*626 ORDER AMENDING OPINION AND DENYING PETITION FOR REHEARING AND PETITION FOR REHEARING EN BANC AND AMENDED OPINION
ORDER
Thе opinion in this case, filed September 14, 2009, is amended by revising the first sentence of the paragraph that starts at the bottom of page 13360 of the slip opinion and extends onto page 13361. The sentence previously started:
In clarifying the standard of review, Abatie also abrogated a line of cases, including Jordan v. Northrop Grumman Corporation Welfare Benefit Plan,370 F.3d 869 (9th Cir.2004), and Bendixen v. Standard Insurance Company,185 F.3d 939 (9th Cir.1999), that had directed reviewing courts....
That sentence is amended as follows:
In clarifying the standard of review, Abatie abrogated a line of cases, including Jordan v. Northrop Grumman Corporation Welfare Benefit Plan,370 F.3d 869 (9th Cir.2004), and Bendixen v. Standard Insurance Company,185 F.3d 939 (9th Cir.1999), to the extent that the cases directed reviewing courts
With this amendment, the panel has voted to deny the petition for rehearing and petition for rehearing en banc.
The full court has been advised of the petition for rehearing en banc, and no judge of the court has requested a vote on it. Fed. R.App. P. 35.
The petition for rehearing and petition for rehearing en banc, filed October 13, 2009, are DENIED. No further petitions for rehearing or for rehearing en banc may be filed.
OPINION
This case presents the question of how a district court should apply the abuse of discretion standard when reviewing a decision by the administrator of an employee benefits plan governed by the Employee Retirement Income Security Act of 1974 (ERISA), 88 Stat. 829, as amended, 29 U.S.C. §§ 1001-1461, when that administrator has a conflict of interest. We conclude that a reviewing court must take into account the conflict and that this necessarily entails a more complex application of the abuse of discretion standard. Specifically, a modicum of evidence in the record supporting the administrator’s decision will not alone suffice in the face of such а conflict, since this more traditional application of the abuse of discretion standard allows no room for weighing the extent to which the administrator’s decision may have been motivated by improper considerations.
Robert Montour appeals the district court’s order granting summary judgment in favor of Hartford Life and Accident Insurance Company in his action challenging Hartford’s decision to terminate his long-term disability benefits as an abuse of its discretion. We reverse and, applying the proper standard of review to the facts of this case, conclude that Hartford abused its discretion because its conflict of interest too heavily influenced its termination decision. Accordingly, we remand to the district court for an order reinstаting Montour’s long-term disability benefits.
I. Background
As an employee of Conexant Systems, Inc. for approximately thirty-seven years, Montour participated in his employer’s group long-term disability insurance plan, which is a welfare benefit plan governed by ERISA. Hartford is both the insurer *627 and the administrator of the Plan. The Plan grants Hartford, as the administrator, discretionary authority to interpret Plan terms and to determine eligibility for benefits, 1 and it places the burden of proving both initial and ongoing disability on the claimant.
In July 2003 Montour took a medical leave of absence from his position as a telecommunications manager after developing symptoms of acute stress disorder. At the time, he was fifty-five years old. In January 2004, following a period of 180 days during which no benefits were payable under the Plan, Hartford accepted Montour’s application for benefits under the Plan and began paying him disability benefits.
At the outset of his psychiatric illness, Montour consulted several times with his primary care physician, Dr. Samuel Park. In September 2003 he began regular psychotherapy sessions with a psychiatrist. His last documented psychotherapy session took place in April 2005.
Meanwhile, in June 2004 Montour consulted Dr. Kenneth Kengla, an orthopedic surgeon, about pain in his right knee and his lower back. Dr. Kengla diagnosed Montour with degenerative changes in both regions and notified Hartford in September 2004 that Montour was at that time also suffering from physical disability that prevented him from returning to the labor force. In October 2004 Dr. Kengla performed arthrosсopic surgery on Mont-our’s right knee. The subject of Mont-our’s back condition did not come up again during their consultations until April 2005. Subsequently, Montour consulted Dr. Kengla about his back pain during appointments in December 2005 and May 2006.
Dr. Kengla consistently maintained to Hartford that Montour remained physically disabled and unable to work in any job as a result of his back and knee impairments. Specifically, he listed the following restrictions on Montour’s physical activities: 1) “no sitting for more than 15-20 min[utes] at a time”; 2) “no prolonged walking”; 3) “no standing greater than 15 min[utes] at a time”; 4) “no lifting or carrying greater than 10 [pounds;]” 4) “no work at or above shoulder level”; 5) “no moderate pushing activities”; 6) “no moderate pulling activities”; and 7) “no driving greater than 30 min[utes] at a time.”
In November and December 2005 Hartford hired two outside companies to conduct surveillance on Montour over the course of four nonconsecutive days. Video footage from this surveillance depicted Montour driving his car to perform occasional errands, such as picking up his grandchildren from school, going to the pharmacy, and getting a haircut. He was observed once bending at the waist to reach into his car.
In March 2006 a Hartford investigator conducted a personal interview with Mont-our at his home, during which Montour listed a “bad back, [an] arthritic right knee, and sleep apnea” as the “disabling medical condition(s)” preventing him from returning to work. He also described an inability to concentrate, which he attributed to the medication he must takе to treat his “constant pain.” The investigator observed that Montour remained alert and responsive during the entire four-and-a-half hour interview, although he called the investigator by the wrong name about two hours into the interview. Montour ac *628 knowledged that the surveillance video footage accurately depicted his level of functionality. He was physically able to complete the interview, but he demonstrated signs of pain in front of the investigator, such as moaning when he stood up or twisted, walking around his house stiffly with a slight limp, and complaining of back pain three times during the final two hours.
In May 2006 a Hartford nurse case manager submitted letters to Dr. Kengla and Dr. Park surmising that Montour was capable of performing “sedentary to light” work and soliciting their agrеement. Dr. Park signed and returned the letter, which signified that he either agreed or found “no contraindications to this work capacity level.” Dr. Kengla, on the other hand, indicated that he disagreed with Hartford’s conclusions, citing Montour’s persistent orthopedic symptoms and physical restrictions.
In July 2006 Hartford hired a consulting physician, Dr. Gale Brown, to conduct a file review. Dr. Brown analyzed Mont-our’s medical records for the 2008-2006 period, including X-rays and MRIs of Montour’s lower back taken in June 2004 and May 2006, Montour’s pharmacy records for the 2004-2006 period, Hartford’s surveillance video and accompanying reports, and the personal interview report. He also spoke with Dr. Kengla on the phone. Dr. Brown concluded that medical evidence suppоrted the existence of a lower back condition called “degenerative spondylostenosis/DDD” but that Dr. Kengla’s offered restrictions were excessive for this “mild to moderate” condition and understated Montour’s demonstrated and admitted physical abilities. Dr. Brown acknowledged that medical evidence supported Montour’s chronic pain but found that Montour was nevertheless capable of working full-time with modest restrictions, such as changing positions every thirty to forty-five minutes.
Hartford next enlisted a vocational rehabilitation expert to compile an Employability Analysis Report, which evaluated Montour’s experience, qualifications, and the physical restrictions identified by Dr. Brown. That expert concluded that Mont-our was capable of working in a high-level managerial capacity in five different fields.
In August 2006 Hartford informed Montour of its decision to terminate his benefits in light of its conclusion that he no longer met the policy’s definition of disability. Montour appealed this decision internally and included a vocational appraisal report by Gene Bruno. The Bruno report concluded that Montour was “not employable in any setting” and that Hartford’s decision was based on numerous mistakes, including a disregard for the fact that the Social Security Administration (SSA) considered Montour to be “totally disabled.”
In response, Hartford hired Dr. Renat Sukhov to conduct a second file review. Dr. Sukhov reviewed Montour’s records for evidence of a physical condition that would prеclude sedentary work and, like Dr. Brown, found none. He noted in particular a lack of objective, clinical data demonstrating the extent to which Mont-our’s pain impacted his functionality. He also noted that Montour’s activities depicted on the surveillance videos exceed the activity requirements of a sedentary job. Dr. Sukhov concluded that Montour could work, at minimum, in a sedentary job with reasonable restrictions, such as avoiding static work posture, not lifting objects weighing more than twenty pounds, and not pushing or pulling loads weighing more than thirty to thirty-five pounds.
In light of concerns raised in the Bruno report, Hartford also requested a vocational specialist to conduct an Employability Analysis Report addendum, which reached *629 the samе conclusion as the Employability Analysis Report regarding the sedentary nature and thus the feasibility of the five proposed managerial positions. In February 2007 a Hartford appeal specialist affirmed the company’s previous decision to terminate Montour’s benefits.
Having exhausted his administrative remedies, in June 2007 Montour and his wife, Tina Montour, filed suit against Hartford in California Superior Court to recover benefits under ERISA § 502(a)(1)(B), 29 U.S.C. § 1132(a)(1)(B). The complaint also alleged breach of fiduciary duty, promissory estoppel, invasion of privacy, intentional infliction of emotional distress, and loss of consortium. Hartford removed the case to the Central District of California and promptly moved to dismiss. The district court dismissed the causes of action for brеach of fiduciary duty and promissory estoppel, remanded the claims for invasion of privacy, intentional infliction of emotional distress, and loss of consortium back to state court, and proceeded to hold a bench trial on the administrative record for the remaining ERISA benefits recovery claim.
In April 2008 the court entered its Findings of Fact and Conclusions of Law ordering Judgment in favor of Hartford. The district court concluded that although Hartford had a structural conflict of interest in its position as both the administrator of the insurance policy and the payor of benefits, it did not abuse its discretion in determining that Montour failed to provide sufficient evidence to demonstrate disability within the meaning of the policy.
Montour timely appealed.
II. Standard of Review
The district court’s findings of fact in a bеnch trial on the administrative record are reviewed under the clearly erroneous standard.
See Pannebecker v. Liberty Life Assurance Co. of Boston,
As for review of a plan administrator’s decision, “[t]he Supreme Court has held that a denial of benefits ‘is to be reviewed under a de novo standard unless the benefit plan gives the administrator ... discretionary authority to determine eligibility for benefits or to construe the terms of the plan.’ ”
Burke v. Pitney Bowes Inc. Long-Term Disability Plan,
In the absence of a conflict, judicial review of a plan administrator’s benefits determination involves a straightforward application of the abuse of discretion standard.
See Sznewajs,
Commоnly, however, the same entity that funds an ERISA benefits plan also evaluates claims, as is the case here.
See Metro. Life Ins. Co. v. Glenn,
— U.S. -,
More particularly, the court must consider numerous case-specific factors, including the administrator’s conflict of interest, and reach a decision as to whether discretion has been abused by weighing and balancing those factors together.
See MetLife II,
The weight the court assigns to the conflict factor depends on the facts and circumstances of each particular case. For example, the Supreme Court has explained that this factor
should prove more important (perhaps of great importance) where circumstances suggest a higher likelihood that *631 it affected the benefits decision, including, but not limited to, cases where an insurance company administrator has a history of biased clаims administration. It should prove less important (perhaps to the vanishing point) where the administrator has taken active steps to reduce potential bias and to promote accuracy, for example, by walling off claims administrators from those interested in firm finances, or by imposing management checks that penalize inaccurate decisionmaking irrespective of whom the inaccuracy benefits.
MetLife II,
Our court has implemented this approach by including the existence of a conflict as a factor to be weighed, adjusting the weight given that factor based on the degree to which the conflict appears improperly to have influenced a plan administrator’s decision.
See Abatie,
In any event,
Abatie
explained that the court should adjust the level of skepticism with which it reviews a potentially biased plan administrator’s explanation for its decision in accordance with the facts and circumstances of the case.
See
In clarifying the standard of review,
Abatie
abrogated a line of cases, including
Jordan v. Northrop Grumman Corporation Welfare Benefit Plan,
For this reason, the district court’s reliance in this case on
Boyd,
another decision in which there was apparеntly no evidence of a conflict of interest, was incorrect.
See
As the district court’s decision did not aрpropriately balance the pertinent factors, and in particular the conflict factor, we proceed to do so here. Judicial review of an ERISA plan administrator’s decision on the merits is limited to the administrative record,
4
so “we are in the same position as the district court....”
See Moapa Band of Paiute Indians v. U.S. Dep’t of Interior,
III. The Merits
We begin with the district court’s comprehensive description of the “signs of *633 bias” exhibited by Hartford throughout its decisionmaking process:
Hartford was both the plan administrator and funder of the Plan, and evidence of this conflict of interest appears throughout the record. For example, in its letters to Plaintiff, Hartford overstates and over-relies on surveillance of Plaintiff. Plaintiff was observed over forty daylight hours on four days in November and December 2005. During this time, he was observed making two twenty minute trips to pick up or drop off his grandchildren from school and one trip of about two and a half hours conducting errands at various stores. He was also observed to be away from his home on two occasions for about an hour and forty minutes. During this time, he was observed bending once at the waist and picking up a small bag of medication.
This observed activity was brief and consistent with Plaintiffs self-reported limitаtions. Plaintiff admitted that he was able to drive for up to thirty minutes, could walk short distances, and could lift objects lighter than five pounds. Yet Hartford claimed that Plaintiffs “self-reported limitations were not consistent with his observed activities.” Hartford strung together a laundry list of discrete activities observed over the course of four days, suggesting that Plaintiff was capable of sustaining those activities throughout the day, as would be required in a sedentary occupation. However, that Plaintiff could perform sedentary activities in bursts spread out over four days does not indicate that he []is capable of sustaining activity in a full-time occupation.
Furthermore, despite its own internal observation that Plaintiff walked “in a slightly stiff and slow manner” and entered his car in “a somewhаt slow and deliberate manner”, Hartford indicates that Plaintiff “did not show any limitation in his movement and demonstrated no sign of physical distress.” This statement is also inconsistent with Hartford’s report on the in-home interview, which noted that Plaintiff showed signs of pain and discomfort, moaning the majority of times that he stood or twisted.
Hartford’s attempts to obtain information from Plaintiffs physicians were marred by this overstatement of the surveillance findings, as well as apparent advocacy for the position that Plaintiff was not disabled. In the letter sent to Drs. Park and Kengla, Hartford again strung together discrete activities observed in short bursts over several days, stated that he was seen “walking at a brisk pace,” and noted that his activities were “performed without apparent difficulty, hesitation, or the use of assistive devices.” Instead of noting the observations of pain seen during Plaintiffs in-home interview, the letter stated that Plaintiff “displayed minimal physical or mental limitation” while participating in the interview.
In response to this letter, Dr. Park, who saw Plaintiff twice a year and was not treating him for his back condition, stated that he either agreed or found no contraindication to the sedentary work capacity suggested by Hartford. Dr. Kengla, who saw Plaintiff once a month for the conditions that rendered him disabled, stated that Plaintiff remained disabled. Dr. Park was apparently less familiar with Plaintiffs condition and thus more apt to be swayed by Hartford’s slanted presentation of the facts. Nonetheless, Hartford found the disagreement between Dr. Park and Dr. Kеngla to be significant, and sent Plaintiffs file to Dr. Brown for independent review.
*634 Although he based his opinion on a number of factors, Dr. Brown relied on both the surveillance and the conflict of opinion between Drs. Park and Kengla in reaching the conclusion that Plaintiff was able to return to work. In conducting a further review of Plaintiffs medical records after Plaintiff appealed the denial of benefits, Dr. Sukhov again overemphasized the surveillance, stating that the “video clearly shows the claimant performing activities above those required for a sedentary job.” As noted above, this was clearly not the case. Thus, there is a common theme, both in Hartford’s communications with Plaintiff and in the assessments of those professionals Hartford hired to evaluate Plaintiffs condition, of presenting evidence of capability in the best possible light, while failing to subject evidence of capability to the same skepticism and rigorous analysis applied to evidence of disability.
(Alterations and internal citations omitted). In other words, Hartford’s bias infiltrated the entire administrative decision-making process, which leads us to accord significant weight to the conflict.
Another factor is Hartford’s failure to present extrinsic evidence of any effort 'on its part to “assure accurate claims assessment!,]” such as utilizing procedures to help ensure a neutral review process.
See MetLife II,
Another factor is Hartford’s decision to conduct a “pure paper” review in this case, that is, to hire doctors to review Montour’s files rather than to conduct an in-person medical evaluation of him. While the Plan does not require a physical exam by a non-treating physician, in this case that choice “raise[s] questions about the thoroughness and accuracy of the benefits determination!,]”
Bennett v. Kemper Nat’l Servs., Inc.,
In its decision to terminate Montour’s benefits, Hartford relied significantly on the conclusions reached by Dr. Brown and Dr. Sukhov. Dr. Brown, who conducted the first file review, diagnosed Montour with a “mild to moderate” back condition that he felt should not cause Montour as much pain as he was reportedly suffering. Likewise, the second file reviewer, Dr. Sukhov, based his conclusion of non-disability in part on the lack of objective medical data to support Montour’s alleged pain levels, in addition to the lack of a self-reported pain scale or- some form of quantification of the
*635
impact of his pain on his functional abilities. It would probably have been unreasonable for Hartford to require Montour to produce objective proof of his pain level, per Dr. Sukhov, or to reject his subjective claims of “excess pain” based solely on Dr. Brown’s observation.
See Smolen v. Chater,
On the other hand, Dr. Sukhov also fixated on the lack of progression
(ie.,
lack of further degeneration) in Montour’s back condition, as evidenced by X-rays and MRIs taken in June 2004 and May 2006, and Hartford noted this in its decision on appeal. It is not clear “why further degeneration is necessary to sustain a finding that [Montour] is disabled.”
Saffon,
One final factor is Hartford’s failure to grapple with the SSA’s contrary disability determination. The Plan requires claimants to apply for social security disability benefits from the SSA and, if denied, to exhaust all possible appeals. Hartford emphasized this requirement to Montour in its initial award letter and then again in a separate letter the following month. In March 2004, the SSA concluded that Mont-our was disabled and awarded him disability benefits retroactively to January 2004. Hartford benefitted from this award significantly, as it received a dollar-for-dollar financial offset, nearly halving its liability. As of April 2005 and December 2006, the SSA considered Montour’s disability to be “continuing.” In August 2006, Hartford nevertheless concludеd that Montour was no longer disabled. Although Montour had immediately forwarded the SSA’s April 2005 notice of continuing disability to Hartford, the plan administrator made no mention of the SSA’s contrary determination in its initial termination decision. In its decision denying Montour’s appeal, Hartford acknowledged the SSA’s decision but did not articulate why the SSA might have reached a different conclusion.
See MetLife I,
While ERISA plan administrators are not bound by the SSA’s determination, complete disregard for a contrary conclusion without so much as an explanation raises questions about whether an adverse benefits determination was “the product of a principled and deliberative reasoning process.”
See MetLife I,
Unlike the SSA, Hartford was not bound by the treating physician rule, which accords “special weight” to the opinions of a claimant’s treating physician. See
Black & Decker Disability Plan v. Nord,
538 U.S.
*636
822, 825, 831, 834,
Ordinarily, a proper acknowledgment of a contrary SSA disability determination would entail comparing and contrasting not just the definitions employed but also the medical evidence upon which the decisionmakers relied.
See, e.g., MetLife I,
Although the Plan places the burden on Montour to submit “written proof’ of his disability, that is, the pertinent documents and information necessary to facilitate a disability determination, regulations promulgated by the Secretary of Labor authorize, if not require, plan administrators working with an apparently deficient administrative record to inform claimants of the deficiency and to provide them with an opportunity to resolve the problem by furnishing the missing information.
See
29 C.F.R. § 2560.503 — 1(f)(3)—(4), (g)(1)(iii);
see also Saffon,
*637 Although the record reflects that Mont-our kept Hartford regularly apprised of his continuing disability status with the SSA, Hartford’s initial decision in this case made no mention of the contrary SSA disability determination and did not advise Montour that further documentation, such as the ALJ’s decision or the underlying administrative record, would facilitate Hartford’s review. At the appeals stage, Hartford’s sole acknowledgment was that the SSA’s contrary determination was “a consideration and part of the totality of the evidenсe!,]” but it then continued with the statement that its decision “must be based on the weight of the evidence in this file
To the extent this latter statement implies that the missing ALJ opinion and underlying record precluded Hartford from attributing much weight to the SSA’s contrary determination, this information came too late.
See Chuck,
Ultimately, Hartford’s failure to explain why it reached a different conclusion than the SSA is yet another factor to consider in reviewing the administrator’s decision for abuse of discretion, particularly where, as here, a plan administrator operating with a conflict of interest requires a claimant to apply and then benefits financially from the SSA’s disability finding.
See MetLife II,
Weighing all of the foregoing factors together, we conclude that Hartford’s conflict of interest improperly motivated its decision to terminate Montour’s benefits. This constituted an abuse of its administrative discretion.
IV. Conclusion
We reverse the district court’s summary judgment in favor of Hartford and remand to the district court to enter summary judgment in favor of Montour and to order the reinstatement of long-term disability benefits in accordance with this opinion and the terms of the Plan.
See Pannebecker,
REVERSED and REMANDED.
Notes
. Specifically, the policy provides:
Who interprets policy terms and conditions?
We [Hartford] have full discretion and authority to determine eligibility for benefits and to construe and interpret all terms and provisions of the Group Insurance Policy.
. We refer to the Supreme Court's decision as "MetLife II" because we also reference the lower court's opinion, infra, which MetLife II affirmed.
. For example, in
MetLife II
the Supreme Court endorsed the approach taken by the Sixth Circuit, which weighed three factors, namely, the medical evidence, a contrary SSA disability determination, and the administrator's inherent conflict of interest, and ultimately concluded that the administrator had abused its discretion in denying benefits to the claimant.
See Glenn v. MetLife,
. In the ERISA context, the "administrative record” consists of "the papers the insurer had when it denied the claim.”
See Kearney,