OPINION

DUNCAN, Circuit Judge:

Pursuant to the provisions of the Telecommunications Act of 1996, 47 U.S.C. § 151 et seq., Sprint Communications Company of Virginia, Inc., and Sprint Communications Company L.P. (collectively "Sprint" or the "Sprint Defendants") entered into interconnection agreements with nineteen incumbent local exchange carriers (collectively "CenturyLink" or the "CenturyLink Plaintiffs") providing for the mutual exchange of telecommunications traffic. When Sprint began to withhold payments under the agreement, CenturyLink brought a breach of contract claim in federal district court. After rejecting Sprint's threshold argument that its jurisdiction was limited to reviewing determinations by State utilities commissions, the district court entered judgment in favor of CenturyLink on the merits. The district court judge subsequently also concluded that a belatedly discovered financial interest in CenturyLink held in a managed Individual Retirement Account did not require his recusal. Sprint appeals all of the district court's rulings; for the reasons that follow, we affirm.

I.

A.

Prior to the Telecommunications Act of 1996 (the "1996 Act"), telephone service within a local calling area was provided by an incumbent local exchange carrier ("ILEC") operating as a state-licensed monopoly. The purpose of the 1996 Act was to create competition within these local telephone markets.

At the core of the 1996 Act is a requirement that ILECs "interconnect" their facilities and equipment with competitive local exchange carriers ("CLECs"), such as Sprint, for the mutual exchange of traffic. Defined as "the linking of two networks for the mutual exchange of traffic," 47 C.F.R. § 51.5, interconnection allows CLEC customers to call ILEC customers and vice versa. While the carriers may reach agreement through arbitration or negotiation, the product of the process is an interconnection agreement (an "ICA"), which must include "a detailed schedule of itemized charges for interconnection and each service or network element included in the agreement." 47 U.S.C. § 252(a)(1).

B.

A brief description of the corporate relationship between Sprint and CenturyLink, and the latter's organizational parentage, provides necessary background for our consideration of the recusal issue. We then set out the underlying facts and procedural history of the appeal before us.

When Sprint sought negotiation of the ICAs at issue in April 2004, it and the nineteen companies that comprise the CenturyLink Plaintiffs were wholly owned subsidiaries of Sprint Corporation.1 The CenturyLink Plaintiffs were a part of Sprint Corporation's local telephone division. In May 2006, Sprint Corporation spun off the CenturyLink Plaintiffs, which then formed a separate company known as Embarq Corporation ("Embarq"). In July 2009, CenturyTel, Inc. ("CenturyTel") acquired Embarq and its subsidiaries. The resulting entity began doing business as "CenturyLink." Between 2004 and 2005 Sprint and CenturyLink executed the nineteen ICAs that are the subject of this dispute, and which were approved by the appropriate State commissions.2

Some general information about telecommunications traffic provides context for our discussion of the issues in this case. There are three ways to place a call: landline, wireless, and Voice-over Internet Protocol ("VoIP"), which, as its name suggests, relies on the internet to originate voice communications. See Vonage Holdings Corp. v. Nebraska Pub. Serv. Comm'n, 564 F.3d 900, 902 (8th Cir.2009) ("VoIP is an internet application used to transmit voice communication over a broadband internet connection."). Likewise, a call placed through any of these three formats can be classified into three categories of traffic, depending on the locational relationship of those speaking: local,3 long distance intrastate,4 and long distance interstate. The facts giving rise to the underlying dispute revolve around the ICA's compensation structure for these three categories of traffic.

In assessing the appropriate compensation for a local, long distance intrastate, and long distance interstate call, the relevant metrics are where a call originated and where it terminated.5 This determination is often referred to in the telecommunications industry as the "jurisdictionalizing" of a call. All the calls at issue in this case originated on Sprint's network and were terminated by one of the CenturyLink Plaintiffs on its local network.

The ICA first addresses local traffic. In salient part, the ICA defines local traffic as traffic "that is originated and terminated within Sprint's local calling area." J.A. 718. The ICA applies to local traffic a practice known as "Bill & Keep" or "reciprocal compensation." As § 38.1 of the ICA explains, "[u]nder Bill and Keep, each Party retains the revenues it receives from end user customers, and neither Party pays the other Party for terminating Local Traffic which is subject to the Bill and Keep compensation mechanism." J.A. 744. In brief, no payments exchange hands for the termination of local traffic.

By contrast, the ICA provides for "access charges"6 for the two categories of long distance traffic discussed above: intrastate and interstate. Under the ICA, the applicable access charge depends on the category of long distance traffic.

All three categories of traffic — local, long distance intrastate, and long distance interstate — travel across "trunks." See 47 C.F.R. § 69.2(x) (defining "trunk" as including "transmission media such as radio, satellite, wire, cable and fiber optic cable means of transmission"). The parties stipulated below that Sprint delivers some traffic to CenturyLink via local interconnection trunks, and other traffic by way of Feature Group D ("FGD") trunks. J.A. 315. FGD trunks carry long distance traffic only. Id. at 316. FGD trunks attach to local interconnection trunks, and thus enable long distance traffic to be terminated on one of CenturyLink's local networks.

The dispute in this case only involves VoIP traffic, which travels over FGD trunks (for a long distance call) and over local interconnection trunks.7 Section 38.4 of the ICA addresses the compensation system for the termination of VoIP traffic: "Voice calls that are transmitted, in whole or in part, via the public Internet or a private IP network (VoIP) shall be compensated in the same manner as voice traffic (e.g. reciprocal compensation, interstate access and intrastate access)." J.A. 746.

Sprint paid CenturyLink access charges for VoIP traffic as set out in § 38.4 from the time of the execution of the ICAs between the parties in 2004 and 2005 until June 2009. At that point, Sprint began filing written disputes with CenturyLink. Although the nature of Sprint's objections changed, its core contention appeared to be twofold: (1) that the ICA did not apply to long distance VoIP traffic that traveled over FGD trunks; and (2) that CenturyLink had billed Sprint at an improperly high rate for VoIP traffic since May 1, 2007.

Instead of following the Dispute Resolution Procedure established in the ICA, Sprint unilaterally reduced the rate for termination of VoIP-originated traffic. It demanded that CenturyLink apply Sprint's recalculated rate going forward and remit portions of previous payments made by Sprint which Sprint deemed to be in excess of what it should have paid. Sprint withheld payments for both VoIP and non-VoIP traffic, although no dispute concerning the latter existed.

C.

In November 2009, CenturyLink filed a complaint in the United States District Court for the Eastern District of Virginia alleging one count of breach of contract based on the facts recounted above. Sprint moved to dismiss CenturyLink's complaint for lack of jurisdiction due to failure to exhaust administrative remedies, or, alternatively, to stay the case under the doctrine of primary jurisdiction. Sprint also filed a counterclaim alleging, inter alia, that CenturyLink breached the North Carolina ICA (the "NC ICA") by billing Sprint for local traffic not subject to access charges.

The district court denied Sprint's motion to dismiss in a lengthy opinion. It concluded that it had federal question jurisdiction under Supreme Court and circuit precedent interpreting the 1996 Act. It then decided that the 1996 Act imposed no requirement for CenturyLink to exhaust its remedies before a State commission. Finally, it declined to stay the case under the doctrine of primary jurisdiction, which allows a court to refer a case within its jurisdiction to an administrative agency.

The district court conducted a bench trial on CenturyLink's breach of contract claim during August and September 2010. It subsequently entered judgment in favor of CenturyLink, and issued a memorandum opinion setting out findings of fact and conclusions of law. The district court concluded that "in refusing to pay the access charges as billed, Sprint breached its duties under the ICAs, which clearly included paying access charges for VoIP-originated traffic according to the jurisdictional endpoints of the calls." Cent. Tel. Co. of Va. v. Sprint Commc'ns Co. of Va., Inc., 759 F.Supp.2d 789, 792 (E.D.Va.2011). Central to the district court's decision was its finding that § 38.4 of the ICA operated to apply the compensation regime for local, long distance intrastate, and long distance interstate traffic to VoIP calls.8 Although the district court did not find the relevant portions of the ICA to be ambiguous, it noted that if ambiguity existed, it would be construed against Sprint as the drafter of the ICA.9 The district court then awarded CenturyLink $23,376,213.76, which consisted of damages under the breach of contract claim and prejudgment interest.

The second phase of the litigation, the bench trial on Sprint's counterclaim, took place over two days in December 2010. Unlike CenturyLink's breach of contract claim, which focused on long distance VoIP traffic, Sprint's counterclaim alleged that CenturyLink had improperly charged it for local VoIP calls.10 Recognizing that the ICA's "Bill and Keep" provision did not provide for access charges for local traffic, the district court rejected Sprint's argument that the ICA did not permit CenturyLink's use of the Billing Telephone Number ("BTN") method for determining whether a call should be deemed local. Under the BTN method, a carrier classifies a call as local or nonlocal for billing purposes based on a unique account number that is assigned to a specific facility. Sprint contended that CenturyLink was required to use another method for "jurisdictionalizing" a call known as the "Calling Party Number" or "CPN." CenturyLink used BTN throughout the period in dispute, in part because at the time the parties executed the NC ICA, CenturyLink did not have the software in place to use the CPN method. See J.A. 490-91.

The district court found that the NC ICA did not specify a method for identifying local calls, but instead incorporated by reference a telecommunications industry publication that explicitly permitted use of the BTN method. Alternatively, the district court again construed any ambiguity in the NC ICA against Sprint as the drafter.

D.

By May 10, 2011, the district court judge had presided over both bench trials, had ruled on Sprint's jurisdiction motion, had entered judgment for CenturyLink on its breach of contract claim, including the issuance of an accompanying 50-page memorandum opinion, and had begun drafting the opinion on Sprint's counterclaim. On that date, as the district judge was preparing his financial disclosure for financial year 2010, he discovered that he had owned eighty shares of CenturyLink in a managed Individual Retirement Account ("IRA") during the time he presided over this case.

Fund managers made decisions to buy and sell stocks in the IRA, which at any given time held shares in as many as 200 companies, without input from the judge. The fund managers initially purchased shares in Embarq, which subsequently converted into forty-seven shares in CenturyTel following its acquisition of Embarq. The fund managers' subsequent purchases of CenturyTel stock in October 2009 and March 2010 brought the district court judge's total holdings up to eighty shares, which represented between 0.52% and 0.81% of the value of the IRA. In July 2010, the district court judge had filed his 2009 financial disclosure form, which reflected that the IRA held eighty shares in CenturyTel.11 Because CenturyTel was not a party to this case, the computerized system used to identify potential conflicts did not register any conflict. J.A. 139-40.

As soon as the district court judge discovered that his IRA owned shares in CenturyLink, he initiated a conference call to inform the parties. He explained the situation, and stated that he was "not qualified under the canons of ethics to sit as a judge in this case." J.A. 129. He then noted: "I don't know that there's any alternative but for me to vacate the orders that I've entered in the case and recus[e] myself and hav[e] the case reassigned to some other judge to make a decision." J.A. 130. Counsel for CenturyLink requested time to research the recusal and vacatur issues before the judge took any action. Hearing no objection from counsel for Sprint, the district court judge gave CenturyLink one week to conduct research and, if desired, file a brief.

On May 16, 2011 CenturyLink moved for the district court judge to divest from CenturyLink, or, in the alternative, to recuse without vacating any of the previous orders and opinions. In the interval between the conference call and the filing of CenturyLink's motion, the district court judge conducted additional research, and came to the conclusion that his earlier statements rested on a misinterpretation of the ethical rules. He therefore had the parties agree to a briefing schedule to address the recusal issue. He also directed the IRA fund manager to sell the shares of CenturyLink immediately, which was done on May 17, 2011.

After considering briefing from the parties, the district court judge concluded that (1) his statements during the conference call did not constitute recusal; (2) allowing briefing of the recusal issue as a response to CenturyLink's request was permissible; and (3) neither relevant subsection of the statutory provision governing recusal, 28 U.S.C. § 455, required recusal or vacatur. He issued a written opinion on the recusal issue on December 12, 2011, and, on the following day, a separate opinion in which he ruled in favor of CenturyLink on the merits of Sprint's counterclaim.

Sprint's appeal followed.

II.

This appeal raises questions of federal and state jurisdiction under the 1996 Act, and also implicates a decision of the Federal Communications Commission (the "FCC"). We therefore requested the FCC to submit an amicus brief.12 We incorporate the FCC's views in our discussion as appropriate.

Sprint advances two reasons why the district court should not have reached the merits of this case. It first argues that the district court had no authority under the 1996 Act to interpret and enforce an ICA because its role is limited to reviewing a State commission determination, and no such determination occurred here. It further contends that the district court judge should have recused himself and vacated all orders and judgments issued in the case. Before considering each argument, we briefly describe pertinent parts of the 1996 Act.

A.

Section 252 of the 1996 Act requires carriers to submit their proposed ICA to the applicable State commission, which "shall approve or reject the agreement, with written findings as to any deficiencies." 47 U.S.C. § 252(e)(1) If a State commission fails to approve or reject a proposed ICA within ninety days of an ICA adopted by negotiation or within thirty days of an ICA adopted by arbitration, the ICA is deemed approved. § 252(e)(4). By contrast, if the State commission fails to "carry out its responsibility under [section 252] in any proceeding or other matter under [section 252]," the FCC must preempt the State commission and assume whatever responsibility it otherwise possessed. § 252(e)(5). The 1996 Act does not define what constitutes a "responsibility" for purposes of § 252(e)(5).

Section 252(e)(6) provides for review of State commission actions in two distinct ways. First, when the FCC preempts a state commission under section 252(e)(5), section 252(e)(6) makes the FCC proceeding and any judicial review of that proceeding the "exclusive remedies for a State commission's failure to act." Second, section 252(e)(6) provides that "[i]n any case in which a State commission makes a determination under this section, any party aggrieved by such determination may bring an action in an appropriate Federal district court to determine whether the agreement or statement meets the requirements of section 251 of this title and this section."

B.

Sprint urges us to read the 1996 Act as permitting only State commissions to interpret and enforce ICAs in the first instance. In Sprint's view, because no State commission considered the ICA dispute before the district court did, the district court lacked authority to do so. Sprint first argues the 1996 Act contains a statutory exhaustion requirement. Even if the text and structure of the 1996 Act do not mandate initial State commission consideration, Sprint contends that we should impose such a requirement as a prudential matter. Both arguments raise legal questions, which we review de novo. See Cavalier Tel., LLC. v. Va. Elec. & Power Co., 303 F.3d 316, 322 (4th Cir.2002). We begin with the former.

1.

In considering whether the 1996 Act bestows on State commissions the exclusive responsibility to interpret and enforce ICAs in the first instance, it is useful to begin by identifying what is not at issue in this case. Following the Supreme Court's decision in Verizon Maryland, Inc. v. Public Service Commission of Maryland, 535 U.S. 635, 122 S.Ct. 1753, 152 L.Ed.2d 871 (2002), and our decision in Verizon Maryland, Inc. v. Global NAPS, Inc., 377 F.3d 355 (4th Cir.2004), it is clear that a federal court has jurisdiction to interpret the terms of interconnection-related compensation provisions in an ICA under 28 U.S.C. § 1331. Sprint does not argue otherwise. Instead, Sprint contends that although the 1996 Act confers federal question jurisdiction over the dispute in this case, it restricts federal courts to reviewing initial determinations made by State commissions.

Sprint's argument assumes that a State commission has the authority to interpret and enforce an ICA. Interestingly, nothing in the 1996 Act's text so provides. Nonetheless, every circuit to have considered the question has concluded that a State commission does have such authority. See Core Commc'ns, Inc. v. Verizon Pa., Inc., 493 F.3d 333, 342-44 (3d Cir. 2007); Sw. Bell Tel., L.P. v. Pub. Util. Comm'n, 467 F.3d 418, 422 (5th Cir.2006); E.SPIRE Commc'ns, Inc. v. N.M. Pub. Regulation Comm'n, 392 F.3d 1204, 1207 (10th Cir.2004); Iowa Network Servs., Inc. v. Qwest Corp., 363 F.3d 683, 691-92 (8th Cir.2004); BellSouth Telecomms., Inc. v. MCImetro Access Transmission Servs., Inc., 317 F.3d 1270, 1277 (11th Cir.2003) (en banc); MCI Telecomms. Corp. v. Ill. Bell Tel. Co., 222 F.3d 323, 337-38 (7th Cir.2000); Sw. Bell Tel. Co. v. Brooks Fiber Commc'ns of Ok., Inc., 235 F.3d 493, 497 (10th Cir.2000).13 The FCC has as well. In re Starpower Commc'ns, LLC, 15 F.C.C.R. 11277, 11280 (F.C.C.2000). Although we have yet to address this issue,14 we need not do so here because we reject Sprint's more particularized claim that a State commission must interpret an ICA before a federal district court can do so.

We begin with the text of the 1996 Act. Just as there is no statutory text granting State commissions the authority to interpret and enforce the provisions of ICAs, there is similarly none granting them the exclusive authority to do so in the first instance. See Core Comm'cns, 493 F.3d at 340 ("[T]he [1996] Act is simply silent as to the procedure for post-formation disputes."). Congress could have made review by a State commission in the first instance an exclusive remedy; it used this very phrase when discussing the review of an FCC preemption action. See 47 U.S.C. § 252(e)(6) ("In a case in which a State fails to act as described in [§ 252(e)(5) ], the proceeding by the [FCC] under such paragraph and any judicial review of the [FCC]'s actions shall be the exclusive remedies for a State commission's failure to act." (emphasis added)). But Congress did not do so.

Notwithstanding the 1996 Act's textual silence on this point, Sprint argues that language in § 252(e)(6) requires a State commission to make a "determination" before a federal district court can act. Specifically, Sprint points to the following: "In any case in which a State commission makes a determination under this section, any party aggrieved by such determination may bring an action in an appropriate Federal district court...." § 252(e)(6). We disagree. The plain import of this language is to provide for federal court review of a State commission's decision to approve or reject a proposed ICA.15 This conclusion is reinforced by the fact that the 1996 Act precludes such review by state courts. See § 252(e)(4) ("No State court shall have jurisdiction to review the action of a State commission in approving or rejecting [an ICA]....").

Nonetheless, Sprint seeks to rely on cases that interpret "determination" in § 252(e)(6) to include a State commission's post-formation interpretation or enforcement of an ICA. See e.g., BellSouth Telecomms., 317 F.3d at 1277; Sw. Bell Tel. Co. at 497 ("We next conclude that the district court had jurisdiction to review the decision of the [Oklahoma Corporation Commission] interpreting the Interconnection Agreement."). Sprint's reliance is misplaced. In recognizing that a federal court's authority to review State commission actions includes the authority to review a State commission's interpretation or enforcement of an ICA provision, these cases neither explicitly nor implicitly hold that parties must bring a dispute concerning an ICA to a State commission in the first instance. Rather, they simply acknowledge that where parties make a State commission the first stop for interpreting or enforcing a provision in an ICA, the language of § 252(e)(6) provides that it need not be the last stop.16

What Sprint terms its "structural" argument — that the structure of the 1996 Act gives State commissions "plenary authority" over the interpretation of an ICA in the first instance, and limits federal courts to a reviewing role — merely repackages the same textual argument. Just as Sprint cannot ground its textual argument in any text, neither can it point to any structural features of the 1996 Act indicating that Congress intended to mandate initial State commission consideration. Nor does Sprint's vague appeal to "cooperative federalism" advance its argument. As we have observed, adhering to the 1996 Act's policy of cooperative federalism requires close attention to the role that Congress (and the FCC) has actually assigned to State commissions. See BellSouth Telecomms., Inc. v. Sanford, 494 F.3d 439, 449 (4th Cir.2007) ("States' continuing exercise of authority over telecommunications issues forms part of a deliberately constructed model of cooperative federalism, under which the States, subject to the boundaries set by Congress and federal regulators, are called upon to apply their expertise and judgment and have the freedom to do so." (emphasis added)). Here, neither the plain text of the 1996 Act nor its structure evince congressional intent to place exclusive authority to interpret an ICA in the first instance in a State commission.

Sprint further contends that even if the text and structure of the 1996 Act do not require initial State commission consideration, deference to the FCC's Starpower decision should lead us to construe the statute as imposing such a requirement.17 In Starpower, the FCC considered whether to preempt the Virginia State Corporation Commission ("Virginia Commission") when the Virginia Commission took no action on a disputed issue arising from an ICA. The FCC concluded first that the Virginia Commission had the authority to interpret and enforce an ICA, and then decided that preemption was proper because "a state commission's failure to `act to carry out its responsibility' under section 252 can in some circumstances include the failure to interpret and enforce existing [ICAs]." Starpower, 15 F.C.C.R. at 11280 (emphasis added).

There is some support for Sprint's argument that Starpower advocates initial State commission consideration of ICA provisions. In Core Communications, the Third Circuit purported to apply Chevron deference to the FCC decision's in Starpower. 493 F.3d at 338-39. In what it acknowledged to be a "broad" reading of Starpower, that court interpreted the FCC's observation that a State commission's responsibility under § 252(e)(6) can "in some circumstances include the failure to interpret and enforce existing [ICAs]," Starpower, 15 F.C.C.R. at 11280, as conferring an exclusive authority on State commissions to resolve disputes over ICAs in the first instance. Core Commc'ns, 493 F.3d at 342.

In its amicus brief in this case, however, the FCC disputed that it had taken such a position in Starpower, describing the Third Circuit's interpretation as "incorrect."18 FCC Amicus Br. at 15. The FCC agrees that Starpower gives State commissions the responsibility to interpret and enforce an ICA "when asked to do so" by the parties, but emphasizes that the decision "does not hold that state commissions are the only entities with that responsibility." Id. Moreover, the FCC argues that reading the Act to grant exclusive authority to State commissions to interpret and enforce ICAs would be "inconsistent with the broad adjudicatory authority that [other sections] of the [1996] Act confer on the FCC and federal district courts." Id. at 11. Finding this interpretation of the proper allocation of decisional authority between federal courts and State commissions consistent with our reading of the 1996 Act, we decline Sprint's invitation to follow the Third Circuit and accord Chevron deference to a position the FCC did not take. Accordingly, we hold that the 1996 Act does not require a State commission to interpret and enforce an ICA in the first instance.

2.

Sprint next argues that even if the 1996 Act does not mandate initial State commission consideration, we should nonetheless impose this step as a prudential exhaustion requirement. Sprint contends that State commissions necessarily bring a level of expertise to the consideration of interconnection issues that federal courts lack. Where, as here, Congress has legislated no explicit exhaustion requirement, we are nonetheless "guided by congressional intent in determining whether application of the [exhaustion] doctrine would be consistent with the statutory scheme." Cavalier Tel., 303 F.3d at 322 (quoting Patsy v. Bd. of Regents of Fl., 457 U.S. 496, 501 n. 4, 102 S.Ct. 2557, 73 L.Ed.2d 172 (1982)). Cognizant of the "virtually unflagging obligation" to exercise that jurisdiction which we possess, Colorado River Water Conservation District v. United States, 424 U.S. 800, 818, 96 S.Ct. 1236, 47 L.Ed.2d 483 (1976), we must use our sound judicial discretion to "balance the interest of the individual in retaining prompt access to a federal judicial forum against countervailing institutional interests favoring exhaustion," Cavalier Telephone, 303 F.3d at 323 (quotation and citation omitted). Here, that balance tips against imposing an exhaustion requirement.

An exhaustion requirement would neither align with congressional intent nor serve one of the exhaustion doctrine's core purposes. As we have noted, the purpose of the 1996 Act was to introduce the benefits of competition into the local telecommunications market. An exhaustion requirement exists in part to promote efficiency. See Woodford v. Ngo, 548 U.S. 81, 89, 126 S.Ct. 2378, 165 L.Ed.2d 368 (2006) ("Claims generally can be resolved much more quickly and economically in proceedings before an agency than in litigation in federal court."). The goals of competition and efficiency would only be disserved by a requirement that the underlying merits of a dispute between Sprint and CenturyLink be considered by multiple individual State commissions with the attendant risk of disparate interpretations and dispositions. Nor are we persuaded that State commissions necessarily possess superior expertise to resolve such disputes. Congress certainly did not agree, as it expressly provided for the FCC to act when a State commission has failed to do so, and for federal courts to review such disputes in any event. For these reasons, we conclude that neither the text of the 1996 Act nor prudential considerations compel federal deference to State commissions in the first instance.

C.

In its second challenge to the district court's authority to decide the merits of this case, Sprint argues that the district court judge's discovery of a financial interest in CenturyLink required recusal and vacatur of all opinions and orders already issued. We review a district judge's recusal decision for abuse of discretion. Newport News Holdings Corp. v. Virtual City Vision, Inc., 650 F.3d 423, 432 (4th Cir. 2011).

Sprint advances two arguments under the judicial recusal statute, 28 U.S.C. § 455. First, in Sprint's view, recusal here was mandatory because the district court judge knew that he had a "financial interest in ... a party to the proceeding." § 455(b)(4). Alternatively, Sprint argues that the district court judge should have recused himself because "his impartiality might reasonably be questioned." § 455(a). We consider each in turn.

1.

Relying in part on the Federal Circuit's decision in Shell Oil Co. v. United States, 672 F.3d 1283 (Fed.Cir.2012), Sprint contends that the 2009 financial disclosure form that the district court judge completed in July 2010 established knowledge of his financial interest in CenturyLink before he took any significant action in this case. See Appellant's Br. at 50. Generally, a "financial interest" is "ownership of a legal or equitable interest, however small ... in the affairs of a party." § 455(d)(4). And "it is well-established that the ownership of stock constitutes a `financial interest'" for purposes of § 455(b)(4). Shell Oil, 672 F.3d at 1289.

Sprint's argument, however, fails to distinguish between direct ownership of securities and ownership of securities in a common investment fund over which a judge exercises no management responsibilities. The judicial recusal statute specifically carves out the latter situation from the definition of a "financial interest": "Ownership in a mutual or common investment fund that holds securities is not a `financial interest' in such securities unless the judge participates in the management of the fund." § 455(d)(4)(i). Congress created this exception to enable judges to hold securities without risking recusal across a broad range of cases. See New York City Dev. Corp. v. Hart, 796 F.2d 976, 980 (7th Cir. 1986) ("When Congress amended § 455 in 1974, it designed § 455(d)(4)(i) as a safe harbor, a way for judges to hold securities without needing to make fine calculations of the effect of a given suit on their wealth."). Moreover, the operation of § 455(d)(4)(i) is "mechanical": "Just as § 455(b)(4) requires disqualification when there is any financial interest, however small, so § 455(d)(4)(i) eliminates any inquiry into the size of the likely effect of a decision on the value of securities held through a mutual fund." Id. (citation omitted).

The safe harbor exception created in § 455(d)(4)(i) applies here. J.A. 514. Although the recusal statute does not define "common investment fund," the core elements, which include assets held together in trust in order to provide "satisfactory diversification" and a "reduction of administrative expenses," see 26 C.F.R. § 1.408-2(b)(5)(ii); see also Federal Tax Coordinator ¶ H-12205 (2d.), 1997 WL 511225 (defining "common investment fund" as "a tax-exempt group trust created to provide a diversification of investments or a reduction of administrative expenses" "into which IRA assets may be commingled"), are found here. The record demonstrates that the district court judge held the CenturyLink shares in an IRA "along with the assets of many others who hold similar accounts." J.A. 514. The record also indicates that "[d]ecisions to buy and sell stocks in the IRA were made by the fund managers without input from the presiding judge." J.A. 128. Thus, the shares at issue here were held in a common investment fund in whose management the district court judge did not participate.19 Accordingly, the district court judge's ownership of shares in CenturyLink does not constitute a "financial interest" in CenturyLink for purposes of § 455(b). See Hart, 796 F.2d at 980 ("Because the underlying assets are not a `financial interest' of the judge it is unnecessary and inappropriate to inquire how a case might affect the value of the fund's assets.").

2.

Sprint alternatively argues that § 455(a) required recusal because the district court judge should have known about his interest in CenturyLink, and a reasonable observer would ascribe such knowledge to him and call into question his partiality. We note at the outset that a judge whose conduct has satisfied the § 455(d)(4)(i) safe harbor will almost certainly have complied with § 455(a) by acting in a reasonable and impartial manner. See Hart, 796 F.2d at 980 (noting that using § 455(a) as a "back door ... inquiry into the substantiality of the effect on the value of assets" held in a managed fund is inappropriate, and holding that "[a] reasonable person would not question the impartiality of a judge who holds nothing but well diversified mutual funds"). Even if one could imagine a scenario where § 455(d)(4)(i) applies but a judge's partiality might reasonably be questioned, such a scenario is not present here. Given the small number of shares the district court judge held, the fact that the CenturyLink shares only came into his portfolio after a series of mergers about which he was unaware, and, as we have noted, that he held the shares in an IRA managed by others, a reasonable observer would have no cause to question his impartiality. His prompt action to inform the parties of his stock when he learned of it, and to divest from CenturyLink shortly thereafter further supports this conclusion.

The district court judge did not violate the recusal statute, and therefore did not abuse his discretion in deciding that neither recusal nor vacatur was appropriate.

III.

Having concluded that the district court properly reached the merits of this case, we now consider whether it decided them correctly. Sprint raises two challenges. First, Sprint asserts that the district court misconstrued the ICA as applying to long distance VoIP traffic. Second, it contends that CenturyLink impermissibly billed Sprint for local calls. When reviewing a district court's judgments after a bench trial, we accept factual findings unless they are clearly erroneous and examine conclusions of law de novo. Plasterers' Local Union No. 96 Pension Plan v. Pepper, 663 F.3d 210, 215 (4th Cir.2011).

As the district court did, we apply Virginia law to CenturyLink's breach of contract claim, Cent. Tel. Co. of Va., 759 F.Supp.2d at 797 n. 4, and North Carolina law to Sprint's counterclaim.20 To succeed on a breach of contract claim, the plaintiff must prove by a preponderance of the evidence that a legally enforceable obligation existed between it and the defendant; that the defendant breached that obligation; and that the plaintiff incurred damages as a result of the breach. See Sunrise Continuing Care, LLC v. Wright, 277 Va. 148, 671 S.E.2d 132, 135 (2009) (citation omitted); Birtha v. Stonemor, N.C., LLC, 727 S.E.2d 1, 9 (N.C.Ct. App.2012). Both CenturyLink's claim and Sprint's counterclaim implicate the first of these elements. We interpret a contract as written and, when its terms are clear and unambiguous, we construe the contract "according to its plain meaning." City of Chesapeake v. States Self-Insurers Risk Retention Grp., Inc., 271 Va. 574, 628 S.E.2d 539, 541 (2006); State v. Philip Morris USA Inc., 363 N.C. 623, 685 S.E.2d 85, 90 (2009).

Where ambiguities arise, however, the "basic contract law principle contra proferentem counsels that we construe any ambiguities in the contract against its draftsman." Maersk Line, Ltd. v. United States, 513 F.3d 418, 423 (4th Cir.2008); Station Assocs., Inc. v. Dare Cnty., 130 N.C. App. 56, 501 S.E.2d 705, 708 (1998) rev'd on other grounds, 350 N.C. 367, 513 S.E.2d 789 (1999). Finally, a settled rule of contract interpretation allows consideration of the "practical construction put by the parties upon the terms of their own contract." First Nat. Exch. Bank of Roanoke v. Roanoke Oil Co., 169 Va. 99, 192 S.E. 764, 771 (1937); see also id. ("No rule for the construction of written instruments is better settled than that which attaches great weight to the construction of the instrument by the parties themselves." (citation omitted)); Century Commc'ns, Inc. v. Hous. Auth. of Wilson, 313 N.C. 143, 326 S.E.2d 261, 264 (1985). Armed with these principles, we turn to Sprint's arguments.

A.

Sprint first argues that the ICA does not apply to long distance VoIP traffic carried over FGD trunks. In its thorough opinion on CenturyLink's breach of contract claim, the district court explained how local calls are subject to "Bill and Keep" reciprocal compensation under § 38.1 of the ICA, while the ICA set out specific access charges for intrastate long distance and interstate long distance traffic. In the district court's view, § 38.4 of the ICA, which provides that "[v]oice calls that are transmitted, in whole or in part, via the public Internet or a private IP network (VoIP) shall be compensated in the same manner as voice traffic (e.g., reciprocal compensation, interstate access and intrastate access)," J.A. 746, applied that existing compensation regime to VoIP traffic.

On appeal, instead of directly challenging this interpretation of the ICA, Sprint argues that the ICA's "plain language" only covers interconnection of local networks, of which FGD trunks are not a part.21 Thus, under Sprint's view, the compensation regime set out in § 38.4 does not apply to VoIP traffic carried over FGD trunks. Appellant's Br. at 35. The "plain language" which Sprint cites includes provisions that (1) define the scope of the ICA as relating to the establishment of "Local Interconnection," § 2.1; (2) set out guidance for the "Local Interconnection Trunk Arrangement," § 37; and (3) define the "Physical Point of Interconnection" as "the physical point that establishes the technical interface, the test point, and the operational responsibility hand-off between CLEC and Sprint for the local interconnection of their networks," § 1.55, see Appellant's Br. at 35. But given that federal law defines interconnection simply as "the linking of two networks for the mutual exchange of traffic," 47 C.F.R. § 51.5, and nothing in the ICA defines it otherwise, none of these provisions advances Sprint's position. Simply put, the ICA is ambiguous as to whether local interconnection includes or excludes long distance VoIP traffic carried over FGD trunks.

This ambiguity,22 along with the parties' previous course of dealings, compels us to reject Sprint's argument. The district court found, and Sprint does not challenge, that it drafted the ICAs at issue here. Had Sprint intended to exclude FGD trunks from the scope of the ICA, a simple provision stating that the established compensation framework for VoIP traffic in § 38.4 did not apply to any VoIP traffic traveling on FGD trunks would have accomplished this goal. Instead, Sprint's argument seeks to cobble together a handful of other ambiguous provisions — and ignore the language of § 38.4 — to support the proposition that VoIP traffic over FGD trunks is excluded from the ICA's coverage. As the district court observed, "only so much can be gained from Sprint referencing other provisions in the ICAs, but ignoring the one provision, Section 38.4, that speaks directly to the issue in dispute — compensation for termination of VoIP-originated traffic." Cent. Tel. Co. of Va., 759 F.Supp.2d at 801. Because we are required to construe ambiguity against the drafter — here, Sprint — we conclude that the ICA must be read to apply to VoIP traffic carried over FGD trunks.

The parties' own longstanding "practical construction" of the ICA bolsters our conclusion. See First Nat. Exch. Bank of Roanoke, 192 S.E. at 771. Sprint paid the CenturyLink plaintiffs for VoIP traffic carried over FGD trunks in accordance with § 38.4 from the execution of the ICAs in 2004 and 2005 until June 2009. Although the fact that the parties were for some of this time both part of Sprint Corporation may explain Sprint's reluctance to challenge this arrangement until the CenturyLink Plaintiffs split off from Sprint Corporation in May 2006, it does not explain Sprint's continued payments until June 2009. The district court's unchallenged factual findings, however, offer an explanation: "In the summer of 2009, Sprint, like many companies at the time, embarked on company-wide cost-cutting efforts. Notably, during this time period, Sprint launched a coordinated effort to contest access charges on VoIP-originated traffic with other carriers across the telecommunications industry." Cent. Tel. Co. of Va., 759 F.Supp.2d at 796. Sprint's change in strategy does not undo the parties' shared interpretation of the ICA over a number of years. When viewed in conjunction with the ambiguity in the ICA's coverage of VoIP traffic over FGD trunks, the parties' course of dealing reinforces our conclusion that the district court did not err in entering judgment for CenturyLink on its breach of contract claim.23

B.

Sprint also challenges the district court's ruling on its counterclaim, which alleges that CenturyLink improperly billed it for local traffic under the NC ICA. Sprint does not dispute that the calls are covered by the ICA, but instead contends that CenturyLink wrongly identified the calls as intrastate long distance, using an impermissible method. Specifically, in Sprint's view, the plain terms of the NC ICA disallowed CenturyLink's use of the Billing Telephone Number (the "BTN") method. We disagree.

As noted above, the "Bill & Keep" compensation regime — under which neither party pays the other — applies to local traffic, whereas the ICA requires Sprint to pay the applicable access charges for intrastate long distance traffic. Thus, while no charge would apply to a call deemed local, Sprint would have to pay access charges for any traffic deemed intrastate long distance. Moreover, there are at least two different methods for determining whether a call is deemed local. The BTN method identifies traffic as local or nonlocal for billing purposes based on a unique account number that is assigned to a specific facility. By contrast, the "Calling Party Number" or "CPN" method identifies the actual originating location of the call in question. In some instances, traffic that the BTN method identifies as intrastate long distance — and therefore subject to access charges — the CPN method would identify as local.

The crux of the question, then, is whether the ICA establishes a method for determining when a call is properly considered local and when it is not. We must therefore determine whether the ICA required CenturyLink to use the CPN method instead of the BTN method it actually used.

In its comprehensive opinion, the district court identified the relevant provisions of the NC ICA at issue. First, § 1.40 defines local traffic under the NC ICA as traffic "that is originated and terminated within Sprint's local calling area." J.A. 795. Importantly, the text of the NC ICA "does not prescribe a specific method of `jurisdictionalizing' traffic as local (subject to `Bill and Keep') or non-local (subject to the applicable access charges) for billing purposes." J.A. 164. Instead, § 42.1 provides that "[e]ach Party shall calculate terminating interconnection minutes of use based on standard AMA recordings made within each Party's network, these recordings being necessary for each Party to generate bills to the other Party." J.A. 830. Section 1.6 of the NC ICA defines "AMA" as "Automated Message Accounting," and notes that "AMA format is contained in the Automatic Message Accounting document published by Telcordia as GR-1100-CORE which defines the industry standard for message recording." J.A. 792. The parties do not dispute that the industry manual known as the Telcordia GR-1100-CORE permits use of the BTN method.

Both Sprint and CenturyLink invoke the express language of the NC ICA in support of their respective positions. Sprint claims that § 1.40's definition of traffic that "is originated and terminated within [a] local calling area," J.A. 795, required CenturyLink to use the Calling Party Number or "CPN" method to jurisdictionalize the originating and terminating points of a call. According to Sprint, application of the CPN method to the calls at issue would transform them from long distance into local calls. And to construe the NC ICA as permitting the BTN method, Sprint contends, would lead to absurd results — akin to treating a cab ride that began and ended in Richmond but traveled along Interstate 95 as "originating" anywhere along that highway, even as far afield as Portland, Maine. CenturyLink, by contrast, argues that § 42.1 in conjunction with § 1.6 incorporated by reference the industry standard set out in the Telcordia GR-1100-CORE document, and therefore explicitly permitted use of the BTN method. The district court credited the latter view. So do we.

Sprint's argument that the NC ICA disallowed use of the BTN method suffers from an insurmountable flaw: the NC ICA nowhere explicitly so provides. As the district court observed, Sprint's reliance on the definition set out in § 1.40 overlooks the more rudimentary question of how a carrier can determine where a call has originated and terminated. At the time the parties agreed to the NC ICA, CenturyLink only had the capability to use the BTN method, and although the CPN method existed, no provision in the NC ICA required that CenturyLink adopt this method at any point in the future. Sprint's argument that the NC ICA only incorporated the Telcordia document for the purpose of calculating minutes, see Appellant's Br. at 47-49, finds no grounding in the text of the relevant provisions. Accordingly, we agree with the district court's conclusion that the NC ICA permitted CenturyLink to identify the origination and termination points of calls using the BTN method, and therefore to bill Sprint at the applicable access charges for those calls identified as nonlocal under that method.

At best, Sprint's arguments render ambiguous the propriety of the BTN method for identifying calls as local under the NC ICA. But as we explained in the context of CenturyLink's breach of contract claim, Sprint must do more than identify an ambiguity in a contract it drafted. In the face of ambiguity, we construe the relevant provisions of the NC ICA against Sprint and in favor of CenturyLink.

IV.

For the reasons discussed above, we affirm.

AFFIRMED