

Affirmed.

ILANA DIAMOND ROVNER, Circuit Judge, dissenting.

Fort Halifax advises that "ERISA pre-emption analysis 'must be guided by respect for the separate spheres of governmental authority preserved in our federalist system.' " Fort Halifax Packing Co. v. Coyne, 482 U.S. 1, 19, 107 S. Ct. 2211, 96 L. Ed. 2d 1 (1987) (quoting Alessi v. Raybestos-Manhattan, Inc., 451 U.S. 504, 522, 101 S. Ct. 1895, 68 L. Ed. 2d 402 (1981)). When we federalize state law causes of action, we deprive the States of opportunities to "address uniquely local social and economic problems." Fort Halifax, 482 U.S. at 19, 107 S. Ct. 2211. Because of this concern, Congress was careful to limit ERISA pre-emption to those instances where employers might be subject to conflicts between the laws of the various States or conflicts between state and federal regulations in relation to benefit plans. See Alessi, 451 U.S. at 522, 101 S. Ct. 1895 (the exercise of federal supremacy is not lightly to be presumed; pre-emption of state law by federal statute is not favored in the absence of persuasive reasons).

Fort Halifax teaches that not every benefit constitutes a benefit plan, and that only plans are controlled by ERISA. 482 U.S. at 11, 107 S. Ct. 2211. The reason for ERISA preemption, the Supreme Court explains, is to avoid subjecting an employer to the diverse regulatory requirements of the various States:

An employer that makes a commitment systematically to pay certain benefits undertakes a host of obligations, such as determining the eligibility of claimants, calculating benefit levels, making disbursements, monitoring the availability of funds for benefit payments, and keeping appropriate records in order to comply with applicable reporting requirements. The most efficient way to meet these responsibilities is to establish a uniform administrative scheme, which provides a set of standard procedures to guide processing of claims and disbursement of benefits. Such a system is difficult to achieve, however, if a benefit plan is subject to differing regulatory requirements in differing States.

Id., 482 U.S. at 9, 107 S. Ct. 2211. One of Congress' concerns was that employers would reduce benefits if faced with the costs of complying with varying requirements by the States. Id., 482 U.S. at 10, 107 S. Ct. 2211.

In Fort Halifax, the Supreme Court deemed not pre-empted a Maine statute that established a severance benefit for employees who lost their jobs due to plant closings or relocations. The Court reasoned that Congress' concern about subjecting employers to differing regulations arose only "with respect to benefits whose provision by nature requires an ongoing administrative program to meet the employer's obligation." 482 U.S. at 11, 107 S. Ct. 2211. ERISA pre-emption thus applies only to plans rather than to benefits because "[o]nly a plan embodies a set of administrative practices vulnerable to the burden that would be imposed by a patchwork scheme of regulation." 482 U.S. at 11-12, 107 S. Ct. 2211. In the case of the Maine statute, a requirement that an employer provide a severance benefit of one week's pay for every year of employment was not a plan because it contemplated only a one-time, lump-sum payout, triggered by a single event. The Court noted that the employer assumed no responsibility to pay benefits on a regular basis and faced no periodic demand on its assets that would create a need for financial coordination and control. Further, the benefit was predicated on the occurrence of a contingency that might never occur, that being the closing or relocating of the plant. The Court thus declined to find pre-emption because "the theoretical possibility of a one-time obligation in the future simply creates no need for an ongoing administrative program for processing claims and paying benefits." 482 U.S. at 12, 107 S. Ct. 2211.

The majority here relies on the existence of multiple severance contracts to conclude that Collins' state law contract action is pre-empted by ERISA. Yet nothing in the record supports the majority's conclusion that there are as many as sixty agreements, each requiring consistent discretionary decisions to be made over time. The majority distinguishes the Golden Cat severance agreements from the statutory benefits in Fort Halifax, positing that "Golden Cat faced the prospect of multiple payments to various managers, at different times and under different circumstances." Supra, at 595. The majority assumes that the company would be expected to process claims like Collins' in a consistent manner, which in turn would require an administrative scheme of procedures and record keeping. Supra, at 596. The most important distinguishing factor, the majority explains, is not the number of agreements per se, but the fact that Golden Cat executed multiple agreements requiring it to make non-clerical judgment calls. Supra at 597.

But only one contract appears in the record, and that is the contract between Collins and Golden Cat. Both Collins and Golden Cat advert to the existence of other contracts, but neither their number nor their terms are part of the record. The district court noted that, although the defendants claimed in their pleadings that there were sixty agreements, they did not present any evidence of the existence or terms of any agreements other than the one at issue. Collins v. Golden Cat Corp. and Ralston Purina Co., No. 3:95-CV-943RM, slip op. at 9 n. 2 (N.D.Ind. May 2, 1996). The court also pointed out that any admissions by Collins regarding the other agreements were not based on independent knowledge and included no information about the terms of the agreements. Id. That should have been the end of the matter, for it was defendants' burden to establish the basis for removal of the case from state to federal court and they clearly failed in that burden.1 But venturing beyond the record, the district court found that there was a second, "almost identical" agreement between Robert Orr, another key employee, and Golden Cat. Apparently, Mr. Orr had also sued Golden Cat for breach of his severance agreement, and his case had also been assigned to Judge Miller. Based on his awareness of that second contract, Judge Miller found that there were "two, and as many as about sixty, agreements substantially similar to that of Mr. Collins." Id. The district court premised its determination of pre-emption on that factual finding. Without any evidentiary support in this record, that conclusion was error, an error that is regrettably repeated here by the majority.

At oral argument, Ralston and Golden Cat pointed to other evidence of multiple contracts in the record, all of which came after the trial court decided that jurisdiction existed, and none of which supports their claim that there were sixty "similar" contracts, each requiring consistent discretionary decisions.2 In particular, they cited the trial testimony of three former key employees and executives who testified that they each signed a severance agreement. The testimony of these executives in fact proves that these separate agreements were not meant to be subjected to a single set of administrative procedures, but contained differing terms and conditions for payment. One executive testified that his agreement was "substantially similar" to and "more lucrative" than Collins' agreement (Tooker, Trial Transcript at I-58, I-82), another that his agreement was "different" and "perhaps" more lucrative than Collins' agreement (Ramey, Trial Transcript at I139), and a third stated that his agreement was "similar" to Collins' agreement except that it provided a smaller benefit (Brew, Trial Transcript at II-245). At the same time, the terms of all of the agreements were designated confidential by Golden Cat and there was no testimony regarding the terms or conditions under which these employees would receive their severance benefits. As far as we know, the other employees entered into agreements that required no discretionary decisions whatsoever. What is clear, however, is that the contracts vary in their terms and necessarily do not envision consistent outcomes.

As the only agreement included in the record, Collins' agreement must provide the sole basis for jurisdiction. But nothing in Collins' agreement implicates the need for an ongoing administrative scheme with a set of procedures to determine payments. Fort Halifax, 482 U.S. at 11, 107 S. Ct. 2211. Nor does his agreement create a periodic demand on the company's assets that would create a need for financial coordination and control. Id., 482 U.S. at 12, 107 S. Ct. 2211. And as in Fort Halifax, the benefit was predicated on the occurrence of a contingency that might never occur, in this case the termination of an employee if and when the company changed hands. This "theoretical possibility of a one-time obligation in the future simply creates no need for an ongoing administrative program for processing claims and paying benefits," and thus is not preempted by ERISA. Id. In fact, if the terms of Collins' agreement are sufficient to establish ERISA pre-emption, then I believe that the line drawn by Fort Halifax has been erased, for any agreement requiring even a single discretionary decision regarding eligibility will trigger preemption, notwithstanding the fact that the contract governs the rights of only one employee.

None of the cases relied upon by the majority involved a single benefits contract between an employer and one employee. The most analogous case cited by the majority is Bogue, where the employer established a "Special Compensation Program for Designated Key Executives." 976 F.2d at 1321. Unlike the individual, quite possibly unique agreement between Golden Cat and Collins, the employer in Bogue implemented a single severance plan governing a number of employees. Under the plan, employees who were terminated or not offered "substantially equivalent" employment were eligible for the payments. Id., 976 F.2d at 1323. The Bogue court found that the plan required a case-by-case discretionary analysis, and thus an administrative scheme, to determine eligibility. Although Collin's contract contains a similar contingency, his is the only contract in the record. Thus, unlike the plan in Bogue, there is no evident need for an ongoing administrative scheme to handle multiple transactions.3 The Bogue court also noted that the termination event could occur more than once and at a different time for each employee. Id.

Since Bogue, the Ninth Circuit has clarified that not every discretionary decision in a contract is enough to trigger ERISA pre-emption. Backing off of Bogue's sweeping language, the court found that a severance contract between an employer and a single employee did not implicate an ongoing administrative scheme, even when the contract required monthly payments over a two-year period. Delaye v. Agripac, Inc., 39 F.3d 235, 237-38 (9th Cir.1994), cert. denied, 514 U.S. 1037, 115 S. Ct. 1402, 131 L. Ed. 2d 289 (1995). The contract obligated the employer to pay the employee his regular salary prorated to the date of his termination if he was terminated for cause, or a fixed monthly amount for twelve to twenty-four months based on a formula if he was terminated without cause. Id., 39 F.3d at 237. The court distinguished Bogue.

In Velarde v. PACE Membership Warehouse, Inc., 105 F.3d 1313 (9th Cir.1997), the Ninth Circuit limited Bogue even further. The court analyzed a severance agreement signed by twenty-five employees, that contained a provision requiring the employer to determine whether the employee was terminated for cause in calculating the size of the payment due. Although this decision required the exercise of managerial discretion, the court noted that "this minimal quantum of discretion [was not] sufficient to turn a severance agreement into an ERISA plan." 105 F.3d at 1317. Contrary to the employer's contention that some "modicum of discretion" was enough to trigger ERISA preemption, the court explained that the "key to our holding in Bogue was that there was 'enough ongoing, particularized, administrative discretionary analysis,' ... to make the plan an 'ongoing administrative scheme.' " Id. (emphasis added by Velarde court). The court concluded that the level of discretion necessary to determine whether an employee was terminated for cause was "slight," and failed to rise to the level of ongoing particularized discretion necessary to transform a simple severance agreement into an ERISA benefits plan. Id. The instant case requires an exercise of discretion that the Ninth Circuit found sufficient to trigger ERISA pre-emption only when it is part of a formal plan covering a large number of employees. Bogue, 976 F.2d at 1323. But Delaye and Velarde demonstrate that the need for some exercise of discretion, even when required on multiple occasions (and when payments are made over an extended period of time), is not always enough to trigger ERISA pre-emption.4 Because we have only Collins' contract to consider, I do not believe we can find the need for an ongoing administrative scheme.

True, this Court has held that a single agreement can constitute an ERISA plan. See Cvelbar v. CBI Illinois Inc., 106 F.3d 1368, 1376 (7th Cir.1997), cert. denied, --- U.S. ----, 118 S. Ct. 56, 139 L. Ed. 2d 20 (1997). But in the only case in which we have so held, the agreement required monthly payouts over a three-year period following termination, implicating the need for financial oversight and planning. Id., 106 F.3d at 1376-77. The plan also involved a number of discretionary decisions. For example, employees terminated for cause were not eligible for benefits. We know from Delaye and Velarde that the Ninth Circuit does not consider this enough to trigger ERISA pre-emption. But the contract also required that the employer review the contract annually to determine if the employee was complying with the terms of a non-compete agreement. Moreover, the amount of the payments could not be calculated mechanically but depended on a number of factors that created a need for administrative involvement. In light of all of these factors, including the extended and periodic demand on assets creating a need for financial oversight, and because of the monitoring of the non-compete pact, we held that the agreement implicated an ongoing administrative scheme. Id.

Collins' agreement, in contrast, required a onetime payout that could be calculated mechanically, similar to the payment in Fort Halifax. The payment would occur, if at all, during a relatively short period of time, because the agreement expired approximately 14 months after it was signed.5 Indeed, as in Fort Halifax, the condition precedent to the payment might never occur. Although Collins' contract purports to require discretionary decisions, Golden Cat did not retain discretion to interpret or apply plan terms. Thus, the contract terms would be strictly construed against Golden Cat as the drafter of the agreement. Moreover, Collins' contract contains no reference to the other contracts, or to any standards that would call for discretion and commensurate deference on our part. The defendants have, in effect, contracted themselves out of ERISA. In any event, even if executing the contract involved some discretionary decisions, nothing in the agreement created the need for an ongoing administrative scheme. Golden Cat was never in danger of being held to two different sets of standards by two different state regulatory schemes, because execution of Collins' agreement never required the ongoing administrative scheme envisioned by the Court in Fort Halifax. ERISA pre-emption is, therefore, inapplicable.

Federal courts are courts of limited jurisdiction, and the party seeking removal to federal court bears the burden of proving that jurisdiction exists. County Collector, 96 F.3d at 895. The defendants here have utterly failed in that burden, and I believe the majority's reliance on the existence of sixty contracts that do not appear anywhere in the record is misplaced. I would remand for the district court to hold a hearing to determine if the defendants could prove facts in support of jurisdiction. If the defendants put the terms of the alleged sixty contracts into evidence and demonstrated a need for an ongoing administrative scheme to conduct the plan, I would agree that, as in Bogue, ERISA pre-empts Collins' state law action. Until Golden Cat presents that evidence, federal jurisdiction is lacking. Therefore, I respectfully dissent.