John E. DOWLING, Appellant v. PENSION PLAN FOR SALARIED EMPLOYEES OF UNION PACIFIC CORPORATION AND AFFILIATES; Roy Schroer, Named Fiduciary-Plan Administration of the Pension Plan For Salaried Employees of Union Pacific Corporation and Its Affiliates; Edwin A. Willis, Delegate of the Named Fiduciary-Plan Administration of the Pension Plan For Salaried Employees of Union Pacific Corporation and Affiliates; The Northern Trust Co in Its Capacity as Trustee of the Pension Plan For Salaried Employees of Union Pacific Corp and Affiliates; Roy Schroer, Administrator of the Supplemental Pension Plan For Officers and Managers of Union Pacific Corporation and Affiliates; Edwin Willis, Delegate of the Administrator of the Supplemental Pension Plan For Officers and Managers of Union Pacific Corporation and Affiliates; Union Pacific Corporation; Supplemental Pension Plan For Officers and Managers of Union Pacific Corporation and Its Affiliates, 871 F.3d 239

OPINION OF THE COURT VANASKIE, Circuit Judge. Retirement plans can be complex documents that span hundreds of pages with numerous peculiarities. But when do a plan’s terms move from merely complex to ambiguous? That is the question in this pension plan dispute. Former Union Pacific employee John Dowling is covered by a 277-page retirement plan composed of introductory material, 19 articles of content, and various appendices—none of which explicitly address Dowling’s precise situation.

When Dowling retired, the plan administrator interpreted the plan to provide Dowling with a lower monthly payment than he expected. Dowling challenged the administrator’s decision as contradicting the plan’s plain language, but' the District Court found the plan ambiguous and the administrator’s

interpretation reasonable. Dowling appealed, and the dispute now centers on three issues: the text of the plan, our standard of review, and whether a conflict of interest alters the outcome. Because the plan’s terminology, silence, and structure render it ambiguous, the plan accords the plan administrator discretion to interpret ambiguous plan terms, and the mere existence of a conflict of interest is alone insufficient to raise skepticism of the plan administrator’s decision, we will grant deference to the plan administrator and affirm. I.

Dowling was hired at age 41 by Appellee Union Pacific Corporation in 1988, where he served in the high-ranking position of Vice President for Corporate Development. Just seven years later, Dowling’s life was dealt a severe blow when he was diagnosed with multiple sclerosis. By 1997, Union Pacific had determined that Dowling possessed a “Total Disability,” because he was “unable to work at any job.” (App. 153, 520.) That decision made Dowling eligible for long-term disability benefits that he could receive for the duration of his disability or until he reached age 65 in 2012, whichever came first.

When Dowling turned 65 in 2012, the long-term disability benefits stopped, and he began to draw on his Union Pacific retirement. His credited years of service for purposes of calculating his pension benefit included the 15 years he. received disability benefits.

Union Pacific’s plan administrator interpreted the plan to require that Dowling’s pension be calculated in accordance with what the administrator saw as applicable to disabled plan participants: Instead of calculating Dowling’s pension based on Dowling’s last ten years of actual work—ending in 1997—the administrator operated as if Dowling had worked and been paid his final base salary—$208,000 per year— for his credited years of service, up until his retirement in 2012, even though Dowling had not in reality worked during that period. Under the administrator’s

interpretation, Dowling was entitled to a monthly pension payment of $7,006.96. Dowling objected to the calculation and filed a claim via the plan’s administrative procedures, asking for a benefit increase. He argued the plan required his pension payment to be based on his ten years of income prior to 1997, when he became disabled and stopped working, and not. a hypothetical income stream for the ten years prior to his 2012 formal retirement date.

If Dowling’s theory about the 1987 to 1997 window were correct, then Union Pacific would owe Dowling a much higher monthly payment because during that earlier period Dowling received significant performance bonuses on top of his base salary. Dowling lost his administrative claim, exhausted his administrative remedies, and filed this action against Union Pacific and the other Appellees in the Eastern District of Pennsylvania. Dowling sought a declaratory judgment stating his rights and liabilities, pursuant to ERISA. 29 U.S.C. § 1132(a)(1)(B). The District Court granted

summary judgment to Union Pacific, holding that the plan administrator’s

interpretation of the plan was not unreasonable, and Dowling appealed. II. Dowling’s retirement is governed by Union Pacific’s “Pension Plan for Salaried Employees.” The plan is a substantial legal document: it opens with seven pages of preliminary information, then continues across 133 pages of content divided into 19 articles. At the back are 137 pages of appendices, schedules, exhibits, and tables. Out of all this material, two key factors largely determine the amount of a plan participant’s pension payment: compensation and service.

The compensation factor is called “Final Average Compensation” and is defined as a plan participant’s average monthly salary during his or her three highest-earning years—the “high-three”— during the ten years “immediately preceding ... the last date on which [the plan participant] is a Covered Employee.” (Plan § 2.35, App. 144.) The service factor is the participant’s “Credited Service,” which refers to the amount of time a plan participant spent as a “Covered Employee.” Thus, for the run-of-the-mill plan participant, pension calculation is easy: it is based on the years the individual spent at work, and his average paycheck during his three highest-earning years of his final ten years of employment.

But the plan treats a disabled participant differently.

For Credited Service, instead of stopping the accumulation of service when the disabled participant stops work, as is the case with the typical participant, the plan permits disabled participants to accumulate service during their pre-retirement, post-disability years, “as if’ they remained Covered Employees until their date of retirement—even though they may have stopped working years earlier. (Plan §§ 4.02(c)(2), 6.05, App. 157, 178; see also Plan § 2.40(a)(5), App. 148 (noting the “Hours of Service” credited to not-working disabled participants).) For Final Average Compensation, the plan’s application to disabled participants is less clear, with the confusion largely centering on the plan’s use of the term “absence.” During an “absence” from work, a plan participant- is “deemed to have received” for the duration of their absence “Compensation at the base pay rate in effect” prior to the absence. (Plan § 2.18(a)(8)(C), App. 139.) Thus, for purposes of pension calculation, the rate of pay during an employee’s unpaid absence is deemed to be their pay prior to the absence.

But does the definition of “absence” extend to time away from work due to disability? The plan is not clear. The lengthy

definitions

section does not define “absence.” (See Plan § 2.01-2.76, App. 131-54 (defining 76 terms, over 23 pages, but providing no definition for “absence”).) The plan does define two particular types of absences—absences for temporary family medical leave, and temporary approved absences, (Plan § 2.10, App. 134-35 (defining “Approved Absence”); Plan § 2.10B, App. 135 (defining “Approved Non-HCE Absence”))—and it references two more types of “absences” in the “Hours of Service” section, which details how much time should be credited in various scenarios. (Plan § 2.40(a)(4), App. 147-48 (listing as examples Approved Absences, temporary lay-offs, military leave, and Approved Non-HCE Absences)).

A departure from work due to disability is not one of the four examples of “absehce” listed. Additionally, a different subsection in the “Hours of Service”

section addresses the hours credited during “Total Disability”; that subsection is directly below the “absences” subsection, and does not mention “absences.” (Plan § 2.40(a)(5), App. 148.) More generally, the plan grants the plan administrator the authority “to determine all questions of ... eligibility, ... to make factual determinations, ... to construe and interpret the provisions of the Plan, to correct defects and resolve ambiguities therein, and to supply omissions thereto.” (Plan § 13.02(f), App. 242.) The plan is funded entirely by Union Pacific; contributions are neither required nor accepted from plan participants. (Plan § 12.01-03, App. 232.) During the times relevant here, the plan’s designated administrators, including Barbara Schaefer [Schaefer isn’t listed as an Appellee], Roy Schroer, and Edwin A.

Willis, were also Union Pacific employees or officers. Schaefer and Schroer each held the title of Vice President for Human Resources, and Willis was Assistant Vice President for Compensation and Benefits. III. The District Court had jurisdiction under 28 U.S.C. § 1331 and 29 U.S.C. § 1132(e). We have jurisdiction under 28 U.S.C. § 1291. IV. Federal courts review the decisions of ERISA plan administrators under standards derived from “principles of trust law,” fin that the plan document itself determines the appropriate level of review. Conkright v.

Frommert, 559 U.S. 506, 512, 130 S.Ct. 1640, 176 L.Ed.2d 469 (2010) (quoting Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 111, 109 S.Ct. 948, 103 L.Ed.2d 80 (1989)). In the default scenario, a plan administrator’s “denial of benefits ... is to be reviewed under a de novo standard.” Id. (quoting Firestone, 489 U.S. at 115, 109 S.Ct. 948). But if the plan document “gives the administrator or fiduciary discretionary authority to determine eligibility for benefits or to construe the terms of the plan,” then the Court reviews the administrator’s decision on a more deferential basis.

Id. (quoting Firestone, 489 U.S. at 115, 109 S.Ct. 948). This case falls in the latter scenario, because the Union Pacific plan explicitly grants the administrator the ability to determine benefit eligibility and to “construe and interpret” the plan’s provisions. (Plan § 13.02(f), App. 242.) In such circumstances, we will not set aside the administrator’s

interpretations of “unambiguous plan language” as long as those

interpretations are “reasonably consistent” with the plan’s text, Fleisher v. Standard Ins. Co., 679 F.3d 116, 121 (3d Cir. 2012) (quoting Bill Gray Enters. v. Gourley, 248 F.3d 206, 218 (3d Cir. 2001)), and we will only disturb the administrator’s

interpretations of ambiguous plan language when those

interpretations are “arbitrary and capricious,” id. (quoting McElroy v. SmithKline Beecham Health & Welfare Benefits Tr. Plan, 340 F.3d 139, 143 (3d Cir. 2003)). Whether plan language is ambiguous or unambiguous is itself a question of law subject to our de novo review, with the definition of ambiguity being language that is “subject to reasonable alternative

interpretations.” Id. (quoting Taylor v. Cont’l Grp. Change in Control Severance Pay Plan, 933 F.2d 1227, 1233 (3d Cir. 1991)). Many cases will therefore turn, as this one does, on whether a proffered

interpretation of plan language is “reasonable.” Courts apply this deferential standard for at least two good reasons. First, courts have an obligation to give effect to the plan-drafter’s intentions, because “ERISA abounds with the language and terminology of trust law,” Firestone, 489 U.S. at 110, 109 S.Ct. 948, and the hallmark purpose of trust law is “to accomplish the settlor’s intentions,” Restatement (Third) of Trusts, Foreword (Am. Law Inst. 2003).

Here, since the plan-drafter explicitly specified that the plan administrator should possess the ability to interpret terms, we must be deferential because the de novo alternative—examining each of the plan administrator’s legal decisions anew—would undermine rather than give effect to the drafter’s wishes. Second, giving deference pays heed to Congress’s concern for not discouraging employers in their adoption of ERISA plans.

Existing federal statutes do not require employers to offer employee-retirement plans, and when Congress passed ERISA to make retirement programs fairer, it also worked to reduce the burdens of its new regulations and to keep in check disincentives that might discourage an employer from offering a retirement plan at all—disincentives such as high “administrative costs” and “litigation expenses.” Conkright, 559 U.S. at 517, 130 S.Ct. 1640 (quoting Varity Corp. v. Howe, 516 U.S. 489, 497, 116 S.Ct. 1065, 134 L.Ed.2d 130 (1996)). To that end, the Supreme Court has recognized that employers often commit the power of

interpretation to a plan administrator because doing so serves the employer’s interests of efficiency, predictability, and uniformity—interests ERISA seeks to protect. Id. Thus, when a court pays deference to the administrator at the request of the plan-drafter, the court acts in congruence with Congress’s wishes. y. We now turn to the debate over the meaning of the plan’s terms.

Union Pacific argues the plan requires the administrator, in light of Dowling’s status as a disabled participant, to “deem” Dowling to have been paid his final base salary from the moment he became totally disabled in 1997 until he retired in 2012, and then calculate his Final Average Compensation from those deemed earnings based on the ten-year window from 2002 to 2012. That is the approach the plan administrator took and the District Court found reasonable. Dowl-ing, on the other hand, argues Union Pacific’s

interpretation involves too many interpretive gymnastics: Dowling stopped working and earning a salary in 1997 and his ten-year window must accordingly look backward from 1997, even though he continued to accrue credited service until 2012. We pass no judgment as to which proffered

interpretation is best, because at least three aspects of the plan combine to make it ambiguous and each party’s

interpretation reasonable. The first aspect is the plan’s use of the word “absence.” Is & person who is not at work due to a disability “absent”? Union Pacific says yes; Dowl-ing says no. If yes, then Plan § 2.18(a)(3)(C)—which “deems” a participant to have been paid during an “absence”—can reasonably be read as requiring the administrator to deem disabled persons to be paid their base salaries for the duration of their disability, up until their retirement date, for purposes of calculating Final Average Compensation.

That would mean that Dowling should be counted as earning his base salary up until 2012, as the administrator found. But on the other hand, if time spent not working due to disability is not an “absence,” then the plan’s language favors Dowling and disabled participants are not “deemed” to receive any pay at all after they leave work, and the only reasonable approach would be to calculate Final Average Compensation by looking backward from the date the person became totally disabled and stopped working—1997 in Dowling’s case.

The plan administrator adopted the former approach, that missing work due to disability does in fact constitute an “absence.” Dowling argues that the administrator went too far in extending the definition of “absence” to cover indefinite departures from work, and that only more limited short-term departures should count. Because “absence” is given no specialized meaning in the plan’s

definitions section, the word must be interpreted in accordance with its generally prevailing meaning, Restatement (Second) of Contracts § 202(3)(a) (Am. Law Inst. 1981) (“Rules in Aid of

Interpretation”), and its generally preváiling meaning is broad—it means nothing more specific than the state of being “[n]ot present,” Absent, Oxford English Dictionary (3d ed. 2009). While a person who misses one day of work can surely be said to be “not present,” and thereby “absent,” so can a person who endures an indefinite departure from work, whether due to disability or some other reason. Michael Jordan’s three-year hiatus from basketball was an “absence,” according to the New York Times Rick Moranis’s 18-year disappearance from film was an “absence” in the eyes of the Hollywood Reporter.

And Miles Davis’s five-year departure from music in the 1970’s was an “absence” as well, as told by National Public Radio. Thus, given that the generally prevailing meaning of “absence” permits the word to be used to refer to indefinite departures from the workplace, we find the administrator’s use of the same word in the same manner to be reasonable. Dowling argues we must reject the administrator’s

interpretation of the word “absence,” because the plan requires a specialized, narrower

interpretation that applies only to shorter departures from work. For authority, Dowling points to Plan § 2.40. That

section has separate subsections for time spent in “Total Disability” and time spent in four specific types of “absences.” (Plan § 2.40(4), (5), App. 148.) Dowling suggests this separate treatment indicates the two terms are mutually exclusive—time spent in “Total Disability” is not an “absence,” and vice versa. But the plan’s text does not go so far.

Section 2.40 does not purport to define “absence”; it just lays out four nonexclusive scenarios that fit the definition of “absence.” Dowl-ing’s argument is not without persuasive force—to the contrary, it is a reasonable one. But it would be a stretch to say his

interpretation is the only reasonable approach, to the exclusion of the plan administrator’s. The second aspect that contributes to the plan’s ambiguity is its silence on how to calculate Final Average Compensation specifically for disabled participants. The plan has a default method of pension-plan calculation and an exception to that default for disabled participants, in two relevant respects: (1) the availability of a pension, and (2) the calculation of credited service.

Compare Plan § 4.02(a), App. 155 (laying out the “General Rule” for “Credited Service”), and Plan § 6.01-03, App. 173-74 (laying out general rules for retirement benefits), with Plan § 4.02(c)(2), App. 157 (providing special rules for disabled participants’ credited service), and Plan § 6.05, App. 178 (providing special rules for disability retirement benefits). The plan, however, leaves a gaping hole as to whether the default-and-exeeption pattern continues for calculation of the Final Average Compensation for disabled participants. Given this gap, the plan administrator faced a choice: (

a) calculate Dowling’s Final Average Compensation in line with the default scheme, even though disabled participants are explicitly treated as sui gen-eris with respect to pension-availability and credited service, or (

b) calculate Dowl-ing’s Final Average Compensation in line with the two disabled-participant exceptions that treat his formal retirement date like his' final day of work, even though nothing in the plan says to do as much. Given the silence, we cannot say that either approach is unreasonable.

Dowling, however, argues the silence can only be read in one way, to foreclose any deviation from the default scheme of determining Final Average Compensation, because of the expressio unius canon of construction: since the plan explicitly provides an exception for disabled participants in two respects, and says nothing explicit for the Final Average Compensation, we must assume that the lack of explicit terms for the latter scenario indicates that no such exception exists.

That argument might win the day if we were reviewing the plan de novo, but the expressio unius canon cuts the opposite way when we are paying deference to a plan administrator, because when a plan administrator interprets a text that contains a “mandate in one

section and silence in another,” the silence “often suggests ... simply a decision not to mandate any solution ..., i.e., to leave the question” open to the reasonable

interpretation of the administrative decisionmaker. Van Hollen, Jr. v. FEC, 811 F.3d 486, 493-95 (D.C. Cir. 2016) (quoting Catawba Cty. v. EPA, 571 F.3d 20, 36 (D.C. Cir. 2009) (finding in the context of Chevron deference that the expressio uni-us canon counsels against the court disturbing an agency’s

interpretation)). Such is the case here. Dowling’s criticism of the plan administrator’s approach is again not necessarily without merit, but when granting deference “we do not demand the best

interpretation, only a reasonable one,” id. at 494, and the plan’s silence suggests the plan administrator’s approach is not unreasonable. The third aspect that renders the plan ambiguous is its structure. See Zheng v. Gonzales, 422 F.3d 98, 114-16 (3d Cir. 2005) (looking to “text and structure” to determine ambiguity). Here, the relevant plan terms are structured into several “Articles,” three of which are relevant here:

Article II provides “Definitions,”

Article IV describes the “Crediting of Service,” and

Article VI lays out “Retirement Benefits.” Across these articles, the plan effectively provides two sets of rules, as noted above: a default scheme for the typical participant, and exceptions for disabled participants. The default scheme and its exceptions are not neatly laid out in one article; they are scattered across all three articles, with bits and pieces in various sections and subsections, and the operation of it all must be determined by cross-referencing the various articles, sections, and subsections, and reading them together. It is quite the legal task.

This buckshot distribution of relevant terms does little to clarify the disputes over the text and contributes to our finding that the at-issue provisions are ambiguous. We also take pause to note two counter-intuitive aspects of Dowling’s proposed

interpretation. First, Dowling wants all the benefits and none of the detriments of an artificial delay in the date he left work. When it comes to Credited Service, he is content that the administrator deemed him to have worked an extra fifteen years of time, from 1997 to 2012, even though he did not, but when it comes Final Average Compensation, he disapproves of the administrator taking the same approach and deeming him to have received a fictional salary over the same unworked period.

It should not be minimized how beneficial the first aspect of this scheme was for Dowl-ing: if he had not received Credited Service for the same post-disability pre-retirement period that he does not want to be deemed to have been paid a salary, one estimate suggests his monthly payment would be only about 75% of the current payment—an amount that Dowling believes already too low. (App. 118-19 (Willis’s 1997 estimates of various scenarios).) The point is that Dowling likes the fictional delay when it benefits him for purposes of Credited Service, but dislikes it when it hurts him on Final Average Compensation.

In other words, Dowling suggests that we read one provision two different ways, both to his advantage. But there is nothing unreasonable about harmonizing Credited Service with the calculation of Final Average Compensation. Additionally, for the typical disabled participant, Dowling’s position is likely worse and the administrator’s better. For an employee whose base salary is the near-total source of income, the salary that employee receives in his or her final year of work may often be the highest of his or her career.

For Dowling, however, that was not the case, due to his high-ranking status and incentive bonuses that made up a hefty portion of his overall compensation. Thus, by deeming him to receive only his base salary and no bonus over the final ten years of his Credited Service, Dowling was deprived of his three actual highest-earning years. But for the typical employee whose pay comes mainly or exclusively from salary, with raises arriving in yearly step increases, it is Union Pacific’s

interpretation that is best. To illustrate, imagine an employee who is paid a base salary, no bonus, and receives a step-salary, increase every year from Year 1 to Year 10. In Year 10, the employee becomes totally disabled and leaves work. In Year 20, he begins to draw on his pension. Under Union Pacific’s theory, he should be deemed to have been paid his Year 10 salary—his highest salary ever—from Years 10 to 20, his high-three will necessarily be equivalent to his Year 10 salary, and his pension payment will increase accordingly.

But under Dowling’s theory, the employee’s ten-year window should be based off Years 1 through 10, and his high-three will be Years 8, 9, and 10, resulting in an average salary that is inevitably lower than what he was paid in Year 10, and what he would have received' under Union Pacific’s approach. We suspect most employees are in situations closer to our hypothetical employee’s than to Dowling’s, and would benefit less from Dowling’s approach than the administrator’s.

While these counterintui-tive aspects of Dowling’s position do not on their own render it unreasonable, they lend support to the reasonableness of the administrator’s

interpretation. VI. Finally, Dowling makes an argument that a conflict of interest requires us to look more skeptically at the administrator’s decision. ERISA plan administrators are fiduciaries, and “if a benefit plan gives discretion to an administrator or fiduciary who is operating under a conflict of interest, that conflict must be weighed as a ‘facto[r] in determining whether’ ” the administrator’s benefits decision should stand'. Firestone, 489 U.S. at 115, 109 S.Ct. 948 (quoting Restatement (Second) of Trusts § 187, cmt. d (Am. Law. Inst. 1959)).

A conflict “clear[ly]” exists when the employer “both funds the plan and evaluates the claims,” because “[i]n such a circumstance, ‘every dollar provided in benefits is a dollar spent by ... the employer; and every dollar saved .... is a dollar in [the employer’s] pocket.’ ” Metro. Life Ins. Co. v. Glenn, 554 U.S. 105, 112, 128 S.Ct. 2343, 171 L.Ed.2d 299 (2008) (quoting Bruch v. Firestone Tire & Rubber Co., 828 F.2d 134, 144 (3d Cir. 1987), affd in part, rev’d in part on other grounds, 489 U.S. 101, 109 S.Ct. 948, 103 L.Ed.2d 80 (1989)).

The mere existence of a conflict is not determinative, however, and a conflict on its own does not change our standard of review “from deferential to de novo.” Id. at 115,128 S.Ct. 2343. A conflict is just another “factor,” and “Firestone means what the word ‘factor’ implies, namely, that when judges review the lawfulness of benefit denials, they will often take account of several different considerations of which a conflict of interest is one.” Id. at 117, 128 S.Ct. 2343.

The conflict may “act as a tiebreaker when the other factors are closely balanced,” or it may mean little at all, depending on the other factors at play. Id. Also, the circumstances of the conflict itself may render it more or less significant, depending on whether those “circumstances suggest a higher likelihood” that the conflict actually affected the benefits decision. Id. For example, if “an insurance company administrator has a history of biased claims administration,” then it is more likely the conflict affected the benefits decision, and .the court may grant less deference to the plan administrator. Id.

On the other hand, if “the administrator has taken active steps to reduce potential bias and to promote accuracy ... by walling off claims administrators from those interested in firm finances, or by imposing management checks that penalize inaccurate decision-making irrespective of whom the inaccuracy benefits,” then the conflict may be said to have been unlikely to infect the administrator’s decision, and may be of “vanishing” significance. Id. The Supreme Court exemplified this only-a-factor approach in the Glenn case. Id. at 118, 128 S.Ct. 2343.

The Glenn administrator-company was subject to a conflict, and evidence also showed it (1) refused to honor the government’s findings as to disability while encouraging the plan participant to pursue government disability benefits, (2) emphasized medical reports that disfavored the claimant while deem-phasizing reports cutting in the claimant’s favor, and (3) failed to provide the pro-claimant reports to medical experts. Id.

On these facts the Sixth Circuit refused to enforce the administrator’s decision and the Supreme Court affirmed—yet the Court took care to note that the conflict of interest alone probably would not have warranted overriding the administrator’s decision, and that the additional bad facts were crucial. Id. (suggesting the Sixth Circuit “would not have found the conflict alone determinative”).

Since Glenn, we have only been willing to disturb an administrator’s decision based on a conflict of interest if'evidence either suggests the conflict actually infected the decisionmaking or if the conflict is one last straw that calls a benefits determination into question. For an example of the last-straw scenario, in Miller v.

American Airlines, Inc., we refused to uphold a benefits determination where the airline operated under a conflict of interest and also (1) doubled back on an initial factual finding that the plan participant was disabled, (2) considered extra factors not called for in the plan, (3) failed to comply with ERISA’s notice requirements, and (4) failed to fully evaluate an earlier diagnosis. 632 F.3d 837, 855-56 (3d Cir. 2011). We gave “significant weight” to the four factors other than the conflict and only “slight weight” to the conflict itself. Id. at 855-56. By comparison, in Fleisher v.

Standard Insurance Co., there was a conflict, but the plan-beneficiary presented no evidence at all that the conflict actually infected the administrator’s decisionmak-ing, and we were still willing to apply deference and affirm, without requesting additional factfinding. 679 F.3d at 122 n.3 (3d Cir. 2012) (finding no evidence the conflict “affect[ed] the analysis of [the] claim”); cf. id. at 130 (Garth, J., dissenting) (noting that the majority found no need to remand for additional factfinding).

Dowling’s case is more like Fleisher than Miller or Glenn, because we know very little about the Union Pacific administrator’s conflict. A conflict does exist— Union Pacific both funds and administers the plan—but that is about all we know.

Dowling has highlighted no evidence suggesting Union Pacific has any sort of negative history of failing to properly exercise its fiduciary responsibilities, and Union Pacific has put forward nothing indicating that it took steps to wall off the plan administrator from the company’s financial decisionmaking or incentivize its staff to make accurate benefits determinations instead of reducing costs. We do have the job titles of the relevant individuals—Willis worked in “Compensation and Benefits” and Schroer and Schaefer were in “Human Resources”—but titles alone do not tell us much.

The one fact that Dowling says cuts in his favor is an early flip-flop by Union Pacific that he analogizes to the problematic reversal in Miller. In 1995, Willis and Dowling corresponded, and Willis told Dowling that Union Pacific would calculate Dowling’s Final Average Compensation the way Dowling now believes to be correct. Then • a year later, in 1996, Willis wrote Dowling again and reversed his initial position, calculating that Dowling’s pension should be what Union Pacific later finally adopted in 2013.

But again, Dowling has presented the Court with nothing more than the bare fact of this reversal; no evidence suggests Willis’s motivation was ulterior, or anything other than a desire to correct what he saw as an errant calculation. Comparing Willis’s reversal to the problematic reversal in Miller also shows just how far Willis’s reversal is from warranting a skeptical take on Union Pacific’s conflict.

Three factors in Miller suggested the reversal was motivated not by a desire to correct an error, but instead by an ulterior cost-cutting motivation that might be attributed to the conflict: First, the Miller reversal came on a factual question—whether the claimant was disabled—even though no new evidence had been received. 632 F.3d at 841-42, 855-56. Here the reversal was on a purely legal question—how the plan’s complex terms should be properly interpreted. Second, in Miller the initial decision resulted in the claimant actually receiving benefits, and the reversal cut off the flow of those benefits. Id.

Here, the initial determination was preliminary and advisory, the reversal came a year later, and no benefits flowed for another 17 years. Third, in Miller the reversal was one of four factors, not including the conflict, that together undermined the Court’s trust in the administrator, the most notable factor being the administrator’s consideration of information not permitted by the plan document, which may have on its own been evidence of an arbitrary and capricious benefits determination. Id. at 855-56, 856 n.5.

Here, by comparison, Willis’s reversal is the only factor—not one of four—that Dowling has marshaled to support his argument that the conflict affected the administrator’s determination. Without more, there is little indication that the' conflict played a role, and its bare existence is not enough to justify disturbing the plan administrator’s otherwise reasonable decision. VII. For the foregoing reasons we will affirm the District Court’s decision sustaining the plan administrator’s calculation of Dowl-ing’s pension benefit. .

Plan § 2.'35 states in pertinent part, " 'Final Average Compensation' shall mean the average of the Participant's monthly Compensation for the 36 consecutive calendar months of highest Compensation within the 120-cal-endar month period immediately preceding ... the last date on which he is a Covered Employee.” (Plan § 2.35, App. 144.) .Plan § 4.02(c)(2) states, "A Disabled Participant who is a Covered Employee on his Disability Date shall be credited with years of Credited Service as if he were a Covered Employee from his Disability Date to the date on which he ceases to be a Disabled Participant as set forth in

Section 6.05.” (Plan § 4.02(c)(2), App. 157.) Plan § 6.05 states in pertinent part, [A] Participant who has a Disability Date shall continue to be credited with years of Vesting Service and Credited Service (to the extent provided in

Section 4.02(c)(2)) while he remains a Disabled Participant. A Disabled Participant shall cease to be such if and when: (

a) he ceases to suffer from a Total Disability; (

b) he ceases to receive benefits under the Long Term Disability Plan of Union Pacific Corporation; (

c) he dies; or (

d) he elects a' Benefit Payment Date. When a Disabled Participant ceases to 'be such, he shall cease 'to be credited with years of Vesting Service and Credited Service, and he shall be entitled to a pension under the other provisions of this

Article (or

Article VII), applied as if his Separation from Service occurred on the date he ceased to be a Disabled Participant .... (Plan § 6.05, App. 178.) . Plan § 2.18(a)(3)(

C) states in pertinent part, (

C) During a period when an Employee receives credit for Hours of Service under

Section 2.40 for a period of absence immediately prior to which- he is a Covered Employee and which Hours of Service are counted in determining his Credited Service under

Section 4.02: (

i) the Employee, if employed on a full-time basis at the start of the absence, shall be deemed to have received Compensation at the base pay rate in effect for him as of the first day of the month in which such period begins and shall also be credited with any Compensation described in (3)(A)(ii) through (iv), above, that is actually paid to him during that period; .... (Plan § 2.18(a)(3)(C), App. 139.) . Plan § 2.10 states, “Approved Absence” shall mean the period during which an Employee absents himself from work without compensation (to the extent evident in personnel records of the Employer or the Affiliated Company), by reason of: (

a) a period of absence for personal or other reasons, provided that such person returns to work for the Employer or such Affiliated Company at such time as the Employer or such Affiliated Company may reasonably require, or (

b) a family or medical leave within the meaning of the Family and Medical Leave Act of 1993, provided, however, that effective for leaves of absence beginning on or after January 1, 1999, such person returns to work after the family and medical leave at such time as the Employer or Affiliated Company may reasonably require. In the authorization of an Approved Absence under subsection (

a) and in the requirements set forth with respect to assuring the return of the Employee to work within a reasonable time, the Employer or an Affiliated Company shall treat all Employees under similar circumstances in a like manner. (Plan § 2.10, App. 134-35.) • Plan § 2.1 OB states, “Approved Non-HCE Absence” shall mean, effective January 1, 2008, the period during which an Employee who is not a Highly Compensated Employee absents himself from work without compensation (to the extent evident in personnel records of the Employer or Affiliated Company), by reason of a period of absence for a purpose described in a leave of absence policy of the Employer or an Affiliated Company, but the duration of which is longer than otherwise permitted under such policy, and with the approval or at the requirement of the Employer or such Affiliated Company, provided that such person returns to work for the Employer or such Affiliated Company at such time as the Employer or such Affiliated Company may reasonably require. (Plan § 2.10B, App. 135.) .

Plan § 2.40 states in pertinent part, “Hour of Service' shall mean, ... (

a) With respect to a Participant (other than a Disabled Participant) whose Separation from Service occurs prior to January 1, 1999 and with respect to a Disabled Participant who ceases to be such prior to January 1, 1999: (4) 10 Hours of Service for each day on which the Employee is absent (

A) on an Approved Absence, (

B) for temporary layoff on account of reduction in force provided there is a return to work at the first available opportunity, (

C) for military service under leave granted by the Employer or Affiliated Company or required by law provided the Employee returns to service with the Employer or Affiliated Company within such period as his right to reemployment is protected by law, or (

D) effective January 1, 2008, on an Approved Non-HCE Absence. (5) 10 Hours of Service for each day of an Employee’s Total Disability. (Plan § 2.40(a)(4)-(5), App. 147-48.) . Plan § 13.02 states in pertinent part, Authority and Responsibility of the Named Fiduciary-Plan Administration. The Named Fiduciary-Plan Administration shall be the Plan "administrator” as such term is defined in section 3(16) of ERISA, and as such shall have the following duties and responsibilities: (

f) to determine all questions of the eligibility of Employees and of the' status of rights of Participants, Surviving Spouses, Beneficiaries and Alternate Payees, to make factual determinations, to construe and interpret the provisions of the Plan, to correct defects and resolve ambiguities therein, and to supply omissions thereto; .... (Plan § 13.02, App. 242.) . Plan §§ 12.01-03 states in pertinent part, Sec. 12.01 Employer Contributions. Subject to

Section 12.06, the Employer shall contribute such amounts as are necessary to satisfy the minimum funding standards required pursuant to ERISA and

section 412 of the Code, as from time to time amended. ... The Employer shall have the right, but not the obligation, to contribute such additional amounts as it, in its sole discretion, deems desirable in any year. All Employer contributions shall be paid to the Trustee. Sec. 12.02 Mandatory Participant Contributions. No contributions shall be required of Participants under the Plan. Sec. 12.03 Voluntary Participant Contributions. No contributions shall be accepted from any Participant under the Plan. (Plan §§ 12.01-03, App. 232.) . A similar rule of construction is advisable under the Restatement (Third) of Property, the recommended restatement for trust-document

interpretation. It recommends that for courts seeking the drafter’s meaning of text in a trust document, "words and phrases” should be "presumed to bear their customary legal terminology” if the "drafter is a legal professional or other person experienced in the use of legal terminology” and there is no "[e]xtrinsic evidence” going to the drafter’s subjective intention. Restatement (Third) of Property: Wills and Other Donative Transfers § 10.2 cmt. e (2003); see also Restatement (Third) of Trusts, Ch. 1 intro, note (referring readers to the Restatement (Third) of Property for "general rules of

interpretation and construction”). . Harvey Araton, Sports of the Times; Jordan, a Bit Older, Comes Up Short, N.Y. Times (Oct. 31, 2001), http://www.nytimes.com/2001/10/ 31/sports/sports-of-thetimes-jordan-a-bit-older-comes-up-short.html. . Ryan Parker, Rick Moranis Reveals Why He Turned Down "Ghostbusters" Reboot: “It Makes No Sense to Me," Hollywood Reporter (Oct. 7, 2015), http://www.hollywoodreporter. com/features/rick-moranisreveals-why-he-829779. .

Fresh Air (NPR radio broadcast Apr. 1, 2016), transcript excerpt available at http:// www.npr.org/2016/04/01/472580940/-miles-aheadshows-a-dissipated-davis-who-s-still-burning-hot. . Union Pacific's Pension Plan capitalizes defined terms. In order to keep better track of them, I adopt that convention as well. See, e.g., Plan § 2.54 (defining "Participant”). There's no dispute that Dowling is a Participant.