BEA, Circuit Judge

Charles Robinowitz (argued), Portland, OR, for Petitioner.

John R. Dudrey (argued), Williams Fredrickson, LLC, Portland, OR, for Respondent Pacific Rim Environmental Resources, LLC.

Kathleen Kim and Mark A. Reinhalter, Office of the Solicitor, United States Department of Labor, Washington, D.C., for Respondent Director, Office of Workers' Compensation Programs.

OPINION

BEA, Circuit Judge:

Richard Shirrod was awarded benefits under the Longshore and Harbor Workers' Compensation Act (“Longshore Act”) for injuries he sustained while working on a barge-refitting project for Respondent Pacific Rim Environmental Resources, LLC (“Pacific Rim”). During the workers'-compensation proceedings, an administrative law judge (“ALJ”) awarded Shirrod $33,581.17 in attorney's fees for work Shirrod's attorney, Charles Robinowitz, performed before the ALJ, as is authorized under a fee-shifting provision of the Longshore Act. See 33 U.S.C. § 928(a). The Benefits Review Board (“BRB”) affirmed the ALJ's fee award. The fee award is based on an hourly rate for Robinowitz's services of $340 per hour. Shirrod contends the formula on which this $340–per–hour rate is based is flawed. We agree. We grant Shirrod's petition for review, vacate the BRB's decision and order, and remand this case for further proceedings.

I

Shirrod sustained permanent injuries to his knee and ankle during the course of his employment with Pacific Rim. Shirrod was working in Oregon on a project to refit two rail cars and a barge to carry solid waste from ocean-going barges to waste dumps. Shirrod filed a claim for workers'-compensation benefits under the Longshore Act, and it is undisputed that he is entitled to such benefits. Shirrod's case was overseen by ALJ Steven Berlin.

The Longshore Act permits a claimant to recover attorney's fees in some circumstances and, in this case, Shirrod requested attorney's fees for work Robinowitz performed before Judge Berlin, mostly during 2010. The fee request totaled $38,786.17, including 86.75 hours for Robinowitz's services at $400 per hour, 8.25 hours of legal-assistant services at $150 per hour, and $2,848.67 in costs. Shirrod submitted several affidavits and legal-industry fee surveys to support the proposed $400–per–hour rate for Robinowitz's services, including the Oregon State Bar's 2007 Economic Survey (“Bar Survey”), which provides billing rates for legal services in several Oregon markets. Robinowitz has been a lawyer in private practice in Portland, Oregon, since 1969.

Judge Berlin approved an attorney's-fee award of $33,581.17; he reduced the rate for Robinowitz's services to $340 per hour but approved Shirrod's fee request in all other respects. In determining the proper rate for Robinowitz's services, Judge Berlin primarily relied on the analysis developed by ALJs in two earlier Longshore Act cases in which Robinowitz served as counsel: DiBartolomeo v. Fred Wahl Marine Constr., ALJ No. 2008–LHC–01249 (Dep't of Labor Oct. 26, 2009) (ALJ Gerald Etchingham), and Castillo v. Sundial Marine Tug & Barge Works, Inc., ALJ No. 2010–LHC–00341 (Dep't of Labor Apr. 22, 2011) (ALJ Jennifer Gee).

The claimants in this case, DiBartolomeo, and Castillo submitted similar affidavits and legal-industry fee surveys to support the hourly rate requested for Robinowitz's services. In each case, the presiding ALJ rejected that evidence as not probative of the market rates for Longshore Act litigation and instead found it necessary to develop a proxy market rate. Both Judge Gee in Castillo and Judge Berlin in this case relied on the $316.42–per–hour rate Judge Etchingham calculated in DiBartolomeo as a proxy market rate.

In DiBartolomeo, Judge Etchingham “estimat[ed] the value of Mr. Robinowitz's services in the Portland market” using data from the 2007 Survey of Law Firm Economics by Altman Weil Publications (“Altman Weil Survey”). He determined that Robinowitz's credentials and performance merited a rate in the 75th percentile. As a result, Judge Etchingham averaged the 75th-percentile hourly billing rates for lawyers practicing in areas he found relevant to Longshore Act litigation—employment law ($365), maritime law ($320), personal-injury law ($335), and workers'-compensation law ($200)—and Oregon lawyers with over 30 years of experience ($325). He thus calculated a proxy market rate of $309 per hour for Robinowitz's services, which he increased to $316.42 to account for inflation.

Judge Etchingham disavowed any reliance on the Bar Survey, even though the Bar Survey includes more detail about billing rates in various Oregon markets, including Portland. Judge Etchingham determined that the Altman Weil Survey was a better source of data because it is published every year and provides billing rates for lawyers practicing employment law and maritime law, whereas the Bar Survey is published every four to five years and does not include separate rates for those practice areas.

Judge Berlin agreed with Judge Etchingham's approach in DiBartolomeo and adopted the $316.42–per–hour proxy market rate Judge Etchingham calculated for Robinowitz's services. Judge Berlin then increased this rate to $320 to account for inflation and to $340 to acknowledge Robinowitz's recent accomplishments. He thus approved a fee award totaling $33,581.17. The BRB affirmed this fee award.

II

We have jurisdiction to review final orders of the BRB. 33 U.S.C. § 921(c). The BRB's decisions are subject to the provisions of the Administrative Procedure Act. See Haw. Stevedores, Inc. v. Ogawa, 608 F.3d 642, 648 (9th Cir.2010). Thus, we must set aside decisions of the BRB that are “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A). The BRB must accept an ALJ's factual findings unless they are contrary to law, irrational, or not supported by substantial evidence. Van Skike v. Dir., OWCP, 557 F.3d 1041, 1045–46 (9th Cir.2009). We independently evaluate the evidence in the administrative record to ensure the BRB adhered to the correct standard of review. See id.; Bumble Bee Seafoods v. Dir., OWCP, 629 F.2d 1327, 1329 (9th Cir.1980).

III

If an employer or insurance carrier denies liability for a Longshore Act claim, it must pay a “reasonable attorney's fee” to the claimant if the claimant successfully prosecutes his claim with the aid of an attorney. 33 U.S.C. § 928(a). A successful claimant may collect attorney's fees for work his attorney performed before ALJs, the BRB, and District Directors in the Office of Workers' Compensation Programs. Id.; 20 C.F.R. §§ 702.132, 802.203.

The term “reasonable attorney's fee” has evolved toward a uniform definition in all federal fee-shifting statutes, including the Longshore Act. See Christensen v. Stevedoring Servs. of Am., 557 F.3d 1049, 1052 (9th Cir.2009). The lodestar method, which requires multiplying a reasonable hourly rate by the number of hours reasonably expended on the case, is the starting point for the calculation of attorney's fees. Id. at 1053 & n. 4. The goal of the lodestar method is to “produce[ ] an award that roughly approximates the fee that been representing a paying client who was billed by the hour in a comparable case.” Perdue v. Kenny A. ex rel. Winn, 559 U.S. 542, 550–51, 130 S.Ct. 1662, 176 L.Ed.2d 494 (2010) (emphasis omitted). As such, an attorney's hourly rate is to be calculated “according to the prevailing market rates in the relevant community” and should be “in line with [the rates] prevailing in the community for similar services by lawyers of reasonably comparable skill, experience and reputation.” Christensen, 557 F.3d at 1053 (quoting Blum v. Stenson, 465 U.S. 886, 895–96, 104 S.Ct. 1541, 79 L.Ed.2d 891 (1984)).

Determining the prevailing market rate for claimants' attorneys in Longshore Act cases has proven difficult, as the Longshore Act prohibits claimants from negotiating fees with their attorneys. See 33 U.S.C. § 928(e); Christensen, 557 F.3d at 1053; 20 C.F.R. §§ 702.132(a), 802.203(f). We have held that fees in Longshore Act cases should be “commensurate with those which [claimants' attorneys] could obtain by taking other types of cases.” Christensen, 557 F.3d at 1053 (quoting Camacho v. Bridgeport Fin., Inc., 523 F.3d 973, 981 (9th Cir.2008)). Prior to our decisions in Christensen and Van Skike, Longshore Act fee awards were typically based on fee awards from earlier Longshore Act cases. Id. at 1054; accord Van Skike, 557 F.3d at 1046–47. In Christensen, we found that self-referential approach improper, because it did not look to rates in an independently operating market and, hence, did not produce “market” rates, as is required by the Supreme Court's decision in Blum v. Stenson. See 557 F.3d at 1054–55. As such, we held that, in Longshore Act cases, the “relevant community” must be defined “more broadly than simply fee awards under the [Longshore Act].” Id.; accord Van Skike, 557 F.3d at 1046–47. Recognizing that the relevant decisionmaker has wide—but not unlimited—discretion when making attorney's-fee awards, see Kenny A., 559 U.S. at 558, 130 S.Ct. 1662, we ultimately left it to the BRB, ALJs, and District Directors to determine the “relevant community” and the prevailing market rates in that community, as long as the decisionmaker provides adequate justification. Christensen, 557 F.3d at 1055.

A. The “relevant community ”

Our review of the fee award must begin by investigating just what was the “relevant community” that Judge Berlin used when he awarded fees for Robinowitz's services. Judge Berlin left the “relevant community” undefined but used language suggesting that the “relevant community” was either the state of Oregon or the city of Portland. The BRB, in its review of the fee award, also made no express mention of the “relevant community.” However, which is the “relevant community” need not necessarily be decided anew in each decision awarding fees. See Christensen, 557 F.3d at 1055. We take the “relevant community” here to be Portland, Oregon, the “relevant community” determined by the ALJs in the two prior cases on which Judge Berlin relied. See DiBartolomeo, ALJ No.2008–LHC–01249, slip op. at 7 (“[T]his Court is left with the task of estimating the value of Mr. Robinowitz's services in the Portland market.”); Castillo, ALJ No.2010–LHC–00341, slip op. at 3 (“I find that the ‘relevant community’ for Mr. Rabinowitz [sic] is the city of Portland.”).

We also agree with Shirrod that Portland is the right “relevant community” in this case. In civil litigation, we typically recognize the forum where the district court sits as the “relevant community” for purposes of fee-shifting statutes. Christensen, 557 F.3d at 1053; Barjon v. Dalton, 132 F.3d 496, 500 (9th Cir.1997). By analogy, a determination of the “relevant community” in Longshore Act cases should focus on the location where the litigation took place. But, because district courts are not involved in cases under the Longshore Act, we must look to other indicia to determine where the litigation took place and, thus, which is the “relevant community.” Here, all factors point to Portland as the location of the litigation: Counsel to Shirrod and Pacific Rim maintain their offices in Portland; hearings before Judge Berlin occurred in Portland. Cf. Newport News Shipbuilding & Dry Dock Co. v. Holiday, 591 F.3d 219, 229 (4th Cir.2009) (noting that the “relevant community” should “turn[ ] on inquiries about the lawyer and client”); 20 C.F.R. § 802.203(d)(4) (“The rate awarded by the [BRB] shall be based on what is reasonable and customary in the area where the services were rendered for a person of that particular professional status.”).

B. Applicability of the proxy market rate to the Portland market

We next examine whether the proxy market rate developed by Judge Etchingham in DiBartolomeo and relied upon by Judge Berlin in this case adequately reflects market rates for Portland, Oregon, the “relevant community.” We conclude that it does not because it is based entirely on data not tailored to Portland, even though reliable information about attorney billing rates in Portland was readily available.

The proxy market rate adopted by Judge Berlin is based on five constituent rates from the Altman Weil Survey: the 75th-percentile rates for lawyers practicing employment, maritime, personal-injury, and workers'-compensation law, and the 75th-percentile rate for Oregon lawyers with over 30 years of experience. See DiBartolomeo, ALJ No.2008–LHC–01249, slip op. at 7–8. The constituent rates do not relate specifically to Portland, Oregon, the “relevant community”; one of the rates is expressly geographically limited-but encompasses the entire state of Oregon-and, although it is not entirely clear from the record, the practice-area rates appear to be national in scope. See id. We hold that it was error for Judge Berlin to apply a rate bearing no direct nexus to the “relevant community.” See Camacho, 523 F.3d at 979 (vacating an attorney's-fee award because the market rate was based “almost exclusively” on data for judicial districts other than the Northern District of California, the “relevant community”).

To be sure, if the “relevant community” were defined differently, or if reliable data on attorney's fees in the “relevant community” did not exist, using the proxy market rate—or the constituent rates on which it depends—could be permissible. Cf. id. (suggesting the market rate may be based on rates for communities “comparable to” the “relevant community”). Here, however, the Bar Survey does provide attorney's-fee information specific to the “relevant community” of Portland, Oregon. Although Judge Etchingham in DiBartolomeo gave two reasons for favoring the Altman Weil Survey over the Bar Survey—which Judge Berlin implicitly adopted in this case—we find those reasons unconvincing.

First, Judge Etchingham stated that the Altman Weil Survey is a better source of information because it is published every year, whereas the Bar Survey is published only every four to five years. DiBartolomeo, ALJ No.2008–LHC–01249, slip op. at 7 n. 2. But if Judge Etchingham—and other ALJs relying on his analysis and methodology—preferred the Altman Weil Survey because it is published annually, we would expect them to calculate proxy market rates for each year using the data published for that year. However, to determine proxy market rates for years after 2007, Judge Etchingham in DiBartolomeo, Judge Gee in Castillo, and Judge Berlin in this case merely adjusted the proxy market rate calculated from the 2007 Altman Weil Survey using inflation data. See id. at 8; Castillo, ALJ No.2010–LHC–00341, slip op. at 11–12 & n. 10. We see no reason why the ALJs could not have made the same inflation adjustments to the figures in the Bar Survey.

Second, Judge Etchingham stated that he relied on the Altman Weil Survey because it includes billing rates for attorneys practicing employment law and maritime law, which the Bar Survey lacks. DiBartolomeo, ALJ No.2008–LHC–01249, slip op. at 7 n. 2. Judge Etchingham did not explain why he disregarded the data for the practice areas—personal-injury law and workers'-compensation law—that the surveys have in common and for which the Bar Survey contains data specific to the Portland market. Relatedly, Judge Etchingham did not explain why the Altman Weil Survey's billing rate for Oregon attorneys with over 30 years of experience was superior to the Bar Survey's billing rate for Portland attorneys with the same amount of experience. See id. Nor did Judge Etchingham attempt to harmonize the data in the two surveys in any other way. Cf., e.g., Ramsey v. Cascade Gen., Inc., No. 11–0875, 2012 WL 3903607, at *2–*4 (Ben.Rev.Bd. Aug. 29, 2012) (affirming an attorney's-fee award that was “based on averages of the rates in the Altman Weil Survey, the Oregon Bar Survey and the average hourly rate for all Portland attorneys with 21–30 years of experience, regardless of practice area”).

Because the lodestar method requires a “reasonable attorney's fee” to be based on market rates in the “relevant community,” we hold that the BRB erred in affirming an attorney's-fee award based on a proxy market rate not tailored to the “relevant community,” which, in this case, Judge Berlin found to be Portland.

C. Use of billing rates for workers'-compensation practice

The proxy market rate applied by Judge Berlin depends in part on the 75th-percentile rate, from the Altman Weil Survey, for attorneys practicing workers'-compensation law. See DiBartolomeo, ALJ No.2008–LHC–01249, slip op. at 7–8. Shirrod contends that it was error to include that rate, because it is artificially low and does not represent a market rate. We agree.

The proxy market rate includes, among its constituent rates, a $200–per–hour rate reported by attorneys practicing workers'-compensation law, which is significantly lower than the other rates on which the proxy market rate depends. See id. Judge Etchingham included that rate as part of the proxy market rate he calculated in DiBartolomeo because he found that workers'-compensation law “employ[s] legal skills similar to those required by Longshore practice,” id. at 7, and the BRB affirmed, see DiBartolomeo v. Fred Wahl Marine Constr., No. 10–0257, 2010 WL 3514186, at *5 (Ben.Rev.Bd. Aug. 30, 2010). Judge Berlin in this case agreed with Judge Etchingham that “Oregon state workers' compensation fees are analogous and relevant to a determination of the billing rate applicable to trial-level work” in Longshore Act cases, and the BRB affirmed this decision, giving scant discussion to the use of rates reported by workers'-compensation lawyers.

Yet, the BRB previously recognized that billing rates reported by workers'-compensation lawyers do not necessarily represent market rates that are usable in a lodestar calculation. In Christensen, we vacated the BRB's decision awarding fees for work Robinowitz performed before it because the fee award was not based on market rates. See 557 F.3d at 1052–55. On remand from us, the BRB initially calculated a proxy market rate “by reference to an average of the rates for workers' compensation, plaintiff personal injury civil litigation, and plaintiff general civil litigation cases.” Christensen v. Stevedoring Servs. of Am., 43 Ben. Rev. Bd. Serv. (MB) 145, 146–47 (2009). However, the BRB reconsidered this decision, excluded the workers'-compensation rate, and recalculated the fee award based only on the average of the other two rates. Christensen v. Stevedoring Servs. of Am., 44 Ben. Rev. Bd. Serv. (MB) 39, 40 (2010). It did so because it found the reported workers'-compensation rate to be artificially low and thus not reflective of market rates: Fees paid to claimants' attorneys are typically capped by state law and often paid out of the compensation award, whereas insurers and employers are able to negotiate discounts with attorneys who defend against workers'-compensation claims because they supply a steady stream of work. Id. In response to the employer's request for reconsideration of this decision, the BRB added that “[f]ees awarded by state administrative law judges are not necessarily based on market considerations, just as rates set by administrative law judges in longshore cases have been held to be non-market-based rates.” See Christensen v. Stevedoring Servs. of Am., 44 Ben. Rev. Bd. Serv. (MB) 75, 75 (2010) (citing Christensen, 557 F.3d 1049; Van Skike, 557 F.3d 1041), aff'd sub nom. Stevedoring Servs. of Am. v. Dir., OWCP, 445 Fed.Appx. 912 (9th Cir.2011).

Although Judge Berlin did not mention the BRB's decision on reconsideration in Christensen by name, he gave two reasons why it would not apply to this case. First, Judge Berlin noted that BRB decisions regarding attorney's fees for appellate work performed before the BRB—such as Christensen—are “less instructive” than BRB decisions reviewing attorney's-fee orders for trial-level work performed before ALJs—such as the BRB's decision affirming Judge Etchingham's attorney's-fee award in DiBartolomeo. We see nothing in the BRB's Christensen decision that would detract from its applicability to attorney's-fee orders for trial-level work. We do not dispute that the skills involved in resolving state workers'-compensation claims are similar to those involved in litigating Longshore Act cases and may be more relevant to trial-level work than appellate work. See, e.g., DiBartolomeo, 2010 WL 3514186, at *5. But, in Christensen, the BRB excluded rates reported by workers'-compensation lawyers not because workers'-compensation practice was irrelevant, but because structural factors caused workers'-compensation lawyers to report unusually low—and non-market—hourly rates. See 44 Ben. Rev. Bd. Serv. (MB) at 40. If reported rates for workers'-compensation practice do not reflect market rates in the “relevant community,” they cannot be used in a lodestar calculation, no matter how similar the skills involved are. See Christensen, 557 F.3d at 1054–55 (rejecting reliance on the hourly rates awarded to attorneys in previous Longshore Act cases); accord Van Skike, 557 F.3d at 1046–47.

Second, and relatedly, Judge Berlin called the cap on fees for claimants' attorneys in Oregon workers'-compensation cases an “open market factor” and suggested that rates reported by workers'-compensation lawyers are market rates because workers'-compensation attorneys “freely choose to represent injured workers in cases very similar to [Longshore Act cases], and they do so knowing of the cap.” But this reasoning contradicts the BRB's decision in Christensen; there, the BRB found rates reported by Oregon workers'-compensation attorneys not representative of market rates in part because the total fee paid to claimants' attorneys is capped by statute. See 44 Ben. Rev. Bd. Serv. (MB) at 40; see also, e.g., Or.Rev.Stat. § 656.308(2)(d). We share the BRB's skepticism that billing rates reported under such a regime reflect rates that are “sufficient to induce a capable attorney to undertake the representation,” Kenny A., 559 U.S. at 552, 130 S.Ct. 1662, because the fee cap may reduce reported hourly rates below expected hourly rates.

For example, suppose that an Oregon workers'-compensation attorney reasonably expects to earn $400 per hour and spend 10 hours on a case. He then spends 15 hours on the case—for a total expected fee of $6,000—but his fee is capped by statute at $4,500. Because of the fee cap, the attorney's average hourly rate for the case is $300, or $100 lower than the hourly rate he expected when he accepted the client. Notwithstanding the operation of the fee cap in that case, the attorney would still expect to receive $400 per hour in his next case. If, in the next case, the attorney spends only 10 hours, basing his fee on the $300–per–hour rate earned in the first case would undercompensate him, as that rate represents a blend of full compensation for some time and no compensation for other time. The lodestar method already requires the relevant decisionmaker to award attorney's fees only for the number of hours reasonably expended on the litigation—and to award no compensation for any unreasonably spent time. See, e.g., Blum, 465 U.S. at 888, 104 S.Ct. 1541. Using a billing rate that is depressed because it assumes that counsel will not be fully compensated would doubly reduce the attorney's-fee award. See Kenny A., 559 U.S. at 546, 130 S.Ct. 1662 (noting that the lodestar method includes the factors relevant to determining a “reasonable attorney's fee” and adjustments to the lodestar calculation may not be made “based on a factor that is subsumed in the lodestar calculation”); see also City of Burlington v. Dague, 505 U.S. 557, 560–67, 112 S.Ct. 2638, 120 L.Ed.2d 449 (1992) (rejecting as impermissible double-counting a lodestar enhancement that accounts for the risk the attorney will not collect a contingent fee).

We recognize that we are not in a position to evaluate whether the billing rate for workers'-compensation practice included in the proxy market rate in fact suffers from the flaw we described or the other defects adverted to by the BRB in Christensen. To the extent that it does not, and if it otherwise reflects market rates in the “relevant community,” the BRB, ALJs, and District Directors may be able to use it. But the BRB's published Christensen decision has precedential value, see Price v. Stevedoring Servs. of Am., 697 F.3d 820, 827 (9th Cir.2012) (“In practice as well as theory, it is the BRB's published decisions ... that are precedential and determine the rights of future parties.”), and it provides good reasons to doubt that the rate for workers'-compensation practice is a market rate for Longshore Act claimants' work. As far as we can tell, the BRB has not repudiated its Christensen decision, and it did not distinguish Christensen or justify a departure from Christensen in this case. We thus find that the BRB acted arbitrarily in allowing partial reliance on a rate reported by workers'-compensation lawyers that, according to the BRB's own decisions, is not a market rate for claimant Longshore Act representation. See Christensen, 557 F.3d at 1054–55; Camacho, 523 F.3d at 979; see also Andia v. Ashcroft, 359 F.3d 1181, 1184 (9th Cir.2004) (“The [Board of Immigration Appeals] acts ‘arbitrarily’ and ‘contrary to law’ if it fails to apply and follow its own prior decisions.”).

IV

For the reasons discussed, we hold that the BRB erred in affirming Judge Berlin's award of attorney's fees for work performed by Robinowitz because the proxy market rate on which the award depends does not adequately represent market rates in the “relevant community,” here, Portland, Oregon. We grant Shirrod's petition for review, vacate the BRB's decision and order, and remand this case for further proceedings.

PETITION FOR REVIEW GRANTED; VACATED and REMANDED.