ORDER GRANTING IN PART AND DENYING IN PART DEFENDANT’S MOTION FOR SUMMARY JUDGMENT
-. Re: ECF No. 76 giLE-D-U-N-DER-SEAL1
JON S. TIGAR, United States District Judge
Before the Court is Defendant Secretary of the Department of Homeland Security’s (“Defendant”) motion for summary judgment. The Court will grant the motion in part, and deny it in part.
I, BACKGROUND
A. Factual Background
The Federal Air Marshal Service (“the Service”) is a federal law enforcement [791]*791agency within the Transportation Security Administration (“TSA”) whose mission “is to protect the nation’s commercial aviation system, by detecting, deterring, and defeating hostile acts targeting U.S. air carriers, airports, passengers and crews.” ECF No. 70 ¶ 2. Before the September 11, 2001- attacks, the Service had only 33 Federal Air Marshals (“FAMs”). ECF No. 70 ¶4. After the attacks, “President George W. Bush ordered the rapid expansion of the [Service]” and “thousands of FAMs were added to the ranks.” ECF No. 70 ¶ 4. Until 2014, each FAM worked out of one of the Service’s 26 field offices across the country. ECF No. 70-2 at 5.
According to Defendant, FAMs implement the Service’s mission by traveling on flights that the Service deems “high-risk.” When a high-risk flight does not originate out of a FAM’s home office, the FAM first takes a lower-risk “feeder flight,” which brings the FAM from her or her duty station to the airport with the high-risk flight. ECF No. 70 ¶ 8. “The use of feeder flights is inefficient and reduces the overall amount of risk that the [Service] is able to cover” in several ways. ECF No, 70 ¶ 9. First, the use of the feeder flights lengthens FAMs duty days. “For example, a FAM, stationed in Cleveland, who needed to first fly to New York, to cover a high risk flight to Paris, would have an approximately 16-hour duty day.” ECF No. 70 ¶ 9. This “generate^] undesirable overtime expense” and can make a FAM “fatigued, and less effective, by the time he or she is on the critical part of the mission-the high, risk flight.” Id. The Service claims that “[t]he use of feeder flights also carries a not-insignificant risk that the higher-risk flight will go uncovered altogether if the feeder flight is delayed or cancelled and the FAM is unable to make the higher-risk connection.” ECF No. 70 ¶ 10.
Defendant argues that in the over fifteen years since the September 11 attacks, which is when most of the Service’s field offices were created, two major factors have led to an unsustainable increase in the use of feéder flights. First, “the threats to the commercial aviation industry [ ]' evolved.” ECF NO. 70 ¶ 5. While domestic flights were initially the Service’s focus, the “threat has become increasingly international.” ECF No. 70 ¶ 5. The airports that see more international flights therefore became more important to the Service’s mission. Second, the early 2000s saw a “significant restructuring” of the U.S. airline industry. ECF No., 70 ¶6. “Major airlines have merged,” changing both “passenger throughput” in certain airports and which airports are “hubs.” ECF No. 70 ¶ 6.
By 2013, Defendant claims that a “consensus” had emerged within the Service that there was a “mismatch between the location of many FAMs and the offices at which they were needed for higher criticality flights.” ECF No. 70 ¶ 13. “There was, for example, chronically a higher number of high-risk flights originating from the airports served by the FAMS’s busiest offices.than FAMs available in those offices to. cover them. The opposite was true in the [Service’s] least busy offices.” ECF No. 70 ¶ 13. Complicating matters, Congress imposed a- hiring freeze on the Service beginning in September 2011, “meaning it was unable to replace the FAMs lost through ordinary attrition with new hires.” ECF No. 70 ¶ 12. The Service also experienced “declining budgets” in 2013 and 2014. Id.
To address the increasing use of iheffi-cient feeder flights, then Regional Director of the Service’s Field Operations Division, Eric P. Sarandrea, “was tasked with analyzing the efficiency of the geographic location of the [Service’s] 26 field offices.” ECF No. 70 ¶ 13, The resulting report, entitled “Reshaping, the FAMS Workforce,” recommended closing six field of[792]*792fices with “only minimal strategic value” and “that the FAMs at those offices be reassigned to field offices that needed additional personnel to cover higher-risk flights.” ECF No. 70 ¶ 16; ECF No. 70-1. On February 13, 2014, Director of . the Service Bob. Bray submitted a memo to TSA Administrator John Pistole recommending that the Service adopt Saran-drea’s Report. ECF No. 70-2..That same day, Pistole approved the closure of the six offices on a staggered basis. ECF No. 70 ¶ 20, 29. A week later, Bray sent an email to staff that contained the following announcement: “Based on the results of the future staffing and field office assessment ..., I have made the decision, with the Administrator’s approval, to close six offices: Cincinnati, Cleveland, Phoenix, Pittsburgh, San Diego, and Tampa.” ECF No. 83-1 at 23. All six offices are now closed. ECF No. 70 ¶ 20, 29.
When the closures occurred, the 287 FAMs “in impacted offices were given the opportunity to transfer” to one often field offices that the Service determined “were in need of additional FAMs.” ECF No. 70 ¶ 23. New assignments were given based on each FAM’s ranked preference and by seniority. ECF No. 70 ¶24. 194 of the FAMs in affected offices were over 40-years old, and of those 194, 51 separated from the Service rather than move to a new field office. ECF No. 83-1 at 147-52. In 2016, the Service’s hiring freeze was lifted, and it hired 326 new FAMS. ECF No. 70 ¶29. Forty-eight of these new FAMs were substantially younger than the FAMs who had left during the closures. ECF No. 83-1 at 158.
Plaintiffs argue that, far from being driven by risk-coverage concerns, the six field office closures were an attempt to drive out older FAMS and replace them with newer, younger employees to address the Service’s budgetary problems. ECF No. 83 at 6. The closures, Plaintiffs claim, were motivated by age discrimination and in fact adversely affected older workers. Id.
According to Plaintiffs, evidence of age discrimination pervaded the closure decision-making process. Plaintiffs note that Sarandrea’s Reshaping the FAMS Workforce Report included information on the age of affected FAMs.2 The Report states that 250 of the 287 affected employees were 40 or older, and in discussing the relocation costs associated with the closures, estimates that “31 FAMs will retire ' or resign from the six affected offices in the next two years.” ECF'No. 70-1 at 13. Similarly, Bray’s memo notes that an “estimated 45 employees will be eligible for retirement or may resign from the six affected offices in the next two years,” potentially reducing relocation costs. ECF No. 70-2 at 7. The closures were also communicated to FAMs by Bray and management officials through in-person meetings. According to two FAMs present at one such meeting, Bray stated that if the Service lost “a lot of senior guys over this” it “would hire more guys and they’d be younger and he’d pay them less.” ECF No. 83-1 at 87; ECF No. 83-3 at 2. Bray also responded to questions about the relocation process. Wdien asked whether medical hardships would be considered in making the reassignments, Bray answered “we all have things going on, people get sick, parents get sick, kids die, you just move on.” ECF No. 83-1 at 74-75. Indeed, Plaintiffs assert that the Service purposefully made the reassignment process “extremely difficult for the FAMS,” with the goal of increasing attrition.- ECF No. 83 at 11. Not [793]*793only did FAMs have to uproot their lives and move to cities with higher costs of living, they also had difficulty obtaining answers to their questions about the relocation process. ECF No. 83-1 at 131.
The parties dispute the reaction to the closures in Congress and elsewhere. While Defendant claims that “Congressional briefings were well received,” ECF No. 70 ¶ 28, Plaintiff emphasizes letters from members of Congress criticizing the closures. E.g., ECF No. 83-1 at 36.
B. Procedural Background
Plaintiff K.H. filed suit on behalf himself and all similarly situated FAMs in the six closed offices on June 18, 2015. ECF No. 1. The operative Second Amended Complaint was filed on May 11, 2016, and alleges both disparate treatment and disparate impact claims under the Age Discrimination in Employment Act, 29 U.S.C. §§ 621-34. ECF No. 50. Class notice went out in August 2016, ECF No. 56, and there are now approximately 200 party-Plaintiffs. ECF No. 61.
On February 3, 2017, Defendant, with the permission of the Court, filed two separate motions for summary judgment: one related to the core liability issues in the case, ECF No. 69, and the second related to two Plaintiffs’ — F.S.’s and Scott Kief-ner’s — individual damages claims, ECF No. 73. Defendant later filed an amended motion for summary judgment as to liability, ECF No.76, and the Court terminated the summary judgment motion relating to individual damages after F.S. and Scott Kiefner dismissed their claims with prejudice. ECF No. 88.
II. LEGAL STANDARD
Summary judgment is proper when a “movant shows that there is no genuine dispute as to any material fact and the movant is entitled to judgment as a matter of law.” Fed. R. Civ. P. 56(a). “A party asserting that a fact cannot be or is genuinely disputed must support the assertion by” citing to depositions, documents, affidavits, or other materials. Fed. R. Civ. P. 56(c)(1)(A). A party also may show that such materials “do not establish the absence or presence of a genuine dispute, or that an adverse party cannot produce admissible evidence to support the fact.” Fed. R. Civ. P. 56(c)(1)(B). An issue is “genuine” only if there is sufficient evidence for a reasonable fact-finder to find for the non-moving party. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248-49, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). A fact is “material” if the fact may affect the outcome of the case. Id at 248, 106 S.Ct. 2505. “In considering a motion for summary judgment, the court may not weigh the evidence or make credibility determinations, and is required to draw all inferences in a light most favorable to the non-moving party.” Freeman v. Arpaio, 125 F.3d 732, 735 (9th Cir. 1997).
Where the party moving for summary judgment would bear the burden of proof at trial, that party bears the initial burden of producing evidence that would entitle it to a directed verdict if uncontroverted at trial. See C.A.R. Transp. Brokerage Co. v. Darden Rests., Inc., 213 F.3d 474, 480 (9th Cir. 2000). Where the party moving for summary judgment would not bear the burden of proof at trial, that party bears the initial burden of either producing evidence that negates an essential element of the non-moving party’s claim, or showing that the non-moving party does not have enough evidence of-an essential element to carry its ultimate burden of persuasion at trial. If the moving party satisfies its initial burden of production, then the non-moving party must produce admissible evidence to show that a genuine issue of material fact exists. See Nissan Fire & Marine Ins. Co. v. Fritz Cos., 210 F.3d 1099, 1102-03 (9th Cir. 2000). “A mere scintilla of evidence [794]*794will not be sufficient to defeat a properly supported motion-for summary judgment; rather, the nonmoving party must introduce some significant probative evidence tending to support the complaint.” Summers v. Teichert & Son, Inc., 127 F.3d 1150, 1152 (9th Cir. 1997) (citation and internal quotation marks omitted). If the non-moving party fails 'to make this showing, the moving party- is entitled to summary judgment. Celotex Corp. v. Catrett, 477 U.S. 317, 323, 106 S.Ct. 2548, 91 L.Ed.2d 265 (1986).
III. ANALYSIS
A plaintiff alleging discrimination under the ADEA may proceed under either of two theories: disparate treatment or disparate impact. Palmer v. United States, 794 F.2d 534, 536-37 (9th Cir. 1986). Here, Plaintiffs argue both theories and the Court addresses each in turn.
A. Disparate Impact
To make out a prima facie case of discrimination under the disparate impact theory the plaintiff generally must show: “(1) the occurrence of certain outwardly neutral employment practices, and (2) a significantly adverse or disproportionate impact on persons of a particular [age] produced by the employer’s facially neutral acts or practices.” Spaulding v. University of Washington, 740 F.2d 686, 705 (9th Cir. 1984), overruled on other grounds by Atonio v. Wards Cove Packing Co., 810 F.2d 1477, 1482 (9th Cir. 1987) (en banc).
“Discriminatory motive need not be shown under the disparate impact theory.” Palmer, 794 F.2d at 536-37 (9th Cir. 1986). The disparate impact theory’s requirements, however, “may be more exacting” than those for disparate treatment. Id. In particular, “[a] disparate impact plaintiff must not merely prove circumstances raising an inference of discriminatory impact; he must prove the discriminatory impact at issue.” Id. (citing Moore v. Hughes Helicoptors, Inc., 708 F.2d 475, 482 (9th Cir. 1983)).
Here, Plaintiffs argue that the closure of six field offices was an “outwardly neutral employment practice” that disproportionally and significantly impacted FAMs aged 40 or older. Plaintiffs’ expert, Craig Moore, made the following finding:
The average age of FAMS in the field offices that allegedly were targeted because they had a higher proportion of older employees was 44.76 years of age. The [sic] average age for all other field offices was 41.59. A statistical comparison of the individuals in these two groups shows that this difference is statistically significant.
ECF No. 71-6 at 5. Another indicator of this difference in impact is that 82.5% of the FAMs in the closed field offices were 40 or older, while in all other offices only 60.7% of FAMs were over 40. Id. at 7. This difference, too, was statistically significant. Id.3 '
Defendant attacks Plaintiffs’ statistical evidence on three grounds. First, Defendant argues that Moore did not compare the FAMs in the closed field offices to “similarly-situated” FAMs. ECF No. 76 at 20-22.. Specifically, Defendant argues FAMs from “all other field offices” are not similarly situated to FAMs in the closed offices because the closed offices were less critical to the Service, had lower passenger throughput, and serviced fewer high risk [795]*795flights. ECF No. 76 at 21. [redacted] ECF No. 77 ¶ 4; see also ECF No. 70-2 at 22 [redacted]). According to Defendant’s expert, when FAMs from the six closed offices are compared with only similar low volume offices, any statistically significant age difference disappears. ECF No. 77 ¶ 5.
In his expert report, Moore acknowledges that he had “not been asked to make any judgment regarding the relevant labor pool or to determine if these employees are similarly situated.” ECF No. 71-6. But in response to Defendant’s motion, Moore submitted a declaration stating that all FAMs have similar “skill, effort, and responsibility” and that “[t]he size of , the operations in various offices is not a factor that determines whether [the FAMs] were similarly situated].” ECF No. 89-2 ¶7-8. As for the first conclusion, the Court agrees with Defendant that Moore cannot submit new expert testimony when the expert deadline has long passed. ECF No. 90 at 14. Nevertheless, Defendant admits in reply that “the FAMs do all have the same job title and similar job responsibilities and skills.” Id. at 15. In other words, the parties agree that the FAMs as individuals are similarly situated.4
As for Moore’s second conclusion, the Court finds that differences in the workloads or strategic value of different field offices do not help determine whether their employees are similarly situated, as long as the employees in question are all doing the "same work. The question is not whether employees doing the same work face different collateral circumstances. The question is whéther the employees are similarly situated “in all material respects.” Moran v. Selig, 447 F.3d 748, 755 (9th Cir. 2006). Defendant’s only case on this point, Cooper v. Southern Company, does not say otherwise. 260 F.Supp.2d 1305 (N.D. Ga. 2003), aff'd, 390 F.3d 695 (11th Cir. 2004). There, the court rejected the. Plaintiffs statistics because they did “not account for differences in the type or level of the employees’ applied skills, both of which are highly related to compensation.” Id, at 1315. Not .so here. Even Defendant admits the FAMs all have the same job title and similar job responsibilities and skills.” ECF No. 90 at 15. In sum, there is no reason to think the comparators Moore analyzed were not similarly situated.
Second, Defendant argues that a, onetime office closure is not the type of policy that is subject to a disparate impact challenge. ECF No. 76 at 23. The Ninth Circuit held otherwise in Pottenger v. Potlatch Corp., 329 F.3d 740, 749 (9th Cir. 2003). There, the court explained that a “disparate impact claim must challenge; a specific business practice” and then determined that the challenged reduction in force “would constitute such a practice.” Id. The office closures here are equivalent to the reduction in force in Pottenger.
The out-of-circuit cases Defendant cites do not require a different conclusion. In Sneed v. Strayer University, for example, the court rejected the Plaintiffs statistical analysis because it measured impact in terms of the number of African-Americans “who were fired across the company” even though the plaintiff had only challenged the decision to "close one specific facility. No. 1:15-CV-0004-GBL-IDD, 2016 WL 1023311, at *9 (E.D. Va. Mar. 8, 2016). Plaintiffs- did not make that error here. Defendant also cites Aliotta v. Bair, for the proposition that one-time office closures [796]*796cannot form the basis of a disparate impact claim. 576 F.Supp.2d 113, 127 (D.D.C. 2008), aff'd, 614 F.3d 556 (D.C. Cir. 2010). But in that case, the problem was' the Plaintiffs’ attempt to frame as a single policy multiple actions (i.e., different sets of involuntary terminations and voluntary buyouts) taken by the defendant over several years. Id. Plaintiffs’ challenge to the Service’s decision to close six field offices does not suffer from that flaw.
Finally, Defendant argues that Plaintiffs disparaté impact claim fails because a 3-yeár age difference is insufficiently substantial.5 EOF No. 76 at 22-23. Here, the Court agrees. In support of this argument, Defendant relies on two disparate treatment cases that addressed what age difference is sufficient to create an inference that an employment decision was based on illegal age bias. In O’Connor v. Consolidated Coin Caterers Corporation, 517 U.S. 308, 116 S.Ct. 1307, 134 L.Ed.2d 433 (1996), the Supreme Court held that an inference of discrimination could not “be drawn from the replacement of one worker with another worker insignificantly younger.” Id. at 313, 116 S.Ct. 1307. Only a “substantially younger” • replacement was “reliable indicator of age discrimination.” Id. The Ninth Circuit was more specific in France v. Johnson, holding that “an average age difference often years or more between the plaintiff and the replacements will be presumptively substantial, whereas an age difference of less than ten years will be presumptively insubstantial.” 795 F.3d 1170, 1174 (9th Cir. 2015), as amended on reh’g (Oct. 14, 2015); see also Diaz v. Eagle Produce Ltd. P’ship, 521 F.3d 1201, 1210 (9th Cir. 2008) (finding that a disparity of 16 years between those hired and those laid off was substantial).
Measured against that benchmark, the less than 3-year age difference reported by Plaintiffs’ expert is insufficient to support an inference of discrimination. Of course, both O’Connor and France were disparate treatment claims, but the Court sees no reason why the standard they set out should not apply in the more “exacting” disparate impact context. Palmer, 794 F.2d at 536-37. Tellingly, Plaintiffs do not cite and the Court could not locate any case in which a court relied on an age difference of less than three years to deny summary judgment on a disparate impact claim. Nor are O’Connor and France distinguishable because those cases involved “replacements” rather than “comparators.” Because both Plaintiffs asserted failure-to-promote claims, the replacements were the comparators. There is no reason not to apply the logic of those cases here just because the comparators are FAMs in unaffected offices.
Not only is the age difference between Plaintiffs and their comparators small, but the average age of FAMs in unaffected offices was 41.59. ECF No. 71-6 at 5. That is above the age 40 cutoff for protected status under the ADEA. Although perhaps not determinative, this too weighs against a finding that the office closures had a “significantly adverse or disproportionate impact on persons of a particular [age].” Spaulding, 740 F.2d at 705; see also Schechner v. KPIX-TV, No. C 08-05049 MHP, 2011 WL 109144, at *5 (N.D. Cal. Jan. 13, 2011)6 (explaining that although a 55-year old replaced by a 40-year old could prevail on a disparate treatment claim, the “focus for a disparate impact claim .,. must be on the impact of. a facially age-blind employment decision on a speciflcally-identi-[797]*797fied protected group to which Plaintiffs belong” and to which the comparators do not).
In sum, the Court concludes that Plaintiffs have faded to make out a prima facie case on their disparate impact claim. The statistical analysis offered by Plaintiffs’ expert, which reveals an age difference between Plaintiffs and their comparators of less than three years, does not show “a significantly adverse or disproportionate impact on persons of a particular' [age] produced by the employer’s facially neutral acts or practices.” Spaulding, 740 F.2d at 705. The Court therefore grants summary judgment as to Plaintiffs’ disparate impact claim.7
B. Disparate Treatment
1. Prima Facie Case
An employee can establish a pri-ma facie ease of age discrimination under the disparate treatment theory by showing he: (i) was a member of the protected class, (ii) was performing his job in a satisfactory manner, (iii) experienced an adverse employment action, and (iv) was replaced by a substantially younger employee with equal or inferior qualifications. Douglas v. Anderson, 656 F.2d 528, 533 (9th Cir. 1981); Peterson v. Hewlett-Packard Co., 358 F.3d 599, 603 (9th Cir. 2004). Defendant does not argue that Plaintiffs fail to make out a prima facie case of disparate treatment, and the Court concludes that they have. First, Plaintiffs are all over 40. Second, Plaintiffs assert that there were no complaints about their job performance and Defendant does not argue otherwise. Third, Plaintiffs suffered an adverse action when they were forced to transfer offices. Fourth, when the Service finally hired new FAMs in 2016,8 they were substantially younger than the ■ FAMs impacted by the office closures.9
2. “But-for” Versus McDonnell Douglas Burden Shifting
Rather than challenge Plaintiffs’ prima facie case, Defendant focuses on the argument that Plaintiffs cannot show that age was the “but for” cause of the office closures. ECF No. 76 at 24-25. But Defendant misstates the standard for disparate treatment claims under the ADEA at summary judgment. In Gross v. FBL Financial Services, Inc., the Supreme Court held that a “plaintiff bringing a disparate-treatment claim pursuant to the ADEA must prove, by a preponderance of the evidence, that age was the ‘but-for’ cause of the challenged adverse employment action,” not just that it was a “motivating factor.” 557 U.S. 167, 180, 129 S.Ct. 2343, 174 L.Ed.2d 119 (2009). Three years later, however, the Ninth Circuit cabined Gross’s holding to the trial context. Shelley v. Geren, 666 F.3d 599, 607-08 (9th Cir. 2012). The Shelley court explained that because “Gross involved a case that had already progressed to trial, it did not address the [798]*798evidentiary framework applicable to a motion for summary judgment” and in fact “explicitly noted that it ‘has not definitively decided whether the evidentiary framework of McDonnell Douglas utilized in Title VII cases is appropriate in the ADEA context.’ ” Id. (quoting Gross, 557 U.S. at 175 n.2,129 S.Ct, 2343). The Ninth Circuit went on to conclude that McDonnell Douglas framework was still applicable at summary judgment and summarized the inquiry as follows:
Thus, to survive summary judgment on his claim for a violation of the ADEA under the disparate treatment theory of liability, Shelley must first establish a prima facie case of age discrimination. If he is successful, the burden of production shifts to the Corps to articulate a legitimate non-discriminatory reason for its adverse employment action. It is then •Shelley’s task to demonstrate that there is a material genuine issue of fact as to whether the employer’s purported reason is pretext for age discrimination. At. trial, he must carry the burden to prove that age was the “but-for” cause of his non-selection.
Id. (internal citations omitted), Under Shelley, Plaintiffs need not show that age was the but-for cause of the field office closures, to defeat , summary judgment; that is their burden at trial.10
3. Legitimate Non-Discriminatory Reason
Because Plaintiffs establish a prima fa-cie case of discrimination, the “burden of production shifts to the [Service] to articulate a legitimate non-discriminatory reason for its adverse employment action.” Shelley, 666 F.3d at 608. Here, the Service submits substantial evidence that it closed the six field offices in order to increase its ability to cover more high-risk flights with greater efficiency. Sarandrea’s “Reshaping the FAMS Workforce” .Report details the ratio of domestic versus international flights at each field office and the number of passengers transiting the airports served by each office. ECF No. 70-1 at 11-12. Based on that data, the Report recommended closing six field offices with “only minimal strategic value” and suggested “that the FAMs at those offices be reassigned to field offices that needed additional personnel to cover higher-risk flights.” ECF No. 70 ¶ 16; ECF No. 70-1. Bray and, Pistole endorsed and executed these recommendations, ECF No. 70-2. Defendant’s risk and efficiency-based justifications for the closures are a legitimate non-discriminatory reason for'its adverse employment action.'
4, Pretext
The remaining question for purposes of summary judgment, therefore, is whether Plaintiffs can “show that the articulated reason is pretextual ‘either directly by persuading the court that a discriminatory reason more likely motivated the employer, or, indirectly by showing that-the employer’s proffered explanation is unworthy of credence.’ ” Chuang v. University of California Davis, 225 F.3d 1115, 1123 (9th Cir. 2000) (quoting Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 256, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981)).11
[799]*799Before addressing Plaintiffs’ evidence, the Court notes Defendant’s argument that there was nothing unlawful about the Service considering budget limits in its closure decision. So far as it goes, the statement is correct. If the Service had employed more FAMs than it needed to achieve its mission, it was entitled to close less important offices to 'conserve resources. What the. Service could not do, however, was purposefully close offices with a greater'number of older FAMs to force them to retire and then hire less expensive replacements. In both situations, the Service’s motives could fairly be described as budget-related. But only the first course of action is permitted under the ADEA.12 . .
Plaintiffs argue that here, Bray’s statements at. meetings to announce the field office closures constitute direct evidence that age bias- motivated the Service’s decision. FAM [redacted] testified that at one such meeting, he asked Bray if he “realize[d] that [he was] going to lose a bunch of senior people” over the closures and Bray replied that if they lost “a lot of senior guys over this” the Service. “would hire more guys and they’d be younger and he’d pay them less.” ECF No. 83-1 at 86-87. Another FAM, [redacted], similarly reported that Bray had told the meeting attendees that the Service “ ‘could hire two younger FAMs for every one of us’ and they could pay them less.” ECF No.' 83 — 3 at 2.13 These comments, if accurately reported, are direct evidence that the age of FAMs in the affected offices contributed to thfe Service’s decision.14
[800]*800Defendant offers several reasons to reject the evidence of Bray’s remarks. None has merit. First, Defendant argues that Bray denies making the statement and several other FAMs testified that they did not recall him saying it. ECF No. 76 at 26-27. Even if that were true, [redacted] and [redactedj’s testimony constitutes “sufficient evidence for a reasonable fact-finder” to believe Plaintiffs’ version of events. Anderson, 477 U.S. at 248-49, 106 S.Ct. 2505. Second, Defendant argues that it would be “illogical” for Bray to make the statement because the Service’s hiring freeze meant that Bray could not have immediately replaced older FAMs with younger ones. At summary judgment, however, the Court “is required to draw all inferences in a light most favorable to the non-moving party.” Freeman, 125 F.3d at 735. That Defendant can now articulate a reason why Bray would not have said that he could “hire more guys and they’d be younger and he’d pay them less,” ECF No. 83-1 at 86-87, does not prove that he did not make the statements. Finally, Defendant argues that Bray’s statements are irrelevant because he “was not the decision maker with respect to the office closures.” ECF No. 76 at 28. This fact, too, is disputed. Plaintiffs point out that Bray, as Director of the Service, recommended that Pistole approve Sarandrea’s Report. ECF No. 70-2. Moreover, in announcing the closure decision via email, Bray wrote: “Based on the results of the future staffing and field office assessment ..., I have made the decision, with the Administrator’s approval, to close six offices: Cincinnati, Cleveland, Phoenix, Pittsburgh, San Diego, and Tampa.” ECF No. 83-1 at 23 (emphasis added). Given this evidence, the Court cannot conclude that it is undisputed that “Mr. Bray was not the decision maker.” ECF No. 76 at 25.
Plaintiffs argue that this direct evidence of discriminatory intent is sufficient to rebut Defendant’s legitimate non-discriminatory reason, citing Ezell v. Potter, 400 F.3d 1041, 1051 (7th Cir. 2005). In Ezell, the plaintiff presented evidence that his supervisor claimed he “had a plan to get rid of older workers and replace them with younger, faster workers.” Id. The court concluded that “Ezell [did] not need the assistance of the burden-shifting test for he produced direct evidence of his supervisor’s discriminatory intent.” Id. The Court finds Ezell’s logic persuasive, but the Ninth Circuit has instructed that “[a]ll of the evidence — whether direct or indirect— is to be considered cumulatively” when considering “whether the proffered reason is pretextual.” Shelley, 666 F.3d at 609. The Court therefore considers Bray’s statements together with Plaintiffs’ other circumstantial evidence.
Where an employer’s proffered explanation is “internally inconsistent,” that evidence tends to show that the explanation is pretext for discrimination. Shelley, 666 F.3d at 609. Here, Plaintiffs argue that, although the Service justifies its decision as an attempt to cover a greater .number of high-risk flights with limited employee resources, ECF No. 76 at 18-19, a 2016 Report by the Government Accountability Office (“GAO”) found that the Service was not considering risk when making budgeting decisions. ECF No. 83-1 at 182. Specifically, the GAO Report contained the following conclusions:
First, FAMS officials reported that they consider risk when selecting specific domestic and international flights to cover, but we found that they did not consider risk when deciding how to initially divide their annual resources between domestic and international flights. Rather, each year FAMS con[801]*801siders two variables — travel budget and number of air marshals — to identify the most efficient way to divide the agency’s resources between domestic and international flights. With this approach, FAMS attempts to maximize the total number of flights it can cover with its budget.
ECF No. 83-1 at 182 (emphasis added). In his deposition, Sarandrea said he agreed with the GAO’s findings. ECF No. 71-1 at 63. How could the Service have based its 2014 closure decisions on risk management, Plaintiffs argue, if two years later the GAO concluded the Service did not consider risk in budgeting?
Defendant argues that Plaintiffs have misinterpreted the GAO Report. First, Defendant emphasizes that, even according to the GAO, the Service did “consider risk when selecting specific domestic and international flights to cover.” ECF No. 83-1 at 182. But that does not undermine Plaintiffs? argument that risk was not considered in budgeting, which is relevant to the closure decision. Second, Defendant casts the GAO Report as a series of recommendations for how the Service can improve its risk assessment, rather than a criticism for not considering risk at all. ECF No. 90 at 12. A supplemental declaration by Saran-drea bolsters this view. ECF No. 95. Indeed, Sarandrea offers a nuanced explanation of the GAO’s recommendation that the Service take a different approach to flight coverage decisions. Id. ¶ 39. Defendant’s problem again, however, is that at summary judgment, the Court must draw all inferences in Plaintiffs’ favor. While one reading of the GAO Report might be consistent with the Service’s justification for the closure decision, Plaintiffs have identified statements in the Report that on their face appear inconsistent with that justification. Read in that light,- the Report does “undermine” the contrary claims of the Service’s witnesses that the closure decision was entirely risk-based. ECF No. 90 at 12. The GAO Report is circumstantial evidence of pretext.
Next, Plaintiffs argue that the inclusion of age data in Sarandrea’s Reshaping the FAMS Workforce Report suggests that age bias contributed to its recommendations. For example, the Report notes that 250 of the 287 affected employees were 40 or older. ECF No. 70-1 at 13.15 Similarly, the Report notes that an “estimated 31 FAMs will retire or resign from the six affected offices in the next two years,” reducing relocation costs from full projections. Id. Defendant claims to have'included this information “so that management would recognize that a number of FAMs would soon have the option of retiring, and there was thus a risk that they would not agree to transfer, as the FAMS wanted them to do.” ECF No. 90 at 10. As Saran-drea stated in his deposition, “I was concerned that we were going to lose employees. [Leadership] needed to know what was at risk, especially, you know, the people that already had their time and grade, could retire. This may be a reason for them — you know, they may not want to go because they have — they’re in that location.” ECF No. 71-1'at 31.
Although the references to age in the report are quite limited and Defendant has offered a logical explanation for their inclusion, the Court nonetheless finds that the Report constitutes circumstantial evidence, though weak, of pretext. Shelley is again [802]*802instructive. There, two supervisors “sought out the retirement dates [of applicants] at the time of their participation in the hiring process for the temporary and permanent positions.” Shelley, 666 F.3d at 61Q.16 The Ninth Circuit held that this fact “shows more than that the decision-makers may have known of the candidates’ ages. It raises an inference that they considered this information relevant to their decisions.” Id. By analogy, the inclusion of age data in the Report suggests that age may have played a role in the closure decisions.17
Finally, Plaintiffs put forward evidence that, in response to the closure decision, 61 FAMs 'separated, from the Service rather than move to a new field office, ECF No. 83-1 at 147-52. Then, in 2016, the Service hired 326 new FAMS after the end of the. hiring freeze, ECF No. 70 ¶29. Forty-eight of these new FAMs were substantially younger (average age of 31) than the 51 FAMs who had left during the closures. ECF No. 83-Í at 158. Viewed in the light most favorable to Plaintiffs, those younger hires make it appear that the Service made good on Bray’s alleged statement that if the Service lost “a lot of senior guys over this” it “would hire more guys and they’d be younger and he’d pay them less.” ECF No. 83-1 at 86-87.
Defendants claim these hires were not “replacements" because the new FAMs were not hired into any of the six closed offices. ECF No. 90 at 17. That elevates form over substance. Nevertheless, this evidence is entitled to less weight because the new hires occurred two years after the Service announced its closure decision. See Rose v. Wells Fargo & Co., 902 F.2d 1417, 1422 (9th Cir. 1990) (finding that replacement with a younger employee “six or seven months after Reed’s discharge ‘substantially weaken[s]’ his claim”) (quoting Simpson v. Midland-Ross Corp., 823 F.2d 937, 941 (6th Cir. 1987)).
In sum, the GAO Report, the references to age in Sarandrea’s Report, and the fact that roughly the same number of younger FAMs were hired to replace those who left after the closures, constitute circumstantial evidence of pretext. Considered' cumulatively with the direct evidence of Bray’s statements, Plaintiffs have raised a genuine issue of material fact as to whether the Service’s preferred reasons for the. closure decision were pretext for age discrimination. The Court therefore denies Defendant’s motion as to Plaintiffs’ disparate treatment claim,
CONCLUSION
The Court grants the motion for summary judgment as to Plaintiffs’ disparate impact claim, and denies it as to the disparate treatment claim.
IT IS SO ORDERED.