LAY, Circuit Judge.

Qwest Corporation ("Qwest") appeals two district court decisions upholding orders of the Minnesota Public Utilities Commission ("MPUC" or "Commission") which set and applied retroactively the rates that telecommunication companies are required to pay when leasing Qwest's network infrastructure. We affirm the district court.

I.

The 1996 Telecommunications Act ("1996 Act") was intended to create competition between carriers in local telecommunication service markets that had been traditionally dominated by a single monopoly carrier. Incumbent local exchange carriers ("ILECs"), such as Qwest, own the network infrastructure necessary to provide local telephone service. The 1996 Act requires ILECs to lease their networks to competitive local exchange carriers ("CLECs") and outlines procedures for determining lease rates. 47 U.S.C. §§ 251, 252. Section 252 allows ILECs and CLECs to negotiate lease terms themselves in the form of interconnection agreements ("ICAs"). § 252(a). However, if the parties cannot successfully negotiate a rate, state public utilities commissions have the authority to arbitrate rates. § 252(b).

FCC rules require state public utilities commissions to establish rates based on a cost methodology called Total Element Long Run Incremental Cost ("TELRIC"). See Implementation of the Local Competition Provisions in the Telecommunications Act of 1996, First Report and Order, 11 F.C.C.R. 15,499 ¶¶ 674-90 (FCC Aug. 8, 1996) (" Local Competition Order"). Under TELRIC, unbundled network element ("UNE") rates are not set based on an ILEC's actual costs in building and maintaining its network. Rather, UNE rates are calculated according to what it would cost today to build and operate an efficient network that can provide the same services as the ILEC's existing network. In this way, UNE rates can provide accurate market signals to inform CLECs' decisions about whether to invest in their own facilities or lease the ILEC's facilities.

In the spring of 2002, MPUC reviewed UNE rates, some of which the Commission had previously set in a generic cost proceeding and were based on TELRIC. A group of CLECs contended that some of these rates no longer complied with TELRIC due to changes in market conditions and technology. MPUC concluded that UNE rates would likely change, and therefore issued an order on April 4, 2002, declaring that all Qwest UNE rates under review would be deemed interim and subject to true-up payments after MPUC set permanent rates ("April 4 Order"). MPUC adopted an Administrative Law Judge's ("ALJ") recommendation on October 2, 2002, and filed an order establishing permanent rates on March 24, 2003. As a result, Qwest owed the CLECs almost $13 million in true-up payments, which began accruing on April 4, 2002.

Qwest brought suit in district court, challenging both the establishment of interim rates in the April 4 Order and the permanent UNE rates set by MPUC. In two separate opinions, the district court denied both of Qwest's challenges and upheld the orders. Qwest now appeals.

Title 47 U.S.C. § 252(e)(6) provides for federal court review of state commission decisions. We review a state commission's interpretation of federal law de novo. Qwest Corp. v. Minnesota Pub. Util. Comm'n, 427 F.3d 1061, 1064 (8th Cir. 2005). However, in recognition of the state commission's superior technical expertise, we review its factual determinations under the arbitrary and capricious standard. Ace Tel. Ass'n v. Koppendrayer, 432 F.3d 876, 878 (8th Cir. 2005).

II. A. Retroactive Ratemaking

Qwest first argues that the April 4 Order violates the rule against retroactive ratemaking. The rule against retroactive ratemaking prohibits a commission from prescribing rates to recoup a utility's past losses for transactions that have already taken place. See, e.g., BP West Coast Prod., LLC v. FERC, 374 F.3d 1263, 1301 (D.C. Cir. 2004); Pac. Gas Elec. Co. v. FERC, 373 F.3d 1315, 1319-20 (D.C. Cir. 2004); Consol. Edison Co. of New York v. FERC, 347 F.3d 964, 969 (D.C. Cir. 2003). The purpose of the rule against retroactivity, and the closely related filed rate doctrine, is to ensure predictability. Pub. Util. Comm'n of California v. FERC, 988 F.2d 154, 163 (D.C. Cir. 1993). Therefore, the rule does not apply in situations where there is "adequate notice that resolution of some specific issue may cause a later adjustment to the rate being collected at the time of service." Natural Gas Clearinghouse v. FERC, 965 F.2d 1066, 1075 (D.C. Cir. 1992); see also OXY USA, Inc. v. FERC, 64 F.3d 679, 699 (D.C. Cir. 1995) ("The goals of equity and predictability are not undermined when the Commission warns all parties involved that a change in rates is only tentative and might be disallowed.").

The April 4 Order expressly put Qwest and all other interested parties on notice that the existing UNE rates were under review and subject to true-up payments when MPUC adopted permanent rates. Rates collected prior to April 4, 2002, are not subject to true-up payments. With these considerations in mind, the April 4 Order "change[d] what would be purely retroactive ratemaking into a functionally prospective process by placing the relevant audience on notice at the outset that the rates being promulgated [were] provisional only and subject to later revision." Natural Gas Clearinghouse, 965 F.2d at 1075 (citation omitted). Therefore, the April 4 Order does not violate the rule against retroactive ratemaking.

B. The Language and Policy of the 1996 Act and FCC Orders

The remainder of Qwest's arguments about the validity of the April 4 Order center around the language and policy of the 1996 Act. Qwest argues that Congress intended "parties to know the rates for network elements prior to the transactions to which they would apply" because under the 1996 Act, rates for network elements must be included in ICAs. See 47 U.S.C. § 252(a)(1), (c)(2). However, merely because state public utilities commissions must arbitrate and "establish any rates for interconnection, services, or network elements," § 252(c)(2), there is nothing in the 1996 Act to suggest that these rates may not be interim.

Qwest further argues that interim rates and true-up payments are inconsistent with the 1996 Act and FCC pricing rules, which were designed to facilitate a market-driven telecommunications industry with a preference for voluntarily negotiated ICAs and UNE rates that send the correct economic signals to CLECs. See Pac. Bell v. Pac-W. Telecomm, Inc., 325 F.3d 1114, 1127 (9th Cir. 2003); Iowa Utils. Bd. v. FCC, 120 F.3d 753, 801 (8th Cir. 1997), rev'd on other grounds, AT T Corp. v. Iowa Utils. Bd., 525 U.S. 366, 119 S.Ct. 721, 142 L.Ed.2d 835 (1999); In re Review of the Commission's Rules Regarding the Pricing of Unbundled Network Elements and the Resale of Service by Incumbent Local Exchange Carriers, Notice of Proposed Rulemaking, 18 F.C.C.R. 18,945, 20,265 ¶ 7 (FCC rel. Sept. 15, 2003) (" TELRIC NPRM"). However, none of the policy goals noted by Qwest changes the state public utilities commissions' responsibility to ensure that the rates charged by ILECs comply with the cost-based standards of the 1996 Act. See Verizon Communications, Inc. v. FCC, 535 U.S. 467, 476, 122 S.Ct. 1646, 152 L.Ed.2d 701 (2002); 47 U.S.C. § 252(d). Rather, the need for correct economic signals actually supports the use of interim rates and true-up payments. While some regulatory lag is inevitable and even helpful to ensure facilities-based competition, see Verizon, 535 U.S. at 504-07, 122 S.Ct. 1646, a "rate that is long out of date . . . frustrates the goals of TELRIC." AT T Communications of Illinois, Inc. v. Illinois Bell Tele. Co., 349 F.3d 402, 411 (7th Cir. 2003). A limited adjustment of UNE rates that are no longer TELRIC-compliant will not discourage CLECs from investing in their own facilities. See Verizon, 535 U.S. at 505-07, 122 S.Ct. 1646.

Moreover, interim rates and true-up payments will not discourage voluntary negotiations by CLECs. Parties may not seek arbitration until after negotiations have failed. § 252(a), (b). The possibility of true-up payments will not reduce the parties' incentive to initially negotiate. Rather, potential true-up payments that may favor either an ILEC or the CLECs provide a disincentive for any party to delay negotiations so that out of date rates may continue to be charged. Nothing in the language or policy of the 1996 Act suggests that interim rates subject to true-up are forbidden when a state public utility commission is uncertain whether the current rates are TELRIC-compliant. In fact, contrary to Qwest's assertions, the FCC Orders actually support the opposite conclusion.

The FCC approved the use of interim rates subject to true-up when state public utilities commissions were initially attempting to arbitrate TELRIC-compliant rates after the passage of the 1996 Act. Local Competition Order ¶ 693. The FCC has also authorized the use of interim rates subject to true-up in proceedings under 42 U.S.C. § 271. ILECs enter into ICAs and make their local network infrastructure available to CLECs so that the ILEC may enter the interLATA market. § 271(c). The FCC has approved several ILECs' entry into the interLATA market under § 271 even though the rates in the relevant ICAs were interim and subject to true-up. See In re Application by SBC Communications Inc., for Authorization to Provide In-Region, InterLATA Services in California, 17 F.C.C.R. 25,650 ¶ 37 (FCC rel. Dec. 19, 2002) (" California 271 Order"); In re Application by SBC Communications Inc., Pursuant to Section 271 of the Telecommunications Act of 1996 to Provide In-Region, InterLATA Services in Texas, 15 F.C.C.R. 18,354 ¶¶ 85-90 (FCC rel. June 30, 2000) (" Texas 271 Order"); In re Application by Bell Atlantic New York for Authorization Under Section 271 of the Communications Act to Provide In-Region, InterLATA Service in the State of New York, 15 F.C.C.R. 3953 ¶¶ 256-60 (FCC rel. Dec. 22, 1999) (" New York 271 Order").

Interim rates are also used in orders by the FCC's Wireline Competition Bureau. If states refuse to arbitrate UNE rates, the FCC will preempt the state commission's jurisdiction and assume responsibility for the matter. See 47 U.S.C. § 252(e)(5). The FCC has delegated this authority to the Wireline Competition Bureau. See In re Procedures for Arbitrations Conducted Pursuant to Section 252(e)(5) of the Communications Act of 1934, As Amended, 16 F.C.C.R. 6231 ¶ 8 (FCC rel. Jan. 19, 2001). When the Wireline Competition Bureau arbitrates UNE rates for ICAs, the FCC requires the Bureau to set the rates as interim subject to true-up in the event the FCC ultimately modifies any of the rates set by the Bureau. Id. ¶ 10. In 2003, the Wireline Competition Bureau arbitrated UNE rates in Virginia. In re Petition of WorldCom, Inc. Pursuant to Section 252(e)(5) of the Communications Act for Preemption of the Jurisdiction of the Virginia State Corporation Commission Regarding Interconnection Disputes with Verizon Virginia Inc., 18 F.C.C.R. 17,722 (FCC rel. Aug. 29, 2003) (" Virginia Arbitration Order"). The Wireline Competition Bureau set UNE rates for Virginia and, in accordance with FCC requirements, ordered that any rate established would be subject to a retroactive true-up if the FCC established different rates upon review. Id. ¶ 26.

Qwest contends that these FCC orders do not support the interim rates set by MPUC in this case. First, Qwest argues that the FCC's approval of interim rates is limited to cases where permanent TELRIC-compliant rates were not already established, or where there was misconduct by the ILEC. Second, Qwest attempts to distinguish the Virginia Arbitration Order by arguing that arbitration by the Wireline Competition Bureau is a unique circumstance because the Bureau's orders are always subject to review by the FCC at a party's request.

We find Qwest's arguments unpersuasive. First, merely because the FCC has allowed (or in the case of the Wireline Competition Bureau, required) interim rates subject to true-up in instances different than the one presented here does not mean the use of interim rates is limited to those circumstances. Rather, it suggests the FCC recognizes that limited use of interim rates can be valuable when it is uncertain whether the current rates are TELRIC-compliant. See California 271 Order ¶¶ 33, 37; Texas 271 Order ¶¶ 85-90; New York 271 Order ¶¶ 258-60; see also In re Application of Verizon New England Inc. for Authorization to Provide In-Region InterLATA Services in Massachusetts, 16 F.C.C.R. 8988 ¶ 34 (FCC rel. Apr. 16, 2001) (stating that interim rates subject to true-up are appropriate "in certain circumstances"). Second, the FCC approved TELRIC-based interim rates in the Texas 271 Order, suggesting that the retroactive adjustment of TELRIC-compliant rates is not disfavored by the FCC. Texas 271 Order ¶ 89. Third, while ILEC misconduct was a contributing factor in the use of interim rates in the California 271 Order, California 271 Order ¶¶ 23, 37 n. 85, ILEC misconduct is certainly not the sole instance where interim rates have been approved by the FCC. Rather, the FCC has approved interim rates in § 271 Orders "so long as an interim solution to a particular rate dispute is reasonable under the circumstances, the state commission has demonstrated its commitment to [FCC] pricing rules, and provision is made for refunds or true-ups once permanent rates are set." Texas 271 Order ¶ 88, quoted in California 271 Order ¶ 37; see also New York 271 Order ¶¶ 258-59.

Qwest argues that the interim rates established by MPUC create too much uncertainty. However, whether rates are interim as arbitrated by the Wireline Competition Bureau or set by a state public utilities commission for the purposes of § 271 proceedings, there is always a significant amount of time where the rates are subject to a refund or true-up in the future. The FCC relies upon a case-by-case approach, where interim rates may be appropriate if they are "reasonable under the circumstances." California 271 Order ¶ 37; Texas Order ¶¶ 85, 88; New York Order ¶ 258. Under the circumstances, we consider the limited uncertainty created by the April 4 Order to be reasonable.

Finally, Qwest argues that the April 4 Order violates the 1996 Act by retroactively changing the terms of binding ICAs. See 47 U.S.C. § 252(a)(1). Qwest bases this argument on the Ninth Circuit's decision in Pacific Bell v. Pac-West Telecomm, Inc., 325 F.3d 1114 (9th Cir. 2003). In Pacific Bell, the California Public Utility Commission ("CPUC") issued a generic order that designated all calls to internet service providers as local traffic, thereby subject to reciprocal compensation provisions in the parties' existing ICAs. Id. at 1120-21. The Ninth Circuit first established that CPUC lacked authority under the 1996 Act to promulgate generic regulations over internet service provider traffic because the FCC had deemed this traffic to be "interstate." Id. at 1125. CPUC's role over interstate traffic is limited to its authority under § 252 to approve new arbitrated ICAs and to interpret existing agreements. Id. at 1125-27. Second, the court stated that the generic order violated § 252 because it effectively changed the terms of all applicable ICAs in California. Id. at 1127. Finally, the Ninth Circuit concluded that CPUC could not be exercising its power to interpret ICAs under § 252 because CPUC failed to consider any specific ICA or reciprocal compensation provision. Id. at 1128.

Unlike CPUC in Pacific Bell, there is no question in the present case that MPUC has authority to arbitrate UNE rates under the 1996 Act. § 252(c), (d), (g). Moreover, the April 4 Order does not purport to interpret the rates or terms of existing ICAs. Rather, MPUC exercised its authority to establish new UNE rates in a generic proceeding. § 252(g). The April 4 Order was not based upon an interpretation of the existing ICAs. Therefore, there was no need for MPUC to review the terms of specific agreements before issuing the April 4 Order.

III.

We must next consider whether MPUC was arbitrary and capricious in setting select permanent UNE rates. Qwest challenges four MPUC actions: (1) the reduction of switching costs; (2) the rates set for high capacity loops; (3) the plant mix estimates; and (4) the adoption of the HAI Model to determine general support assets. After considering the record, we hold that MPUC was not arbitrary and capricious.

A. Reduction of Switching Costs

Switches route telephone calls to their destinations. Switches can either be modern, computerized digital loop carrier ("DLC") systems, or older and more expensive analog switches. To comply with TELRIC, MPUC had to establish switch rates that included the percentage of digital switches that a hypothetical efficient network in Minnesota would utilize. MPUC adopted the HAI Model, proposed by AT T/MCI, which uses the switching investment calculated by the FCC Synthesis Model. See In re Forward-Looking Mechanism for High Cost Support, Tenth Report and Order, 14 F.C.C.R. 20,156 (FCC rel. Nov. 2, 1999) (" Inputs Order"). The HAI Model makes a reduction to the calculated switching investment to reflect the savings that would result from an increased use of DLC systems. The HAI Model assumes that the FCC Synthesis Model accounts for only 18.3% DLC and that in Minnesota DLC penetration in a hypothetical efficient network would actually be 57.5%. Therefore, the MPUC adopted a DLC adjustment to account for the higher DLC penetration and resulting savings.

Qwest argues that the DLC adjustment is arbitrary and capricious because there is no evidence that the FCC Synthesis Model assumed only 18.3% DLC, or that the FCC Synthesis Model did not already reflect the savings associated with digital switches. The FCC refused to adopt a DLC adjustment for this very reason in the Inputs Order. See id. ¶ 327. However, the FCC cost model was developed to determine federal universal support and the FCC explicitly cautioned parties against using the federal input values to make claims regarding UNE rate determinations in other proceedings. Id. ¶ 32. Therefore, MPUC was not arbitrary and capricious merely because it reached a different conclusion than the FCC.

AT T's expert testified that the FCC inputs assumed approximately 18.3% of the switches were digital. Moreover, AT T's expert and an expert for the Minnesota Department of Commerce ("Department of Commerce" or "Department") testified that the FCC inputs from 1999 did not adequately reflect the savings that a forward-looking network would achieve with DLC switches. Qwest provided no evidence to the contrary. Therefore, the ALJ rejected Qwest's argument that the DLC adjustment should be set to zero, as in the FCC Synthesis Model, because this would overstate the switching investment for a forward-looking network. The 1996 Act requires state public utilities commissions to make their decisions based upon the best evidence available, and does not mandate that they independently acquire data to build an evidentiary record. See § 252(b)(4); GTE S., Inc. v. Morrison, 199 F.3d 733, 748 (4th Cir. 1999). Even assuming the HAI Model contains flaws, Qwest provided the ALJ with no satisfactory alternative. As a result, MPUC was not arbitrary and capricious in making the DLC adjustment based upon the evidence presented.

B. Rates for High Capacity Loops

Qwest next argues that MPUC improperly adopted the HAI Adjunct Model proposed by AT T and MCI to calculate the rates for high capacity loops. High capacity loops are larger wires that connect high-volume customers to a telephone company wire center. The FCC requires TELRIC studies to "explain with specificity" how "the associated costs are developed." Local Competition Order ¶ 691. When offered in a spreadsheet, a model should reveal the underlying formulae and should include data or inputs that are easily verifiable. TELRIC NPRM ¶ 41. Qwest objects to the Adjunct Model because it provided only "aggregate per line investment figures," rather than specifying the individual cost of the particular components that made up that input.

While the ALJ acknowledged the model would have been stronger with more price documentation, the Adjunct Model uses per-line costs based on proprietary information from Qwest's competitors and the vendors that supply the relevant equipment. As a result, the data concerning costs for each particular company was not disclosed. The data collected from the individual competitors and vendors was used to compute an average per line investment for the relevant equipment and operations. The inputs were verified by expert witnesses who sponsored the model, and were independently verified by the Department of Commerce's technical expert.

Qwest asserts that there was insufficient documentation to support the conclusion of the experts; however Qwest failed to provide the ALJ with any evidence that the prices used in the Adjunct Model were inadequate. Moreover, the ALJ and MPUC rejected Qwest's proposed model and proffered alternative per-line costs because they relied on the structure of Qwest's existing network, see 47 C.F.R. § 51.505(d)(1); Local Competition Order ¶ 705, and overstated loop costs. MPUC is required by the 1996 Act to resolve the arbitrated issues in a timely manner, using the best information available. See § 252(b)(4). As a result, MPUC must choose the model it feels most accurately reflects an efficient, forward-looking network. See GTE S., Inc., 199 F.3d at 748. In light of the evidence presented to the ALJ, we cannot say MPUC was arbitrary and capricious in choosing the Adjunct Model to calculate high capacity loop rates.

C. Plant Mix Determination

Qwest also challenges MPUC's plant mix determinations. "Plant mix" refers to the ratios of different categories of telephone lines. Aerial lines are strung between telephone poles, while buried and underground lines are located below ground. Buried lines are placed in a trench and covered with earth. Underground lines are placed in a buried conduit that is accessible by manholes. The ALJ and MPUC adopted plant mix ratios recommended by the Department of Commerce, rejecting ratios suggested by Qwest and the CLECs in competing models. The ALJ concluded that the CLEC model included too high a percentage of aerial plant, given that communities were moving away from aerial plant due to aesthetic concerns, while the Qwest model overstated the amount of underground plant. Therefore, the ALJ adopted the Department's recommendation that increased buried plant and reduced investment in underground and aerial plant. Qwest is challenging the reduction in underground lines, arguing that MPUC improperly relied upon the unsupported testimony of the Department's expert.

In a prior proceeding, MPUC had adopted the Department's conclusion that 14.5% of plant was underground. Based upon the testimony of the Department's expert, the ALJ concluded that this percentage was no longer accurate and established a new weighted average of 5.8% underground plant. The Department's expert reached this conclusion in part by relying upon a survey conducted for a Department of Commerce study for a different proceeding which reflected that telecommunications companies were installing very little aerial or underground plant. Qwest argues that MPUC improperly relied upon the unsupported testimony of the Department's expert because this survey was not included in the record and the expert failed to explain his methodology.

However, the survey was only part of the expert's rationale. The Department's expert explained to the ALJ that the most accurate forward-looking estimates of plant mix have changed since he made his initial calculation of 14.5% underground plant. He explained that this prior estimate improperly relied upon carriers' actual use of plant in embedded networks and that there would be a significant difference between these numbers and what a telecommunications company would chose when "starting from scratch." As a result, he used his prior estimates as a benchmark, which he adjusted to account for what an efficient carrier deploying a forward-looking network would do today. With that in mind, the Department's expert explained that buried plant would be the preferred placement method in a forward-looking network because it is less expensive to place than underground plant and easier to maintain and more accepted by communities than aerial plant.

Qwest again objects to the ALJ's rejection of its proposal, which was based upon Qwest's embedded numbers. Qwest's embedded numbers are not dispositive. See 47 C.F.R. § 51.505(d)(1). While Qwest may actually employ a much higher percentage of underground plant than the adopted proposal, MPUC could conclude that a forward-looking network would employ a smaller percentage of underground plant. As discussed above, MPUC was engaged in time-sensitive decision making. After MPUC determined that neither the Qwest nor the CLEC models was appropriate to determine plant mix, it was not arbitrary and capricious in choosing to adopt a middle ground proposed by the Department of Commerce. See GTE S., Inc., 199 F.3d at 748.

D. Determining General Support Assets

Finally, Qwest objects to MPUC's adoption of the HAI model to calculate general support asset ("GSA") expenses. GSA expenses are the costs for the necessary items that keep a network running and make telephone service possible, without actually being part of the network. These include the costs of computers, desks, buildings, vehicles, tools, and other non-network assets. Under FCC rules, these costs are treated as common and are recovered through a percentage markup on UNEs. See Local Competition Order ¶¶ 676, 682; see also 47 C.F.R. § 51.505(a). GSA expenses include the support costs a telephone company assumes when delivering service to both retail and wholesale customers. However, state public utilities commissions may not set UNE rates that include retail-only GSA costs because CLECs do not utilize an ILEC's retail support services to provide telephone service to their own customers. See 47 C.F.R. § 51.505(c)(2)(ii), (d)(2). Therefore, MPUC was required under the FCC's pricing rules to calculate the GSA expenses that an efficient forward-looking carrier would incur in the wholesale provision of UNEs. See id.

MPUC based GSA costs on the HAI Model proposed by AT T/MCI. Under the HAI Model, GSA expenses are adjusted through two allocation factors that remove retail-only costs. Qwest argues that the HAI Model's calculation of GSA expenses already allocates to Qwest's retail operations an appropriate share of costs, and therefore the further adjustment under the allocators was unnecessary. Qwest also claims there is a lack of evidence explaining the methodology used to remove retail costs under the allocation factors.

While Qwest objects to the evidentiary support given to the HAI Model and the allocation factors, at its heart, Qwest's argument focuses on its disagreement with the methodology chosen to remove retail costs. Qwest argues that the preadjusted GSA costs allocated an appropriate share of such costs to Qwest's retail operations, and that further reduction under the allocators would result in Qwest's retail operations bearing a disproportionate share of GSA costs. Given that Qwest's retail operations should not merely be bearing a share of the retail GSA costs, but rather all of Qwest's retail costs, see 47 C.F.R. § 51.505(c)(2)(ii), (d)(2), we cannot see how MPUC erred in relying upon expert testimony that the allocators were necessary to remove retail GSA costs.

Qwest argues that there was insufficient explanation of the allocator methodology. However, Qwest's expert challenged the allocators in a sophisticated manner, suggesting that Qwest understood the HAI Model's methodology, and merely disagreed with it. While the ALJ and MPUC could have provided more detail in explaining their decision to adopt the HAI Model, they were not arbitrary and capricious merely because they chose the explanation and methodology of one expert over another.

Qwest argues that the Wireline Competition Bureau rejected a similar secondary adjustment in the Virginia arbitration. See Virginia Arbitration Order ¶ 151. However, the Virginia arbitration adopted a different cost model than the HAI Model adopted by MPUC. The Bureau rejected the inclusion of a secondary adjustment that removed costs associated with special access and toll in the universal service support context because there was nothing to indicate that this adjustment correlated to a reduction in retail expenses in the GSA context. Id. Relevant to the HAI Model, however, the ALJ noted that the model was previously adopted by MPUC in another proceeding and that Qwest presented its same objections to two other state public utilities commissions, both of which accepted the HAI Model as properly excluding retail GSA expenses.

Therefore, we conclude that MPUC was not arbitrary and capricious in its adoption of the cost models relevant to the present case.

IV.

For the foregoing reasons, we affirm the decision of the district court.